**Identification of Cold Spots Using Non-Destructive Hyperspectral Imaging Technology in Model Food Processed by Coaxially Induced Microwