**A Review on Individual and Combination Technologies of UV-C Radiation and Ultrasound in**

### **Okon Johnson Esua 1,2, Nyuk Ling Chin 1,\*, Yus Aniza Yusof <sup>1</sup> and Rashidah Sukor <sup>3</sup>**

**Postharvest Handling of Fruits and Vegetables**


Received: 24 September 2020; Accepted: 29 October 2020; Published: 10 November 2020

**Abstract:** Ultraviolet-C radiation and ultrasound technology are widely accepted and continuously being appraised as alternatives to conventional thermal techniques for decontamination of fruits and vegetables. However, studies in these areas have presented challenges related to quality, safety, limited capability, and cost of energy. This review paper presents an up-to-date summary of applications of ultraviolet-C radiation and ultrasound technology for postharvest handling of fruits and vegetables from relevant literature. The limitations associated with applications of ultraviolet-C radiation and ultrasound technology individually has prompted their combination alongside other antimicrobial strategies for enhanced bactericidal effect. The combination of ultraviolet-C radiation and ultrasound technology as a hurdle approach also provides enhanced efficiency, cost effectiveness, and reduced processing time without compromising quality. The review includes further scope of industrial-led collaboration and commercialization of ultraviolet-C radiation and ultrasound technology such as scale-up studies and process optimization.

**Keywords:** deterioration; cavitation; dosage; hurdle technology; microorganisms; nonthermal; decontamination

#### **1. Introduction**

Fruits and vegetables are an essential component of our diet and their consumption increases yearly due to perceived interest of consumers in healthy and functional foods [1,2]. However, they are susceptible to postharvest moisture loss, improper handling, mechanical damage, and microbial contamination. Postharvest moisture loss affects maturation, which can be inferred from evaluation of important quality parameters such as weight, texture, acidity, sugars, carotenoids, vitamins, and phenolic compounds. Microbial contamination is accompanied by postharvest losses and prevalence of foodborne disease outbreaks. Data suggest that about one-third of food (1.3 billion tons) fit for human consumption is wasted yearly, with 44% attributed to fruits and vegetables, while numerous foodborne illnesses and deaths linked to fresh produce contamination have been reported in the last two decades [3–6]. Food losses and wastes pose serious concern at a global level; hence, the United Nations agenda has initiatives with development investments towards preventing and reducing postharvest losses [6,7].

The common practices to reduce spoilage and enhance shelf life of fruits and vegetables include washing with sanitizers, refrigeration, controlled atmosphere storage using natural and synthetic preservative agents, and drying [8–14]. The majority of these approaches are marginally effective and may alter nutritional properties while the corrosive nature of some sanitizers has been linked with environmental and health concerns [15,16]. As consumer demands for minimally processed products with fresh-like characteristics has increased in recent times, this preoccupation has encouraged continuous evaluation of non-thermal technologies for postharvest handling of fruits and vegetables.

The non-thermal technologies of ultraviolet-C (UV-C) radiation and ultrasound (US) technology have been extensively studied for surface decontamination of fruits and vegetables, and are considered highly efficient, non-toxic, and environmentally friendly [17]. UV-C radiation is generated at wavelengths of 250–280 nm and is reported to disrupt functionality and integrity of microorganisms' DNA in addition to being linked with generation of reactive oxygen species that regulates physiological processes to induce secondary metabolite production [18]. On the other hand, US technology is a form of pressure waves beyond human hearing (>20 kHz), and high intensity US at low frequencies of 20–100 kHz can induce acoustic cavitation that generates reactive hydroxyl radicals associated with microbial inactivation and stimulation of secondary metabolite accumulation in fruits and vegetables [17,19,20].

Nevertheless, UV-C radiation can only penetrate up to several millimeters and US technology consumes high energy coupled with long treatment time, which limits their applications. This has encouraged the use of UV-C radiation and US technology in combination with other techniques for postharvest handling of fruits and vegetables with encouraging results [1,19,21–23]. The current review, drawn from selected past and current literature, presents research progress of the individual and combination approaches of UV-C radiation and US technologies in postharvest handling of fruits and vegetables. The types of disinfecting systems typically utilized for postharvest UV-C radiation and US technology decontamination of fruits and vegetables is also summarized. This review also included discussion on limitations associated with individual applications of UV-C radiation and US technology for postharvest decontamination of fruits and vegetables. The potential of a hurdle technology utilizing UV-C radiation and US technology in combination or simultaneously for maximizing and harnessing their advantages is also presented.

#### **2. Applications of UV-C Radiation in Postharvest Handling of Fruits and Vegetables**

#### *2.1. Disinfecting Systems and Factors A*ff*ecting E*ffi*ciency*

The use of UV-C radiation for food preservation was discovered in the 1930s, and the first studies on fruits and vegetables was recorded in 1977 when resveratrol and viniferins were induced in grape vine from UV-C exposure [24,25]. The efficacy of UV-C radiation for decontamination of fruits and vegetables is heavily dependent upon the type of disinfection systems, mode of applications, pulse intensity, morphology and location of microorganisms on fruits and vegetables, distance from irradiation source, and type of microorganism.

Typical systems for exposing fruits and vegetables to UV-C radiation consist of self-contained chambers with fluorescent germicidal lamps and have been widely reported in the literature. However, modifications to the conventional self-contained chambers have been reported to induce improved microbial reductions and maintain quality properties even at low UV-C radiation doses. For instance, Collazo et al. [23] showed that low doses of 0.1–0.3 kJ/m<sup>2</sup> from water-assisted UV-C radiation equipment with pressurized water sprinklers were enough to reduce respiration, maintain quality, and provoke 0.9–2.0 log CFU/g (log colony forming unit/g) reductions of *Listeria monocytogenes* (*L. monocytogenes*) and *Salmonella enterica* (*S. enterica*) in lettuce and spinach (Figure 1a). The study revealed that dual action of irradiation by immersion and simultaneous actions of process water from modification of the conventional treatment chambers with pressurized water sprinklers accounted for higher efficiency. Yan et al. [26] demonstrated that it is possible to reduce large variations and maximize UV-C radiation dose on fruit surfaces in disinfecting systems. In this study, conveyor belts were placed at a distance of 14 cm from UV-C lamps to convey and rotate fruits between two treatment chambers connected by an inclined belt (Figure 1b). This inclusion to the conventional UV-C lamp self-contained

chambers ensured dose uniformity such that fruits received an average dose of 1 kJ/m2, compared with 0.2 kJ/m<sup>2</sup> received by fruits without rotation. The modified rotation device ensured 1.3–1.8 and 1.0–1.2 log CFU/fruit reductions in *Escherichia coli* O157:H7 (*E. coli* O157:H7) and *E. coli* ATCC 25922, respectively, compared with 0.5–0.7 log CFU/fruit reduction obtained without the rotation device, revealing the importance of dose uniformity for maximum effect of UV-C radiation in disinfection systems. Likewise, a benchtop UV-C radiation system that allowed upward and downward movement of fruits from a rotating roller was employed by Taze and Unluturk [27] to ensure uniformity of intensity (Figure 1c). They observed 3.0 and 2.38 log CFU/g reductions in total aerobic mesophilic bacteria and yeast and mold, respectively, at doses of 31.01 and 7.75 kJ/m2, and reported that light intensity was almost constant at the lamp center and lower at both ends. Ensuring the uniformity of exposure on food surfaces due to the low penetrative ability of UV-C radiation is seen as the most important challenge to commercialization, and these studies have provided an avenue to ensure dose uniformity for maximum impact.

The applied dosage of UV-C radiation, the type of microorganism, and their location on fruits and vegetables have been reported to have significant effect on decontamination efficiency of UV-C radiation. Mukhopadhyay et al. [28] showed that increasing UV-C radiation doses from 0.6 to 6.0 kJ/m<sup>2</sup> resulted in 2.2–3.1 and 2.3–3.5 log CFU/fruit reductions, respectively, for cocktail mixtures of *S. enterica* and *E. coli* O157:H7 on the surfaces of tomato. Lower reductions of 1.9–2.8 and 1.7–3.2 log CFU/fruit, respectively, were recorded on the stem scar under the same conditions, indicating that specific location of microorganisms on produce may influence efficacy of treatment and should be considered during the design of UV-C radiation disinfection systems. Further analysis showed that color and texture of tomato were not affected by the increasing dosages. In contrast, Lim and Harrison [29] observed a reduction range of 3.22–4.39 log CFU/g in *S. enterica* for a dosage range of 0.223–1.785 kJ/m2, irrespective of the contamination location on tomato placed in a close end reflector lined with aluminum foil. The high reduction from such low doses was attributed to the bacterial strain and low initial inoculum level utilized in the study, which corroborates the assertion that the sensitivity of bacteria to UV-C radiation may vary among species and strains of the same species.

The mode of application of UV-C radiation, pulse intensity, and distance of samples from irradiation source have also been reported to influence the actions of UV-C radiation during decontamination of fruits and vegetables. For example, short pulses of incident UV-C irradiance was not effective in reducing microbial population, which could be linked to ability of resistant strains of microorganism to produce protein protection mechanism against UV-C radiation treatment, especially at short treatment duration [30]. The study showed that there was no significant difference in log reductions (1.0 log CFU/g) of resistant strains of *Enterococcus faecalis* on zucchini squashes when a UV-C radiation system was operated in continuous incidence radiation of 0.011 kJ/m<sup>2</sup> or pulse irradiation of 0.067 kJ/m<sup>2</sup> in a discontinuous fluence. In addition, a continuous UV-C incidence radiation of 0.011 kJ/m2 for 90 s did not have any significant effect on the reduction of *Deinococcus radiodurans* on zucchini squashes. However, higher inactivation of *E. coli* was reported on tomatoes and lettuce when the intensity of UV-C radiation was increased from 6.5 to 16 W/m2 in a treatment cabinet [31]. Enhanced reductions of 2.8 and 1.7 log CFU/g were also observed in tomatoes and lettuce, respectively, for samples closer to the radiation source at 31 cm compared with samples placed at a distance of 70 cm. In the same vein, Cote et al. [32] observed that increasing intensity for a given dose could maximize the benefits of UV-C radiation. For an applied dosage of 4 kJ/m2, only 5% of tomatoes irradiated with higher intensity of 33 W/m<sup>2</sup> showed incidences of postharvest rot caused by *Botrytis*, compared with 8% from lower intensity of 3 W/m2, after 9 days of storage. On the other hand, 12% of strawberries irradiated with higher intensity showed decay symptoms after 5 days, compared with 46% of strawberries irradiated with lower intensity. The treatment also delayed ripening and maintained better quality of the fruits, while titratable acidity, soluble solids, and antioxidants properties were unaffected.

**Figure 1.** Disinfecting systems for UV-C radiation decontamination of fruits and vegetables. (**a**) Water-assisted UV-C set-up. Reproduced with permission from Collazo et al., International Journal of Food Microbiology; published by Elsevier, 2019. (**b**) Schematic diagram of a commercial UV-C treatment chamber. Reproduced with permission from Yan et al., Postharvest Biology and Technology; published by Elsevier, 2014. (**c**) Benchtop UV-C set-up with (i) tray system, (ii) rotating roller bearing, and (iii) UV-C lamps and cooling system. Reproduced with permission from Taze and Unluturk, Scientia Horticulturae; published by Elsevier, 2018.

The influence of fruit surface characteristics on the inactivation efficiency of UV-C radiation has also been studied, with results demonstrating that microorganisms on fruits' surfaces responded differently to UV-C radiation [33]. It was observed that microbial inactivation was lower for fruits with rougher surfaces (strawberry, cantaloupe, and raspberry) compared to less hydrophobic fruits with smoother surfaces (pear and apple). Reductions of 2.9 and 2.1 log CFU/g were recorded for *E. coli* O157:H7 on apple and pear surfaces, respectively, at 0.9 kJ/m<sup>2</sup> for 60 s compared with 2.0 and 1.1 log CFU/g recorded for strawberry and raspberry at 7.2 and 10.5 kJ/m<sup>2</sup> for 480 and 720 s, respectively. On the other hand, *L. monocytogenes* showed more resistance than *E. coli* O157:H7, revealing 1.6 and 1.7 log CFU/g reductions, respectively, on apple and pear at 3.75 kJ/m2 and 11.9 kJ/m2. The log reduction of *L. monocytogenes* was 1.0 log CFU/g on both cantaloupe and strawberry at 11.9 kJ/m2. The findings from this study were corroborated by Syamaladevi et al. [34] by using a similar device when they observed that trichomes on peach surfaces and wounds on pears protected and shielded microorganisms from irradiation, while smaller surface roughness and spreading coefficient of pear surfaces compared with peach likely caused a more uniform distribution of microbial cells on pear surfaces. Maximum reductions of 3.70 log CFU/g were recorded for *E. coli* on intact pear surfaces, with lower reductions of 3.10 and 2.91 log CFU/g achieved on wounded pear and peach surfaces, respectively, after 4 min at 7.56 kJ/m2.

#### *2.2. E*ff*ects of UV-C Radiation on Microorganism Inactivation and Quality of Fruits and Vegetables*

Table 1 lists studies that have used UV-C radiation to inactivate microorganisms and subsequent effects on quality parameters for a wide variety of fruits and vegetables. Collectively, these studies demonstrate the potential of UV-C radiation at low doses of 0.00176–1 kJ/m<sup>2</sup> for enhancing bioactive compound production, reducing enzymatic activities, extending shelf life, and having degradative effects on biological process-regulating proteins responsible for deteriorative changes of postharvest produce [33,35–39]. The range of doses were also responsible for 1.2–2.9 log CFU/g reduction in *E. coli* O157:H7, *Salmonella* spp., and *Listeria* spp., as well as reduced rot and decay development. In contrast, Wang et al. [40] observed that a low dose of 0.3 kJ/m2 had no effect on total bacteria number and browning-related enzymes of fresh cut lotus during storage, and resulted in a high degree of browning, but the authors maintained that a moderate dose of 1.5–3 kJ/m<sup>2</sup> significantly retarded microbial growth and suppressed the activities of browning-related enzymes. Pinheiro et al. [41] also reported that low doses of 0.97 kJ/m2 was not adequate for metabolic degradation delay in tomatoes but presented 2.1 log CFU/g reduction in mesophilic load, delayed red color development, and caused increases in firmness and phenolic content. The variations in results could be attributed to the factors discussed earlier, in addition to the type of disinfection systems used.

It is apparent that increasing dosage applications could result in increased microbial reduction. For instance, Sommers et al. [42] observed a progressive increase from 2.59 to 3.79 log CFU/g in the inactivation of *L. monocytogenes*, *Salmonella* spp., and *Staphylococcus aureus* (*S. aureus*) on Roma tomatoes and jalapeño peppers when the dosage was increased from 5 to 40 kJ/m2. Similarly, successive increases in UV-C radiation doses evoked greater reductions of *E. coli*, *S. aureus*, *S.* Enteritidis, and *Listeria innocua* (*L. innocua*) on tomato, lettuce, and strawberry, with no modifications in texture and appearance [19,43]. Doubling UV-C radiation dose from 32 to 72 kJ/m2 also doubled the reduction of human adenoviruses from 0.92 to 2.22 and 1.26 to 3.98 logs for tomato and strawberry, respectively.



#### *Processes* **2020**, *8*, 1433



*Processes* **2020** , *8*, 1433



bacteria, APB: aerobic

psychrotrophic

 bacteria.

The synergistic or additive effects of combining UV-C radiation with other techniques for improved efficiency have also been studied. Enhanced reductions of 3 log CFU/g were recorded in the populations of *E. coli* and *S.* Enteritidis when UV-C radiation was combined with either neutral electrolyzed water or peroxyacetic acid for broccoli decontamination, compared with 1.3–2.2 log CFU/g obtained for individual application of UV-C radiation at 7.5 kJ/m2 [11,47]. Combination treatment also maintained quality, retained bioactive and fatty acid contents, and was seen as a potential eco-friendly alternative to conventional NaClO application. Similar enhanced reductions in *E. coli* on blueberry, and *L. monocytogenes*, *E. coli*, and yeast and mold on cauliflower were also reported when low UV-C doses were combined with ozone and antimicrobial formulations [13,46]. The effects of combining UV-C radiation and ClO2 gas was also observed to be significantly greater than the sum of individual inactivation for the decontamination of spinach and tomato [16]. Individual UV-C radiation application at 0.848 kJ/m<sup>2</sup> reduced *E. coli* O157:H7 and *Salmonella* Typhimurium by 1.85 and 2.02 log CFU/g, respectively, while combined treatment reduced the population by 5.17 and 5.41 log CFU/g on spinach and 5.62 and 5.46 log CFU/g on tomato. Enhanced microbial reduction was attributed to cell membrane damage and changes to membrane permeability as a result of the synergistic actions of UV-C radiation and ClO2 gas. This position was also maintained by Kim and Song [58] during the sequential application of ClO2 gas and UV-C radiation on plum. The synergistic effect produced 4 and 3.27 log CFU/g reductions in *L. monocytogenes* and *E. coli* O157:H7, respectively, compared with 1.62 and 2.07 log CFU/g reductions obtained for individual UV-C radiation application. The authors however upheld that sequential applications of ClO2 gas, UV-C radiation, and fumaric acid, which produced 5.48 and 6.26 log CFU/g reductions, respectively, in *E. coli* O157:H7 and *L. monocytogenes*, was the most effective combination treatment.

#### **3. Applications of US Technology in Postharvest Handling of Fruits and Vegetables**

#### *3.1. Disinfecting Systems and Factors A*ff*ecting E*ffi*ciency*

The antimicrobial effect of US technology was first documented 80 years ago, but its prospective practical application for microbial inactivation began in the 1960s [63]. The chemical reactivity of a system is affected by US technology through actions of cavitation and sonolysis from mechanical vibrations, typically produced from transformation of high-frequency electrical energy by piezoelectric transducers in probe or tank systems [15,19]. The threshold of cavitation bubble in the tank or probe systems is usually influenced by US frequency, power, intensity, and exposure time. The sonochemist typically uses frequencies of 20–50 kHz, as it is easier to obtain cavitation at these frequencies [64]. Together with the type of microorganism, mode of application in terms of pulse or continuous modes, and dual-mode or mono-mode frequency, these factors affect the efficiency of US technology during decontamination of fruits and vegetables.

For the US probe systems, a pulse mode application of 10 s on and5soff from a 15 mm diameter, 20 kHz, 400 W, and 226 W/cm2 probe system submerged in water (Figure 2a) was capable of reducing initial total number of colonies and yeast and mold on cucumber by 0.49–1.02 and 0.41–0.84 log CFU/g, respectively, with increasing treatment time from 5 to 15 min under modified atmosphere packaging storage [65]. Treatment up to 10 min effectively reduced losses of firmness, ascorbic acid, color, soluble solids, and weight, and maintained cell wall integrity. Further increase in treatment time led to deterioration of quality properties. Similarly, Millan-Sango et al. [66] reported 2.56 log CFU/cm2 reduction in *E. coli* O157:H7 on lettuce leaves at pulse mode application of 10 s on and 6 s off from a 14 mm probe, 26 kHz, 90 μm, and 200 W US system operating in both continuous and pulsed mode configurations. Results showed no significant differences between continuous and pulsed mode applications for up to 25 min of application, suggesting that decontamination efficacy was dependent upon treatment time rather than the configuration of the probe system utilized.

**Figure 2.** Typical disinfecting systems for ultrasound (US) technology decontamination of fruits and vegetables. (**a**) Schematic representation of US probe system. Reproduced with permission from Fan et al., Ultrasonics Sonochemistry; published by Elsevier, 2019. (**b**) Schematic diagram of US vacuum impregnation system. Reproduced with permission from Yilmaz and Bilek, Ultrasonics Sonochemistry; published by Elsevier, 2018. (**c**) Continuous-flow US bath equipment depicting (i) different transducers T1, T2, and T3, and (ii) washing area, W. Reproduced with permission from Zhou et al., Innovative Food Science and Emerging Technologies; published by Elsevier, 2012.

Increasing US intensity in a probe system has also been reportedly associated with enhanced antimicrobial activities while higher intensities had negative effects on product quality during storage. For example, reduced decay incidence of gray mold on strawberry with increasing intensity from 10.6–31.8 W/cm<sup>2</sup> was reported by Aday et al. [67] from a 1.9 cm diameter probe system operating at 20 kHz. Results also show that pH, color, soluble solids, and sugar content of strawberry treated with 10.6–21.2 W/cm2 were better when compared with those treated with 31.8 W/cm2, indicating that higher US intensity was detrimental to the quality of strawberry. In the same vein, Wang et al. [68] showed that the population of total bacteria and yeast and mold on tomato reduced to a range of 0.42–1.04 and 0.41–0.93 log CFU/g, respectively, with increasing US intensity from 66.64 to 145.74 W/L. Likewise, lower intensity levels inhibited ethylene production and respiration rates, and maintained firmness and antioxidant properties, while higher intensity had negative effects on tomato quality, demonstrating that appropriate US intensity can inhibit decay and maintain nutritional properties and flavor of tomato during storage.

For the US bath systems, research has demonstrated that it is more challenging to obtain cavitation at higher frequencies, with US penetration inversely proportional to frequency, while dual-mode frequency can be associated with improved cavitation when compared with mono-mode frequency applications. In this manner, Zhou et al. [63] fabricated a closed-tank continuous-flow system that allowed for frequency and nominal power variation and observed that 25, 40, and 75 kHz transducers dissipated 95, 85, and 45% rated power, corresponding to intensities of 79.41, 68.95, and 42.36 W/L, respectively (Figure 2b). The use of agitation ensured spatial uniformity of US treatment, and the populations of *E. coli* O157:H7 on spinach leaves were reduced by 4.45, 4.21, and 2.42 log CFU/g, respectively, from 25, 40, and 75 kHz transducers, suggesting that non-uniform cavitation from ultrasonic field variations may contribute to variations in microbial inactivation and cross-contamination in the US bath system. In a recent study, Mustapha et al. [20] showed that the application of dual-mode frequency of 20/40 kHz during US treatment of tomato presented higher microbial reductions of 1.3–2.6 log CFU/g for mesophilic bacteria and molds and yeasts, compared with <1 log CFU/g obtained when mono-mode frequency of 20 or 40 kHz was applied. There were no

detrimental effects on the bioactive compounds and physicochemical properties of tomato during storage, substantiating the enhanced cavitation characterized by dual-mode frequency applications.

It has also been demonstrated that substances can be incorporated into samples from the US bath system for better effects. Yilmaz and Bilek [69] combined vacuum impregnation and US (72–840 W, 35 kHz) in custom-made equipment, observing that calcium and black carrot phenolics can be incorporated into apple tissues to produce natural colorant-fortified products without disturbing cellular integrity (Figure 2c). Treatment reduced psychrophilic and mesophilic bacteria on apples by 1.0–2.6 log CFU/g during storage and led to increases in firmness and bioactive compounds, while higher power levels resulted in apple cell rupture, which was evident from increased ion leakage in apple tissues. Wiktor et al. [70] showed that immersion treatment from a 40 kHz, 180 W bath system presented apples with lower mass loss of 0.5–1.2%, dry matter of 0.115 kg dm/kg, higher bioactive compound contents, and less color change when compared with contact treatment from a 24 kHz, 100% amplitude ring sonotrode probe system. The surrounding water during immersion treatment played a significant role in water gain and cooling of apples, hence lower weight loss, while localized oxygen on apple tissue pores and enhanced enzymatic activities during contact treatment could degrade bioactive compounds. Although this may appear to suggest that treatment using the US bath system has shown a better quality of sample properties when compared with the US probe system, this does however need more detailed and careful comparison in terms of same usage of US energy frequency and intensity.

#### *3.2. E*ff*ects of US Technology on Microorganism Inactivation and Quality of Fruits and Vegetables*

Table 2 presents a summary of some studies that have applied US technology to inactivate microorganisms and subsequent effects on quality parameters for a number of fruits and vegetables. The majority of the applications were carried out in combination with sanitizers so as to avoid possible cross-contamination, since microorganisms are usually released into the wash water during US decontamination. However, among the individual US applications, Gani et al. [71] reported 2.0 and 1.2 log reductions in bacterial counts and yeast and mold, respectively, in addition to retention in color, firmness, vitamin C, and antioxidant activity. Prolonged exposure time after 40 min, 33 kHz, and 60 W did not further extend shelf life but may have caused injuries to strawberry samples. In another study, the action of US at 25 kHz and 26 W/L was reported as similar to the actions of abiotic stress such as wounding in triggering and eliciting defense systems in Romaine lettuce [72]. A response mechanism was suggested for such actions (Figure 3a), and treated lettuce displayed enhanced antioxidant activity, higher firmness, delayed enzymatic browning, and absence of deterioration during storage. Preservative treatments are usually known to be associated with the degradation of important freshness-related attributes that define the quality of fresh produce. This perception needs to be critically considered, especially the effects on health-stimulating phytonutrients and other related compounds during the design and selection of processing parameters.

For combination applications, the synergistic effects of US and chemical sanitizers, surfactants, organic acids, and electrolyzed water for improved decontamination efficiency and enhanced quality attributes have been reported [10,73–84]. Increased efficiency was observed without synergistic effect, and also with both synergistic and antagonistic effects in some cases. Park et al. [79] reported antagonistic effects against *Cronobacter sakazakii* when US at 37 kHz and 380 W was combined with 150 ppm of NaOCl. Although the plausible reason for antagonistic effect was not clear, synergistic combination mostly eliminated bacteria cells from the surface of lettuce and in the stomata, as observed in the scanning emission spectroscopy (Figure 3b), yielding 4.44 log CFU/g reduction of *C. sakazakii* compared with 0.01–2.71 log CFU/g reduction for individual applications of US and NaOCl. These studies demonstrated that US alone was not greatly effective, even with long treatment time, while the cavitation effect of US detached microorganisms from fresh produce, especially at inaccessible regions, increasing their susceptibility to sanitizers. Thus, the chemical composition of the liquid in the US system can influence decontamination actions. The US system also promoted the rapid dispersion

of organic matter, which increased their contact time with sanitizer constituents, leading to minimal residues in post-treatment solutions. The presence of microorganisms in treatment solutions were also very low and, in some cases, nonexistent, indicating that US and adequate sanitizer concentrations could prevent and/or reduce cross-contamination.

**Figure 3.** (**a)** Proposed mechanism of secondary metabolism response in Romaine lettuce during US. Reproduced with permission from Yu et al., Food and Bioprocess Technology; published by Springer Nature, 2016. (**b**) Field emission scanning electron microscopy of lettuce head showing aggregation and elimination of *Cronobacter sakazakii* during treatment: (i) control, (ii) US treatment, (iii) NaOCl treatment, and (iv) combined US and NaOCl treatment. Reproduced with permission from Park et al., Food Control; published by Elsevier, 2016.


 **2.** List of fruits and vegetables treated with US technology.

**Table**




**Table 2.** *Cont.* VI: vacuum impregnation, BPW: buffered peptone water, PAL: phenylalanine ammonia lyase, PA: peroxyacetic acid, SA: salicylic acid, POD: peroxidase, SAEW: slightly acidicelectrolyzed water, NNEW: near neutral electrolyzed water, MAP: modified atmosphere packaging, HAdV: human adenovirus.

The systems for nano-ZnO coating application on kiwifruit after US treatment have also been evaluated [89]. Although individual US treatment of kiwifruit was somewhat effective, sequential application of US at 40 kHz, 350 W, and 1.2 g/L nano-ZnO coating was more effective in senescence delay, mass and firmness loss reduction, and storage life extension. The addition of mild heat to US treatment have also been explored, where a more pronounced effect of US (35 kHz, 120 W, and 21.4 W/L) was reported at 65 ◦C on red bell pepper, watercress, and strawberry, with improved quality retention and enhanced reductions of 7.43, 8.24, and 8.10 log CFU/g for *L. innocua*, total mesophiles, and total coliforms, respectively [56]. Similarly, the efficacy of US at 40 kHz, 400 W/L combined with SAEW was also enhanced at 40 and 60 ◦C, respectively, for fresh cut kale and bell pepper [21,86,87]. Browning and off-odor were under acceptable limits for fresh cut kale, while hardness and color parameters for control and treated samples were not significantly different for bell pepper. Enhanced bactericidal efficacy led to 3.32, 3.11, 3.00, 3.97, 3.66, 3.62, and 3.24 log CFU/g reductions in *E. coli*, *L. monocytogenes*, S. *Typhimurium,* total bacterial counts, *Enterobacteriaceae*, *Pseudomonas* spp., and yeast and mold counts, respectively.

#### **4. Combination of UV-C Radiation and US Technology**

Some studies have begun to examine the development of hurdle techniques involving UV-C radiation and US technology, either in sequential or simultaneous application. Sequential application involves treatment with one technology immediately followed by others and has shown improved microbial reductions and minimized quality changes in zucchini squash [26]. Alternative blanching treatments of UV-C radiation (1.17–8.2 kJ/m2) followed by thermosonication (20 kHz, 90 ◦C) induced 3 log reductions in *Enterococcus faecalis* and *Deinococcus radiodurans*, compared with 1 log reduction achieved from individual treatment with UV-C radiation. Birmpa et al. [19] demonstrated that sequential application of US at 30 W/L followed by UV-C radiation at 24 kJ/m2 evoked better log reductions (2–3 log) of human adenovirus (hAdVs) with less processing time, compared with individual results of UV-C radiation (2.13, 0.92, 1.25 log) and US (0.85, 0.36, 0.53 log) on lettuce, cherry tomato, and strawberry, respectively.

Simultaneous application, on the other hand, applies both technologies at the same time with enhanced results in disinfection efficacy and functional properties. A constant US intensity of 13.87 W/L from a 40 kHz, 1 kW transducer and UV-C radiation dosage of 0.72–10.76 kJ/m<sup>2</sup> revealed increasing reduction in the population of total aerobic bacteria on tomato and up to 2.33 log reduction, while yeast and mold were undetected at higher dosage levels [22]. Enhanced biosynthesis and extractability of phytochemicals were also reported in tomato, with 36, 60, 30, and 90% increases in antioxidant activities, vitamin C, total phenols, and lycopene contents, respectively [1,92]. The combined effects of UV-C radiation and US cavitation were observed to disrupt the structural integrity of macromolecules and chemical bonds of tomato phytochemicals, which stimulated their accumulation and facilitated their extraction. Collectively, these few studies present the potential of a hurdle technology involving UV-C radiation and US technology for the postharvest decontamination of fruits and vegetables, with this area of focus warranting further exploration.

The theory elucidating the mechanism of the synergy between US technology and UV-C radiation can be explained from their individual mechanisms of operation. It is possible that the resultant mechanical effect of the collapse of transient cavitation from US technology, which includes high heating and cooling rate (10<sup>9</sup> K/s), shock wave formation, high temperature (5000 K), and pressure (1000 atm), ruptures microbial cell envelopes [93]. The ruptured cell envelopes then facilitate access of UV-C radiation to light-sensitive DNA of the microorganisms within a short processing time. The formation of cyclobutane thymine dimmers from the action of UV-C radiation combined with chemical effects of OH- and H<sup>+</sup> species formation from US technology eventually offers a more potent bactericidal effect that is capable of depleting microbial growth, leading to the eventual death of microorganisms [42,93]. However, this theory needs to be properly tested from more studies and relevant tools.

#### **5. Future Perspectives and Industrial Relevance**

The low penetrating capacity of UV-C radiation into opaque substances, high energy consumption, and long treatment time associated with US technology has limited their individual usage. Thus, research in the applications of UV-C radiation and US technology for postharvest decontamination of fruits and vegetables is focused on hurdle technology of combination applications, which have shown enhanced decontamination efficiency. The hurdle technology has shown potential of overcoming limitations and harnessing advantages of each individual technology. As studies reported on this hurdle technology are relatively few, it is necessary to have more validation towards achieving industrial-scale applications. Scale-up studies to larger proportions would reveal important considerations to ensure uniformity of US cavitation and UV-C exposure on food surfaces due to variations in ultrasonic field and non-penetrative ability of UV-C radiation, which have been observed as the most important challenges to the commercialization of US technology and UV-C radiation for fresh produce decontamination [63,94]. The use of agitation in free-flowing US baths and application of mixed-mode US frequencies can reduce standing-wave effects and creation of different cavitation bubble size to ensure uniform cavitation [20,63]. Wide variations in UV-C radiation intensity can be reduced so that energy reaches the samples more effectively [26]. The radiochromic film dosimetry approach could also be introduced to UV-C treatment to measure or evaluate exposure of radiation dose. Film dosimetry is reportedly associated with measuring and verifying the dosage level on food substances during irradiation [94,95]. Once non-uniformity of irradiation or wide variations in dose is established, a system to rotate irradiated samples as described by Yan et al. [26] can be introduced to ensure UV-C dosage uniformity on the samples for improved efficiency, especially in high loading industrial applications. For combination approaches of US technology and UV-C radiation, we can incorporate sanitizer solutions as the washing medium in US baths lined with UV-C radiation lamps to enhance efficiency and reduce cross-contamination. The heat generated from simultaneous combination energies, which may cause dehydration and damage the appearance of fresh produce, may be overcome by incorporating a water circulatory system that simultaneously introduces and drains water in the tank to help control temperature rise and ensure minimal cross-contamination. Identification of important factors influencing the decontamination process can be enhanced through better understanding of microbes' behavior where more accurate inactivation of target microorganisms can assist optimization of the process parameters for cost-effective practical applications.

#### **6. Conclusions**

The research on the postharvest decontamination of fruits and vegetables is rapidly evolving due to increased safety awareness and consumer demands for minimally processed foods with fresh-like characteristics. UV-C radiation and US technology have presented encouraging results in postharvest decontamination of fresh produce with minimal effects on quality characteristics. However, long exposure time, high US energy consumption, and limited penetration capability of UV-C radiation have plagued research in these areas. The potential for synergistic or complementary effects for improved efficiency have also been established when UV-C radiation or US technology is applied in combination with other antimicrobial techniques. The hurdle technology involving UV-C radiation and US technology offer substitute benefits towards ensuring overall safety and wholesomeness of fruits and vegetables in comparison with individual applications or their combinations with conventional approaches. More studies are required on this hurdle technology, while scale-up and optimization of the process will provide opportunities for industry-led collaborations and chart the path towards effective commercialization.

**Author Contributions:** Conceptualization, O.J.E. and N.L.C.; investigation, O.J.E.; resources, O.J.E. and N.L.C.; data curation, O.J.E.; writing—original draft preparation, O.J.E.; writing—review and editing, N.L.C.; supervision, N.L.C., Y.A.Y., and R.S.; project administration, N.L.C.; funding acquisition, O.J.E and N.L.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Universiti Putra Malaysia, grant number GP-I/2014/9439000, and the first author was funded by Niger Delta Development Commission, Nigeria (NDDC/DEHSS/2014PGFS/AKS/014), for postgraduate study.

**Acknowledgments:** The authors are grateful to Daniel Onwude for providing some insights and discussion of the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Impact of Process Parameters and Bulk Properties on Quality of Dried Hops**

**Sharvari Raut 1,\*, Gardis J. E. von Gersdor**ff **1, Jakob Münsterer 2, Klaus Kammhuber 3, Oliver Hensel <sup>1</sup> and Barbara Sturm 1,4,5**


Received: 26 October 2020; Accepted: 18 November 2020; Published: 20 November 2020

**Abstract:** Hops are critical to the brewing industry. In commercial hop drying, a large bulk of hops is dried in multistage kilns for several hours. This affects the drying behavior and alters the amount and chemical composition of the hop oils. To understand these changes, hops of the var. Hallertauer Tradition were dried in bulks of 15, 25 and 35 kg/m2 at 60 ◦C and 0.35 m/s. Additionally, bulks of 25 kg/m<sup>2</sup> were also dried at 65 ◦C and 0.45 m/s to assess the effect of change in temperature and velocity, respectively. The results obtained show that bulk weights significantly influence the drying behavior. Classification based on the cone size reveals 45.4% medium cones, 41.2% small cones and 8.6% large cones. The highest ΔE value of 6.3 and specific energy consumption (113,476 kJ/kgH2O) were observed for the 15 kg/m<sup>2</sup> bulk. Increasing the temperature from 60 ◦C to 65 ◦C increased the oil yield losses by about 7% and myrcene losses by 22%. The results obtained show that it is important to define and consider optimum bulk and process parameters, to optimize the hop drying process to improve the process efficiency as well the product quality.

**Keywords:** bulk weight; hop cones size distribution; chemical analysis; energy consumption

#### **1. Introduction**

Hop (Humulus lupulus) is a perennial climbing plant that is used in the brewing industry to impart a bitter taste and provide a "hoppy" aroma to beer [1]. About 97% of the world's hop crop is used for beer making [2]. Initially, hops were primarily used to provide stability to beer and increase shelf life. However, with extensive research and knowledge acquired over time, hops have also increasingly been used for other purposes such as creation of foam characteristics [3]. The bittering acids, tannins, chemical and aroma compounds are situated in the lupulin glands of the hop cone [1,4]. As of 2016, hops can be spilt into three categories namely aroma hops, bitter hops, and dual-purpose hops. Aroma hops are responsible for contributing fruity, piney, citrusy, and tropical aromas to the beer while bittering hops are associated with its bitterness [5,6]. Dual-purpose hops are described as bitter hops with special aroma characteristics [6]. However, freshly harvested hop cones have a high water content, which in turn promotes microbial spoilage and thus, cannot be stored for an extended

period of time [7]. Therefore, to ensure minimal post-harvest losses and consistent product quality, hop cones need to be processed immediately after harvesting. In the Hallertau region, 22 hop varieties are grown [6]. As harvest lasts for a maximum of eight weeks a year, and with harvesting period for each variety lasting for less than two weeks, the efficient processing of harvested hops is of great economic importance to growers [8].

To extend shelf life and increase stability of hops, convection drying is used as a preservation technique by hop growers. Hop drying is a complicated process due to both the physical and chemical characteristics of the hop cone. The anatomy of the hop cone consists of strig as its main axis, followed by bracteole which is made up of sheets of fine petals adhering and covering the strig and contain the lupulin glands and finally the bract as the outer structure of the hop cone which consists the hop leaves. Compared to the mass, the bracteole and bract account for relatively larger surface area of the cone. The strig, however has an extremely small surface area. During the drying process, the strig is shielded from the drying air due to the exterior bract and thus leading to high-water content retention. Furthermore, the cone size and shape, strig shape, arrangement of the cone leaves and density of the capillaries in the bracteole differ in individual hop varieties and thus further complicate the drying process [9]. According to [9], at the end of the drying process if a hop cone has a water content of 8–9%, the water content within the bracteole is 4–7%, while that within the strig is up to 20–30%. The larger amount of water within the strig is released to the bracteole and bracts after the drying process during ventilation in the conditioning chamber [9]. This stage allows for the establishment of constant moisture equilibrium between the strig, the bracteole and bract [10]. Therefore, the stage of conditioning is of utmost important in commercial drying of hop cones. Optimum process parameters during hop drying are of paramount importance as over drying of hops (moisture content ≤ 7%) can increase lupulin losses due to the shattering of hop cones [11]. Conversely, stopping the drying process at moisture contents ≥ 13% would initiate microbial spoilage. Furthermore, the temperature of the drying process also influences quality parameters such as color, oil yield and oil composition of the produce.

Traditionally, hops were dried in a single layer (3–5 cm). The drying time required was largely dependent on the weather conditions and usually ranged between two to ten days [7]. With improvement in technology and understanding of the main characteristics for hop drying, this traditional method was replaced by multistage kiln and belt driers. In belt driers, the hops are dried in three layers in a continuous process. The fresh hops are fed onto the uppermost belt, while the dried hops remain on the lowest belt until the final moisture content is achieved. For optimum drying using belt driers, it is important to evenly feed fresh hops over the entire belt width. Unlike the continuous process of belt driers, multistage kiln drying is a batch process [10]. The drier consists of three to four superimposed perforated trays, through which hot air is passed (bottom to top) to reduce the moisture content within hops. Freshly harvested hop cones are filled within the topmost tray and as the drying progresses, the bulk of hop cones is transferred to the trays below. Once the drying process is ended, the dried hop cones are transferred into a conditioning chamber. Depending on the variety in focus and the drying system, the drying time with multi stage kilns can vary between five to eight hours [7]. One of the biggest problems with kiln driers is potential uneven drying of the hop cones due to the formation of so called "nests". Nest generation is the formation of compacted zones within the hop stack and can lead to varying air resistances from the beginning of the drying process. Parameters such as air velocity, bulk height and density as well as air temperature are the influencing process parameters that need to be considered and optimized to avoid the formation of nests [9].

With hops being so important to the brewing industry, the need for optimisation of the drying process and understanding of the changes within the product during drying is imminent. Currently, to determine moisture content, temperature and humidity sensors are placed within the hop stack. However, this technique lacks in providing information of the overall stack and thus, leads to not detecting potential variations of moisture content within the hop stack [10]. With advancement in technology, smart drying systems that combine non-invasive measurement techniques with process

control have also been developed. Implementation of non–invasive techniques such as computer-aided vision (CAV), thermography (TI), laser backscattering (LB), and hyperspectral imaging (HSI) have demonstrated the ability to analyse the physical and chemical properties, colour, shrinkage, texture, shape, water content, porosity, defects and firmness of the product [12]. A recent investigation performed also used color and HSI imaging as non-invasive techniques to monitor quality parameters such as color of hops of different bulk weights, provide spectral correlations and build models for prediction of selected chemical components. This study has shown promising results for novel real time monitoring during the drying of hops [3]. Applications of thermal imaging in kilns have also been used to understand the drying behavior of the hop cones [9,13] and further potentially combine them with control systems for real time product based control of the drying process.

Drying is an energy intensive process and in case of hop drying, fresh air is heated to 63 ◦C–68 ◦C using fuel oil or gas burners [7]. On average, 44 litres of fuel oil is currently required per 100 kg of dried hops. With increasing fuel prices, alternative energy sources and heat recovery systems are becoming of larger interest [14]. In the context of hop drying, options such as preheating of air, use of waste heat from power generator, use of combined solar-building heat and heat recovery from exhaust air are some of the considered alternatives. However, the economic efficiency of these alternative energy sources largely depends on the acquisition costs, operating costs, and equipment life expectancy [9]. Another option to reducing heating oil consumption was investigated by performing experiments with inclusion of a heat recovery system and optimisation of drying process parameters. The results for the performed experiments reveal a potential of 20% savings in heating oil [7,9]. The results from these experiments successfully led to the installation of heat recovery systems on industrial hop kilns and further led to the development of a pilot plant scale dryer that considers better control of the process parameters and gentler kilning of the hop cones [7].

The experiments conducted within this study were performed on the aforementioned pilot plant dryer and aimed to understand the effect of varying bulk weights and hop size distributions on the drying behavior of hops and the associated changes in quality parameters such as color, hop oil yield and hop oil content. As energy is a crucial aspect of the drying process, specific energy requirement for the different bulk weights were also calculated.

#### **2. Materials and Methods**

Investigations were performed at the Bavarian State Research Center for Agriculture (LfL), Hop Research Centre Hüll, Wolnzach. Organic aroma hops of the variety Hallertauer Tradition were used for investigation. Based on the harvest dates for this variety in 2018, the experiments were conducted for a week starting from end of August to beginning of September. Hop vines were freshly harvested prior to each investigation and further mechanically de-vined using a hop cone harvester (Hopfenpflückmaschine Wolf WWE 220, Giesenfeld, Germany).

#### *2.1. Dryer*

The pilot plant scale dryer described by [3,7] was used for conducting the experiments within this study (Figure 1). The dryer consisted of two drying layers and a slip case which is also considered as a drying stage. However, in case of this experimental investigation, the second drying layer was not used, and the slip case was only used to remove the hop cones. Thus, the overall drying process was conducted using the top drying layer itself. This means that the hop cones were not tipped into the second drying layer but were dried continuously on the top layer and tipped directly into the slip case at the end of drying. Temperature and humidity of the air entering and exiting the top drying layer was measured using Testo 174 H data loggers with an accuracy of ±0.5 ◦C for temperature and ±3% for relative humidity (Testo SE & Co. KGaA, Lenzkirch, Germany).

**Figure 1.** Schematic of hops dryer (adapted and revised from [9]).

Experimental investigations were conducted for hops with bulk weights of 15, 25 and 35 kg/m<sup>2</sup> at 60 ◦C and 0.35 m/s, respectively. The 25 kg/m2 bulk was considered as the control bulk and changes to the velocity to 0.45 m/s and temperature of 65 ◦C were only performed at this bulk weight. Thus, allowing for better comparison for further analysis. The drying time for each bulk weight was estimated based on the set final moisture content of - 7–8% (equivalent moisture ratio [MR] - 0.10–0.12). For reproducibility and repeatability, three repetitions were performed for all experiments. Table 1 provides a summary of the experimental plan for the investigations conducted.


**Table 1.** Experimental plan for drying of Hallertauer Tradition.

#### *2.2. Sampling*

For further analysis, the samples were extracted from the front section of the dryer at defined time intervals. At each time interval, a small set of approximately 50 g was extracted from the dryer. To dissuade the formation of artificial holes due to sample extraction, the hop cones were redistributed throughout the bulk. The initial moisture content of hops for the different bulk weight varied between 78% and 80%. For moisture content analysis, the samples were collected at a 20 min interval from the start for the first hour, followed by a 30 min interval for the next hour, and a 60 min interval thereafter. End time for the drying process varied depending on the weight of the bulk. Table 2 provides details of the sampling intervals for the different experiments.


**Table 2.** Sampling intervals for different experiments.

The samples extracted at different intervals were further placed in an oven at 105 ◦C for 24 h in accordance with the official Association of Official Analytical Collaboration (AOAC) method [15] to obtain the final moisture content of the samples. The moisture content on dry basis (MCdb) was calculated using Equation (1).

$$\text{MC}\_{\text{db}} = \frac{\text{m}\_{\text{t}}}{\text{m}\_{\text{dm}}} - 1 \tag{1}$$

where mt is the total mass of a sample at a given time and mdm is the absolute dry matter in a sample. In addition to MCdb, the moisture ratio (MR) of the samples was also calculated as it allows to compare the different samples with each other by normalizing the moisture contents to their initial moisture contents and thus displaying them in values ranging from one to zero. MR was calculated using Equation (2).

$$\text{MR} = \frac{\text{MC}\_{\text{db},1} - \text{MC}\_{\text{c}}}{\text{MC}\_{\text{db},0} - \text{MC}\_{\text{c}}} \tag{2}$$

where MCdb,1 is the dry basis MC at a given time, MCdb,0 the dry basis MC before drying and MCe is the equilibrium moisture content. As the influence of MCe is relatively small (≤2%) [16], the MR equation is further simplified to Equation (3).

$$\text{MR} = \frac{\text{MC}\_{\text{db},1}}{\text{MC}\_{\text{db},0}} \tag{3}$$

Sampling intervals similar to those described for moisture content analysis were used for color acquisition. A detailed explanation on the color measurements is provided in Section 2.3.

Hop cone size varies both between and within the varieties. This variation is evidently observed from Figure 2 wherein hop cone sizes for two varieties namely Hallertauer Tradition and Hallertauer Herkules is shown.

**Figure 2.** Hop cone size variation in varieties (**a**) Hallertauer Tradition, (**b**) Hallertauer Herkules [9].

Therefore, to understand the distribution of the hop cone size, 50 g of fresh hops were extracted from the bulk and sorted based on their length and diameter. Cones smaller than 2 cm in length were classified as small, between 2 cm–3.2 cm as medium and larger than 3.4 cm as large. The classified cones were further weighed to establish their ratios within the 50 g mix. The investigation of hop cone size distribution was limited to fresh hop cones (i.e., 0 min) as the aim of this study was to investigate the effect of hop cone size on the drying process.

For distillation of essential oils, samples were collected prior to drying (fresh) and at the end (dried) of each trial. Distillation of essential oils was carried out only for the control bulk i.e., 25 kg/m2 bulk dried at air velocity of 0.35 m/s and at 60 ◦C and 65 ◦C, respectively, to assess the effect of temperature on the overall hop oil yield and hop oil composition. Finally, to mimic the commercial drying process, it was ensured that the bulk was thoroughly mixed using a rake prior to sample collection.

#### *2.3. Imaging*

During the drying process, hop cones undergo significant color change. To understand the development of color, hop samples were placed in a photo box (Life of Photo 60 cm LED Light Cube, Weiwa Foto, Gummersbach, Germany). The photo box includes an array of 160 LED lamps as the illumination and a window for the camera. To capture the images, an RGB Camera (model no. 61BUC02, The Imaging Source Europe GmbH, Bremen, Germany) was placed at a distance of 29 cm from the surface of the product [17] using the camera window of the photo box. A schematic of the imaging system used is presented in Figure 3.

**Figure 3.** Schematic of RGB camera imaging system for hops.

The acquisition of the images was realized using the IC Capture software (the Imaging Source Europe GmbH, Bremen, Germany). The captured images were further processed in MATLAB® (The MathWorks, Inc., Natick, MA, USA) to convert RGB values to CIE L\*a\*b\* values. For this purpose, RGB images were converted to gray and further into binary images (0, 1) using appropriate threshold factors. In the binary images, the values 0 and 1 were assigned to the background and the sample, respectively. To remove any undesirable objects, the function bwareaopen was applied. The images were then masked to ensure that all values now belong to the image and no values are being assigned to the background. Finally, the function rgb2lab was used to convert the RGB values into CIE L\*a\*b\* values for the images captured [18]. The function rgb2lab within MATLAB® uses a two-step process. In the first step, RGB values are converted into XYZ space using linear matrix in Equation (4) [19].

$$
\begin{bmatrix}
\mathbf{X} \\
\mathbf{Y} \\
\mathbf{Z}
\end{bmatrix} = \begin{bmatrix}
0.412453 & 0.357580 & 0.180423 \\
0.212671 & 0.019334 & 0.072169 \\
0.019334 & 0.119193 & 0.950227
\end{bmatrix} \begin{bmatrix}
\mathbf{R} \\
\mathbf{G} \\
\mathbf{B}
\end{bmatrix} \tag{4}
$$

The XYZ values thus obtained are further normalized to the D65 white points. In the final step, the XYZ values are converted to CIE L\*a\*b\* values using Equations (5)–(7) [20].

$$\begin{array}{l} \text{L}^\* = 116 \times \left( \frac{\chi}{\text{T}\_n} \right)^{\frac{1}{5}} - 16 \text{ if } \left( \frac{\chi}{\text{T}\_n} \right) > 0.008856\\ 116 \times \left( \frac{\chi}{\text{T}\_n} \right) \text{if } \left( \frac{\chi}{\text{T}\_n} \right) > 0.008856 \end{array} \tag{5}$$

$$\mathbf{a}^\* = 500 \times \left[ \left( \frac{\chi}{\chi\_n} \right)^{\frac{1}{\xi}} - \left( \frac{\chi}{\chi\_n} \right)^{\frac{1}{\xi}} \right] \tag{6}$$

$$\mathbf{a}^\* = 200 \times \left[ \left( \frac{\mathbf{Y}}{\mathbf{Y}\_\mathrm{n}} \right)^{\frac{1}{3}} - \left( \frac{Z}{Z\_\mathrm{n}} \right)^{\frac{1}{3}} \right] \tag{7}$$

where L\*, a\* and b\* represents lightness, red/green and yellow/blue, respectively. Xn, Yn, and Zn are the values of the blank references.

To compare the non-invasive method to a gold standard color measurement method, a CR-400 Chromameter Konica Minolta Sensing Europe B.V., München, Germany) in combination with the SpectraMagic NX Software (Konica Minolta Sensing Europe B.V., München, Germany) was also used. For calibration, a ColourChecker Classic chart (X-rite Gmbh, Planegg-Martinsried, Germany) was used to compare the values obtained from the chromameter and the RGB camera for the 24 color patches.

Finally, the values obtained from each of the repetitions were further averaged to calculate the total color difference (ΔELab) using Equations (8)–(11):

$$
\Delta \mathbf{E}\_{\rm LB} = \sqrt{\left(\Delta \mathbf{L}^\*\right)^2 + \left(\Delta \mathbf{a}^\*\right)^2 + \left(\Delta \mathbf{b}^\*\right)^2} \tag{8}
$$

$$
\Delta \mathbf{L}^\* = \mathbf{L}\_1^\* - \mathbf{L}\_0^\* \tag{9}
$$

$$
\Delta \mathbf{a}^\* = \mathbf{a}\_1^\* - \mathbf{a}\_0^\* \tag{10}
$$

$$
\Delta \mathbf{b}^\* = \mathbf{b}\_1^\* - \mathbf{b}\_0^\* \tag{11}
$$

where L\*1, a\*1, b\*1 is the lightness, red/green and blue/yellow at the specific time interval and L\*0, a\*0, b\*0 is the initial lightness, red/green and blue/yellow [21].

#### *2.4. Gas Chromatography*

Gas Chromatography (GC) analysis was performed using a Dani GC with flame ionisation detection (FID; DANI Instruments SpA, Milan, Italy) with a defined temperature program. The gas flow was divided in a ratio of 1:25 (spilt) with helium being used as the carrier gas. The injector and detector temperature were set to 200 ◦C. The GC unit was also calibrated using the internal standard developed by the Hops Research Center, Wolnzach. Myrcene, linalool, ß-caryophyllene and humulene—representing the most valuable oils related to Hallertauer Tradition—were the main components analysed in the GC unit. Mathematical calculation for quantity of hops required for distillation of essential oils was determined using the method described in [17].

#### *2.5. Energy Consumption*

To calculate the specific energy consumption per bulk weight it was essential to first calculate the amount of dry matter at the end of the drying time. To this end, Equations (12) and (13) were used to calculate the dry matter in the bulk:

$$\text{DM}\_{\text{i}} = \frac{100 - \text{MC}\_{\text{f}}}{100} \tag{12}$$

$$\rm{DM}\_{\rm{f}} = \rm{DM}\_{\rm{i}} \times \rm{W}\_{\rm{i}} \tag{13}$$

where DMi is the amount of dry matter per kg, MCf is the final moisture content, DMf is the amount of dry matter in the bulk (kg) and Wi is the initial weight of bulk.

The specific energy requirement per bulk weight was calculated using Equations (14) and (15) [22,23]:

$$\mathbf{E\_t = A \times \\$} \times \boldsymbol{\Psi} \times \boldsymbol{\Delta T \times \rho\_\mathbf{a} \times C\_{\mathbf{p},\mathbf{a}} \times \mathbf{t}} \tag{14}$$

$$\mathbf{E\_s} = \frac{\mathbf{E\_t}}{\mathbf{W\_i} - \mathbf{D}\mathbf{M\_f}} \tag{15}$$

where Et is the total energy consumed (kJ), A is the area of the drying chamber, v is the velocity of the air entering the drying chamber (m/s), ΔT is the temperature difference between the inlet air and surrounding air (K), ρ<sup>a</sup> is the density of air (kg/m3), Cp,a is the specific heat capacity of air (kJ/kg K), t is the run time (s), Es is the specific energy required (kJ/kg) and Wi is the initial weight of the bulk.

#### **3. Results and Discussion**

#### *3.1. Drying Behavior*

Figure 4 presents the temperature and the humidity profile of the air exiting the drying chamber. From the extracted results, it is observed that the humidity for all bulks other than that for 15 kg/m2 is close to 100% at the initial phase of drying. As the drying continues, the temperature of the air exiting (top layer of the hop bulk) begins to increase and the humidity to decrease. The time required for the temperature to reach the set drying temperature significantly depends on the temperature as well as the bulk weight. The lower the temperature and the higher the bulk weight the longer the time required for drying and vice versa.

**Figure 4.** Temperature (symbol and line) and humidity (symbol) profile for different bulk weight and different process settings.

Figure 5 presents the drying curves for different bulk weights of hops dried at 60 ◦C and 0.35 m/s air velocity. The bulk weight had a significant influence on the drying time required to attain the set moisture ratio of 0.12. Hops of 15 kg/m<sup>2</sup> bulk weight required 180 min while 25 kg/m<sup>2</sup> and 35 kg/m2 bulks required 210 min and 240 min, respectively. Similar results were also obtained by [3] for hops with bulk weights of 12 kg, 20 kg and 40 kg dried at 60 ◦C and 0.35 m/s. Overall, the development of the drying curve for all three bulks is characteristic to the curves obtained from convective drying process; with rapid losses in moisture content during the initial phase of drying and eventual reduction as it transitions from the first to the second phase of drying. In the case of hops, this development of drying behavior can be associated to two factors: structure of hop cone and the variety in focus. As mentioned previously, the structure of the hop cone plays a crucial role in the drying behavior. The strig which contains high water content and is surrounded by the bracteole and bract of the hop cone, only begins to dry once the bracteole (leaves of the hop cone) has reached a specific moisture content. It is believed that the development of high suction tension acts as the driving force to draw the moisture from the strig [3] and thus, resulting in specific drying behavior. Furthermore, the weight percentage ratio between the strig and the bracteole also depends on the variety in focus. In case of Hallertauer Tradition, this ratio varies from 9–10% [24]. In addition to the variety, harvesting conditions such as environmental temperature, humidity, location of harvest, soil quality, hop cone size and shape also affect the overall hop cone drying behavior [3]. The effect of hop size, shape and distribution on the overall drying behavior has been further explained in Section 3.2.

**Figure 5.** Moisture ratio (MR) vs. time (min) for hops with three different bulk weights dried at 60 ◦C and 0.35 m/s.

Figure 6 presents the results obtained from changing the velocity (Figure 6a) and changing the temperature (Figure 6b) for 25 kg/m2 bulk of hops.

No significant impact on the drying time between air velocity of 0.35 m/s (210 min) and 0.45 m/s (200 min) was observed. This is contradictory to several investigations performed through the years that identify air velocity to be an influencing factor for reducing the drying time [20,25]. For most of these studies, a thin layer convective drying of products is investigated as compared to a bulk of 25 kg/m2. The insignificant influence of air velocity on the bulk could be due to the air being completely saturated with water vapor prior to reaching the top of the bulk. Thus, the deficit in saturation of the drying air acts as the limiting factor as compared to the air velocity [26]. This further explains the similarity in drying behavior for the first 90 min and then the observed rapid moisture losses for 0.45 m/s for the rest of the drying period. However, increasing the air velocity is also not an ideal solution to reduce the drying time as increasing the velocity leads to faster drying at the surface as compared to the center, in turn stiffening the surface and causing uneven shrinkage of the product [27]. Furthermore, increasing velocity also increases the overall energy demand for the drying process [9].

**Figure 6.** Moisture ratio (MR) vs. time (min) for hops dried at (**a**) 0.35 m/s and 0.45 m/s at 60 ◦C (**b**) 60 ◦C and 65 ◦C at 0.35 m/s.

Changing the drying temperature, the overall drying time is observed to have reduced from 210 min for 60 ◦C to 180 min for 65 ◦C. This is analogous to the several investigations performed on the influence of drying temperature on the drying time [20,23,28–30]. From the drying curve presented in Figure 6b, it is also observed that the moisture losses were rapid for 60 ◦C than 65 ◦C for the initial phase (up to 60 min), indicating that the 60 ◦C bulk dried faster than the 65 ◦C bulk. Within the 7-day investigation period, a severe thunderstorm on Day 3 led to harvesting of wet hops, thus increasing the overall moisture content of the hop cones. Furthermore, as the hop cones were harvested just prior to the beginning of each drying trial, it was observed that those harvested in the morning had a higher moisture content due to the morning dew as compared to those harvested after midday. The trial period within the experiments were randomized to ensure that each repetition was performed on a different day and at a different time. However, the difference in the environmental conditions and time for 65 ◦C trials could have led to the observed results. A study performed on the impact of environmental conditions on the hop cones during drying, shows that weather conditions such as heavy rains, morning dew and sunny days significantly influence the drying process and the overall product quality [3].

#### *3.2. Hop Cone Size*

Results from the classification of average hop cone sizes within a hop stack is presented in Table 3.


**Table 3.** Average hop cone size distribution within a hop stack.

For the Hallertauer Tradition, the distribution of cone size lies between small to medium with small hop cones accounting for 41.2%, medium for 45.4% and large for 8.6% of the total measured weight. A total of 50 g of hops were weighed for classification of which 4% were lost due to the cones either being broken or shattered. In addition to the size, a linear increase in the average weight of one cone is also observed. A study conducted on drying of coarse lignite particles of varying particle size indicates that larger particle sizes require longer drying times as compare to the small particles [31]. Particles smaller in size have a larger specific surface area which results in faster heating rate and better transfer of heat from the drying medium to the center. Additionally, the moisture transfer from inside

to the surface is also faster within smaller particles, while smaller diameter also provides less resistance to the internal heat transfer [32], thus, leading to faster drying processes. However, bulks with a large number of small particles such as that investigated within the current study, tend to increase resistance within the bulk due to a decrease of average particle size and an increase of void fraction, and thus the associated increase in pressure drop. Investigations performed on grain drying showed that the grain surface characteristics can potentially tend to increase the airflow resistance [33]. For locust beans at low moisture content, the cotyledons tend to become hard and smooth which in turn increases the frictional losses due to high resistance to airflow [34].

Furthermore, the degree of compaction of the hop cones also increases with bulk height. This results in a rapid decrease in the air flow rate and the drying capacity while simultaneously increasing the retention time of the hop in the top layer. As drying progresses, the cones tend to become solid and maintain their individual shape and their position in relation to each other in the top layer. This results in the creation of air channels and formation of zones with uneven air flow [9]. In the current investigation, the proportion of small size cones is significantly high and hence the degree of compaction within the stack is also high. Although no air channels formation was observed in 25 or 35 kg/m2 stack, relatively many air channels were observed to have formed in the 15 kg/m<sup>2</sup> bulk. Figure 7 shows the formation of air channels within the bulk.

**Figure 7.** Air channel formation in a 15 kg/m<sup>2</sup> bulk of hops.

The overall bulk height of a 15 kg/m2 is not high. Therefore, bulks with low heights such as those for 15 kg/m<sup>2</sup> combined with high numbers of small sized cones and set air flow rate can also lead to the formation of air channels. A study performed on pistachio nuts also revealed a linear increase in air flow resistance with increasing depth of the pistachio column. The study also adds that the air flow rate significantly influenced the pressure drop within the column as compared to other parameters such as moisture content, particle interaction and filling method [35]. The inclusion of a higher number of small cones resulted in faster drying but also led to the formation of air pockets or so called "nests" within the drying chamber which in turn led to uneven drying [10]. Investigations performed on storage of hops revealed that hops stored for a longer amount of time undergo oxidation which lead to formation of compounds that interfere with the aromatic smell associated with hops [17]. Bulks with unevenly distributed moisture content can also undergo a similar degradation and thus leading to significant losses of both physical and chemical attributes of the product. Therefore, in commercial hop drying units, to avoid formation of large air pockets, low bulk weights such as 15 kg/m<sup>2</sup> are not usually

considered. Furthermore, in the commercial process, the air velocity is also lowered as the drying progress so as to ensure even drying of hop cones and deter the formation of air channels. Moreover, during extraction of samples at specified intervals, a number of small hop cones were observed to have exited the drying chamber. This could be due to two reasons. Firstly, as mentioned earlier, small cones have large surface area and hence dry faster. Secondly, the pressure difference within the drying chamber and the surrounding environment could have also led to the dried small hop cones to exit the dryer.

As the size of the hop cone is an uncontrolled parameter, a solution to optimize the drying performance, and reduce the air resistance, is to fill low bulks followed by tipping at shorter interval. Thus, aide in loosening the compacted layers within the bulk and reducing air resistance which increases the air flow and the drying performance [9].

#### *3.3. Chemical Analysis*

#### 3.3.1. Oil Content

The variety Hallertauer Tradition is a product of cross breeding between the variety Hallertauer Tradition Gold and a male breeding line. It has a slightly flowery, herby-spicy aroma with an amount of oil ranging between 0.005 L/kg–0.007 L/kg [6,9]. Within the experimental investigation conducted, the amount of oil extracted for fresh hops ranged between 0.0025–0.0027 L/kg while those for dried hops ranged between 0.0023–0.0024 L/kg. Table 4 summarizes the amount of oil extracted and the standard deviations for the investigations conducted.


**Table 4.** Amount of oil extracted in L/kg with standard deviations (SD) for prior and after drying process.

As compared to the dried samples, significant deviation is observed within the fresh samples which could be due to the natural heterogeneity of hops within and between the different bulks. Figure 8 presents the amount of oil extracted prior to drying and after drying at 60 ◦C and 65 ◦C, respectively.

The overall amount of oil extracted for fresh hops was lower than that reported in literature (0.007 L/kg) [6]. Factors such as location, environmental conditions at the location and harvest time have been shown to have a significant effect on the total amount of oil [36,37]. In case of the Hallertauer Tradition, a five-week investigation conducted for three continuous years for essential oil content reveals that the amount of oil extracted is lower at the initial harvest time (about 0.005 L/kg) and increases over time (about 0.02 L/kg) until it reaches an optimum point (roughly Week 3). Harvesting after this period has also shown a decrease in the overall amount of oil [38]. The investigation within this study was performed in the first week of the harvesting period for Hallertauer Tradition. The early harvest combined with associated environmental conditions could have led to the lower amounts of oil extraction.

Drying temperature also had a significant effect on the overall hop oil content. Within the current investigation it was observed that increasing the drying temperature decreased the total amount of oil extracted after drying. An investigation performed on the Mandarina Bavaria variety of hops revealed losses in the total oil extracted after drying as compared to prior to drying [17]. Furthermore, drying of Laurel (*Laurus nobilis* L.) leaves at either 45 ◦C or 65 ◦C accounted for losses close to 83% [39]. An experimental study performed on drying the aerial parts of sweet wormwood (*Artemisia annua* L.) also revealed a decrease in the yield of oil with an increase in drying temperature [40].

**Figure 8.** Amount of oil in L/kg hops extracted for samples extracted at 0 min and 210 min for hop cones dried at 60 ◦C and 65 ◦C at 0.35 m/s.

#### 3.3.2. Gas Chromatography of Essential Oils

The effect of drying is quite significant on the overall hop oil content with losses accounting for 30–40% [41] and are mainly associated to the high water vapor volatility [42]. Table 5 provides a summary with standard deviation values of the essential oils namely myrcene, linalool, β-caryophyllene and humulene analyzed through gas chromatography for 25 kg/m2 bulk dried at 60 ◦C and 65 ◦C, respectively.

**Table 5.** Essential oils with standard deviation (SD) prior to and after drying process extracted using GC-FID.


In line with the results on hop oil yield (Section 3.3.1), large deviations in the essential oils is observed. As mentioned previously, hops are a heterogeneous product and the aroma/essential oil content significantly varies within and between the bulks. Furthermore, as mentioned previously the environmental conditions during harvest (dry, dew or rain wet) between the repetitions also varied significantly. Factors such as environmental conditions and harvest time, influences the amount on essential oil components within hops [43]. With harvest lasting for less than two weeks, variations in harvesting conditions during this period can further add to the large variations in oil components [44]. In additions to these factors, the overall sample size was small and hence large deviations were observed. Thus, making it a poor estimator for representation of deviations within the samples. Figure 9 represents the myrcene content prior to and after drying for 60 ◦C and 65 ◦C respectively. Losses in myrcene content after drying are observed in both cases. The losses are relatively higher for 65 ◦C as compared to 60 ◦C.

**Figure 9.** Results obtained from gas chromatography analysis for myrcene in g/kg hops at 0 min and 210 min for hop cones dried at 60 ◦C and 65 ◦C at 0.35 m/s.

Myrcene is a terpene hydrocarbon and is one of the most important and sensorially dominant component of fresh hop cones [41]. It is a key contributor to the hop aroma and encompasses 17–37% of the total hop oil [45]. During the drying process, myrcene undergoes oxidation and polymerization which leads to cyclic reactions that form products such as terpenoids which include linalool, and geraniol [46]. Furthermore, myrcene is highly volatile at high temperature. With hop drying temperatures ranging between 55 ◦C and 65 ◦C, significant losses of myrcene are expected. Studies conducted on essential oil losses show that myrcene losses range between 25% and 30% [44,47]. Within the present study myrcene losses of 7.74% and 30.27% were observed for 60 ◦C and 65 ◦C, respectively. Experimental investigation conducted on Laurel (*Laurus nobilis* L.) leaves also reveal losses in monoterpene hydrocarbons such as myrcene due to volatility with steam or potential chemical rearrangement of compounds [39].

Linalool is one of the key contributors to the hoppy flavor within beer. It provides a citrusy, flowery and fruity aroma to the beer [4,48]. During the drying process linalool content also undergoes significant changes but unlike myrcene, the losses associated to the linalool are much lower [47]. The results from GC analysis (Figure 10) reveal that after drying at 60 ◦C, no significant losses in the linalool content were observed. However on increasing the temperature to 65 ◦C, the losses were much higher as compared to 60 ◦C. Investigation conducted by [47] show that increasing the temperature increases the losses in linalool content. Too high temperatures such as 80 ◦C or 90 ◦C have shown to increase the linalool content. This increase is associated to presence of glycosidically bound linalool that are still active at high temperature and indirectly increase the linalool content. Low temperatures such as 50 ◦C have shown to retain maximum linalool content as compared to those at 65 ◦C [47]. Similar losses in linalool content were observed for the Mandarina Bavaria variety dried at 60 ◦C [17].

β-caryophyllene and humulene belong to the sesquiterpenes group and are both hydrophobic in nature; thus have limited solubility in aqueous solutions such as beer [46]. During the drying process, both compounds undergo oxidation and form various other products. Oxidation of β-caryophyllene leads to formation of caryophyllene oxide, 14-hydroxy-b-caryophyllene, caryolan-1-ol,4S-dihydrocaryophyllene-5-one, (3Z)-caryophylla-3, 8(13)-diene-5a-ol, caryophylla-4(12), and 8(13)-diene-5a/b-ol [46,49]. The dominant products formed are caryophyllene oxide and 14-hydroxy-b-caryophyllene that have musty, floral, spicy, cedar, woody odor. Humulene theoretically oxidizes to three humulene monoepoxides however only two, namely mono and diepoxides, are found to be present in hops and have hay-like, moldy, cedar like odor [46]. From the results obtained for β-caryophyllene (Figure 11) and humulene (Figure 12), the values after drying are observed to have

remained either similar or increased as compared to fresh hops. As mentioned previously, the Hallertauer Tradition has a typical herby, spicy aroma and oxidation of β-caryophyllene and humulene could have led to the potential increase of the byproducts of these main compounds. Thus, explaining the corresponding increase in the β-caryophyllene and humulene content after drying.

**Figure 10.** Results obtained from gas chromatography analysis for linalool in g/kg hops at 0 min and 210 min for hop cones dried at 60 ◦C and 65 ◦C at 0.35 m/s.

**Figure 11.** Results obtained from gas chromatography analysis for β-caryophyllene in g/kg hops at 0 min and 210 min for hop cones dried at 60 ◦C and 65 ◦C at 0.35 m/s.

**Figure 12.** Results obtained from gas chromatography analysis for humulene in g/kg hops at 0 min and 210 min for hop cones dried at 60 ◦C and 65 ◦C at 0.35 m/s.

#### *3.4. Color Development*

Figure 13 provides the results for the CIE L\*a\*b\* values measured using a chromameter and the values obtained from processing the acquired RGB images. R2 values of 0.96, 0.90 and 0.95 were obtained for L\*, a\* and b\* respectively. However, significant differences in the actual values were observed especially in case of L\* values. The illumination system used for the RGB imaging was observed to have lower intensity as compared to the illumination within the chromameter and could potentially have affected the lower values from the RGB camera system.

**Figure 13.** Correlation between CIE L\*a\*b\* values between chromameter and RGB camera system.

The color of fresh hop cones varies between slight dark green to yellow green on the exteriors depending on the variety and harvest time. During the drying process, these hop cones undergo discoloration due to influencing factors such as drying time, temperature, and air velocity. The results obtained for the change in color during the drying process under various conditions is presented in Figure 14.

**Figure 14.** ΔE as a function of moisture ratio (MR) for hops dried under different process settings and different bulk weights.

The highest ΔE values were observed for hops with bulk weight of 15 kg/m<sup>2</sup> dried at 60 ◦C and the lowest ΔE values for hops with bulk weight of 35 kg/m2 dried again at 60 ◦C. The associated discoloration is caused during the initial drying phase when the hop cones are colder than the surrounding air. Thus, leading to recondensation of water on the top layers of the hop stack [8,50]. Furthermore, the formation of compact zones or nests in the 15 kg/m<sup>2</sup> bulk led to uneven drying of hop cones which in turn increased the required specific drying time. As drying time is also an influencing factor for color changes, the longer drying time combined with lower bulk weight, could have led to the observed discoloration in the 15 kg/m<sup>2</sup> bulk. The overall results for the total change in color is in contradiction to the results obtained by [3]. According to investigation performed on Mandarina Bavaria with bulk weight of 12, 20 and 40 kg at 60 ◦C, the highest color changes were observed for 40 kg bulk and lowest color changes for 12 kg bulk [3]. The contradiction between the two studies could potentially be due to two factors. Firstly, the RGB system set up within this study was outside the drying chamber as compared to the system setup of [3] wherein the camera was placed above the bulk within the drying chamber. Secondly, the study conducted by [3] considers the variety Mandarina Bavaria while the current study focuses on the variety Hallertauer Tradition. As each variety has slight differences in color as well as the initial moisture content values, it is possible that opposite results are obtained for experimental investigations performed using the same equipment but slight parameter variations. Furthermore, the varying cone size and the natural heterogeneity in hop cone color could have also influenced the color measurements.

Comparing the results for 25 kg/m<sup>2</sup> bulk dried at 60 ◦C and 65 ◦C, it is observed that the ΔE values for samples dried at 65 ◦C are lower than those dried at 60 ◦C. Drying time and drying temperature are two of the influencing factors with degradation of color [51]. An increase in either of two factors can lead to higher degradation of color. The exposure time for the hop cones dried at 65 ◦C was lower than that of those dried at 60 ◦C. Investigations on 3 mm thick carrot slices have shown maximum degradation at 50 ◦C where the exposure time is higher and are concurrent to the findings from the current study [30]. Increasing the velocity has also shown to reduce the color change within the hops. Similar results for apple drying under increased velocity were also obtained [20].

Drying temperature has also shown to influence the lupulin color within the hops. Under optimum conditions, the color tends to remain lemon yellow while at higher temperature and longer exposure time the color degrades to brown [52].

#### *3.5. Energy Consumption*

As shown in Figure 15, the highest specific energy consumption was required by the 15 kg/m2 bulk (113,476 kJ/ kgH2O), followed by the 25 kg/m<sup>2</sup> bulk (91,955 kJ/ kgH2O) and finally the 35 kg/m<sup>2</sup> bulk (84,794 kJ/ kgH2O). The 15 kg/m2 bulk required 180 min for drying, while the 25 kg/m2 and 35 kg/m2 bulk required 210 min and 240 min, respectively. During the drying process for the 15 kg/m2 bulk, air pockets were observed to have formed after the first 60 min. These air pockets might have caused the hops cones to clump in one section and thus requiring a much longer time to dry. For 25 kg/m<sup>2</sup> and 35 kg/m2, the overall bulk weight and the height of the bulk dissuaded the formation of air pockets during the drying process and hence required comparatively lower energy to dry the hop cones.

**Figure 15.** Specific energy requirement for three different hop bulks dried at 60 ◦C and 0.35 m/s.

On increasing the velocity of the 25 kg/m<sup>2</sup> bulk to 0.45 m/s, the overall energy consumption is also observed to increase as compared to 0.35 m/s. The difference between the drying time is insignificant and, thus, does not drastically affect the consumption values. The factor influencing the increased energy consumption in this case is the velocity. Results of energy consumption with varying velocity is presented in Figure 16a.

Lower energy consumptions were observed for 65 ◦C (Figure 16b). The decrease in drying time for higher temperatures compensates for the overall energy demand. The effect of temperature was observed to have a significant influence on the specific energy consumption and is in agreement with previously conducted investigations by [23,53].

**Figure 16.** Specific energy requirements for 25 kg/m<sup>2</sup> bulk dried at (**a**) 0.35 m/s and 0.45 m/s at 60 ◦C (**b**) 60 ◦C and 65 ◦C at 0.35 m/s.

#### **4. General Discussion**

This study investigated the effects of varying bulk weights and process settings on the drying behavior and product quality attributes such as color, hop oil yield, hop oil components and energy consumption. The study also extended to analyze the influence of hop cone size on the overall drying behavior and its extended effects on quality.

Results obtained from the drying curves show that low temperature in combination with high bulk weight lead to an increase in the overall drying time. Increasing the velocity or the temperature does not significantly influence the drying behavior. Factors such as the environmental conditions during harvest as well as during the drying process, the hop variety in focus and the cone size distribution influence the overall drying behavior. Results obtained from classifying hop cones by size reveals that for the variety Hallertauer Tradition, the majority of the cones lie between small to medium size. The results reveal that such distribution of cone can lead to increase in the air resistance, void fraction and hence increase the pressure drop. In case of 15 kg/m2 bulk, this further led to the formation of air channel which affect the moisture content distribution within the hop stack. Thus, the 15 kg/m2 bulk requiring a comparatively longer specific drying time than expected. Hop cone size is a natural parameter and varies according to the variety in focus. In order to reduce the resistances, considering bulk density rather than bulk weights/height has been observed to be a better solution [3,9,54].

In terms of color changes, the highest discoloration of ΔE = 6.83 was obtained for the 15 kg/m2 bulk, followed by the 25 kg/m<sup>2</sup> at ΔE = 6.16. The lowest discoloration was observed for the 35 kg/m2 bulk at ΔE = 3.58. The higher discoloration of the 15 kg/m2 bulk can be associated to the formation of nests or compact zones which increased the required amount of drying time. As temperature and drying time are influencing factors for color changes, an increase in either of these factors, can cause significant effects on the color quality of the product. Temperature and time are also influencing factors for the specific energy consumption, with the 15 kg/m<sup>2</sup> bulk consuming the high amount of energy per unit of product. In order to optimize and reduce the energy costs, it is important to consider alternative solutions that can be implemented with the current drying techniques [9].

Drying temperature is not only an influencing factor affecting color, but also affects the yield and chemical compositions within hops. The drying process has a significant effect on the hop oil yield and composition, and thus a decrease in the values after completion of drying is expected. On increasing the temperature from 60 ◦C to 65 ◦C, increased losses for yield were observed. This is analogous to the results of previous studies [3,17]. In terms of composition, the expected trend of losses was observed for myrcene and linalool, however, an increase in β-caryophyllene and humulene was observed at higher temperature of 65 ◦C. The increase in β-caryophyllene and humulene was previously observed in storage of hops for 24 h prior to drying wherein it was believed the formation

of other by products led to an overall increase in these components [17]. As no storage tests were conducted within the current study, the results obtained for β-caryophyllene and humulene are in contradiction. However, hops in general have a large heterogeneity in the aroma components which fluctuate based on various conditions such as hop variety, hop garden, soil, environmental factors at harvest, etc. [44]. The increase in the β-caryophyllene and humulene content could potentially also be associated to the other influencing factors.

#### **5. Conclusions**

The investigations conducted within the current study demonstrated that process and product quality in hop drying depend on multiple factors such as harvest time, drying conditions, and bulk characteristics (weight and cone size distribution). An optimization of the process would require consideration of multiple potential target conflicts, i.e., energy demand vs product quality, reduction of drying time with bulk reduction vs additional labor demand and machine times caused by additional harvesting schedules.

The study, therefore, indicates the need for development of systems that consider various factors such as hop variety, harvest maturity, environmental conditions at harvest (dry, rain wet) and bulk quality (e.g., cone size distribution), for optimum process settings. This includes the necessity for the development of non-invasive measurement systems in combination with integration of adaptive control systems. With advances in optical sensors, computational models and with the advent of internet of things (IoT) integrated processing systems, it will be possible to further develop tailored systems which consider all above aspects. In addition to integration of measurement and control systems, the study also highlights the need for further work with bulk density rather than bulk weight/height in combination with pressure drop measurements as this will not only aide in optimizing the process parameters but also help in reducing the energy consumption and, thus, the associated processing costs.

**Author Contributions:** Conceptualization, S.R., G.J.E.v.G., J.M. and B.S.; data curation, S.R., G.J.E.v.G. and B.S.; formal analysis, S.R. and G.J.E.v.G.; funding acquisition, J.M., O.H. and B.S.; investigation, S.R., G.J.E.v.G., J.M. and K.K.; methodology, S.R., G.J.E.v.G. and B.S.; project administration, J.M., O.H. and B.S.; resources, J.M. and B.S.; software, S.R.; supervision, J.M., O.H. and B.S.; validation, S.R.; visualization, S.R. and B.S.; writing—original draft, S.R. and G.J.E.v.G.; writing—review and editing, S.R., G.J.E.v.G., J.M., K.K., O.H. and B.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was funded by transnational funding bodies, being partners of the CORE Organic Cofund, and the cofund from the European Commission. This study is part of the SusOrgPlus project within the framework of the CORE Organic Cofund Programme and is supported by funds of the Federal Ministry of Food and Agriculture (BMEL) based on a decision of the Parliament of the Federal Republic of Germany via the Federal Office for Agriculture and Food (BLE) under the BÖLN programme (Project number: BLE–2817OE005).

**Acknowledgments:** The authors would like to thank Sebastian Wambach for his support during the experimental investigations.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Evaluation of Postharvest Processing of Hazelnut Kernel Oil Extraction Using Uniaxial Pressure and Organic Solvent**

**Gürkan Alp Ka ˘gan Gürdil 1,\*, Abraham Kabutey 1, Kemal Ça ˘gatay Selvi 1, Cestm ˇ ír Mizera 1, David Herák <sup>1</sup> and Adéla Fra ˇnková <sup>2</sup>**


Received: 8 July 2020; Accepted: 6 August 2020; Published: 8 August 2020

**Abstract:** Uniaxial loading and organic solvent are small-scale oil expression methods used to evaluate the mechanical behavior, oil content, and oil efficiency of oil-bearing materials aimed at designing a low-cost mechanical pressing system. Bulk kernels of pressing height 40 mm were heated from 40 to 60 ◦C and compressed at maximum force of 60 kN and speeds from 4 to 8 mm/min. Relaxation times between 3 and 12 min were applied to assess the kernel oil efficiency. The kernel oil point was identified at deformation levels between 15 and 25 mm at a speed of 4 mm/min using a litmus test. The kernel oil was analyzed for peroxide value and free fatty acid. Kernel oil content was determined by Soxhlet extraction. Increased speed caused a serration effect on the force–deformation curve leading to lower oil yield. Lower and upper oil point forces at 6.21 ± 0.58 and 10.61 ± 0.71 kN were observed to be useful for predicting the pressure for maximum output oil. The peroxide value and free fatty acid content of kernel oil decreased with increasing temperature, indicating its quality usage. The relaxation time of 12 min after compression increased kernel oil efficiency of 15.6%. In designing new presses, there is a need to consider compression and relaxation processes to reduce the residual kernel cake oil.

**Keywords:** bulk hazelnut kernels; mechanical properties; heating temperature; oil efficiency; relaxation process

#### **1. Introduction**

Hazelnut (*Corylus avellana* L.) is a fruit consisting of a hard shell with a kernel inside it that is edible [1]. The edible kernel is covered by a removable thin fibrous pellicle, with the internal tissue of the cotyledons consisting of parenchyma cells separated by very small intercellular spaces [2,3]. The tree crop belongs to the family Betulaceae and is grown in the Mediterranean Sea regions, mainly in Turkey, Spain, and Italy [4,5]. Turkey is the main producer and exporter with 420,000 tons/year (contributing around 70% of the total global production), followed by Italy with 120,572 tons/year, while Spain produces approximately 15,300 tons/year [6–8]. Portugal, France, United States (Oregon and Washington), New Zealand, China, Azerbaijan, Chile, Iran, Georgia, Kirgizstan, Poland, and Croatia are other producing countries [9–11].

Hazelnuts are rich in antioxidant compounds such as tocopherols and polyphenols, unsaturated fatty acids, flavan-3-ols (flavanols) and proanthocyanidins, essential amino acids, dietary fibers, vitamins (E and B), and complex and essential minerals (Ca, Mg, P, K), which have a wide range of

health benefits, including reducing oxidative stress and risk of cancer, stroke, inflammation, and other neurodegenerative diseases [7,12–14]. Hazelnuts are usually used in dairy, bakery, coffee, spreads, confectionery products, and salads due to their nutritious quality and unique flavor [14,15]. In addition, the hazelnut oil is used in cooking, cosmetics, and pharmaceutical industries [16,17]. Hazelnut is primarily composed of oil content (58–65%), protein (11–16%), fat content (60%), and carbohydrates (15–18%) [18,19].

The roasting process is performed to remove the pellicles of kernels, inactivate enzymes, destroy microorganisms, and reduce moisture [11,20]. In addition, roasting is used to improve the sensory features (color, crispy texture, appearance, and flavor) of the product [11,21,22]. Postharvest practices, such as handling, processing, and storage, could affect the biochemical composition and sensory properties of hazelnuts [23–25]. Against this background, knowledge of the physical and mechanical properties of the nuts is useful for the design and development of systems for handling, processing, packaging, storage, and transportation [26–30]. The physical properties of the nuts—mostly the average length, width, thickness, geometric mean diameter, sphericity, unit mass, volume, bulk density, true density, porosity, projected area, terminal velocity, and static and dynamic coefficients of friction—have been studied as a function of moisture content [30–33] where they tend to increase and/or decrease with increases in moisture content. The mechanical properties, on the other hand, include the rupture force, deformation at rupture point, deformation ratio at rupture point, hardness, and energy [1,34,35]. According to [30,36,37], the rupture force is the minimum force required to break the sample. The deformation at rupture point is the deformation at loading direction which can be used for the determination of the gap size between the surfaces to compress the fruit or nut for dehulling. Deformation ratio at the rupture point is the axial strain at the rupture point of the sample. Hardness is the ratio of rupture force and deformation at rupture point. The energy for rupture is the energy needed to rupture the sample, which is characterized by the area under the force–deformation curve. It is important to note that moisture content, speed, heating temperature, heating time, pressure, varieties, storage conditions, and vessel diameter thus affect the mechanical properties and force–deformation curves of oil-bearing material [38].

Based on the available literature, there is a lack of information on the postharvest processing of bulk hazelnut kernels in relation to processing factors such as speed, heating temperature, and relaxation time. To aid the design and development of optimal oil processing technology, it is necessary to enhance the knowledge of the uniaxial compression process. The uniaxial compression involves placing the bulk oil-bearing material in a pressing vessel with holes of about 0.5 mm around the bottom to allow the oil discharge while retaining the seed/kernel cake under a given load and speed using a universal testing machine [39,40]. The output data based on the preset pressing variables are usually the force–deformation and relaxation curves, which can further be analyzed to determine the mechanical properties and oil expression efficiency of the oil-bearing material under consideration [36,41,42]. The kernel oil extraction efficiency is determined based on the ratio of kernel oil recovered to that of the kernel oil content determined by Soxhlet extraction. The Soxhlet/solvent extraction technique involves repeated solvent distillation through a solid sample to remove the oil or analyte of interest [43]. Consequently, the study aimed to examine the effects of heating temperature, speed and relaxation time on hazelnut kernel oil processing in uniaxial pressure and to evaluate the chemical properties (peroxide value and free fatty acid) of the kernel oil in relation to heating temperatures.

#### **2. Materials and Methods**

#### *2.1. Samples and Moisture Content Determination*

Samples of bulk hazelnut kernels (Figure 1a) were from Ordu province of northeast Turkey, called the East Black Sea region of Turkey (40◦58.8 N, 37◦52.2 E). The Cakildak variety was used. The moisture content of 3.46 ± 0.11% (w.b.) was determined using the standard procedure where samples were randomly selected and dried at a temperature of 105 ◦C for 24 h [44]. An electronic

balance (Kern 440–35, Kern & Sohn GmbH, Balingen, Germany) having an accuracy of 0.01 g was used to measure the mass (g) of the samples before and after oven drying. To determine the moisture content value, Equation (1) was applied [45].

$$\text{MCC} = \left[ \left( \frac{m\_b - m\_a}{m\_b} \right) \cdot 100 \right] \tag{1}$$

where *MC* is the moisture content % (w.b.) and *ma* and *mb* are the masses of the samples before and after oven drying (g).

**Figure 1.** Set up of compression test and Soxhlet extraction: (**a**) hazelnut kernels and seed cake after compression; (**b**) Soxhlet extraction setup for kernel oil content determination; (**c**) a pan holding the pressing vessel with a plunger placed on two separated metal bars with a piece of white paper for oil point identification similar for the compression test; and (**d**) hazelnut kernel oil from the compression test and Soxhlet extraction (covered with cotton wool).

#### *2.2. Determination of Percentage Kernel Oil Content*

The percentage kernel oil content (Figure 1b) was determined to be (64.67 ± 0.01%) using the Soxhlet extraction method [46–48]. Based on the procedure, the sample mass of 8 g was ground and packed into a thimble. The thimble was placed in a Soxhlet extractor attached to a 150 mL round-bottom flask containing 100 mL of petroleum ether. The oil extraction process was allowed for 24 h involving several extraction cycles. The residual water and solvent in the extracted oil were removed by drying at 105 ◦C for 5 h. The kernel oil content was calculated based on the dry weight of the sample [45].

#### *2.3. Heating of Samples*

Samples were heated at temperatures between 40 and 60 ◦C using the oven device (MEMMERT GmbH + Co. KG, Schwabach, Germany) at a heating time of 30 min. The samples laboratory temperature of 25 ◦C served as the control.

#### *2.4. Compression Tests*

The universal compression machine (Tempos, ZDM 50, Czech Republic) together with the pressing vessel of diameter, *D* (60 mm) with a plunger (Figure 1c) were used to record the compression and relaxation curves of bulk hazelnut kernels at constant initial pressing height, *H*, of 40 mm (mass 58.31 <sup>±</sup> 0.23 g and volume 11.31·10−<sup>5</sup> m3). The preset load of 60 kN was determined from a preliminary test where the load was initially set at 100 kN and speed 5 mm/min, which indicated an undulation pattern on the force–deformation curve (ejection of kernel cake through the holes of the pressing vessel). Four separate experiments were performed. The first experiment varied the speed from 4 to 8 mm/min at force 60 kN without sample pretreatment. The second experiment considered pretreatment of the samples at temperatures between 40 and 60 ◦C for 30 min and compressed at a force of 60 kN and speed of 4 mm/min. The third experiment varied the relaxation time from 0 to 12 min for the samples with a heating temperature of 60 ◦C, speed of 4 mm/min, and force of 60 kN to determine maximum oil output. The speed and temperature values were selected based on the results of the preliminary study. The fourth experiment determined the first leakage of the oil (oil point) of the samples before and after pretreatment based on the litmus test procedure [39]. For identifying the lower and upper oil points with the corresponding strain, force, and energy values, the sample deformation levels between 15 and 25 mm were examined at a speed of 4 mm/min.

#### *2.5. Determination of Peroxide Value (PV) and Free Fatty Acid (FFA)*

The hazelnut kernel oil obtained from the compression test (Figure 1d) was evaluated based on peroxide value and free fatty acid. For peroxide value determination, an oil sample of mass of 5 g was measured into a volumetric flask and dissolved in 30 mL of a chloroform and glacial acetic acid mixture of ratio 2:3. Saturated potassium iodine (KI), 1 mL, was added to the solution and left in a dark place for 5 min. Distilled water (40 mL) was added and the solution was titrated with 0.1 M Na2S2O3 until the yellow color ceased to be visible. Subsequently, 1 mL of 1% starch was added, and the solution was titrated with Na2S2O3 until complete discoloration. For FFA determination, 5 g of the oil sample was measured into a titration flask, then a 100 mL of neutralized ethanol (warmed up to 60–65 ◦C) and 2 mL of 1% phenolphthalein were added. Titration was performed directly with ethanolic KOH (0.1 g/mol) until the color was light pink. Peroxide value was expressed as (μg/g of active oxygen) and free fatty acid was expressed as (mg KOH/g) [1,48,49].

#### *2.6. Compression Tests Calculated Parameters*

The deformation values were obtained directly from the compression tests. The oil yield (%) and oil expression efficiency (%) were calculated according to Equations (2) and (3), as given by [50,51].

$$
\Delta OY = \left[ \left( \frac{m\_o}{m\_s} \right) 100 \right] \tag{2}
$$

where *OY* is the oil yield (%), *mo* is the mass of oil obtained as the difference of mass of seed cake and initial mass of the sample *ms* (g).

$$OEE = \left[ \left( \frac{OY}{O\_s} \right) \text{100} \right] \tag{3}$$

where *OEE* is the oil expression efficiency (%) and *Os* is the percentage of oil content (%). The energy (kJ) was calculated based on the area under the force–deformation curve, which is given by Equation (4) [52,53].

$$E = \sum\_{n=0}^{n-i-1} \left[ \left( \frac{F\_{n+1} + F\_n}{2} \right) \cdot (\mathbf{x}\_{n+1} - \mathbf{x}\_n) \right] \tag{4}$$

where *E* is the deformation energy (kJ), *Fn*+<sup>1</sup> + *Fn* and *xn*+<sup>1</sup> − *xn* are the compressive force (kN) and deformation (mm), *n* is the number of data points, and *i* is the number of sections in which the axis deformation was divided.

#### *2.7. Statistical Analyses*

All experiments were repeated twice based on prior knowledge of the statistical techniques applied and the reliability of the results of the preliminary experiment. The data were statistically analyzed using STATISTICA software (Version 13) [54] by employing the ANOVA (San Francisco, CA, USA), correlation, and regression techniques.

#### **3. Results**

#### *3.1. E*ff*ects of Speed, Temperature, and Relaxation Time on Responses*

The effects of speed, temperature, and relaxation time on deformation, oil yield, oil expression efficiency, and energy of hazelnut kernels are presented in Tables 1–3, respectively. The speed and temperature did not greatly influence the deformation values. The lower speed of 4 mm/min combined with a higher temperature of 60 ◦C increased the oil expression efficiency by 4.68%. A higher speed rather increased the energy for recovering the kernel oil compared to a higher temperature. This could be attributed to the softening of the kernels and occurrence of the serration effect, which increases the area under the force–deformation curve, which is defined as the energy. Lower speed and higher temperature recorded an equal amount of energy of 0.21 kJ. The relaxation time of 12 min after compression increased the oil expression efficiency, meaning that the residual kernel oil of 15.6% was recovered from the kernel cake after compression.

**Table 1.** Effect of speed on deformation, oil yield, oil expression efficiency, and energy.


**Table 2.** Effect of heating temperature on deformation, oil yield, oil expression efficiency, and energy.


\* Sample compressive force of 60 kN and speed of 4 mm/min.

**Table 3.** Effect of relaxation time on sample heating temperature of 60 ◦C.


\*\* Compressive force of 60 kN and speed of 4 mm/min.

#### *3.2. Force–Deformation Curves, Strain, Oil Point Force, and Oil Point Energy*

The force–deformation curves for different speeds and heating temperatures are illustrated in Figures 2 and 3. The strain, oil point force, and oil point energy are also given in Tables 4 and 5. The maximum applied force of 60 kN was based on a preliminary experiment where a higher force using a pressing vessel diameter of 60 mm caused a serration effect characterized by the ejection of kernel cake through the holes of the pressing vessel. The strain is the ratio of deformation to the initial height of the sample and is useful for identifying the oil point force with the corresponding energy. The lower and upper oil points were observed at deformation values of 20 and 25 mm at a speed of

4 mm/min with the corresponding strain values of 0.5 and 0.625 (−) for both heat-treated kernels and those without heat treatment. The oil point force and energy of the samples before and after heating were determined from the deformation levels between 15 and 25 mm (Figure 4). The lower oil point denotes the minimum force required for the first drop of kernel oil whiles the upper oil point describes the maximum force needed to recover maximum kernel oil.

**Figure 2.** Force–deformation curves of bulk hazelnut kernels for different speeds.

**Figure 3.** Force–deformation curves of bulk hazelnut kernels for heating temperatures.

**Table 4.** Strain and oil point force of deformation levels of kernels at a speed of 4 mm/min.


\* For initial pressing height, *H* = 40 mm; <sup>a</sup> Lower oil point; <sup>b</sup> Upper oil point; \*\* (Control at 25 ◦C).

**Table 5.** Strain and oil point force of deformation levels of kernels at a speed of 4 mm/min.


\* For initial pressing height, *H* = 40 mm; <sup>a</sup> Lower oil point; <sup>b</sup> Upper oil point; \*\*\* Heating temperature of 60 ◦C.

**Figure 4.** Oil point identification at deformation levels of bulk hazelnut kernels.

#### *3.3. Peroxide Value and Free Fatty Acid*

In Figure 5 is illustrated the box plot of values for peroxide and free fatty acid of kernel oil at different heating temperatures. The peroxide values of hazelnut kernel oil increased from 40 to 50 ◦C and then decreased at 60 ◦C but the free fatty acid values decreased linearly with heating temperatures.

**Figure 5.** Box plot of peroxide value and free fatty acid of hazelnut kernel oil under heat treatment temperature.

#### *3.4. ANOVA, Correlation, and Regression Results*

The results of ANOVA, correlation, and regression of the effects of speed, heating temperature, and relaxation time are given in Tables 6–10. The ANOVA results show that the increase in speed and heating temperature did not significantly (*p* > 0.05) decrease or increase the deformation values of hazelnut kernels. Oil yield and oil expression efficiency significantly (*p* < 0.05) decreased with increased speed while they increased significantly (*p* < 0.05) with heating temperature. The speed and heating temperature combined did not indicate a significant effect (*p* > 0.05) on energy. The correlation results also showed no relationship between deformation and speed as well as deformation and temperature. Oil yield and oil expression efficiency correlated with speed and heating temperature effects with high correlation efficiency between 97 and 99% respectively. Energy showed no correlation with the effects of speed and heating temperature (*p* > 0.05). The oil point force, oil expression efficiency, and energy were statistically significant (*p* < 0.05) with a coefficient of determination (R2) of 87%. Peroxide value was not significant (*p* > 0.05) with heating temperature but free fatty acid proved significant (*p* < 0.05) with a correlation efficiency of 95%. The regression equations representing oil yield, oil expression efficiency, and free fatty acid as functions of speed, heating temperature, and relaxation time effects, respectively, indicated significance (*p* < 0.05). The energy and peroxide value against speed and heating temperature revealed a non-significant (*p* > 0.05) effect.


**Table 6.** ANOVA and correlation results of the effect of speed and heat treatment temperature on the dependent variables.

Compressive force of 60 kN and speed 4 mm/min; <sup>a</sup> Speed; <sup>b</sup> heating temperature; Significant (*p*-value < 0.05 or F-value > *p*-value); Non-significant (*p*-value > 0.05 or F-value < *p*-value).

**Table 7.** ANOVA and correlation results of the effect of relaxation time on oil expression efficiency.


\* Heating temperature of 60 ◦C; compressive force of 60 kN; speed of 4 mm/min.

**Table 8.** ANOVA and correlation results of oil point force, oil expression efficiency and energy.


\* Speed of 4 mm/min; <sup>a</sup> Control at 25 ◦C; <sup>b</sup> Heating temperature of 60 ◦C.

**Dependent variables R R<sup>2</sup> F-Value** *p***-Value** Peroxide value, *PV* (μg/g of active oxygen) <sup>−</sup>0.43 0.64 2.70 <sup>&</sup>gt;0.05 Free Fatty Acid, *FFA* (mg KOH/g) <sup>−</sup>0.95 0.99 108.37 <sup>&</sup>lt;0.05

**Table 9.** ANOVA and correlation results of *PV* and *FFA* in relation to heating temperature.

**Table 10.** Regression results of the effects of speed, heating temperature, and relaxation time.


*Tp*: Heating temperature (◦C); *Sp*: Speed (mm/min); *Rt*: Relaxation time (min); Significant (*p*-value < 0.05 or F-value > *p*-value); Non-significant (*p*-value > 0.05 or F-value < *p*-value).

#### **4. Discussion**

The experimental and statistical results of the effects of speed, heating temperature, and relaxation time on deformation, oil yield, oil expression efficiency, force–deformation curves, peroxide value, and free fatty acid of postharvest processing of hazelnut kernels and oil under uniaxial compression have been described in the preceding sections. The results are herein elaborated.

A lower speed of 4 mm/min recovered maximum kernel oil compared to a higher speed of 12 mm/min, where the minimum oil yield was realized. Increases in speed increased the energy but decreased the oil yield and oil expression efficiency whereas increases in heating temperature showed a significant increase in oil yield and oil expression efficiency but not in energy. Similar results were reported by [55] for jatropha kernels heated at 80 ◦C. It has been reported that the increase in heat will not always increase the oil yield [56]. The oil yield, however, is dependent on the quality of the oil-bearing material, moisture content, and the oil extraction process. Theoretically, higher screw speed means more seed material throughput and higher oil content residual in the press cake due to less time being available for the oil to discharge from the solids. A lower energy input also means lower oil recovery efficiency, higher oil residue in the press cake, and higher seed material throughput, and higher pressure will lead to higher temperature generation and higher oil recovery efficiency [57–59].

The relaxation time for kernels heated at 60 ◦C and compressed at a force of 60 kN and speed of 4 mm/min increased the oil yield and oil expression efficiency. A significant increase in oil yield and oil expression efficiency occurred at 3 and 6 min of relaxation time in comparison with 9 and 12 min of the relaxation time. However, the percentage increase of 10.1% and 15.6% of oil yield and oil expression efficiency was observed at 12 min relaxation. The correlation efficiency between oil expression efficiency and relaxation time was 91%. The compression and relaxation processes are important for evaluating the oil yield and energy at the compression area and oil expression efficiency without energy at the relaxation phase, which is dependent on several factors such as the volume of kernels, moisture content, vessel diameter, force, and speed, among other factors [48]. Further, from the principles of viscoelasticity, the relaxation time shows how fast the material dissipates after receiving a sudden deformation, that is, when the maximum force is attained, and it can also be used to characterize the elastic and viscous components in the behavior of the material. The relaxation data can be used to develop rheological models useful in the design of food engineering processes by applying the concepts of momentum, energy, and mass balances [60,61].

The lower oil point force of 6.21 ± 0.58 kN and the upper oil point force of 10.61 ± 0.71 kN of hazelnut kernels were observed. Based on these values, a maximum force of 60 kN was applied to recover the maximum kernel oil. However, it was also observed that the oil point forces at both lower and upper oil points for heat-treated kernels were lower compared to kernels without heat treatment. This indicates that kernels without heat treatment are more resistant to breakage and, hence, higher energy than those with heat treatment. The oil point force decreased with increased heating temperature. The results are in agreement with [62] based on cashew kernels. The oil point pressure/force indicates the threshold pressure at which oil emerges from the seed or kernel during mechanical or uniaxial compression [63–65]. The effective pressure for the oil to discharge from oil-bearing cells is determined by first identifying the oil point pressure. The oil point pressure improves the oil expression efficiency and provides useful information for the design and performance evaluation of oil expellers [62,65]. Nevertheless, the oil point force is dependent on the initial moisture content, heat treatment, heating time, speed, vessel diameter, and kernel pressing height and varieties. This suggests that the optimal processing factors for identifying the oil point pressure/force with corresponding energy and for recovering higher percentage kernel oil need to be determined to help in the design of energy-efficient oil processing system.

Peroxide value and free fatty acid of hazelnut kernel oil decreased at the highest heating temperature of 60 ◦C. This indicates that the quality of the oil at the abovementioned heating temperature might be within the acceptable limit. The explanation is that higher temperature leads to the higher acid value, resulting in high free fatty acid. The lower the acid value, the lower the free fatty acid hence, the better the oil usage quality [55]. Peroxide value and free fatty acid are strongly correlated [66]. Generally, peroxide values in edible oils indicate lipid oxidation whilst free fatty acids indicate lipid hydrolysis [67–69]. Oxidation of edible oils reduces oil quality and human health benefits and, therefore, it is important to minimize lipid oxidation and rancidity [70]. Peroxide values depend on the cultivar, harvest time, delaying of storage after harvest, ecological conditions, cultural management during the growing season, and handling procedures after harvest [71]. On the other hand, enzymatic lipid hydrolysis produces free fatty acids which are driven by a reaction between moisture and oil [72,73].

In addition to the abovementioned facts, mechanical pressing of oil from oil-bearing materials involves the application of pressure by using either a hydraulic or screw press to recover the oil [52,64]. Improving the performance of the oil expression process and/or designing an optimum universal pressing system requires understanding of the mechanical and relaxation behaviors of the oil material under uniaxial compression. The results of the present study thus provide supplementary information to the available literature.

#### **5. Conclusions**

A laboratory-scale study was conducted to examine the mechanical behavior, oil content, and oil efficiency of hazelnut kernels with different speeds, heating temperatures, and relaxation times. The quality of the kernel oil was assessed in terms of peroxide value and free fatty acid. The following findings were observed. For a pressing vessel of diameter 60 mm and sample pressing height of 40 mm, kernel oil efficiency was achieved at a maximum force of 60 kN and speed of 4 mm/min. The maximum force applied was based on a preliminary experiment, which showed that a greater force under the same processing conditions demonstrated the serration effect, which is not energy efficient for recovering the maximum kernel oil. The lower and upper oil points with strain values of 0.5 and 0.625 with the corresponding forces of 6.21 ± 0.58 and 10.61 ± 0.71 kN were in reference to the maximum force applied for recovering the kernel oil. Peroxide value, *PV* (μg/g of active oxygen), and free fatty acid, *FFA* (mg KOH/g), of hazelnut kernel oil decreased with heating temperature, indicating that the quality of the oil at 60 ◦C is acceptable for use. Regression models representing oil yield, oil expression efficiency, and free fatty acid were described based on the predictors of speed, heat treatment temperature, and relaxation time. The coefficients of the model predictors were statistically significant (*p* < 0.05) for describing the responses. The relaxation time of 12 min after compression increased the kernel oil expression efficiency of 15.6%, specifying that the compression and relaxation processes for each pressing chamber along the screw configuration need to be considered when designing a screw press to reduce the additional energy for recovering the residual kernel cake oil.

**Author Contributions:** Conceptualization, G.A.K.G.; A.K.; C.M. and D.H.; Funding acquisition, D.H.; Methodology, ˇ G.A.K.G.; A.K.; C.M. and A.F.; data curation, A.K.; ˇ C.M.; K.Ç.S. and A.F.; Writing—original draft, G.A.K.G.; ˇ A.K. and K.Ç.S.; Writing—review and editing, G.A.K.G.; A.K., K.Ç.S.; C.M. and D.H. All authors have read and ˇ agreed to the published version of the manuscript.

**Funding:** The research through the project "supporting the development of international mobility of research staff at CULS Prague, was funded by EU, Managing Authority of the Czech Operational Programme Research, Development and Education" Grant Number: CZ.02.2.69/0.0/0.0/16\_027/0008366.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Physico-Chemical Changes, Microbiological Properties, and Storage Shelf Life of Cow and Goat Milk from Industrial High-Pressure Processing**

#### **Sok Fern Tan 1, Nyuk Ling Chin 1,\*, Tuan Poy Tee <sup>2</sup> and Sook Kuan Chooi <sup>3</sup>**


Received: 21 March 2020; Accepted: 11 April 2020; Published: 16 June 2020

**Abstract:** Industrial high-pressure processing (HPP) was conducted on cow and goat milk in comparison to conventional heat pasteurization. No significant changes were found in the physico-chemical properties of the treated milk except for pH, where pasteurized cow milk experienced a decrease while goat milk's pH increased for both pasteurized and HPP treated. HPP-treated cow and goat milk both achieved microbial shelf life of 22 days at 8 ◦C storage with no increase in *Bacillus cereus*, mesophilic aerobic spores, coliform, yeast and mold but slight increase in psychrotrophic bacteria and total plate count. Pasteurized goat milk was spoilt at the end of storage with exceeding count of psychrotrophic bacteria (9.0 <sup>×</sup> 10<sup>8</sup> CFU/mL) and total plate count (3.5 <sup>×</sup> 108 CFU/mL). HPP-treated cow milk exhibited higher physico-chemical stability than goat milk as evidenced by non-significant change of titratable acidity but goat milk experienced an increase of 0.04% averagely.

**Keywords:** milk quality; high pressure processing; pasteurization; milk storage; shelf life

#### **1. Introduction**

Milk is well-known as a major source of dietary energy, protein, and fat. It contributes 134 kcal of energy/capita per day on average [1]. Apart from cow milk, goat milk has also started to gain interest worldwide in recent years [2]. Goat milk contains lower amount of caseins by 89% compared to cow milk, hence gives a greater digestive utilization compared to cow milk [3]. However, the nature of milk and its chemical composition render it as one of the ideal culture media for microbial growth and multiplication [4]. Hence, the major challenge in milk industries is to ensure delivery of safe and healthy milk products of consistent quality to an ever-increasing demand of milk.

Thermal treatments such as pasteurization high temperature short time (HTST–72 ◦C/15 s), low temperature long time (LTLT–63 ◦C/30 min), and ultra-high temperature (UHT–135 ◦C/3 s) are usually used to treat raw milk in order to be rendered safe for consumer and increase its shelf life [5]. Studies have shown that microbial population counted is significant beyond 20 days of storage for pasteurized milk [6–8]. UHT-treated milk typically has up to 4–6 months of commercial shelf life at ambient temperature [9]. However, these conventional heat treatments have detrimental effects on the nutrient content of milk. Studies have proven that thermal treatments caused 5–15% of whey protein denaturation, 10–25% of water soluble vitamins destruction, disturbed structure of casein micelles, and also enhancement of allergenic properties of milk proteins [5,10,11].

High pressure processing (HPP) is a promising alternative to conventional thermal treatments as it is capable to inactivate foodborne pathogens while minimizing the loss of nutrients, such as water-soluble vitamins and maintaining the fresh-like characteristics of food products [12–14]. During HPP, pressure is allowed to remain in contact with the product for a specific holding time to activate the destruction of spore and microbes, and formation of non-covalent bonds of food components to gelatinize the enzyme [13,15,16]. Pressure regime of 200–600 MPa with holding time of 3–10 min is usually applied in HPP of milk. Generally, higher pressure resulted in higher rate of microbial destruction as well as longer shelf life. Pereda et al. [17] reported a shelf life of 14–18 days for milk treated at 200 MPa at storage temperature of 4 ◦C. Mussa and Ramaswamy [18] reported that pressure of 350 MPa renders a milk shelf life of 18 days at similar storage temperature. When milk was treated at 600 MPa, shelf life of 28 days was achieved at similar storage temperature [14]. Storage temperature of milk also plays a role in milk's shelf life. Rademacher and Kessler [19] reported that milk subjected to pressure treatment of 400 MPa for 15 min or 600 MPa for 3 min could only achieve 10 days of shelf life when stored at 10 ◦C, which is 18 days shorter than that of storing at 4 ◦C.

While studies about the HPP effects on nutrient content like fat, protein, and moisture are very limited, several studies have proven that HPP is able to preserve vitamins in milk better than pasteurization and UHT [20,21]. From the microbiological perspective, a more pronounced microbial inactivation was obtained by increasing HPP pressure levels and holding time. Alexandros et al. [14] reported that HPP above 550 MPa and 3 min reduced the number of *E. coli*, *Salmonella* spp., and *L. monocytogenes* to below limit of detection while pressure regime of 400 MPa for 1 min did not result in statistically significant differences in reduction levels. Yang et al. [22] reported that holding time of 30 min at 300 MPa is sufficient to inactivate *Salmonella* spp., *E. coli*, *Shigella,* and *S. aureus*.

Most HPP studies conducted on milk are based in countries with seasonal climates. For a hot tropical country like Malaysia, HPP for milk is a new technology. Because of the high consumption of pasteurized UHT milk, this study was conducted to investigate industrial HPP effects on local cow and goat milks in terms of its physico-chemical and microbiological properties. The changes in milk properties were measured at storage temperature of 8 ◦C following commercial handing practices for 22 days and compared with pasteurized milk.

#### **2. Materials and Methods**

#### *2.1. Milk Sampling and Treatment*

Total of 10 liters of fresh cow milk was collected from the university farm, Ladang 16, Universiti Putra Malaysia while fresh goat milk was collected from a private farm, Janggut Ternak Farm, Dengkil, Selangor, Malaysia. The milk was maintained below 10 ◦C during transportation to the HPP facility (Kara F&B Productions, Sdn. Bhd.) within 1 h for bottling into HPP grade plastic bottles of 250 mL after a thorough stirring. Three bottles of milk sample, each for cow and goat was used directly for analysis as control or fresh milk samples.

The two HPP pressures applied were 450 MPa and 600 MPa for two durations, 5 min or 7 min at 15 ± 2 ◦C. The three HPP treatments combinations were 450 MPa/7 min, 600 MPa/5 min, and 600 MPa/7 min giving a total of 27 milk samples, i.e., 3 HPP treatments × 3 storage durations of 0, 11-, and 22-days × 3 analysis which include the physico-chemical and microbiological properties. Both cow and goat milk samples were loaded into the pressure chamber of HPP unit with 55 L volume (Hiperbaric 55, Hiperbaric, Spain) at the same time for the same treatment.

Pasteurization of cow and goat milk were done at 72 ◦C for at least 15 sec (HTST–72 ◦C/15 s). Milk was poured into beakers which were pre-heated with boiling water to ensure sterile condition and heated in the water bath (BS-21, Jeio Tech, South Korea) with temperature monitored using clean thermocouples with regular stirring. After pasteurization, the milk was cooled rapidly to room temperature in an ice cooler box filled with ice cubes. Then, the milk was bottled into 9 plastic bottles (3 storage days × 3 analysis) each for cow and goat milk. All treated milk samples were kept in a chiller (RV710, Hitachi, Japan) at 8 ± 2 ◦C for storage until analysis.

#### *2.2. Physico-Chemical Analysis*

The pH values of both cow and goat milk samples were determined using a portable pH meter (Mi805, Milwaukee, Hungary). The pH meter was calibrated to the pH of the buffer solution at 25 ◦C. Titratable acidity was determined following AOAC Official Method 947.05. The specific gravity was determined following the method of AOAC 925.22 using a pycnometer. Total protein was determined using Pyne's method which is known as formaldehyde titration [23]. This method was proven to be having the same accuracy as compared to Kjeldahl method, as the differences were insignificant between protein contents measured by using both methods [24]. The total protein was calculated using factor of 1.74. The total solid of both cow and goat milk was determined using the oven drying method (UM500, Memmert, Germany) in accordance to AOAC Official Method 16.032 where total solid in percentage was calculated as follow:

$$\text{Total solid } (\%) = \frac{\text{Weight of disk} + \text{dired sample} - \text{Weight of empty disk}}{\text{Weight of sample}} \times 100\tag{1}$$

The total fat percentage was determined following the AOAC Official Method 905.02, which is also known as the Roese-Gottlieb method. The total fat was calculated as follows:

$$\text{Total Fat} \left( \% \right) = \frac{\text{Weight of flask containing fat} - \text{weight of flask after fat washing}}{\text{Weight of sample}} \times 100 \qquad (2)$$

The milk solid-non-fat (SNF) was determined by subtracting the total solid (%) with the total fat content (%).

#### *2.3. Microbiological Analysis*

Microbiological analysis was conducted in an aseptic environment to prevent cross contamination. The total plate count (TPC) and total yeast and mold count in milk were determined following methods in FDA Bacteriological Analytical Manual (BAM) Chapter 3 and 18, respectively. The total coliform count was determined following AOAC Official Method 991.14. The mesophilic, thermophilic aerobic spore count and psychrotrophic bacteria count were determined following Compendium of Methods for the Microbiological Examination of Foods (CMMEF) Chapter 23, 26, and 13, respectively. *Bacillus cereus*, *Staphylococcus aureus,* and *Listeria monocytogenes* were determined following FDA BAM Chapter 14, 12, and 10, respectively. *Escherichia coli* and *Clostridium perfringens* were determined by AOAC Official Method 991.14 and 976.30, respectively. *Salmonella* spp. was determined using enzyme-linked immunosorbent assay (ELISA) method.

#### *2.4. Experimental Design and Statistical Analysis*

All measurements were done in duplicates with samples prepared from the same batch of milk. Minitab software (Version 16, Minitab Statistical Software, United States) was used to perform analysis of variance (ANOVA). Tukey's test was performed to compare the differences among the mean values at a confidence level of 0.05.

#### **3. Results and Discussion**

#### *3.1. E*ff*ects of HPP on Physico-Chemical Properties*

Physico-chemical properties of cow and goat milk before and after the treatment are presented in Table 1. The pH of pasteurized cow milk decreased significantly (*p* < 0.05) from 6.59 ± 0.01 to 6.52 ± 0.00. In contrast, the pH of pasteurized goat milk increased significantly (*p* < 0.05) from 6.34 ± 0.00 to 6.38 ± 0.01. Ma and Barbano [25] reported that pH of cow skim milk decreases linearly with increasing temperature during treatment. However, José et al. [26] reported that pasteurization increased pH of bovine milk because of the lower whey protein associated with the casein micelles. Tamime [10]

explained that the increase of pH is due to the reduced casein micelles where heat denatured its whey protein. For HPP-treated milk, pH for cow milk only changed slightly but statistically insignificant (*p* > 0.05). Nonetheless, pH of goat milk has increased significantly (*p* < 0.05). This increment is mainly due to the alteration of the distribution of minerals by HPP, which includes calcium, phosphate, and other ionized minerals such as calcium [27]. Other studies which have observed increase in the milk pH after HPP include Liepa et al. [28] who found that pressure of 500–600 MPa for 15 min increased the pH of cow milk; while Zobrist et al. [29] reported that pressure of 100–600 MPa for 30 min also caused an average 0.08 units of pH increment in cow milk.


**Table 1.** Physico-chemical properties of cow and goat milk before and after treatment.

Note: Mean ± standard deviation in same row with different superscripts letters are significantly different at *p* < 0.05.

In general, the results show that the titratable acidity, specific gravity, total protein, total fat, total solid, and solid-non-fat in both cow and goat milk were not influenced by both HPP and pasteurization. The values changed slightly after HPP treatment but they were statistically insignificant when compared with untreated samples (*p* > 0.05).

#### *3.2. E*ff*ects of HPP on Microbiological Properties*

Table 2 summarizes the microbiological properties of cow and goat milk before and after the treatment. According to the Malaysian Standard for pasteurized milk (MS 410:1995), the total plate count (TPC) should not be more than 105 CFU/mL. In this study, the initial TPC of raw cow and goat milk was 1.5 <sup>×</sup> <sup>10</sup><sup>5</sup> CFU/mL and 6.9 <sup>×</sup> <sup>10</sup><sup>6</sup> CFU/mL, respectively. Both HPP and pasteurization produced desirable results of decreased TPC in both milk samples. The pressure regime of 600 MPa/5 min led to the highest TPC reduction in both milk, i.e., 99.98% for the cow milk and 99.99% for the goat milk. The lowest TPC reduction was found in the pressure regime of 450 MPa/7 min (96%) for the cow milk and for goat milk, it was from the pasteurization process (99.98%). This is in line with the findings from Alexandros et al. [14] who found that HPP (600 MPa/3 min) led to a more pronounced decrease of TPC than that of pasteurization.


**Table 2.** Microbiological properties of cow and goat milk before and after treatment.

Psychrotrophic bacteria are undesirable for prolonged storage of milk because they are able to grow below 7 ◦C even though their optimum temperature ranges from 20 ◦C to 30 ◦C. They also produce heat resistant enzymes, including proteolytic and lipolytic enzymes at low temperatures, which are able to hydrolyze milk fat and protein leading to the formation of off-flavors [30]. According to Malaysians Food Act 1983 and Food Regulations 1985, psychrotrophic bacteria count in milk should not exceed 10<sup>5</sup> CFU/mL. In this research, the initial psychrotrophic bacteria count in raw cow and goat milk were 9.2 <sup>×</sup> 10<sup>4</sup> CFU/mL and 8.0 <sup>×</sup> 10<sup>5</sup> CFU/mL, respectively. Both HPP and pasteurization have successfully decreased psychrotrophic bacteria count of both milk samples. The lowest psychrotrophic bacteria count in cow milk was recorded in the pressure regime of 600 MPa/7 min where bacteria was undetectable (<1 CFU/mL) while the highest count was found in the pressure regime of 450 MPa/7 min (8.3 <sup>×</sup> 10<sup>2</sup> CFU/mL). For goat milk HPP treatment at all pressure regimes have reduced the bacteria count to undetectable level (<1 CFU/mL). The highest psychrotrophic bacteria count recorded was in the pasteurized sample which is 3.6 × 10 CFU/mL.

Mesophilic spores are known as the most prevalent spore-former found in bulk milk processing tank. In spore form, these organisms are capable of surviving environmental stresses including low pH, exposure to sanitizers, high pressure and temperature [31]. Mesophilic spores were undetected in all goat milk samples both before and after treatments. For cow milk, the initial mesophilic aerobic spore count was 70 CFU/mL. All pressure regimes in HPP successfully inactivated the spores. However, pasteurization only managed to reduce it to 20 CFU/mL. McGuiggan et al. [32] explained that these aerobic spores have the ability to survive in high temperature as well as pasteurization.

Coliforms are Gram-negative and non-spore forming bacteria that ferment lactose with the production of acid at 35 ◦C within 48 h [33]. According to MS 410:1995, the total coliform should be absent in 0.1 mL of sample. In this study, the total coliform count for untreated cow milk was 37 CFU/mL, but after treatment, the coliform became undetected in all HPP-treated and pasteurized samples. While all pressure regimes of HPP were able to reduce the total coliform count to undetectable level in goat milk, pasteurization was capable to reduce the counts from 2.1 <sup>×</sup> 105 CFU/mL to 26 CFU/mL (99.98% reduction).

Yeast and mold are a large and diverse group which include several hundred species. Although yeast and mold are relatively HPP sensitive [12], mascospores of heat-resistant molds such as *Byssochlamys*, *Neosartorya,* and *Talaromyces* are generally considered to be extremely HPP resistant. In this research, the total yeast and mold count of untreated cow and goat milk were 60 CFU/mL and 80 CFU/mL, respectively. Both HPP and pasteurization were effective in inactivating yeast and mold in cow and goat milk fully. Similar result was reported by Evrendilek [34], where most vegetative yeast and mold were inactivated within 5–10 min by 300–400 MPa at 25 ◦C.

Aerobic spore-forming bacteria of the genus *Bacillus* are commonly present in raw milk. Their spores survive pasteurization and subsequently germinate, outgrow, and multiply [35]. In this study although *B. cereus* was not found in raw cow milk, it was present in HPP-treated and pasteurized milk with the highest count of 20 CFU/mL in milk subjected to 450 MPa for 7 min. The unexpected presence of *B. cereus* in treated cow milk may be due to several factors. McClements et al. [35] reported that *B. cereus* spores were more resistant to pressure than vegetative cells. The contamination, germination, and outgrowth of *Bacillus* spores in milk can occur during simple transfer during milking and bottling when hygiene conditions are not fully observed. Post-pasteurization *Bacillus cereus* adherence to stainless steel contact surfaces in the form of biofilm formation has been reported to occur on milk contact surfaces [36]. Nutrient germinant such as amino acids and unsaturated fatty acids, levels of some indigenous antibacterial factors such as lysozyme, lactoferrin, and lactoperoxidase, somatic cells and metal ions have been suggested as factors influencing the germination of *Bacillus* species [32,37]. *B. cereus* was not found in both raw and treated goat milk.

*C. perfringens*, *S. aureus*, *E. coli*, *L. monocytogenes*, *Salmonella* spp., and thermophilic spores were undetected in both cow and goat milk. The absence of these bacteria and spores is due to good milking hygiene, breeds, climate and geographic location. In India, Chaturvedi and Shukla [38] reported that 51.61% of raw milk samples were found to be highly contaminated with *Clostridium* species due to unhygienic milking practice. Mohamed et al. [39] observed no germs of *S. aureus* and spore of *Clostridium* spp. in all cow and goat milk samples at Djibouti. However, in the same research, the *E. coli* count of 2.58 log CFU/mL was significantly higher than standard value of 2 log CFU/mL in their milk samples. Chye et al. [40] reported that only 1.9% and 1.4% of cow milk samples in Malaysia were contaminated by *L. monocytogenes* and *Salmonella* spp., respectively. Higher incidence of *Listeria* was obtained from eastern region while *Salmonella* spp. was found more prevalent in central region of Peninsular Malaysia. Hence, the presence of these bacteria is highly dependent on the location and hygiene.

#### *3.3. Changes in Physico-Chemical Properties in Milk during Storage*

The pH for both HPP-treated and pasteurized cow and goat milk decreased significantly (*p* < 0.05) during a 22-day storage period (Figure 1). Similar results were obtained by Siddique et al. [41] who reported pH declination at the end of 90 days storage of UHT milk. Brodziak et al. [42] mentioned about statistically significant pH reduction of open drinking cow milk after 7 days of storage. The decreasing trend in pH during milk storage is due to the acidity increase in milk caused by lactic acid and fatty acid [43]. It is obvious that pH of goat milk was always lower than cow milk which may be due to its higher acidity content as shown in Figure 2. This is in agreement with Mahmood and Usman [44] who also found lower pH in goat milk (6.48) than cow milk (6.59).

Generally, the titratable acidity of both cow and goat milk increased during the 22-day storage (Figure 2). The increment found in all HPP-treated cow milk was statistically insignificant (*p* > 0.05) as compared to significant increment (*p* < 0.05) for pasteurized cow milk with highest acidity of 0.23% at day 11. For goat milk, all HPP and pasteurized samples titratable acidity was found to have increased significantly, from 0.25 to 0.26% and 0.30%, respectively at the end of 22-day storage from the range of 0.21 to 0.23% at day 0. It is thus obvious that both pasteurized cow and goat milk deteriorated faster than HPP-treated milk. Brodziak et al. [42] also reported increase of titratable acidity in open drinking cow milk after 7-day storage in pasteurized milk, followed by micro-filtered milk and UHT milk. The increase in titratable acidity is mainly due to increasing lactic acid concentration after degradation

of lactose caused by predominant spoilage lactic acid bacteria (LAB) under low oxygen, temperature, and acidic condition [41,45].

**Figure 1.** Changes in pH of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk, and dotted lines represent fresh milk before treatment.

**Figure 2.** Changes in titratable acidity of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk, and dotted lines represent fresh milk before treatment.

The specific gravity of both cow and goat milk increased gradually when approaching day 22 of storage (Figure 3). All HPP-treated and pasteurized samples showed a statistically significant increase (*p* < 0.05) in specific gravity except for goat milk in the pressure regime of 450 MPa/7 min which shows significant decrease (*p* < 0.05). The increase in specific gravity agrees with Aldubhany et al. [46]'s work who reported increment in specific gravity in cow milk samples after 6 months of storage at various storage temperatures. The specific gravity of milk could be influenced by the proportion of its constituents such as fat and lactose which degrade during storage [45]. Ahmad et al. [47] proved that milk specific gravity decreased due to mastitis. The authors explained that presence of mastitis contributes to the decrease in lactose and increase in chloride contents as specific gravity is positively correlated with lactose and negatively correlated with chloride.

**Figure 3.** Changes in specific gravity of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk and dotted lines represent fresh milk before treatment.

Figure 4 shows that the total protein content in both cow and goat milk samples from both HPP and pasteurization treatment did not change significantly (*p* > 0.05) after 22 days. This is supported by Brodziak et al. [42] who found that storage duration of milk had no statistically significant effect on its protein content. Figure 5, however shows that the total fat content of both cow and goat milk samples from pasteurization treatment have reduced significantly (*p* < 0.05) starting from day 11 for cow milk and day 22 for goat milk while no significant difference (*p* > 0.05) was found for HPP samples. The total fat for both pasteurized cow and goat milk have reduced from 3.32% to 2.94% while for goat milk from 4.34% to 3.11%. HPP has been proven to have no damage on the milk fat globule membrane where lipolysis incidence could be prevented, thus the milk fat content could be well preserved [34,48]. Zajác et al. [43] has reported the significant decrease in fat content of raw cow milk after 24 h of storage at a temperature of 4 ◦C to be due to the lipolysis of milk fat initiated by both indigenous milk lipases and also microbial lipases which usually grew in number during storage. Significant increase of free fatty acids as a consequence of lipolysis in UHT cow milk after three weeks of storage was also confirmed by Janstová et al. [49]. These analyses have also shown that goat milk has relatively higher total protein and fat contents than cow milk.

**Figure 4.** Changes in total protein of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk and dotted lines represent fresh milk before treatment.

**Figure 5.** Changes in total fat of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk, and dotted lines represent fresh milk before treatment.

Figures 6 and 7 show a mimicking trend of total solid and solid non-fat contents of cow and goat milk over the fat content trend, i.e., no significant difference during 22 storage for HPP treated samples and significant decrease (*p* < 0.05) for pasteurized samples. The significant decrease in fat content could affect milk composition in terms of total solid and non-fat contents [43]. Omer and Eltinay [50] have reported similar results where total solid and non-fat content of camel milk reduced significantly after two weeks of storage at 7 ◦C.

**Figure 6.** Changes in total solid of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk, and dotted lines represent fresh milk before treatment.

#### *3.4. Changes in Microbiological Properties in Milk during Storage*

Table 3 presents the count of two microbiological measurements that turned positive during 22 days of storage. The psychrotrophic bacteria population in both cow and goat milk grew progressively throughout the storage period. At the end of 22 days, the best treatment was HPP at 600 MPa/7 min giving the smallest colony count in cow milk, followed by 600 MPa/5, pasteurization and 450 MPa/7 min. All these counts, however, are still far below the limit of 10<sup>5</sup> CFU/mL set by the Malaysians Food Act 1983 and Food Regulations 1985. In the case for goat milk, the largest colony count of 9.0 <sup>×</sup> <sup>10</sup><sup>8</sup> CFU/mL was observed in the pasteurized sample. It exceeded the permissible limit thus considered spoilt at 22 days of storage. The results are in line with Doll et al. [51] who found that 90% of pasteurized cow milk samples stored at 8 ◦C contained 106 to 10<sup>7</sup> CFU/mL of psychrotrophs at the end of 24 days storage. Pressure regime of 600 MPa/7 min presented the best result as the psychrotrophic bacteria in goat milk remained undetectable at day 22. Garcia-Risco et al. [52] reported that the bovine milk sample treated at 400 MPa for 3 min has a psychrotrophic bacteria count of 106 at day 30 of storage at 7 ◦C.

**Figure 7.** Changes in solid-non-fat of (**a**) cow and (**b**) goat milk during storage period of 22 days. Solid lines represent HPP-treated milk (450 MPa/7 min, 600 MPa/5 min, 600 MPa/7 min), dashed lines represent pasteurized milk, and dotted lines represent fresh milk before treatment.


**Table 3.** Psychrotrophic bacteria count and total plate count (CFU/mL) in milk for 22 days of storage.

For TPC, progressive increase was observed during storage period of 22 days. The lowest TPC observed in cow milk at day 22 was in the pressure regime of 600 MPa/7 min, followed by 600 MPa/5 min, then pasteurization and lastly 450 MPa/7 min. Despite the increase, the TPC was still below the limit set by MS 410:1995, which states that the TPC in pasteurized milk should not have TPC above 105 CFU/mL. For goat milk at day 22, the lowest TPC of 8.8 <sup>×</sup> 101 CFU/mL was milk treated at 600 MPa/7 min. Pasteurized milk had the highest TPC of 3.5 <sup>×</sup> 108 CFU/mL, which exceeded the permissible limit. These results are in agreement with Alexandros et al. [14] who found that TPC for HPP milk was always lower compared to pasteurized where TPC in HPP milk reached 107 CFU/mL after 28 days compared to pasteurized milk which took only 14 days to reach >107 CFU/mL. From both the psychrotrophic bacteria population and the TPC microbial count, HPP pressure regime of 600 MPa/7 min is the best for shelf life extension of both milk, renders them the shelf life of more than 22 days.

Other monitored microbes which include *C. perfringens*, *S. aureus*, *E. coli*, *L. monocytogenes*, *Salmonella* spp., and thermophilic aerobic spore remained undetected throughout the 22 days of storage. Yeast and mold that has been inactivated by HPP or pasteurization also remained undetected throughout the storage period. Although the *Bacillus cereus* was found to be present in low number in cow milk at Day 0 after treatment (Table 2), they were not present in all samples on storage Day 11 and

22. The same situation goes to coliform and mesophilic aerobic spores, which were also present in low number in some pasteurized milk samples at Day 0 after treatment; they were undetected on storage Day 11 and 22. This inconsistent result is most probably due to the randomized sampling and testing of milk from different bottles for each analysis on respective experimental day. The storage results from Table 3 have actually reaffirmed the unexplained presence of coliform and mesophilic aerobic spores after HPP and pasteurization treatments (Table 2) suggesting either it was experimental randomization error or that microbes were inactivated during storage at low storage temperature which may have prevented microbial growth [52] or reduced the microbial count to undetectable levels.

#### **4. Conclusions**

The results recognized that HPP treatment of 7 min at 450 MPa is sufficient to render both cow and goat milk a shelf life of 22 days at 8 ◦C, although higher pressure of 600 MPa can give a better safety margin with best treatment results in the order of 600 MPa/7 min, 600 MPa/5 min, and lastly 450 MPa/7 min. The HPP treatment has shown better capability of preserving milk than pasteurization where pasteurized goat milk was considered spoilt during storage since both TPC and psychrotrophic bacteria count exceeded the permissible limit. For storage studies, the HPP-treated milk, particularly the cow milk displayed better physico-chemical properties stability compared to the pasteurized. The HPP-treated milk did not show any significant changes in total fat, total solid, and SNF but pasteurized cow and goat milk demonstrated significant changes in pH, titratable acidity, specific gravity, total fat, total solid, and SNF. In terms of microbiological properties, although both HPP and pasteurization were effective in reducing the numbers of TPC, psychrotrophic bacteria, mesophilic spores, coliform, yeast, and mold to far beyond the permissible limit, pasteurization however did not fully inactivate the mesophilic spores and coliform while HPP did. Pressure of 600 MPa for holding time of both 5 and 7 min have resulted in very low TPC and psychrotrophic bacteria count which is favorable to prolong milk shelf life. The HPP treatment is suggested as a viable alternative treatment for milk if cost is not a consideration, as cost of HPP treatment is much higher.

**Author Contributions:** S.F.T. designed the experiments, collected and analyzed the data, and wrote the manuscript. N.L.C. conceptualized the study, supervised the research, edited and revised the manuscript. T.P.T. supervised the experiments and S.K.C. provided technical know-how in running the industrial HPP unit. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Classification, Force Deformation Characteristics and Cooking Kinetics of Common Beans**

**Ebenezer M. Kwofie 1,2, Ogan I. Mba <sup>1</sup> and Michael Ngadi 1,\***


Received: 4 September 2020; Accepted: 29 September 2020; Published: 1 October 2020

**Abstract:** Post-harvest characteristics of common beans influences its classification, which significantly affects processing time and energy requirements. In this work, ten bean cultivars were classified as either easy-to-cook (ETC) or hard-to-cook (HTC) based on a traditional subjective finger pressing test and a scientific objective hardness test. The hardness study used seed coat rigidity to explain the structural deformation observed during cooking. The result shows that the average hardness of raw dry ETC and HTC beans was 102.4 and 170.8 N, respectively. The maximum seed coat resistance is observed within the first 30 min of cooking regardless of the classification. The results show that a modified three-parameter non-linear regression model could accurately predict the rate of bean softening (R<sup>2</sup> = 0.994–0.999 and RMSE = 3.3–14.7%). The influence of bean softeners such as potassium carbonate (K2CO3) and sodium chloride (NaCl) to reduce cooking time was also investigated. The results showed that the addition of K2CO3 to the cooking water significantly reduced the cooking time by up to 50% for ETC and 57% for HTC.

**Keywords:** common beans; beans classification; hard-to-cook; bean softening

#### **1. Introduction**

Common beans are an essential food and export commodity in many regions of the world. They are unique sources of plant protein and energy in the diets of more than 300 million people in parts of East Africa and Latin America. Beans are low in fats and rich in proteins, complex carbohydrates, and micronutrients [1]. Despite the nutritional benefits, the presence of flatulence oligosaccharides, antinutrients, and the hard-to-cook (HTC) behavior reduces its potential global consumption [2]. This HTC phenomenon necessitates cooking beans for long hours and requires very high energy demand. There is evidence that the long cooking time is partly influenced by the properties of the beans, such as seed size, variety, and storage conditions, as well as pre-cooking treatments and cooking methods [3–5].

Cooking time is, therefore, an important criterion for evaluating bean cooking quality [6]. Several authors have proposed different theories, including the middle lamella-cation-phytate-phytase theory [7] and the lignification theory [8,9], to explain the hard-to-cook behavior of beans. Several techniques have been deployed to soften beans, reduce the cooking time, and increase consumption. These include soaking [10], ultrasound-assisted hydration [11], salts addition [12], blanching [13], autoclaving, radiation, and extrusion [14–16]. Of all these techniques, the most common approach for household bean processing has been soaking. The typical rural community approach has been to soak common beans overnight or to add table salt or potash to the water during the cooking of the beans. There is evidence that depending on the salt added, the cooking time may reduce significantly or take

a much longer time. For instance, Njoroge, Kinyanjui [17] found that cooking stored dried beans in deionized and Na2CO3 took 257 and 193 min, respectively. However, they reported that thermally treated beans with CaCl2 did not cook at all within an experimental time of 300 min.

Although beans generally begin to lose their hardness with the onset of cooking and continue until the desired softness is achieved, the force deformation during cooking and impact of the physical parameters with respect to cooking time seems to be missing in the bean cooking narrative. Understanding and optimizing the softening phenomenon through the kinetics of cell wall rigidity and overall structural deformation at different stages during cooking is important. The objectives of this work were, therefore, to (1) classify ten common bean cultivars as either ETC or HTC based on the typical household subjective mechanism and a scientific objective penetration tests and by examining the impact of physical parameters in the classification; (2) examine the bean structure deformation of ETC and HTC during cooking; (3) study the influence of bean softening techniques during cooking of selected ETC and HTC bean cultivars; (4) model the softening behavior of ETC and HTC common beans samples during cooking.

#### **2. Materials and Methods**

#### *2.1. Materials*

Ten cultivars of common beans cultivated in the Chitipa district of Malawi were used for this work. The bean cultivars included Boma, Mandondo, Sugar, White, Satana, Msiska, Masusu, Kabulangeti, and Magungulu. Before the experiments, the beans were thoroughly cleaned and sorted. All cultivars were from the same farming season and within the same geographical area. The bean samples were part of the lot from the March–April 2018 harvest season. Cooking grade sodium chloride (NaCl) and ACS grade potassium carbonate (K2CO3) were supplied by Fisher Scientific (Fair Lawn, NJ, USA).

#### *2.2. Common Beans Classification*

#### 2.2.1. Cooking of Beans

One hundred grams of dry unsoaked beans from each of the ten cultivars were selected for the cooking test. The beans were cooked in 800 mL distilled water in a water bath at 96 ◦C until fully cooked. At 30 min intervals, randomly selected ten-gram seeds were cooled in an ice bath for 1 min prior to the cookability test. Bean cultivars that were fully cooked within 120 min were classified as ETC. Above 120 min, the beans were considered HTC. The 120 min benchmark mimicked the length of cooking practiced by the bean-growing communities in Malawi.

#### 2.2.2. Subjective Cookability Evaluation

The local subjective cookability test to assess softness is to press the cooked beans between the thumb and the index finger. Like other researchers, this "pinching test" was adopted to determine the beans' cookability (softening) after 120 min cooking [18,19]. The beans were considered cooked if the cotyledons disintegrated upon pressing.

#### 2.2.3. Objective Cookability Evaluation

An objective cookability test was determined using a TA-HD Plus texture analyzer (Stable Micro Systems Ltd., Surrey, UK). The hardness of the samples was measured by a penetration test and recording the peak of maximum force (hardness) as further described in Section 2.4. For each variety, a predetermined average compression force (3.75 ± 0.88 N) for fully cooked beans was used to differentiate the uncooked and cooked beans. The cookability data generated were used to generate cooking curves for each bean variety.

#### *2.3. Bean Texture Measurement*

Following the method described by Gowen, Abu-Ghannam [13], changes in bean hardness (in Newtons) and seed coat rigidity of raw, soaked, and cooked beans were determined as a function of time and temperature using the TA-HD Plus texture analyzer (Stable Micro Systems Ltd., Surrey, UK). At return-to-start (RTS) mode, a 2 mm cylindrical stainless-steel probe (P2) and 50 kg load cell were used to measure the compression force. The selected probe can impact the beans' tegument, which helps differentiate similar samples [20]. Bean samples were compressed axially to 75% of their original height, applying a crosshead speed of 1.0 mm/s and a pre-test and post-test speed of 1 mm/s [21]. Due to the significant variation of individual bean hardness, ten-gram bean seeds were chosen to represent each treatment. For consistency, the orientation of the seeds on the texture analyzer platform was kept uniform. The seed coat rigidity was determined as the distance the probe traveled before penetrating the cotyledon.

#### *2.4. Bean Softening Experiments*

#### 2.4.1. Soaking

The procedure described by Shafaei, Masoumi [22] was adopted for the soaking of representative bean cultivars. Ten grams of randomly selected seeds of three cultivars representing ETC (Masusu), intermediate (Magungulu), and HTC (Msiska) were used. The seeds were soaked in 50 mL of distilled water at room temperature for 10 h. Preliminary experiments showed that the time for equilibrium water absorption in all cultivars was 8–10 h. The soaking experiments were conducted in triplicates. The water absorption capacity was evaluated using Equation (1) [22,23].

$$\mathcal{W}\_a = \frac{\mathcal{W}\_f - \mathcal{W}\_i}{\mathcal{W}\_i} \times 100\tag{1}$$

where *Wa* is the water absorption (d.b.%), *Wf* is the final weight of seeds after soaking (g), and *Wi* is the initial weight of seeds prior to soaking (g).

#### 2.4.2. Cooking in Salt

For each cooking test, one hundred gram dry unsoaked seeds of Masusu, Magungulu, and Msiska) were separately cooked in 800 mL of 5% table salt (NaCl) or potash based (K2CO3) salt solution. NaCl and K2CO3 are commonly used for cooking and softening, respectively, in the local communities. The control bean samples were cooked in distilled water only. All the cooking was done in a water bath at 96 ◦C until the beans were fully cooked. At 30 min intervals, ten-gram seeds were randomly selected and cooled in an ice bath for 1 min prior to the texture evaluation. The beans were considered cooked when the predetermined cooked bean softness for a cultivar was reached.

#### *2.5. Statistical Analysis*

The parameters of the proximate composition analysis were stated as means ± SD (standard deviation, n = 3). The results were statistically analyzed by analysis of variance (ANOVA), and the means were compared using Tukey-Kramer HSD (α = 0.05) on JMP software.

#### *2.6. Kinetics of Bean Softening*

The rate of bean softening was determined using a 3-parameter non-linear exponential regression model shown in Equation (2). The 3-parameter model is a modified version of those used by [2,13,24].

$$F = F\_{\varepsilon} + F\_{d} \exp^{-K\varepsilon t} \tag{2}$$

where *F* and *Fe* represent the force (N) at a given cooking time (min) and the equilibrium force during cooking, respectively. *Fd* is the differential softening, representing the total degree of softening over the cooking period (i.e., the difference between the initial and final force), and *Fd* represents the rate constant of bean softening.

For simplicity, the experimental data were refitted to a simplified 2-parameter non-linear exponential regression model as shown in Equation (3). The force at any given time is defined in terms of only the initial force prior to cooking, *KF*, and the cooking time, *t*.

$$F = F\_o \exp^{-K\_I t} \tag{3}$$

#### **3. Results and Discussion**

#### *3.1. Bean Classification*

The result of the subjective finger-pressing test is shown in Figure 1. The results showed that without any softening technique employed, cooking of common beans will take 2–4 h, depending on the cultivar. Using the locally accepted cooking time classifier of two hours (designated as "subjective bean classification line" in Figure 1), the bean varieties were classified as either easy-to-cook beans (ETC) or as hard-to-cook (HTC). The plot shows that Masusu, Satana, and Sugar beans could be classified as ETC. These results confirmed what the communities had known for years. These varieties sell quickly in the communities because they are cooked in a relatively shorter time and required less fuel. Other cultivars, including Maine, Magungulu, White, Kabulangeti, Mandondo, Boma, and Msiska beans, classified as hard-to-cook (HTC), take up to 4 h to be fully cooked.

**Figure 1.** Subjective hardness test of the bean cultivars. (The red line represents the subjective bean classification line. Bean cultivars below the line are easy-to-cook (ETC), while those above the line are hard-to-cook (HTC).

The cooking profile of the common beans by the objective compression test showed that the maximum force showed similar categorization as the pinching test classifying Masusu, Satana, and Sugar beans as ETC (Figure 2a) and the other cultivars as HTC (Figure 2b). The average hardness of raw dry ETC and HTC beans was 102.4 and 170.8 N, respectively. Reduction in hardness was observed in the first 30 min of cooking for all the cultivars. In the ETC cultivars, the average hardness reduction was 77%, while 87% was recorded for the HTC cultivars. Nevertheless, the ETC reached their cooking texture much faster. Among the HTC cultivars, White and Msiska cultivars took the longest time to cook. This may not be unexpected as their average hardness in the dry unsoaked state was 36–47% higher than all the ETC beans. When compared to the other HTC cultivars, the average hardness was 16–36%. This can be attributed to the inherent seed characteristics such as cell wall structure, seed composition, and compactness of cells in the seeds [25,26].

**Figure 2.** Objective hardness test of the bean cultivars (**a**) ETC cultivars (**b**) HTC cultivars.

Based on the physical parameters, it is expected that Boma and Magungulu cultivars with the higher equivalent diameter and seed volume would have a higher rate of water absorption and cook faster compared to the other beans. However, these physical properties had no significant impact on their softening characteristics. This may be due to the high temperature of cooking water. High temperature tends to diminish the influence of chemical composition, pore structure, and physical properties on water absorption and by extension textural deformation [27]. This result implies that in communities where the cost of energy (especially firewood) for domestic thermal applications is high, preference would be given to ETC rather than the HTC cultivars.

#### *3.2. Bean Structure Deformation during Cooking*

Uncooked beans are generally hard and strong. The bean-cooking process leads to some structural deformation that significantly reduces the hardness and produces a softened edible product. The typical force deformation curves for dry and cooked beans are shown in Figure 3. The sequence of deformation occurs in stages. Stage A represents when the probe just touches the bean. Stage B represents the force required for compressing the seed coat and outside layers of the cotyledon. The graphs show that in the uncooked beans (Figure 3a), AB and BC have the same slope. However, in the cooked sample (Figure 3b), BC had a steeper gradient than AB. This can be attributed to some level of plasticity attained by the seed coat during high-temperature cooking with water. A similar effect was observed for red kidney beans immersed in boiling distilled water at 100 ◦C for 1.5 min [28]. At stage C, the probe penetrates the seed coat and further compresses the bean. A rapid reduction in force was observed in the dry bean (Figure 3a) compared to the cooked beans. Between D and E, a constant force was observed as the probe moved through the bean. The higher force required to penetrate the dry bean resulted in breakage and fragmentation. As the probe traveled through the empty space, a near-zero force was recorded (stage DE). Similarly, DE was relatively constant in the cooked beans as the probe traveled after penetrating the seed coat (Figure 3b).

The force deformation curve changes significantly during cooking regardless of the cultivar. The variation in deformation curves of ETC and HTC bean cultivars is shown in Figures 3c and 3d, respectively. The plots show that the maximum force to penetrate the bean decreased with increasing cooking time. The maximum distance the probe traveled prior to penetration decreased with increasing cooking time. This observation suggested that the seed coat rigidity or fracturability may considerably determine the softening behavior of beans. The figures also show that the force after penetration of the seed coat leveled with cooking time. This may be attributed to the fact that starch gelatinization, protein denaturation, and other phase changes due to cooking become uniform with time.

The evaluation of the seed coat rigidity/fracturability (Figure 4) shows that the dry seed coat is less rigid. The rigidity of the seed coat increased as it absorbed moisture and presented increased resistance to the probe during compression. This rigidity declined as the cooking time increased. The bean seed coat is composed of a thin cuticle above a layer of palisade cells. These palisade cells are prismatic and thick-walled contiguous cells [29]. The thickness of the palisade cells and the lignin and cellulose content of the seed coat are essential determiners of the cooking quality [29]. Seed coat thickness, seed volume, and protein content are known structural features responsible for water absorption [30]. It has been reported that beans with thinner seed coats absorb water more rapidly during the initial soaking of beans [31]. At higher temperatures such as those used in cooking, absorption could be higher, leading to higher initial resistance of the probe through the seed coat.

**Figure 3.** Variation of force deformation in common beans: (**a**) typical uncooked beans (**b**) typical cooked beans (**c**) cooked ETC cultivar (Masusu), and (**d**) cooked HTC cultivar (Msiska).

**Figure 4.** Variation of seed coat rigidity of common beans during cooking (ETC – Masusu, HTC -Msiska, Intermediate – Magungulu)

#### *3.3. Modeling Bean Texture as a Function of Time*

The experimental data were fitted to the three-parameter model (Equation (2)). Using non-linear regression analysis, the model constants were determined and presented in Table 1. The results showed that the model accurately predicted softening characteristics of the cultivars considered in this study with R2 of 0.994–0.999 and RMSE of 3.3–14.7%. Apart from Msiska and White beans, whose equilibrium maximum forces were 0.5 and 3.5%, respectively, above the experimentally determined cooking threshold, the final force determining fully cooked beans were accurately predicted. However, these variations were not statistically significant (*p* < 0.05). The rate constants revealed that Masusu and Sugar bean cultivars, already classified as ETC, had a relatively higher softening rate. The softening rates of Msiska, Boma, and Mandondo cultivars were relatively slower. It can be deduced that the model is useful as a classification tool to differentiate ETC and HTC cultivars and provide meaningful insight on bean softness.

**Table 1.** Model parameters of the 3-parameter and simplified 2-parameter models.


Fe = equilibrium force during cooking; Fd = differential softening representing the total degree of softening over the cooking period; KF = rate constant of bean softening; Fo = initial force prior to cooking. (Tukey–Kramer HSD (α = 0.05)).

The model parameters of the simplified two-parameter model (Table 1) showed that the model could accurately predict the softening characteristics of common beans with R2 of 0.975–0.994 and RMSE of less than 30%. When compared with the three-parameter model, the rate of softening is 14.2–15.1% lower. Considering the standard deviation of texture measurement for each cultivar, these variations in the rate of softening were significantly different (*p* < 0.05).

Cooking kinetics of cereal and legumes have largely relied on Peleg's hydrodynamic model [32]. The hydrodynamic models used in cooking kinetics assume a diffusion process; however, there is evidence that liquid transport during cooking cannot solely be described by the diffusion mechanism [33]. The proposed two-parameter model could provide the benefit of predicting the softening characteristics using the initial bean hardness, which can easily be measured using hardness testing equipment including simple handheld texture analyzers.

#### *3.4. Softening Techniques and Their Impact during Cooking*

The effects of the different softening techniques on the cookability of the bean samples are presented in Figure 5.

#### 3.4.1. Effect of Soaking on the Texture of Common Beans

The comparative cookability profile of soaked and unsoaked beans is shown in Figure 5a. The maximum penetration force representing the hardness was significantly lower for soaked beans compared to unsoaked beans. Cooking time was reduced both for ETC varieties like Masusu and HTC varieties like Msiska. The reduction was 42, 50, and 50% for Msiska, Magungulu, and Masusu, respectively. This result seems to suggest that the rate of cooking time reduction due to soaking does not depend on the classification. This reduction may be attributed to increased cell wall hydration as

enough water is absorbed, causing the softening of the cell wall and enhancing the breakdown of the middle lamella of the cotyledon [34].

**Figure 5.** Effect of softening techniques on common beans cookability profile: (**a**) soaking before cooking, (**b**) cooking in NaCl solution, (**c**) cooking in potash solution, and (**d**) comparison of softening techniques for HTC cultivar (Msiska).

3.4.2. Effect of Salt Addition on Texture Profile during Cooking

Salt addition has been a common technique for softening. The results show that the addition of NaCl had no softening impact (Figure 5b). It had no influence on the HTC cultivars and even increased the hardness of the ETC after 60 min. During the cooking of whole (unhulled) beans, early addition of common salt NaCl before the seed coat has absorbed enough cooking water appears to make the seed coat impervious, toughen the beans middle lamella, and impede further water absorption. These lead to the failure of underlying cells to achieve the needed hydration to separate upon cooking and interfere with the solubility of the pectic substances during cooking. This phenomenon has been mostly associated with the ability of Ca2<sup>+</sup> and Mg2<sup>+</sup> to form water-resistant pectin complexes with the pectin-rich intercellular middle lamella [4,35]. Na<sup>+</sup> is recognized to share similar chemical and biochemical reaction properties with Ca2<sup>+</sup> and Mg2<sup>+</sup>. This finding is consistent with the report of Kinyanjui, Njoroge [18], especially with respect to the HTC cultivars. On the other hand, the K2CO3 salt had a pronounced effect on the hardness of both the ETC and HTC cultivars (Figure 5c). The addition of K2CO3 salt reduced the cooking time of ETC and HTC beans by up to 50 and 57%, respectively. As shown in Figure 5d, the addition of potassium salt resulted in the highest level of bean softening. The impact of increased potassium load on the functional and nutrient quality of the product must be balanced with the desired degree of bean softening.

#### **4. Conclusions**

This study correlated the classification of ten beans cultivar as easy-to-cook (ETC) or hard-to-cook (HTC) based on a subjective pinching test with an objective penetration test. The penetration test showed that three cultivars, namely Sugar, Satana, and Masusu, stood out as ETC cultivars, while the other seven cultivars could be described as HTC cultivars. Force deformation curves and seed coat rigidity/fracturability tests were used to monitor the textural changes during cooking. The average hardness of raw dry beans was found to be 102.4 and 170.8 N for ETC and HTC, respectively. Seed coat rigidity plays important role in bean softening during cooking. It was found that the seed coat resistance increased sharply during the initial 30 min and gradually declined until the bean was fully cooked. The results also showed that the three-parameter non-linear regression model and its simplified version can be used to adequately and satisfactorily model the softening characteristics of common bean during cooking. Potassium-based aqueous solutions and pre-soaking can considerably reduce cooking time up to 57%.

**Author Contributions:** E.M.K. performed the experiments with O.I.M. This research was led by M.N., who was also instrumental in the initiation, conceptualization, and planning of this work. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the International Fund for Agricultural Development (IFAD) for providing financial assistance through IFAD project grant 2000000974.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### *Processes* **2020**, *8*, 1227


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Application of Computational Intelligence in Describing the Drying Kinetics of Persimmon Fruit (***Diospyros kaki***) During Vacuum and Hot Air Drying Process**

#### **Alfadhl Yahya Khaled 1,\*, Abraham Kabutey 1, Kemal Ça ˘gatay Selvi 1, Cestm ˇ ír Mizera 1, Petr Hrabe <sup>2</sup> and David Herák <sup>1</sup>**


Received: 23 March 2020; Accepted: 28 April 2020; Published: 7 May 2020

**Abstract:** This study examines the potential of applying computational intelligence modelling to describe the drying kinetics of persimmon fruit slices during vacuum drying (VD) and hot-air-drying (HAD) under different drying temperatures of 50 ◦C, 60 ◦C and 70 ◦C and samples thicknesses of 5 mm and 8 mm. Kinetic models were developed using selected thin layer models and computational intelligence methods including multi-layer feed-forward artificial neural network (ANN), support vector machine (SVM) and k-nearest neighbors (kNN). The statistical indicators of the coefficient of determination (R2) and root mean square error (RMSE) were used to evaluate the suitability of the models. The effective moisture diffusivity and activation energy varied between 1.417 <sup>×</sup> <sup>10</sup>−<sup>9</sup> m2/<sup>s</sup> and 1.925 <sup>×</sup> 10−<sup>8</sup> m2/s and 34.1560 kJ/mol to 64.2895 kJ/mol, respectively. The thin-layer models illustrated that page and logarithmic model can adequately describe the drying kinetics of persimmon sliced samples with R<sup>2</sup> values (>0.9900) and lowest RMSE (<0.0200). The ANN, SVM and kNN models showed R2 and RMSE values of 0.9994, 1.0000, 0.9327, 0.0124, 0.0004 and 0.1271, respectively. The validation results indicated good agreement between the predicted values obtained from the computational intelligence methods and the experimental moisture ratio data. Based on the study results, computational intelligence methods can reliably be used to describe the drying kinetics of persimmon fruit.

**Keywords:** persimmon fruit; drying methods; computational intelligence methods; artificial neural network model; support vector machine model; k-nearest neighbors

#### **1. Introduction**

Persimmon (*Diospyros kaki*) is an edible fruit with a sweet taste and rich in vitamin A, C, calcium, condensed tannins, carotenoid, phenolic compounds and iron [1–3]. Besides the nutritional value of persimmon, it has many health benefits such as therapeutic effect on cardiovascular system disease, helps effectively to reduce cholesterol and blood pressure, strengthens the immune system, prevents cancer and remedy for digestion [4]. Persimmon has high moisture content resulting in susceptibility to spoilage even at refrigerator temperatures. Thus, it has to be preserved by proper drying processes to increase the shelf life [5,6].

Drying is considered one of the commonly used postharvest preservation techniques. The drying process of harvested agricultural products assists in reducing spoilage, increasing shelf-life and reducing the bulk weight of products during transportation. Drying of agricultural products causes the enzymatic reactions to be inactivated as a result of heat and mass transfer leading to a reduction of the moisture content inside the product [7]. It also helps the extraction of bioactive compounds from food products. Drying methods such as hot-air drying (HAD), freeze-drying (FD), vacuum drying (VD), microwave drying (MWD) and infrared drying (IRD) have been used in drying agricultural crops [7–10]. Amongst these drying methods, the HAD and VD are the most common commercially used drying methods as they provide more uniform dried product, naturally harmless and nontoxic [11]. For the VD method, the use of low temperatures in the absence of oxygen can preserve heat-sensitive and easily oxidizable foods [12]. Consequently, discoloration and decomposition of the flavor and some nutritional substances can be prevented [10]. However, using HAD and VD without appropriate operating parameters can negatively affect the essential properties of food products such as nutritional and phytochemical properties. Hence, the determination of the optimum operating parameters, drying conditions using suitable drying models are indispensable for achieving quality along with minimum product cost and maximum yield [7,13,14].

Previous studies have explored several mathematical thin-layer drying models (empirical, semi-theoretical and theoretical) to describe the drying kinetics of fruits and vegetables for enhancing the overall performance of the drying process [5,15–17]. These models are derived from the physical laws that govern the process such as mass, reaction kinetics and thermodynamics. The models accuracy of a thermal process are limited to other factors like physical properties, which may vary during the thermal process. Besides that, these models can achieve satisfying regression results in comparison with experimental data in specific conditions. Nevertheless, mathematical thin-layer models are empirical in nature, do not give the physical interpretation of the drying process and they are product dependent [18].

Computational intelligence tools such as artificial neural networks (ANN), support vector machine (SVM) and k-nearest neighbors (kNN) are considered as complex tools for complex systems and dynamic modelling [19]. The application of ANN, SVM and kNN offer many advantages compared to conventional modeling techniques due to the learning ability, increased flexibility, online non-destructive measurements, reduced assumptions, suitability to the non-linear process and tolerance of incomplete data [7,13,14]. For example, ANN is inspired by the biological neural system as a useful statistical tool for nonparametric regression [19]. SVM is highly recognized in terms of its superior performance of the regression data due to its excellent performance capability when working with multi-dimensional data [20]. The kNN is an easy-to-implement algorithm that can be used to solve prediction problems [21].

Computational intelligence modelling methods can be applied as a potential alternative to mathematical thin-layer models in the drying of fruits and vegetables due to the stability and precision in case of uncertainties in the input parameters. ANN and SVM have been successfully applied in modelling and optimizing the drying processes of fruits and vegetables such as pomelo [22], ginkgo biloba seeds [13], mushroom [23], tomato [24], wood [25], celeriac slices [26], pumpkin [11], pepper [27], eggplant [28] and tea leaves [29]. The application of computational intelligence modelling methods in many areas such as cybersecurity and health-care has proven their unique ability to learn such complex data interactions [30,31]. Therefore, including these models in the drying techniques can enhance the prediction and optimization of the drying kinetics of agricultural crops. They provide a significant advantage to overcome the large errors in predictions, by automatically adapting the models until the minimum error rate. In view of this, ANN, SVM and kNN techniques can be applied to predicting and optimizing the drying kinetics of persimmon fruit under VD and HAD. It can help in the evaluation of drying parameters in real conditions, the optimization of processing conditions and the increase in the overall drying efficiency. However, this knowledge is limited in the literature regarding the application of ANN, SVM and kNN techniques in modelling the drying kinetics of persimmon fruit under VD and HAD. Application of these models can optimize energy and quality for persimmon fruit and further minimize the required time and production cost.

The objectives of this study are to investigate the drying characteristics of persimmon fruit at different temperatures using VD and HAD, to evaluate the feasibility of applying ANN, SVM and kNN modelling as a non-destructive technique in describing the drying behavior of persimmon fruit under different drying conditions and to compare the results to mathematical thin-layer models.

#### **2. Materials and Methods**

#### *2.1. Samples Preparation*

Persimmon fruit bought from a market in Prague, Czech Republic, was used for the experiment. A total of 30 persimmon fruits were selected based on similar physical appearances (shape, color and size). Before the drying experiments, the persimmon fruits were stored in a refrigerator at 5 ◦C until further processing. Prior to each experiment, samples were peeled and washed under running tap water. Afterwards they were sliced into two thickness levels of 5 mm and 8 mm with a diameter of 56 mm using a Sencor slicer (SFS 4050SS, Prague, Czech Republic). The initial moisture content of the fresh samples was determined to be 3.98 kg/kg (dry basis) based on the ASABE standard by drying 25 g of selected samples at 70 ◦C for 24 h using the conventional oven [32].

#### *2.2. Drying Experiments*

#### 2.2.1. Vacuum Drying (VD) Technique

The VD technique was carried out using a laboratory-scale drying unit (l 450 Gold brunn, Poland). The vacuum was regulated at a 50 mbar ultimate pressure and 2 L/s pump speed by a vacuum pump (VE 135, RoHS, Shanghai, China). However, it is worth mentioning that the pressure was monitored through the vacuum gauge which was unstable during the drying process. This problem was solved manually by fixing the pressure at approximately 50 mbar. This meant that when the pressure increased above or decreased below 50 mbar, the pump was opened for the adjustment. The VD operates by heating the samples with a conduction heat from a heater plate in the container. The vacuum pump reduces the pressure around the sample to be dried and further ensures less atmospheric pressure. This decreases the boiling point of the water inside the product and thereby increases the rate of evaporation. The sliced samples were dried at three temperatures (50 ◦C, 60 ◦C and 70 ◦C). Prior to the experiments, the VD was set-up to the required temperature for 30 min to enable the dryer temperature to reach equilibrium with the surrounding air temperature. The weight of the persimmon sample was measured at 1 h interval using a digital scale (HR-250AZ, A&D Company Limited) weighing balance of 252 g 0.1 mg−<sup>1</sup> precision. All the drying experiments were carried-out in triplicates and the average values were used for further analyses.

#### 2.2.2. Hot-Air Drying (HAD) Technique

A laboratory-scale convective HAD (UF 110, Memmert, Germany) was used. Similar to VD, three temperatures (50 ◦C, 60 ◦C and 70 ◦C) at a constant air velocity of 1.10 m/s was attained until constant weight between two successive readings. The air velocity was measured using a Thermo-Anemometer (Model 451104, EXTECH Instruments, Tainan, Taiwan; with the accuracy of 62% velocity). Before starting the experiments, the HAD was set-up to the required temperature for a half-hour to enable the dryer temperature to reach equilibrium with the surrounding air temperature. Similar to VD, the weight of the persimmon sample was measured 1 h interval using a digital scale weighing balance; whereby the samples were removed from the dryer and measured and then returned to the dryer. The experiments were conducted in triplicate and the average values used in further analyses.

#### *2.3. Drying Kinetics*

The variation in moisture content during VD and HAD techniques was expressed in the form of moisture ratio (dimensionless) as described in Equation (1).

$$\text{MR} = \frac{(M\_t - M\_{\text{c}})}{(M\_o - M\_{\text{c}})} \tag{1}$$

where *Mt*, *Me* and *Mo* are the moisture content of the samples at time *t*, equilibrium moisture content and initial moisture content, respectively. According to Aghbashlo et al. [33], *Me* values did not change because they were relatively low compared to *Mt* and *Mo* values, resulting in negligible error during simplification, thus, in this study, the moisture ratio was expressed as shown in Equation (2):

$$\text{MR} = \frac{\text{M}\_{\text{f}}}{\text{M}\_{\text{o}}} \tag{2}$$

#### *2.4. E*ff*ective Moisture Di*ff*usivity*

To effectively assess the behavior of the VD and HAD methods of samples, it is important to understand the mechanisms of moisture movement within the samples during drying. Fick's diffusion equation as a dimensional approach was applied due to its simplicity to describe the mass transfer of drying samples. The effective moisture diffusivity of samples for VD and HAD methods was estimated using Crank's solution [33] of Fick's diffusion equation as described in Equation (3) [16].

$$\frac{\partial \mathcal{M}\_{\rm t}}{\partial \mathbf{t}} = \nabla \cdot \left( \mathbf{D}\_{eff} \, \nabla \, \mathcal{M}\_{\rm t} \right) \tag{3}$$

Assuming constant diffusion and uniform initial moisture distribution, the Crank's solution for cylindrical shaped sample is shown in Equation (4).

$$\text{MR} = \frac{8}{\pi^2} \sum\_{\mathbf{n}=1}^{\infty} \frac{1}{\left(2\mathbf{n}+1\right)^2} \exp\left(-\frac{\left(2\mathbf{n}+1\right)^2 D\_{eff}}{\mathbf{r}^2}\right) \tag{4}$$

where *De*ff is the effective moisture diffusivity (m2/s), n is the positive integer, r is the radius of the sample (m) and t is the drying time (s). For the sake of mathematical simplicity, Equation (4) was restricted to the first term, resulting in Equation (5).

$$\text{MR} = \frac{8}{\pi^2} \exp\left(-\frac{\pi^2 D\_{eff} \cdot \mathbf{t}}{\mathbf{r}^2}\right) \tag{5}$$

#### *2.5. Activation Energy*

In general, activation energy is the minimum energy needed in order for drying to occur. The activation energies for VD and HAD methods were calculated from the relationship between effective moisture diffusivity and the average temperature of the samples based on the Arrhenius equation as shown in Equation (6) [34].

$$D\_{eff} = D\_o \exp\left(-\frac{E\_d}{\mathcal{R}(\mathcal{T} + 273.15)}\right) \tag{6}$$

where *Do* is the pre-exponential factor, *Ea* is the activation energy (kJ/mol), R is universal gas constant (8.3143 <sup>×</sup> <sup>10</sup>−<sup>3</sup> kJ/mol) and T is the average temperature of the sample (K). The values of *Ea* for VD and HAD methods for different persimmon thickness levels were measured from the resulting slope values by plotting the fitting curve between *ln*D and 1/(T + 273.15) (Equation (7)).

$$\text{Slope} = -\frac{E\_a}{\text{R}}\tag{7}$$

#### *2.6. Mathematical Thin-Layer Modelling*

The experimental drying data measured were fitted to ten selected mathematical thin-layer drying models. The selected mathematical models are listed in Table 1 namely Newton, Page, Modified page, Logarithmic, Two-term, Two-term exponential, Henderson and Pabis, Modified Henderson and Pabis, Midilli et al. and Hii et al. The coefficients of the mathematical models were determined based on non-linear least squares regression analysis using Sigma plot software (Version12.0, Systat Software Inc., California, USA). The application of these models gives a better prediction with fewer assumptions [35].



#### *2.7. Computational Intelligence Methods*

#### 2.7.1. Artificial Neural Network

The structure of a neural network is in the form of interconnected layers [44,45]. Haykin [46] divided an ANN into three groups of structures based on their connection namely, single layer feed-forward network, the multi-layer feed-forward network and the recurrent network. Among these structures, the multi-layer feed-forward network is widely applied in modelling of agricultural and food systems. The feed-forward neural network has an input layer (n), an output layer (m) and one or more hidden layers (h). The number of neurons in the input and output layers is representative of the number of independent variables (input) and dependent variables (output) respectively. Each of the nodes has connected weight to all the nodes in the next layer calculated to give the sum of the nodes (*x*) representing an activation input value function of the node. The value of x is computed first followed by computing the activation function of the node until the output nodes activation function is acquired. Hidden layer with different nodes is used to process the information received by the input nodes through activation function. In this study, a multilayer feed-forward network structure was used with three input parameters (temperature, thickness and drying time), 1–3 hidden layers and one output parameter (moisture ratio) as shown in Figure 1. A back-propagation algorithm was applied in training of the model because it is stable when a small learning rate is used and sigmoid function was used in all cases as illustrated in Equation (8) [47].

$$\mathbf{f}(\mathbf{x}) = \frac{1}{1 + \mathbf{e}^{-\mathbf{x}}} \tag{8}$$

This algorithm passed through four steps; initialization, activation, weight training and iteration to train data. In this case, the error minimization can be obtained by several procedures; gradient descent, conjugate gradient and Levenberge-Marquardt. Several methods for speeding up back-propagation algorithm have been used like using a variable learning rate and adding a momentum term. In this process, the parameters of learning rate, momentum and number of epoch were set as 0.3, 0.2 and 500, respectively. It is worth noting that the training time in this study depended on the number of iterations which was chosen to be 200. This is because at 200 iterations, the minimum error rate

was reached. The datasets were prepared by randomly dividing the data into training (70%) and testing (30%) [48,49]. The chosen hidden layer architectures were (3), (6), (9), (3, 3), (6, 6), (9, 9), (3, 3, 3), (6, 6, 6) and (9, 9, 9) matrix, where for example, (3, 3) and (3, 3, 3), represent the 2 and 3 hidden layers with 6 and 9 neurons each (Figure 1). The number of the hidden layers and neurons are affected directly by the simplicity of the ANN topology, while few numbers of hidden layers and neurons are recommended for simplicity. The software (Weka 3.6, Hamilton, New Zealand) was used to analyze the ANN model. Overall, ANN has the ability to performing the neural fitting and prediction and in this case, could be used for future predictions without the need for the neural network tools for the specific drying conditions.

**Figure 1.** Artificial Neural networks topology with three hidden layers and different number of neurons.

#### 2.7.2. Support Vector Machine

Support Vector Machine (SVM) is a supervised learning model which combines theoretical solutions with numerical algorithms used for classification and regression methods [50]. In 1999, Vapnik [51] developed the SVM algorithm and other important feature information and patterns. SVM as a regression method is considered an effective approach due to its capability of capturing non-linear relationships in the feature space.

The SVM used for the moisture ratio values was determined by the SMOreg sequence in the Waikato Environment for Knowledge Analysis (WEKA) software, whereby, the SMOreg implements SVM for regression. In this regard, the parameters of SVM learned by setting the RegOptimizer with ReqSMOImproved as a learning algorithm [52]. Figure 2 shows the workflow of using the SVM according to ReqSMOImproved learning algorithm.

The three input variables used for the SVM model were temperature, thickness and drying time with the output as the moisture ratio. Two filter types were applied, namely normalize and standardize in order to determine how/if the data need to be transformed. Additionally, three different kernel models: polynomial, Pearson universal and Radial Basis Function (RBF) were used to construct the predictive model of the calculated moisture ratio values. The data required to be optimized in the three kernels' parameters for maximum performance were obtained using the Grid Search technique. Similar to ANN, the number of iterations used was 200 due to the minimum error rate limit. The three kernels' parameters were computed mathematically as described in Equations (9)–(11):

$$\mathbf{f}(\mathbf{x}, \mathbf{y}) = \frac{\left( (\mathbf{x} \times \mathbf{y}) + 1 \right)^{\mathrm{d}}}{\sqrt{\left( (\mathbf{x} \times \mathbf{y}) + 1 \right)^{\mathrm{d}} \left( (\mathbf{y} \times \mathbf{y}) + 1 \right)^{\mathrm{d}}}} \tag{9}$$

$$\mathbf{f}(\mathbf{x}, \mathbf{y}) = \frac{1}{\left[1 + \left(\frac{2 \times \sqrt{\|\mathbf{x} - \mathbf{y}\|\_2 \sqrt{2^{(1/\omega)} - 1}}}{\sigma}\right)^2\right]^{\omega}} \tag{10}$$

$$\mathbf{f}(\mathbf{x}, \mathbf{y}) = \|\mathbf{e}^{-\mathbf{y}\|(\mathbf{x} - \mathbf{y})\|^2} \tag{11}$$

where x represents the feature vector, y is referred to as the label for each x, d is the degree of polynomial, ω and σ are Pearson width parameters and 'Υ is the kernel dimension.

**Figure 2.** The workflow of support vector machine (SVM).

#### 2.7.3. k-Nearest Neighbors

The kNN is a simple algorithm that predicts the test samples category according to the k training samples, which are the nearest neighbors to the test sample and classifies it to the category that has the largest category probability (Zhang and Zhou, 2005). kNN can select the appropriate value of k based on cross-validation and also kNN can perform distance weighting. In this study, the numbers of neighbors were used to be 3, 5, 7, 9 and 11, in order to select the best k-value in kNN based on the highest results of statistical indicators. The parameters of debug are meanSquared, crossValidate, distanceWeighting, nearestNeighbourSearchAlgorithm, windowSize and doNotCheckCapabilities did not change in Weka software. The value of the batchSize parameter in the kNN classifier was set to be 150 since the highest accuracy was observed at this value. The number of iterations was set to be 200 due to the minimum error rate. The number of decimal places used for the output of numbers in the model was 2 (numDecimalPlaces = 2).

#### *2.8. Color Measurements*

Color is one of the most important quality evaluation attributes for fruits and vegetables during drying. For the color measurements, first, the images of fresh (reference) and dried samples obtained from VD and HAD methods were captured using a smartphone camera (oppo F7, Dongguan, China). The smartphone camera is equipped with a 16 mega-pixel charged-coupled device (CCD). Samples

were put in a glass plate located on the white paper as a background during image capture for more focus and the distance between sample and smartphone camera was set-up to be 18 cm vertically. Then the images were transferred to ImageJ software for determining the color parameters: lightness (L\*), redness/greenness (a\*) and yellowness/blueness (b\*) (http://rsb.info.nih.gov/ij/). The total color difference (ΔE) was estimated based on Equation (12). For each sample, three replications were performed.

$$\mathbf{E} = \sqrt{\left(\mathbf{L}^\* - \mathbf{L}\_o^\*\right)^2 + \left(\mathbf{a}^\* - \mathbf{a}\_o^\*\right)^2 + \left(\mathbf{b}^\* - \mathbf{b}\_o^\*\right)^2} \tag{12}$$

where L∗ o, a<sup>∗</sup> <sup>o</sup> and b<sup>∗</sup> <sup>o</sup> indicate the reference values of fresh sample.

#### *2.9. Statistical Analysis for Mean Comparison*

Statistical analysis was performed using the Statistical Analysis System software (SAS version 9.2, Institute, Inc., Cary, NC, USA). ANOVA at 5% level of significance and 95% confidence interval was performed using the Duncan test to compare the mean significant differences between quality attributes (L\*, a\*, b\* and ΔE) for different sample thickness levels of 5 mm and 8 mm, drying time intervals between (0 and 600 minutes) and drying techniques (VD and HAD). The results were presented as mean ± standard error values. The fit accuracy of experimental data to the mathematical thin-layer and computational intelligence (ANN, SVM and kNN) models was determined by the statistical indicators: coefficient of determination (R2) and root mean square error (RMSE) which are described mathematically in Equations (13) and (14):

$$\mathbf{R}^2 = 1 - \frac{\sum\_{i=1}^{N} \left(V\_{pred} - V\_{exp}\right)^2}{\sum\_{i=1}^{N} \left(V\_{pred} - V\_m\right)^2} \tag{13}$$

$$\text{RMSE} = \sqrt{\frac{\sum\_{i=1}^{N} \left(V\_{\text{pred}} - V\_{\text{exp}}\right)^2}{N}} \tag{14}$$

where *Vpred* is the predicted value, *Vexp* is the actual observation from experimental data, *Vm* is the mean of the actual observation and *N* is number of observations. From the values of R<sup>2</sup> and RMSE, the higher the value of R<sup>2</sup> and the lower the RMSE value, the better the goodness of fit.

#### **3. Results and Discussion**

#### *3.1. Drying Process Behavior*

The variations of moisture ratio with time for VD and HAD techniques at different temperatures (50 ◦C, 60 ◦C and 70 ◦C) and samples thicknesses (5 mm and 8 mm) are presented in Figure 3. From the plot, the moisture ratio of the sliced samples for all techniques decreased with an increase in drying time. The drying rates for VD and HAD methods occurred in the falling rate period. Based on Figure 3a, it is clear that the drying time thus reduces as the drying temperature increases. The moisture ratio values of 0.18 and 0.28 were determined at drying time of 170 min and temperatures of 50 ◦ C and 60 ◦C. At a drying time of 360 min at 70 ◦C was found the moisture ratio of 0.22. As shown in Figure 3b, as the drying time increased, for example, the moisture ratio values at 300 min at 50 ◦C and 70 ◦C were 0.44 and 0.12 respectively. Similar results were observed for HAD at samples thickness levels of 5 mm and 8 mm as illustrated in Figure 3c,d. For example, the final time found from HAD of 5 mm sample thickness at 50 ◦C, 60 ◦C and 70 ◦C was 420 min, 300 min and 240 min respectively. The results indicate that the moisture transfer rate from the inner layers to its surface thus increases as the drying air temperature increases. The rate of moisture evaporation at the surface of dried material to the atmosphere also increases as the temperature increase leading to a higher drying rate. The results are in agreement with other researchers on the drying behavior of various varieties of persimmon [1,2,5].

**Figure 3.** Drying characteristics of persimmon fruit sliced samples; (**a**) vacuum drying (VD) of sample thickness of 5 mm; (**b**) VD of 8 mm; (**c**) hot-air drying (HAD) of 5 mm; (**d**) HAD of 8 mm.

#### *3.2. Results of E*ff*ective Moisture Di*ff*usivity*

The values of the effective moisture diffusivity (*De*ff) are presented in Table 2. The Deff values were varied between the range of 1.417 <sup>×</sup> 10−<sup>9</sup> m2/s and 1.925 <sup>×</sup> 10−<sup>8</sup> m2/s. As it can be seen from Table 2, the *De*ff increased significantly with increasing drying temperature from 50 ◦C to 70 ◦C. The HAD with a 5 mm thickness of persimmon samples illustrated the highest values of *De*ff as compared to VD. The result is attributed to the increase in water molecules activity at higher temperatures leading to higher moisture diffusivity [53]. Moreover, the *De*ff values of samples thickness of 5 mm were higher than samples of 8 mm thickness for VD and HAD drying techniques (Table 2). This is because of the increased heat arising from the increase in drying temperature of the product resulting to the water molecules activities compared to samples of bigger thickness. The values of *De*ff obtained in this study were within the general range of 10–6 to 10–12 m2/s for drying of food materials [5,54]. The values of *De*ff are agreement with published works for strawberry drying (2.40–12.1 <sup>×</sup> <sup>10</sup>−<sup>9</sup> <sup>m</sup>2/s), apple drying (2.27–4.97 <sup>×</sup> <sup>10</sup>−<sup>10</sup> m2/s), persimmon slices (2.60–5.40 <sup>×</sup> <sup>10</sup>−<sup>10</sup> <sup>m</sup>2/s) and pumpkin drying (1.19–4.27 <sup>×</sup> <sup>10</sup>−<sup>9</sup> <sup>m</sup>2/s) [17,55–57].


**Table 2.** Values for effective moisture diffusivity, *De*ff.

#### *3.3. Results of Activation Energy*

The activation energy (*Ea*) of persimmon fruit samples was calculated from the values of effective moisture diffusivity, *De*ff. The relationship between *Ea* and *De*ff was described by an Arrhenius-type equation (Equation (6)). The values of activation energy were obtained by plotting *ln*(*De*ff) versus 1/(T + 273.15) for VD and HAD methods. The activation energy is equal to the slope times the universal gas constant (R) as shown in Figure 4. The activation energy values for the VD and HAD for samples thicknesses (5 mm and 8 mm) were further estimated as given in Table 3. The values of *Ea* for the VD method were 64.2895 kJ/mol and 34.1785 kJ/mol for 5 mm and 8 mm respectively. While for the HAD method, the *Ea* was 34.1560 kJ/mol at 5 mm thickness, and that of 8 mm was 46.0715 kJ/mol at temperatures between 50 ◦C and 70 ◦C. The results found in this study are similar to those reported for persimmon sliced samples with activation energy between 30.64 and 43.26 kJ/mol for blanched and control respectively [5]. The *Ea* values of persimmon samples dried using VD and HAD methods were consistent with those in the literature for different fruits and vegetables, for example,30.46–35.57 kJ/mol in strawberry, 25.26–72.47 kJ/mol in yams and 22.66–30.92 kJ/mol in apples [17,58,59]. The values of *Ea* were within the acceptable range from 12.7 to 110 kJ/mol for various food materials [60].

**Figure 4.** Arrhenius-type relationship of effective moisture diffusivity versus temperature for VD and HAD methods at different samples thickness levels.


**Table 3.** Activation energy of persimmon fruit samples for VD and HAD at different thickness levels.

#### *3.4. Comparison of Mathematical Thin-Layer Models*

The mathematical thin-layer models were used to describe the drying kinetics of persimmon fruit samples for VD and HAD methods. Tables 4–7 show the selected mathematical models that fitted the experimental moisture content data in relation to samples thickness levels of 5 mm and 8 mm. Although all the selected ten models adequately fitted the experimental data, the logarithmic model sufficiently described the drying kinetics of persimmon samples with R2 values (> 0.9900) and lowest RMSE values (< 0.0200) for VD technique at all drying temperatures as given in Table 4. Similar results were obtained for VD at 8 mm thickness indicating that the logarithmic model showed the best model to fit the experimental data with the highest R<sup>2</sup> of > 0.0990 and the lowest RMSE of < 0.0100 at all the three temperatures (50 ◦C, 60 ◦C and 70 ◦C). For HAD technique, the Page, Logarithmic, Midilli et.al and Hii et al. models significantly described the drying kinetics of persimmon samples of thicknesses (5 mm and 8 mm) with R2 of more than 0.9990 and lowest RMSE of less than 0.0010 (Tables 6 and 7). For example, the validation of the logarithmic model by comparing the predicted moisture data and those obtained from the experiments is shown in Figure 5. The moisture ratio data predicted using the logarithmic model lied closely along a straight regression line for different drying conditions indicating the suitability of the model for describing the VD and HAD behaviors of persimmon fruit samples. Onwude et al. [61] also reported the adequacy of page, logarithmic, Midilli et al. and Hii et al. models for predicting the drying kinetics of sweet potato. Similarly, Younis et al. [62] indicated the

appropriateness of page, logarithmic, Midilli et al. and Hii et al. models for describing the drying performance of garlic slices.


**Table 4.** Statistical evaluation of the mathematical drying models for persimmon samples of 5 mm thickness for VD.

**Figure 5.** Predicted versus experimental moisture ratio data for logarithmic model at 50 ◦C and 8 mm samples thickness for VD and HAD methods.



**Table 6.** Statistical evaluation of the mathematical drying models for persimmon samples of 5 mm thickness for HAD.



**Table 7.** Statistical evaluation of the mathematical drying models for persimmon samples of 8 mm thickness for HAD.

#### *3.5. Results of Artificial Neural Network*

The parameters namely time, temperature and samples thickness were used to predict moisture ratio using ANN model for VD and HAD techniques. Tables 8 and 9 show the statistical results with respect to training and validation of the multilayer feed-forward network structure of samples drying data for VD and HAD. The training data set were used to assess the optimum number of neurons and hidden layers for multilayer neural network modelling for determining the best predictive power. For VD, it was found that the architecture with 1 and 2 hidden layers with 9 and 12 (6, 6 neurons), obtained the best results for training set (R<sup>2</sup> = 0.9991) and test set (R2 = 0.9881) as compared to those of 1 hidden layer (3 and 6 neurons), 2 hidden layers (6 and 18 neurons) and 3 hidden layers (9, 18 and 27 neurons), respectively (Table 8). Moreover, the networks were found to be susceptible to the number of neurons in their hidden layers. Thus, smaller neurons led to under fitting, while too many neurons contributed to overfitting. In addition, Figure 6 showed the predicted and experimental moisture ratio values for the optimal ANN for training and testing sets.

Table 9 presented the statistical results of drying kinetics of persimmon samples for the ANN model using HAD. Similar to VD, the results showed that the architecture with 1 and 2 hidden layers with 9 and 6 (3, 3) neurons, obtained the best results as compared to those of 1 hidden layer (3 and 6 neurons), 2 hidden layers (12 and 18 neurons) and 3 hidden layers (9, 18 and 27 neurons), respectively. The highest result of R2 values found for training and testing set were 0.9994 and 0.9979, respectively. Therefore, to compare the results obtained from the thin-layer mathematical models and ANN results, it can be seen that the ANN model is very close to the highest model in the theoretical mathematical models with R<sup>2</sup> and RMSE values of 0.9994 and 0.0124, respectively as compared to those of the logarithmic model (R<sup>2</sup> = 0.9998 and RMSE = 0.0044). Additionally, the predicted and experimental moisture ratio values for the optimal ANN for training and testing data sets are illustrated in Figure 7. Zenoozian et al. [63] demonstrated that 1 and 2 hidden layers and 30 neurons for ANN adequately predicted the moisture changes of pumpkin during osmotic dehydration. ANN with 1 and 2 hidden layers have also been successful in predicting the drying behavior of other fruits and vegetables such as pepper, apple slices, mushroom during microwave-vacuum drying [36,64,65].


**Table 8.** Statistical results of drying kinetics of persimmon fruit samples for the artificial neural network (ANN) model using VD.

**Table 9.** Statistical results of drying kinetics of persimmon samples for the ANN model using HAD.


**Figure 6.** Predicted and experimental moisture ratio using VD from training and testing data set of ANN model.

**Figure 7.** Predicted and experimental moisture ratio using HAD from training and testing data set of ANN.

#### *3.6. Results of Support Vector Machine*

The statistical results related to the training and validation of the SVM of persimmon drying experimental data using VD and HAD are given in Tables 10 and 11. Similar to the ANN model, the training data set was used to evaluate the best filter and kernel type modeling for determining the best predictive power. For VD, the training data set at standardize filter with Pearson universal kernel type found the best result of R<sup>2</sup> and RMSE values of 1.0000 and 0.0004 as compared to those of normalize with polynomial, Pearson universal kernel and RBF kernel and also standardize with polynomial and RBF kernel, respectively. For the testing data set, standardize filter with polynomial kernel type obtained the best results for R<sup>2</sup> and RMSE values of 0.9996 and 0.1213 (Table 10).


**Table 10.** Statistical results of drying kinetics of persimmon samples for SVM model using VD method.


**Table 11.** Statistical results of drying kinetics of persimmon samples for SVM model using HAD method.

However, for HAD, the standardize filter type with RBF kernel obtained the best results for training and testing data set of R<sup>2</sup> and RMSE values of 1.0000, 0.0004, 0.9690 and 0.0912 respectively (Table 11). From the results, it is clear that the SVM model showed the highest results as compared to the theoretical mathematical models and also ANN (Tables 7 and 9). Few studies used SVM as a model in drying techniques. Das and Akpinar [50] applied SVM to investigate pear drying performance by different ways of convective heat transfer. The authors applied normalization and standardization filter to the target attribute with three kernel models (polynomial kernel, Pearson universal kernel and RBF kernel). They found that the polynomial kernel showed the lowest RMSE of value 0.3351. This indicates that the applied algorithm for SVM prediction for the moisture ratio during the drying process for both VD and HAD techniques can be promised.

#### *3.7. Results of k-Nearest Neighbors*

Tables 12 and 13 summarized the R<sup>2</sup> and RMSE of kNN tool developed with training and testing data set for VD and HAD methods. For VD method, the prediction model developed using a k-value of 3, produced the highest results for training and testing data set of R<sup>2</sup> values of 0.9327 and 0.8782 and lowest RMSE values of 0.1271 and 0.1829 respectively. Meanwhile, k-values of 11 and 9 indicated the lowest R<sup>2</sup> values of 0.8355 and 0.5638 and the highest RMSE values of 0.2214 and 0.6399 for training and testing data set, respectively (Table 12). For HAD technique, the results of the training set showed R2 values between 0.6347 and 0.8638 and RMSE values between 0.2145 and 0.2907. The k-value of 7 showed the highest result compared to other k-values (3, 5, 9 and 11). The prediction model developed using the testing data set indicated R<sup>2</sup> values of ranging from 0.4877 to 0.7873 and RMSE values from 0.2385 to 0.3099. The k-value of 5 showed the highest values (Table 13).

**Table 12.** Statistical results of drying kinetics of persimmon samples for k nearest neighbors (kNN) model using VD method.


**Table 13.** Statistical results of drying kinetics of persimmon for kNN model using HAD method.


#### *3.8. Comparison between Computational Intelligence and Mathematical Thin-Layer Models*

The highest results obtained from the computational intelligence models (ANN, SVM and kNN) and the top three mathematical thin-layer models (page, logarithmic and Midilli et al.) for moisture ratio prediction are summarized in Table 14. The R<sup>2</sup> value of 0.9991 and RMSE value of 0.0213 at 1 hidden layer with 9 neurons by applying ANN model showed the best results for VD method. For the HAD method applying ANN model, the R<sup>2</sup> value of 0.9994 and RMSE value of 0.0124 were found. For kNN, the results of R<sup>2</sup> and RMSE were 0.9327, 0.1271, 0.8638 and 0.2320 for VD and HAD respectively. On the other hand, the mathematical thin-layers modelling found the range of R<sup>2</sup> values from 0.9963 to 0.9999 and RMSE from 0.0205 to 0.0031 for both VD and HAD methods.


**Table 14.** Statistical results of drying kinetics of persimmon samples for computational intelligence and mathematical thin-layer models using VD and HAD.

It can be seen that the prediction of moisture ratio with the model developed using SVM gave the highest R<sup>2</sup> and the lowest RMSE values of 1.0000 and 0.0004 compared to the models developed using other computational intelligence methods (ANN and kNN) and mathematical thin-layers models; page, logarithmic and Midilli et al. (Table 14). The results indicate that SVM gave the best results which are in agreement with the findings of prior studies [19,66] that used SVM as a data prediction method to improve outcomes. In summary, the results from this study demonstrate the application of computational intelligence methods for the drying processes of persimmon samples at different conditions. Thus, applying computational intelligence methods in drying generally improves the drying performance by controlling the drying input parameters for optimizing energy, quality and production cost.

#### *3.9. Results of Color Measurements*

The color properties of persimmon samples of 5 mm and 8 mm thickness for VD and HAD methods were determined based on the change in color parameters of L \*, a \* and b \* and the total color change ΔE as given in Table 15. It can be seen clearly from Table 4 that the lightness (L \*) of persimmon samples dried using VD and HAD methods decreased significantly (*P* ≤ 0.05) compared to the fresh samples. However, there was no significant difference between the lightness of samples using HAD-60 for 5 mm and HAD-60 and HAD-70 for 8mm, with those of the fresh samples. The VD method significantly (*P* ≤ 0.05) reduced the lightness of samples compared to HAD. The VD-70 for 8 mm showed the lowest lightness value of 33.479 ± 6.888. The redness/greenness (a \*) of all dried persimmon samples reduced significantly (*P* ≤ 0.05) compared to fresh samples, except VD-50, which resulted in higher a\* value than the fresh. Generally, the values of redness/greenness (a \*) found using HAD were lower than the values found using VD. The lowest value of redness/greenness (a \*) was found to be 6.032 ± 1.740 using HAD-60. Similar results were found for the yellowness/blueness (b \*), where all dried persimmon samples were reduced significantly (*P* ≤ 0.05) compared to fresh samples, except HAD-60, which resulted in higher b\* than the fresh. In addition, there was no significant difference between the b\* fresh samples with those of dried samples using VD-50, VD-60, VD-70 for 5 mm; VD-60 for 8 mm; HAD-50, HAD-60, HAD-70 for 5 mm and HAD-50, HAD-60, HAD-70 for 8 mm. Dried samples using VD-50 and VD-70 methods for 8 mm showed different groups of values of 45.250 ± 2.542 and 38.733 ± 6.948 respectively. The values of total color change using VD method showed higher values compared to the HAD method for the persimmon samples. The highest value was found using VD-70 for 8 mm of 38.733 ± 6.948. Generally, the change of color properties of the fresh samples under different drying methods is due to the increased sample temperature resulting from increased enzymatic and non-enzymatic chemical reactions of the product [67,68].


**Table 15.** Drying methods and color parameters of persimmon fruit samples.

Different letters at the same column indicates statistical difference for Duncan test, *P* < 0.05.

#### **4. Conclusions**

This study investigated the potential of using computational intelligence as a modelling tool for predicting the drying process of persimmon fruit samples. The performance of vacuum drying (VD) and hot-air-drying (HAD) techniques for drying persimmon fruit samples was evaluated. The results showed that VD and HAD had a significant effect on the drying kinetics, moisture diffusivity, activation energy and color properties of persimmon fruit samples. An increase in drying temperature and samples thickness influenced the drying kinetics and moisture diffusivity of samples. The effective moisture diffusivity and activation energy varied between 1.417 <sup>×</sup> <sup>10</sup>−<sup>9</sup> m2/s and 1.925 <sup>×</sup> <sup>10</sup>−<sup>8</sup> m2/s and 34.1560 kJ/mol to 64.2895 kJ/mol respectively. Persimmon fruit samples using HAD showed significant color attributes compared to VD method.

The mathematical thin-layer modelling results showed that page and logarithmic models can adequately (R<sup>2</sup> = 0.9999) describe the drying kinetics of persimmon fruit samples. The highest R<sup>2</sup> values of 0.9994, 1.0000 and 0.9327 were observed for ANN (2 hidden layers with (3, 3) neurons), SVM (standardize filter and RBF kernel) and kNN (k-value of 3) models, respectively. SVM tool as a computational intelligence method produced higher results compared to mathematical thin-layer. Therefore, SVM models are able to describe a wider range of experimental data whereas the application of theoretical models is limited to specific experimental conditions in most cases. Thus, computational intelligence models may be considered as a suitable alternative modelling method for describing the drying behavior of persimmon sliced samples. On the other hand, computational intelligence methods can be successfully applied to industrial drying processes and operations as well as online monitoring and control. However, further study is required to ascertain the suitability of ANN, SVM and kNN for predicting the nutritional composition of fruits and vegetables during drying.

**Author Contributions:** Conceptualization, A.Y.K. and D.H.; methodology, A.Y.K.; software, A.Y.K.; validation and investigation, A.K., K.Ç.S.; formal analysis, A.Y.K.; resources, D.H.; writing—original draft preparation, A.Y.K.; writing—review and editing, A.K., K.Ç.S., C.M. and P.H.; supervision and project administration, D.H. All ˇ authors have read and agreed to the published version of the manuscript.

**Funding:** The research was funded by EU, Managing Authority of the Czech Operational Programme Research, Development and Education through the project "supporting the development of international mobility of research staff at CULS Prague", Grant Number: CZ.02.2.69/0.0/0.0/16\_027/0008366. The APC was funded through the project "supporting the development of international mobility of research staff at CULS Prague".

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Some Physical Properties and Mass Modelling of Pepper Berries (***Piper nigrum* **L.), Variety Kuching, at Di**ff**erent Maturity Levels**

#### **Puteri Nurain Megat Ahmad Azman 1, Rosnah Shamsudin 1,2,\*, Hasfalina Che Man <sup>3</sup> and Mohammad E**ff**endy Ya'acob <sup>1</sup>**


Received: 9 August 2020; Accepted: 16 September 2020; Published: 19 October 2020

**Abstract:** Pepper berry (*Piper nigrum* L.) is known as the king of spices and has sharp, pungent flavour and aroma. In this study, the physical properties (weight, dimensions, sphericity, volume, surface area, and projected area) were measured, and the mass of pepper berries of the Kuching variety at different maturity levels (immature, mature, and ripe) was predicted using four models: linear, quadratic, s-curve, and power. When the models were based on volume and projected area, the mass could be predicted with maximum precision. The Quadratic model was best fitted for mass prediction at all mass maturity levels (immature, mature, and ripe). The results showed that mass modelling based on the actual volume of pepper berries was more applicable compared to other properties with the highest determination coefficient, 0.995, at the 1% probability level. From an economical point of view, mass prediction based on actual volume in the Quadratic form, *<sup>M</sup>* = 0.828 <sup>−</sup> 0.015 *<sup>V</sup>* <sup>+</sup> 7.376 <sup>×</sup>10−5*V*2, is recommended. The findings of physical properties and mass modelling of the berries would be useful to the scientific knowledge base, which may help in developing grading, handling, and packaging systems.

**Keywords:** piper nigrum; dimensions; mass; maturity levels; modelling

#### **1. Introduction**

Pepper (*Piper nigrum* L.) is a perennial climber belonging to the family Piperaceae. The leafy pepper tree is comparable to an almond tree that is able to grow tall, reaching a height of 10 m or more. When the main stem matures, many side shoots start growing on it to make a dense canopy. Pepper has a sharp, pungent aroma and flavour. It is widely planted in Vietnam, Malaysia, Indonesia, Sri Lanka, and other areas. According to the Food and Agriculture Organisation of the United Nations Statistics [1], Malaysia was the world's seventh largest pepper producer in 2017. About 98% of pepper production comes from Sarawak, the largest national producer; the other 2% is produced in Johor and other states in Malaysia.

A few recommended local pepper varieties include Kuching, Semongok Emas, and Semongok Aman [2]. Kuching pepper is the most commonly grown cultivar in Sarawak and Johor, Malaysia, due to its high and stable yield. It also has a denser canopy and thinner pericarp, which is the preferred variety for white pepper, production as compared to others. Semongok Emas and Semongok Aman are often used for black pepper production due to their pungent flavour, high yield, and thicker pericarp. In the harvesting process, the selection of pepper berries depends on the maturity levels such as immature, mature, and ripe. Immature pepper berries are described as light green. Mature

pepper berries are dark green and yellowish-green, and ripe pepper berries are red. Black pepper and white pepper are the most common and well-known worldwide. For producing black pepper, dark green pepper berries (mature pepper berries) with soft peppercorns and sharp, pungent aroma and flavour are used. They will undergo the drying process without removing the outer skin (pericarp). Yellowish-green and red pepper berries (mature and ripe pepper berries) are used for white pepper production due to the hardness of the peppercorn and thinner pericarp. They are produced by removing the outer skin (pericarp) after a soaking process, and the peppercorns will undergo a drying process. In general, the average diameter of mature and ripe pepper berries for all varieties is about 5 to 6 mm. Overall, the size of pepper berries is similar at all maturity levels. Furthermore, pepper, referred to as black pepper and white pepper, is used in Ayurvedic remedies for health benefits (appetite stimulant, breathing aid, cough therapy, anaemia, and others), keeping food fresh by combining with salt to cure and flavour a wide variety of meats (early food-preservation techniques), and reducing pain perception and inflammation. It is also rich in vitamins and minerals.

Although the berries are often graded by size, weight-based grading is essential in packaging and handling because it may be more economical. Therefore, the physical properties and the relationships must be determined to design and optimise a machine for handling, cleaning, conveying, and storing [3]. The most crucial properties in the design of the grading system are dimension, mass, volume, and projected area [4,5]. According to Pradhan et al. [6] and Pathak et al. [7], basic data on the physical properties of bio-materials are useful to all engineers, industrial processors, scientists, pharmacists, crop breeders, researchers, and others who may discover their advanced usage. Analysis of regression produces an equation to describe and predict the statistical relationship between one or more predictor factors and the variable of reaction. Some regression relationships, such as Linear, Quadratic, S-curve, and Power, were mostly used in previous studies. The statistical significance in the regression relationship shows that the dependent variable modifications in the dependent variables correlate with shifts. Therefore, a high coefficient of determination (*R*2) indicates that the model explains a decent percentage of the variability in the dependent variable. The correlation determination technique between mass and its physical properties is more specific for automatic classification of most berries and fruits.

There is a lack of previous research on mass prediction by using model equations for berries compared to other fruits. A few differences between berries and other fruits can be observed. The berries are fleshy fruits produced from a single ovary, while other fruits can be produced from single or multiple ovaries. The berries are edible when the entire ovary wall is ripened and yet, it may not be the same for all fruits. The study of mass modelling of pepper berries has not been carried out in detail in published works. Thus, this research aims to determine some physical properties such as the dimensions (major, medium and minor axis), volume, aspect ratio, geometric mean diameter, sphericity, surface area, and projected areas of immature, mature, and ripe pepper berries. It also serves to determine the most applicable model for predicting the mass of pepper berries at different maturity levels to form an essential database for other researchers. To the authors' knowledge, two previous published works concerned the mass modelling of berries based on their physicochemical properties such as raisin berries [8] and sea buckthorn berries [9]. In other research, many valuable studies were concerned about the mass modelling of fruits based on their physicochemical properties that included oranges [10], apples [11], pomegranates [12], onions [13], fava beans [3], persimmons [5], pomelos [14], dates [15], Kinnow mandarin [16], chebula fruits [7], and banana [17]. In summary, the study findings may help in developing grading, handling, and packaging systems based on the physical properties at different maturity levels of pepper berries.

178

#### **2. Materials and Methods**

The Kuching variety of pepper berries was obtained from a farm located in Johor, south of Malaysia (Johor, Malaysia). The fresh pepper berries were immediately transported to the laboratory within 4 h. The pepper berries were selected and divided into three maturity levels: immature, mature, and ripe. For sample selection, colour is an essential visual parameter to differentiate pepper based on the maturity levels. Light green and red berries were labelled as immature and ripe pepper berries, respectively [2]. Dark green and yellowish-green berries were labelled as mature pepper berries [2]. One hundred samples of pepper berries from each maturity level were selected and used for physical property measurements [18,19] as shown in Table 1. All the experiments were conducted at room temperature in the laboratory.

#### *2.1. Measurements of Physical Properties*

Pepper berry mass (*M*) was determined with 0.01 g sensitivity of an electronic balance. Three linear dimensions, namely, the major axis (*L*), medium axis (*T*), and minor axis (*W*), were measured by using a digital vernier calliper with 0.01 mm sensitivity to determine the average size of the samples. The method of water displacement [1,3,12,20,21] was used to determine the actual volume (*V*). The uniform presumed shape of the fruit can be an aspect to determine the fruit volume accuracy [5]. The aspect ratio (*AR*), geometric mean diameter (*Dg*), and sphericity (Φ) were calculated by using the following respective formulas [5,20–22]:

$$AR = \frac{L}{W} \tag{1}$$

$$D\_{\mathcal{R}} = (L\mathcal{W}T)^{\frac{1}{5}}\tag{2}$$

$$
\phi = \frac{(LWT)^{\frac{1}{2}}}{L} \tag{3}
$$

The surface area is defined as the total three-dimensional shape areas of all surfaces. Equation (4) was used to calculate the spheroid surface area of pepper berries.

$$SA\_{sp} = 4\pi r^2\tag{4}$$

The projected areas of pepper berries (*PAL*, *PAT*, and *PAW*) in three perpendicular directions to the dimensions (major axis, medium axis, and minor axis) and the criteria projected area (*CPA*) were calculated by using Equations (5)–(8), respectively. These equations were suggested by Mohsenin [4] and Nur Salihah et al. [14] and are defined as follows:

$$PA\_L = \frac{\pi LW}{4} \tag{5}$$

$$PAT = \frac{\pi TW}{4} \tag{6}$$

$$PA\_W = \frac{\pi WW}{4} \tag{7}$$

$$\text{CPA} = \frac{PA\_L + PA\_T + PA\_W}{3} \tag{8}$$



**Table**

volume; *SAsp*, spherical surface area; *PAL*, projected area perpendicular to major axis; *PAT*, projected area perpendicular to medium axis; *PAW*, projected area perpendicular to minor *CPA*, criteria projected area. Different letters in a row indicate statistically significant differences at *p* < 0.001. Means that do not share a letter are significantly different. Tukey's testappliedwith95%confidenceintervals.

 was

#### *Processes* **2020** , *8*, 1314

#### *2.2. Regression Analysis and Mass Modelling*

In order to predict the mass of pepper berries based on measured physical properties (dimensional characteristics, volume, surface area, and projected area), the considerations of model classifications are as follows:


Four models, namely, Linear, Quadratic, S-curve, and Power, were used and fitted with the results obtained from the experiments. These models are presented in Equations (9)–(12), respectively [3,5]:

$$M = a + bX \tag{9}$$

$$M = a + bX + cX^2 \tag{10}$$

$$M = a + \frac{b}{X} \tag{11}$$

$$M = aX^{\flat} \tag{12}$$

where *M* = *mass* (g); *X* = the value of an independent parameter (physical properties) to find the relationship of it with mass; *a*, *b*, and *c* = curve fitting parameters which are different in each equation.

#### *2.3. Statistical Analysis*

Data analysis and mass modelling prediction were performed by using statistical software such as SigmaPlot (Version 12.0) with significance at the 1% probability level. Coefficient of determination (*R*2) and standard error of the estimate (*SEE*) were selected as the criteria to evaluate the applicability of the regression models. The applicable models were selected as those with higher *R*<sup>2</sup> and lower *SEE* values [21].

#### **3. Results and Discussion**

#### *3.1. Physical Properties of Pepper Berries*

Table 1 summarizes the results of the physical properties of pepper berries based on the maturity levels. The mean values of the major axis (*L*), medium axis (*T*), and minor axis (*W*) for immature pepper berries were 4.75 mm ± 0.38, 5.09 mm ± 0.50, and 4.74 mm ± 0.34, respectively. Therefore, the medium axis had the highest mean value as compared to the major and minor axes of immature pepper berries. The mature pepper berries had mean values of 5.73 mm ± 0.32, 6.16 mm ± 0.44, and 5.87 mm ± 0.34 for major axis, medium axis, and minor axis, respectively. Again, the medium axis had the highest mean value as compared to the major and minor axes of mature pepper berries. Furthermore, the mean values of the major axis, medium axis, and minor axis for ripe pepper berries were 5.55 mm ± 0.66, 5.60 mm ± 0.58, and 5.67 mm ± 0.51, respectively. For the results of dimensional characteristics, the minor axis had the highest mean value as compared to the major and medium axis of ripe pepper berries. As can be seen, mature pepper berries had the highest values of the major, medium, and minor axes among these three maturity levels.

The mean value of the aspect ratio (*AR*) of immature pepper berries was 1.00 with a standard deviation of 0.03. The mature pepper berries had a mean value of 0.98 with a standard deviation of 0.02 for the aspect ratio. The mean value of the aspect ratio of ripe pepper berries was 0.98 with a standard deviation of 0.13. However, the value of the aspect ratio for immature, mature, and ripe pepper berries was significantly in same range as shown in Table 1.

The mean value of the geometric mean diameter (*Dg*) of immature pepper berries was 4.85 mm with a standard deviation of 0.38. Mature pepper berries had a mean value of 5.92 mm with a standard deviation of 0.35 for the geometric mean diameter. Furthermore, the mean value of the geometric mean diameter of ripe pepper berries was 5.61 mm with a standard deviation of 0.46. Thus, the mean value of the geometric mean diameter of mature pepper berries was the highest among all three maturity levels.

Sphericity can be described a solid shape formed relative to that of the same volume of a sphere. The ideal shape of the sphere is a solid with a high sphericity value [4,18]. A sphericity value of 1 is an ideal sphere. The mean sphericity value for immature pepper berries was 1.02 with a standard deviation of 0.02. The mature pepper berries had a mean value of 1.03 with a standard deviation of 0.02 for sphericity. As for ripe pepper berries, the mean value of sphericity was 1.01 with a standard deviation of 0.08. Thus, from the mean values of sphericity for all three maturity levels, it can be considered that the shapes of immature, mature, and ripe pepper berries were an ideal sphere. The mean weight of immature pepper berries was 0.10 ± 0.01 g. The mature pepper berries had a mean weight of 0.14 ± 0.01 g. The mean weight of ripe pepper berries was 0.15 g ± 0.04 g. However, the weight of immature, mature, and ripe pepper berries was significantly in the same range as shown in Table 1.

Based on Table 1, other physical properties of pepper berries included actual volume. The immature pepper berries had mean actual volumes of 96.67 mm<sup>3</sup> <sup>±</sup> 5.77. The mean actual volume of mature pepper berries was 120 mm<sup>3</sup> <sup>±</sup> 10.00. Furthermore, the ripe mature pepper berries had mean values of 120 mm<sup>3</sup> <sup>±</sup> 21.60 for the actual volume. Overall, the mature and ripe pepper berries had the highest mean actual volumes when compared to the immature pepper berries.

The mean value of the spheroid surface area of immature pepper berries was 74.65 mm<sup>2</sup> <sup>±</sup> 11.43. The mature pepper berries had a mean value of 110.49 mm2 <sup>±</sup> 13.66 for the spheroid surface area. The mean value of the spheroid surface area of ripe pepper berries was 99.39 mm2 <sup>±</sup> 16.66. Thus, the highest mean value of the surface area for the spheroid was obtained for mature pepper berries as shown in Table 1.

Table 1 shows the results of mean projected areas, perpendicular to the major axis (*PAL*), medium axis (*PAT*), and minor axis (*PAW*). The results obtained for immature pepper berries were 17.78 mm2 <sup>±</sup> 2.75 (*PAL*), 19.04 mm2 <sup>±</sup> 3.08 (*PAT*), and 17.73 mm2 <sup>±</sup> 2.57 (*PAW*). The mean values of *PAL*, *PAT*, and *PAW* for mature pepper berries were 26.50 mm<sup>2</sup> <sup>±</sup> 3.28, 28.51 mm2 <sup>±</sup> 3.93, and 27.16 mm<sup>2</sup> <sup>±</sup> 3.45, respectively. As for the ripe pepper berries, the mean values were 24.78 mm2 <sup>±</sup> 4.27, 25.10 mm<sup>2</sup> <sup>±</sup> 4.51, and 25.45 mm2 <sup>±</sup> 4.56 for *PAL*, *PAT*, and *PAW,* respectively. Therefore, the mature pepper berries had the highest mean values of *PAL*, *PAT,* and *PAW* when compared to the other two maturity levels of pepper berries. The criteria of projected area were determined by using the results of *PAL*, *PAT*, and *PAW* as indicated in Equation (11). The mean values of projected area criteria for immature, mature, and ripe pepper berries were 18.19 mm<sup>2</sup> <sup>±</sup> 2.73, 27.39 mm2 <sup>±</sup> 3.52, and 25.11 mm<sup>2</sup> <sup>±</sup> 4.13, respectively. Thus, the projected area criteria for mature pepper berries had the highest mean value in Table 1.

Projected area values are fundamental information in the design and development of machine vision-based grading systems [16]. Projected area is also useful for estimating the respiration rate, maturity index, and gas permeability to predict optimum harvest time, water loss, and heat and mass transfer during drying and cooling [7,16]. However, some ellipsoidal shapes of ripe pepper berries were observed among the samples. Also, the measured distinction in the physical properties may be due to the inherent variation in fruit dimensional characteristics.

#### *3.2. Mass Modelling*

The average dimensions, volumes, weight, surface areas, and projected areas of pepper berries obtained were used in mass modelling. Tables 2–4 show the best models obtained and their coefficients of determination, *R*2, and *SEE* to predict mass by using the measured average dimensions, volumes, weight, surface areas, and projected areas of pepper berries. The correlations of physical properties with pepper berry mass as shown from the results obtained were significant at the 0.01 probability level. The regression mass was evaluated by using the coefficient of determination (*R*2), where the best fit model was shown with a higher *R*<sup>2</sup> value (near 1.00).

#### *3.3. Models Based on Dimensions*

According to Table 2, the major axis (*L*), medium axis (*T*), minor axis (*W*), and geometric mean diameter (*Dg*) showed that the Quadratic model was the best-fit model to calculate and evaluate the mass of immature, mature, and ripe pepper berries. Table 2 shows the fitted models based on dimensions such as *L*, *T*, *W*, and *Dg* with the values of *R*<sup>2</sup> and *SEE*.

For immature pepper berries, *Dg* had a highest value of *R*<sup>2</sup> and the lowest value of *SEE,* which were 0.938 and 0.002, respectively, as indicated in Table 2. Equation (13) shows the equation of the Quadratic model obtained.

$$M = 0.225 - 0.071 \, D\_{\%} + 0.009 \, D\_{\%} \, ^2 \tag{13}$$

According to Table 2, the mature pepper berries had *W* and *Dg* values with the highest *R*<sup>2</sup> (0.960) and the lowest *SEE* (0.001). The model equation obtained for these parameters was Quadratic. Equations (14) and (15) were determined for the parameters *W* and *Dg*.

$$M = 0.260 - 0.053\ W + 0.005\ W^2\tag{14}$$

$$M = 0.196 - 0.032 \, D\_{\mathcal{S}} + 0.004 \, D\_{\mathcal{S}} \, ^2 \tag{15}$$

*L* had the highest *R*<sup>2</sup> (0.980) and the lowest *SEE* (0.007) values for ripe pepper berries when compared to the others as given in the Quadratic model equation (Equation (16)).

$$M = 0.351 - 0.107 \, L + 0.012 \, L^2 \tag{16}$$

A similar model for onion and a Malaysian variety of pomelo fruit in another study was suggested and reported by Ghabel et al. [13] and Nur Salihah et al. [14], where the best model for mass determination based on *L* was a Quadratic model following Equations (17) and (18):

$$M = 36.137 - 1.64 \, L + 0.35 \, L^2, \, R^2 = 0.96 \tag{17}$$

$$M = 28500.33 - 417.191 \text{ L} + 1.587 \text{ L}^2, \; R^2 = 0.993 \tag{18}$$

For the entire dimensions, the S-curve model was reported to have lower *R*<sup>2</sup> values as compared to other fitted models. The lower *R*<sup>2</sup> values could be due to the non-uniform mass of pepper berries corresponding to their size. Thus, the sizing of pepper berries based on length is recommended.

**Table 2.** Mass models of pepper berries based on dimensions.


*Processes* **2020**, *8*, 1314



Data are expressed as *L*, major axis; *T*, medium axis; *W*, minor axis; *Dg*, geometric diameter; *R*2, coefficient of determination; *SEE*, standard error of estimate; a, b, c, constantscurve fittings.

 of


Mass models of pepper berries based on volume and surface

area.

**Table**

**3.**

*Processes* **2020** , *8*, 1314 **Table 4.** Mass models of pepper berries based on projected

area.


*Processes* **2020**, *8*, 1314



*CPA*, criteria projected area; *R*2, coefficient of

determination; *SEE*, standard error of estimate; a, b, c, constants of curve fittings.

#### *3.4. Models Based on Volume*

In Table 3 which shows the mass prediction results of the pepper berries based on actual volume (*V*), the Quadratic model based on *V* (Equation (19)) was found to be the best fit when compared to the other models. It had the highest *R*<sup>2</sup> of 0.995 and the lowest *SEE* of 0.006 for ripe pepper berries.

$$M = 0.828 - 0.015\,\text{V} + 7.376 \times 10^{-5}\,\text{V}^2\tag{19}$$

For immature pepper berries, the Quadratic model based on *V* was suitable with the highest values of *R*<sup>2</sup> and *SEE,* which were 0.925 and 0.002, respectively, as shown in Table 3. Equation (20) was obtained for the pepper berries at the immature maturity level.

$$M = 0.081 - 5.726 \times 10^{-6} V + 3.703 \times 10^{-6} V^2 \tag{20}$$

The mature pepper berries had a Quadratic model based on *V* as the best model with the highest *R*<sup>2</sup> of 0.988 and the lowest *SEE* of 0.001, shown in Table 3. The Quadratic model equation (Equation (21)) is as follows:

$$M = 0.795 - 0.012\,\text{V} + 5.164 \times 10^{-5} \text{V}^2\tag{21}$$

Thus, the ripe pepper berries had the highest value of *R*<sup>2</sup> and lowest *SEE* for actual volume among the maturity levels. Therefore, a quadratic form was shown as the suggested mass model-based volume, similar to the prediction of the Fava bean mass with *R*<sup>2</sup> = 0657 [3].

#### *3.5. Models Based on Surface Area*

As shown in Table 3 for the results of mass prediction of pepper berries based on surface area (*SAsp*), the Quadratic model was the best based on the highest value of *R*<sup>2</sup> compared to the other models. For the best fit model, the Quadratic model based on *SAsp* of ripe pepper berries had the highest value of *R*<sup>2</sup> and lowest *SEE* of the surface area-assumed shape; the respective values were 0.984 and 0.006 (as shown in Equation (22)). It was also the best fit among the models of other maturity levels of pepper berries.

$$M = 0.118 - 0.001 \text{ SA}\_{\text{sp}} + 7.900 \times 10^{-6} \text{ SA}\_{\text{sp}}^2 \tag{22}$$

The Quadratic model was suitable for *SAsp* which had the highest *R*<sup>2</sup> of 0.936 and lowest *SEE* of 0.002 for immature pepper berries (as shown in Equations (23)). The suitable model of immature pepper berries is shown in the following equation:

$$M = 0.097 - 6 \times 10^{-4} \text{ SA}\_{sp} + 7.825 \times 10^{-6} \text{ SA}\_{sp} \text{ 2} \tag{23}$$

For mature pepper berries, the suitable model equation was Quadratic. Based on Table 3, the parameter *SAsp* had the highest value of *R*<sup>2</sup> and lowest *SEE,* which were 0.960 and 0.001, respectively. The model equation was formed as indicated in Equation (24).

$$M = 0.124 - 1 \times 10^{-4} \text{ SA}\_{sp} + 1.715 \times 10^{-6} \text{ SA}\_{sp} \text{ 2} \tag{24}$$

Among these maturity levels, the ripe pepper berries had the highest value of *R*<sup>2</sup> and lowest *SEE* for the spheroid surface area in Quadratic form. Therefore, the Quadratic form was shown as the suggested mass model.

#### *3.6. Models Based on Projected Area*

Among the models based on the projected area (*PAL*, *PAT*, *PAW*, and *CPA*), the Quadratic model comprising *PAT* was the best fit with the highest *R*<sup>2</sup> of 0.71 and lowest *SEE* of 0.001 for mature pepper berries as shown in Table 4. Equation (25) shows the model equation obtained. The Quadratic model based on *PAT* (Equation (26)) was suitable, with *R*<sup>2</sup> of 0.946 and *SEE* of 0.002, with respect to immature pepper berries. The Quadratic model based on *PAW* (Equation (27)) also achieved a suitable *R*<sup>2</sup> of 0.942 with *SEE* of 0.012 for ripe pepper berries.

$$M = 0.123 - 0.002PA\_T + 4.808 \times 10^{-5} PA\_T^{\prime 2} \tag{25}$$

$$M = 0.114 - 0.004PA\_T + 2 \times 10^{-4}PA\_T^{-2} \tag{26}$$

$$M = 0.207 - 0.010PA\_W + 3 \times 10^{-4}PA\_W^2 \tag{27}$$

Furthermore, the Quadratic model of the projected area based on the criteria projected area (*CPA*) shown in Table 4 was preferred as the best model to calculate the mass of pepper berries. Due to the high *R*<sup>2</sup> value of 0.966, a model based on *CPA* (Equation (28)) for ripe pepper berries was found as the best fit with *SEE* of 0.009. The mature pepper berries had values of *R*<sup>2</sup> and *SEE* (Equation (29)) which were 0.962 and 0.001, respectively. The immature pepper berries had the lowest *R*<sup>2</sup> and *SEE* values for the Quadratic model based on *CPA* (Equation (30)), which were 0.880 and 0.003, respectively, compared to other maturity levels of pepper berries.

$$M = 0.221 - 0.011 \text{CPA} + 3 \times 10^{-4} \text{CPA}^2 \tag{28}$$

$$M = 0.130 - 0.001 \text{CPA} + 3.55 \times 10^{-5} \text{CPA}^2 \tag{29}$$

$$M = 0.102 - 0.003 \text{CPA} + 2 \times 10^{-4} \text{CPA}^2 \tag{30}$$

Thus, all three projected areas are necessary to be specified and applied in grading the pepper berries by using this model.

#### **4. Conclusions**

In the current study, the mass of ripe pepper berries based on volume (Quadratic model) is the recommended equation, as the nonlinear form *<sup>M</sup>* = 0.828 <sup>−</sup> 0.015*<sup>V</sup>* + 7.376 <sup>×</sup>10−5*V*<sup>2</sup> had the highest *R*2, 0.995, and lowest *SEE*, 0.006, when compared to other maturity levels of pepper berries. This shows a very good relationship between the mass and actual volume of pepper berries. The model predicting the mass of pepper berries considered as spheroid was found to be the most applicable (Quadratic model is recommended). Finally, the Quadratic model is applicable to all properties due to its economical viewpoint. The mass model of pepper berries based on actual volume in the obtained results is recommended for designing and optimizing machines for handling, cleaning, conveying, and storing.

**Author Contributions:** P.N.M.A.A. conducted the experiments, collected and analyzed the data of results, and wrote the manuscript. R.S. supervised the research and revised the manuscript. H.C.M. and M.E.Y. supervised the experiments. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to express their appreciation to Universiti Putra Malaysia for the financial and technical support given during this research work.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Susceptibility of** *Tribolium castaneum* **(Coleoptera: Tenebrionidae) to the Fumigation of Two Essential** *Satureja* **Oils: Optimization and Modeling**

**Asgar Ebadollahi 1,\*, Ebrahim Taghinezhad 1, William N. Setzer 2,3 and Guangnan Chen 4,\***


**Abstract:** Due to the numerous side effects of synthetic pesticides, including environmental pollution, threats to human health, harmful effects on non-target organisms and pest resistance, the use of alternative healthy, available and efficient agents in pest management strategies is necessary. In this paper, the susceptibility of the cosmopolitan, polyphagous, stored-product pest *Tribolium castaneum* (red flour beetle) to the fumigation of the essential oils of two important medicinal and food additive plants, *Satureja hortensis* and *S. intermedia*, was investigated. The insecticidal properties of the essential oils were modeled and optimized using response surface methodology. It was found that a maximum significant mortality of 94.72% and 92.97% could be achieved within 72 h with the applications of 55.15 μL/L of *S. hortensis* (with the linear model) and 58.82 μL/L of *S. intermedia* (with the quadratic model), respectively. There were insecticidal terpenes and phenylpropanoids in both essential oils, including thymol (50.8%), carvacrol (11.2%) and *p*-cymene (13.4%), in the *S. intermedia* and estragole (68.0%) and methyl eugenol (5.6%) in the *S. hortensis*. It was suggested that the essential oils of *S. hortensis* and *S. intermedia* could be offered as promising pesticidal agents against *T. castaneum* for further studies in the management of such pests instead of detrimental synthetic pesticides.

**Keywords:** biorational pesticides; chemical profile; fumigant toxicity; modeling; optimization; *S. hortensis*; *S. intermedia*

#### **1. Introduction**

The globally distributed insect pest, the red flour beetle *Tribolium castaneum* Herbst (Coleoptera: Tenebrionidae), attacks several stored grain products such as beans, cereals, chocolate, flour, meal, nuts, seeds and spices [1]. The adults of *T. castaneum* are long-lived; they can live for up to three years [2]. In addition to qualitative nutritional damage, contaminated products also have reduced quality due to the creation of an unpleasant odor by the secretion of benzoquinone compounds from abdominal glands and remains of various molting stages and cadavers [3]. Furthermore, *T. castaneum* can transmit phytopathogenic microbial agents such as *Aspergillus*, *Pseudomonas* and *Staphylococcus* to infested storage products [4,5]. Bosly and Kawanna [6] demonstrated that *T. castaneum* can actually carry and distribute some *Aspergillus* species, particularly *A. flavus*, the main producer of carcinogenic aflatoxins in agricultural products [7], in stored flour.

Although the use of synthetic pesticides is the principal approach in pest management, the widespread use of these compounds has resulted in numerous side effects, such as environmental contaminations, the accumulation of hazardous residues in food and high toxicity to non-target organisms [8–10]. In addition, the resistance of pests to frequently used chemical pesticides is increasing. For example, the resistance of *T. castaneum* to

**Citation:** Ebadollahi, A.; Taghinezhad, E.; Setzer, W.N.; Chen, G. Susceptibility of *Tribolium castaneum* (Coleoptera: Tenebrionidae) to the Fumigation of Two Essential *Satureja* Oils: Optimization and Modeling. *Processes* **2021**, *9*, 1243. https://doi.org/ 10.3390/pr9071243

Academic Editor: David Fernández-Calviño

Received: 18 June 2021 Accepted: 13 July 2021 Published: 19 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

phosphine, as one of the commonly used fumigants against stored grain insect pests, was reported [11–13]. Therefore, the introduction of low-risk, available and effective pesticides is essential to alternate with synthetic chemicals in the management of such pests.

The effectiveness of plant-derived essential oils in insect pest control has been demonstrated in numerous studies in recent years [14,15]. Along with the prospective insecticidal effects of essential oils on insect pests from various species, genera, families and orders, their biodegradable nature, safety compared to synthetic chemicals and multiple modes of action in targeted pests indicate that they can be suitable alternatives to detrimental chemical pesticides [16,17]. In fact, the volatile oils obtained from different species of the *Satureja* genus were enriched with terpenes such as 1,8-cineole, γ-terpinene, thymol and β-caryophyllene, which were grouped into four primary classifications: monoterpene hydrocarbons, oxygenated monoterpenes, sesquiterpene hydrocarbons and oxygenated sesquiterpenes [18–20]. Promising insecticidal effects of *Satureja* essential oils against insect pests of storage products were also reported. These included, for example, insecticidal activities of *S. bachtiarica* Bung and *S. khuzestanica* Jamzad essential oils against *T. castaneum* [21], *S. hortensis* L. essential oil against the bean weevil (*Bruchus dentipes* (Baudi)) [19] and the cowpea weevil (*Callosobruchus maculatus* (Fabricius)) [22] and *S. spicigera* Boiss essential oil against the granary weevil (*Sitophilus granarius* (L.)) [23] and the maize weevil (*Sitophilus zeamais* Motschulsky) [24].

Response surface methodology (RSM) is a set of mathematical and statistical tools for the optimization of independent factors and modeling of dependent factors [25]. The optimization of agricultural and chemical processes requires the simultaneous optimization of several objective functions [26]. Therefore, RSM has been used in numerous agricultural and pharmacological fields [27–30]. For example, RSM was used to find mathematical models and optimized conditions for the toxicity of *Eucalyptus globulus* Labill and *Teucrium polium* L. essential oils against *T. castaneum* [31,32].

Given the importance of using natural and low-risk compounds in pest management, the main purpose of the present study is to investigate the possibility of using *S. intermedia* C. A. Mey and *S. hortensis* L. essential oils in the control of *T. castaneum*. The insecticidal properties of essential oils are modeled and optimized using response surface methodology. Due to changes in essential oil composition under different environmental and geographical conditions [18], the chemical profiles of the *S. intermedia* and *S. hortensis* essential oils were also assessed.

#### **2. Materials and Methods**

#### *2.1. Plant Materials and Extraction of the Essential Oils*

*S. intermedia* was gathered from its wild populations in the Heiran regions in Ardebil Province, Iran (38◦230 N, 48◦350 E and an elevation of 910 m). *S. hortensis* was collected from Parsabad, Ardebil Province, Iran (39◦38 N, 47◦52 E and an elevation of 52 m). The *Satureja* species were identified based on the work of Jamzad [33]. The fresh leaves and flowers of *S. intermedia* and *S. hortensis* were used for essential oil extraction. All specimens were allowed to dry in the shade over a 10-day period. For each plant sample, 100 g of plant material was added to a 2-L flask of the Clevenger apparatus and subjected to 3 h hydrodistillation. The extracted essential oils were then individually poured into glass vials, covered with aluminum foil, dried over anhydrous Na2SO4 and kept in the refrigerator at 4 ◦C until use.

#### *2.2. Chemical Profiles of the Essential Oils*

The *Satureja* essential oils were analyzed by gas chromatography–mass spectrometry using an Agilent 7890B GC with an Agilent 5977A mass selective detector (MSD), operated in the EI mode (electron energy = 70 eV, scan range = 50–550 amu and scan rate = 3.99 scans/s). The GC column was an HP-5ms fused silica capillary column (30 m in length, 0.25 mm internal diameter, (5% phenyl)-polymethylsiloxane stationary phase and 0.25 μm film thickness). Helium was the carrier gas, with a 52.8 kPa column head pressure and 1.0 mL/min flow

rate. The inlet temperature was 200 ◦C, and the interface temperature was 280 ◦C. The GC oven temperature was programed to hold for 1 min at 50 ◦C and then increase at a rate of 6 ◦C/min to a final temperature of 290 ◦C, with a total run time of 50 min. A 10% *w*/*v* solution of the sample in methanol was prepared, and 1 μL was injected using a 100:1 split ratio. Identification of the oil components was based on their retention indices, determined by reference to a homologous series of *n*-alkanes and by comparison of their mass spectral fragmentation patterns with those reported in the databases [34–36].

#### *2.3. Insect Rearing*

Adult *T. castaneum* specimens were collected from contaminated wheat grains in warehouses in Parsabad, Ardabil Province, Iran (39◦38 N, 47◦52 E and an elevation of 52 m). The adults were reared on wheat grains in cylindrical glass containers (720 mL) covered by a fine mesh cloth for ventilation. Contaminated grains were kept in the incubator at 27 ± 2 ◦C and 60 ± 5% RH in the dark. Newly emerged adult insects 1–10 days old were selected for the bioassays.

#### *2.4. Fumigant Toxicity*

A series of concentrations from 20.59 to 58.82 μL/L and from 21.00 to 55.15 μL/L was selected to assess the fumigant toxicity of *S. intermedia* and *S. hortensis* essential oils, respectively. Filter papers (Whatman No. 1; 2 × 2 cm) separately treated with the concentrations of the essential oils were glued to the inner surface of the screw caps of fumigant glass containers (340 mL). Twenty unsexed adults (1–10 days old) were transferred to containers before screwing their caps. Then, the caps were sealed to be impermeable to air with parafilm. The bioassay was repeated four times, and in the control groups, all steps were performed except for adding essential oils. The fumigant containers were kept in an incubator with 27 ± 2 ◦C and 60 ± 5% RH, and insect mortality was recorded after 24, 48 and 72 h of exposure. The mortality in the control groups was corrected by the following formula: *P*t = [(*P*o − *P*c)/(100 − *P*c)] × 100, in which *P*t is the corrected mortality percentage, *P*o is the mortality of the insects treated by essential oil concentrations and *P*c is the mortality of the insects in the control groups [37].

#### *2.5. Modeling and Optimization by RSM*

Using RSM under the historical data design by the statistical software Design Expert 8.0.6 (Stat-Ease, Inc., Minneapolis, MN, USA), various concentrations of essential oils were analyzed in five levels, with the exposure times in three levels as the independent variables and with the mortality of *T. castaneum* as the dependent variable.

Multiple linear regression analysis for the interactions of the independent and dependent variables was used to achieve the statistically appropriate mathematical model in the following second-order polynomial equation [38]:

$$y = \beta o + \sum\_{i=1}^{k} \beta\_i X\_i + \sum\_{i=1}^{k} \beta\_j X\_j + \sum\_{i=1}^{k} \sum\_{j=1}^{k} \beta\_{ij} X\_i X\_j + \sum\_{i=1}^{k} \beta\_{j\bar{j}} X\_j^2 \tag{1}$$

where *y* is the mortality of *T. castaneum* (dependent variable), *Xi* and *Xj* are the exposure times and essential oil concentrations, respectively (independent variables), *k* is a number of independent variables (2), *β<sup>o</sup>* is the intercept of the model, *β<sup>i</sup>* and *β<sup>j</sup>* are the coefficients of the linear parameters and *βij* specifies the quadratic parameter coefficient. The statistical relationship between the independent and dependent variables was evaluated by the correlation coefficient of determination (*R*2), adjusted *R*<sup>2</sup> and predicted *R*2. Optimization of the insect pest mortality caused by the fumigation of *S. intermedia* and *S. hortensis* essential oils was performed to the maximum desirability using Design Expert software with RSM. The statistical significance of the independent variables on the response variables was examined at a 95% confidence level (*p* < 0.05), and only the variables with a significant effect on the response variable were used in the proposed regression equation.

#### **3. Results**

The results of the GC-MS analyses of the essential oils of *S. intermedia* and *S. hortensis* are presented in Table 1. It can be seen that the essential oil of *S. intermedia* was rich in the phenolic monoterpenes thymol (50.8%) and carvacrol (11.2%) along with monoterpenes *p*-cymene (13.4%) and 1,8-cineole (4.3%). The essential oil of *S. hortensis*, on the other hand, was dominated by phenylpropanoids, especially estragole (68.0%) as well as methyl eugenol (5.6%) and (*E*)-*p*-methoxycinnamaldehyde (4.4%) (Table 1 and Figure 1).




**Table 1.** *Cont.*

RIcalc = Retention index, determined with respect to a homologous series of n-alkanes on an HP-5 ms column; RIdb = Retention index from the databases.

**Figure 1.** The structures of the major components in the essential oils of *Satureja intermedia* and *S. hortensis*.

Table 2 shows the results of the analysis of variance (ANOVA), which indicated that the mortality of *T. castaneum* adults was significantly affected by A (exposure time) and B (essential oil concentration) in the *S. hortensis* treatment and by A, B, and B2 in the *S. intermedia* treatment (*p* < 0.01). The lack-of-fit test for both essential oils was non-significant, demonstrating the validation of treatments (Table 2).

In Table 3, the fitting effect of different levels of the essential oil concentration and exposure time on the mortality amount is presented. The model's fitting was evaluated on the basis of the coefficient of determination (R2), adjusted R2 and prediction R<sup>2</sup> as well as the coefficient of variation (CV). It can be seen that all the R<sup>2</sup> values were high (>0.94), meaning that the response surface methodology models were suitable. Furthermore, the coefficient of variation for almost all parameters was below 8.4%. This means that the results demonstrated good accuracy and precision with the reliability of experiments. As

is shown in Table 3, it was found that the linear effects of the dependent variables on the mortality amount were significant (*p* < 0.05) under the essential oil of *S. hortensis*. The effect on the mortality value under the essential oil of *S. intermedia* was significant according to the quadratic equation (*p* < 0.05). The mortality was strongly influenced by the essential oil concentrations, based on the higher coefficient for B in the equation of Table 3.

**Table 2.** Results of analysis of variance for prediction of the fumigant toxicity of *S. intermedia* and *S. hortensis* essential oils against *T. castaneum*.


A: exposure time (h); B: essential oil concentrations (μL/L); NS: non-significant.

**Table 3.** RSM modeling results for predicting the mortality percentage under different concentrations of essential oils and exposure times.


A: exposure time (h); B: essential oil concentrations (μL/L).

Figure 2 presents the effect of different concentrations of essential oils (in five levels) and exposure times (in three intervals) on the mortality of *T. castaneum* adults. It can be seen that the susceptibility or the mortality of the insect pest was increased by increasing the concentrations of both the essential oils and the exposure times. According to the color points in Figure 2, the mortality of *T. castaneum* was increased from 15% to 95% and from 20% to 100% by increasing the concentrations of the *S. hortensis* and *S. intermedia* essential oils from 21.00 to 55.15 μL/L and from 20.59 to 58.82 μL/L, respectively, and the exposure times from 24 to 72 h.

Plots of the residuals versus the predicted mortality of *T. castaneum* adults caused by the fumigation of *S. intermedia* and *S. hortensis* essential oils are shown in Figure 3. It can be seen that the introduced models for predicting *T. castaneum* mortality were accurate and consistent with the variation hypothesis.

The optimized conditions for the significant maximum mortality of *T. castaneum* adults treated by the essential oils of *S. intermedia* and *S. hortensis* after different exposure times are shown in Table 4. After 72 h of exposure time, the maximum mortality of the pest (92.973%) may be achieved with 58.820 μL/L of *S. intermedia* essential oil with a desirability of 0.9121, while a concentration of 55.150 μL/L would be sufficient to attain the maximum mortality (94.72%) with *S. hortensis* essential oil with desirability of 0.997%. This indicated that the *T. castaneum* adults were more susceptible to the essential oil of *S. hortensis* than the *S. intermedia* (Table 4).

**Figure 2.** Three-dimensional diagrams of the mortality of *T. castaneum* caused by the fumigation of *S. intermedia* and *S. hortensis* essential oils.

**Figure 3.** Plots of residual vs. predicted mortality of *T. castaneum* caused by the fumigation of *S. intermedia* and *S. hortensis* essential oils.


**Table 4.** Optimization of the mortality of *T. castaneum* caused by the fumigation of *S. intermedia* and *S. hortensis* essential oils.

*S. hortensis* 50.000 61.952 29.233 1.000 \* The maximum significant mortality percentage based on high desirability, calculated by Design Expert software.

94.723 \* 72.000 55.150 0.997

#### **4. Discussion**

In the *S. intermedia* essential oil, 95.1% of the identified compounds were terpenes, in which oxygenated monoterpenoids such as thymol (50.8%) and carvacrol (11.2%) had high concentrations. In contrast, the terpenes had only 14.1% of the total *S. hortensis*

essential oil, and this essential oil was rich in benzenoid aromatics such as estragole (68.0%), methyl eugenol (5.6%) and (*E*)-*p*-methoxycinnamaldehyde (4.4%). Therefore, the essential oils studied in the present study were completely different in terms of chemical components. It was reported that γ-terpinene (42.3%), carvacrol (32.8%), *p*-cymene (8.1%), β-pinene (2.3%), β-caryophyllene (2.2%), α-thujene (1.9%) and α-pinene (1.5%) were the main components of the essential oil isolated from the aerial parts of *S. intermedia* at the flowering stage from Romania [39]. In the present study, only carvacrol (0.8%) and α-pinene (1.3%) in low concentrations were found in *S. hortensis* at the pre-flowering stage. In contrast, estragole and methyl eugenol, identified in high amounts in the present study, were not detected in the essential oil investigated by Chambre et al. [39]. The chemical composition of *S. intermedia*, as one of the *Satureja* species endemic to Iran, was assessed [40], and γ-terpinene (37.1%), thymol (30.2%), *p*-cymene (16.2%), limonene (3.9%), α-terpinene (3.3%) and β-myrcene (2.5%) were reported as the main components. In the present study, γ-terpinene, thymol and *p*-cymene were also identified, but there were no traces of limonene, α-terpinene or β-myrcene. Accordingly, differences in the chemical compositions of essential oils may be related to the different species and stages of plants as well as the temperature, humidity, height and other geographical conditions of the cultivation locales [18,41].

The insecticidal effects of the essential oils extracted from *S. hortensis* and *S. intermedia* against stored-product insect pests were also reported in some recent studies. The fumigant toxicity of *S. hortensis* essential oil against *C. maculatus*, *S. zeamais* and the Mediterranean flour moth (*Ephestia kuehniella* Zeller) was studied [22,24,42]. Additionally, the fumigant toxicity of *S. intermedia* essential oils against the lesser grain borer (*Rhyzopertha dominica* (Fabricius)) and the khapra beetle (*Trogoderma granarium* Everts) was observed in one of our previous works [43]. These results are consistent with the current findings about the insecticidal potential of essential oils of the *S. hortensis* and *S. intermedia* species. Overall, the insecticidal effects of plant essential oils are related to their active compounds such as terpenes and phenylpropanoids [44,45]. As is shown in Table 5, the strong insecticidal effects of the components were identified in high percentages in the *S. intermedia* and *S. hortensis* essential oils, including 1,8-cineole, carvacrol, estragole, methyl eugenol, *p*-cymene, thymol and α-pinene (in both essential oils). The insecticidal activity of *S. intermedia* and *S. hortensis* essential oils may be highly associated with these active compounds.

**Table 5.** Review of insecticidal effects for the main compounds identified in *S. intermedia* and *S. hortensis* essential oils.


Furthermore, according to the study of Kim et al. [44], estragole had more fumigant toxicity against *S. zeamais* (LC50 = 0.004 mg/cm3) and *T. castaneum* (LC50 = 0.013 mg/cm3) adults than the terpenes limonene, linalool, β-myrcene, α-pinene and α-humulene. Therefore, having a high percentage of estragole in *S. hortensis* essential oil can be the main reason for its higher toxicity than *S. intermedia*. However, the synergistic and antagonistic effects between all essential oil components should be considered. For example, the toxicity of 1,8-cineole and thymol against the third instar larvae of diamondback moths (*Plutella xylostella* (L.)) were synergistically enhanced in combination with another terpene, pulegone, while the pulegone alone had no significant toxicity to the pest [47].

According to our recent studies, the *T. castaneum* adults were susceptible to the fumigation of *S. hortensis* and *S. intermedia* essential oils [43,52]. However, the optimization and modeling of the fumigant toxicity of these essential oils, through RSM, are reported for the first time in the current study. The use of RSM in the optimization and modelling of the insecticidal efficiency of essential oils can be found in a few other investigations. For example, optimization of the fumigant toxicity of *Thymus vulgaris* L. essential oil indicated that a concentration of 25.86 μL/L and a 59.00-h exposure time were sufficient to achieve 50% mortality of *R. dominica* adults [53]. In another work on the essential oil of *Thymus kotschyanus* Boiss. & Hohen, the optimized condition for 50% mortality against *R. dominica* adults was 24.62 μL/L and a 57.98-h exposure time [54]. Based on the above-mentioned results, the essential oil of *T. kotschyanus* was more toxic than the *T. vulgaris* oil. Regarding the toxicity of *S. hortensis* and *S. intermedia* essential oils against *T. castaneum*, 29.23 and 33.13 μL/L of essential oils after 61.95 and 31.57 h, respectively, can result in 50% mortality of the pest. In one of our other studies, modeling and optimization of the fumigant toxicity of *Teucrium polium* L. essential oil against *T. castaneum* adults was evaluated through RSM, and it was found that 97.97% mortality of the insect pest could be attained by a 20 μL/L essential oil concentration after 72 h with the model +0.71 − 0.047A − 8.84E − 3B + 3.89E − 4AB + 3.27E − 3A2 + 8.38E − 5B2 (A and B are the time and essential oil concentration, respectively) [31]. In comparison with these results, and after 72 h of exposure time, 55.15 and 58.82 μL/L of *S. hortensis* and *S. intermedia* essential oils could result in 94.72% and 92.97% mortality of *T. castaneum*, respectively. The differences may be associated with different tested essential oils and subjected insect pests. In addition, the above-mentioned and present findings indicated that RSM is a useful method for the optimization and modeling of the insecticidal effects of essential oils.

The disadvantage of such essential oils (i.e., low persistence in environmental conditions) can be improved by micro-and nano-encapsulation formulations based on controlled release techniques [55,56].

#### **5. Conclusions**

The present study has shown that the essential oils isolated from the aerial parts of *S. intermedia* and *S. hortensis* had strong fumigant toxicity against the adults of the red flour beetle *T. castaneum*. GC-MS analyses indicated that there was a high percentage of insecticidal components such as estragole and methyl eugenol in the *S. hortensis* essential oil and thymol, carvacrol, *p*-cymene and 1,8-cineole in the *S. intermedia* essential oil. However, additional studies are required to determine which components have efficient insecticidal activity. Based on RSM, it was found that the best model for the prediction of mortality was linear and the quadratic equation for the essential oil of *S. hortensis and S. intermedia*, respectively. Up to 94.72% and 92.97% mortality of the insect pest, as the maximum significant mortality, can be attained with 55.15 and 58.82 μL/L of *S. hortensis* and *S. intermedia* essential oils, respectively, after 72 h. The essential oils of *S. hortensis* and *S. intermedia* can be proposed for further research in the management of *T. castaneum* and probably other stored-product insect pests.

**Author Contributions:** Conceptualization, A.E. and E.T.; methodology, A.E. and E.T.; software, A.E. and E.T.; formal analysis, E.T. and W.N.S.; investigation, A.E.; writing—original draft preparation, A.E. and E.T.; writing—review and editing, A.E., E.T., W.N.S. and G.C.; funding acquisition, G.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the University of Mohaghegh Ardabili (grant number 99/d/9/28635).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data can be made available upon request to the authors.

**Acknowledgments:** This research project was supported by the University of Mohaghegh Ardabili in Ardabil, Iran, which is greatly appreciated. W.N.S. participated in this work as part of the activities of the Aromatic Plant Research Center (APRC, https://aromaticplant.org/, accessed on 16 July 2021).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Processes* Editorial Office E-mail: processes@mdpi.com www.mdpi.com/journal/processes

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com

ISBN 978-3-0365-2374-3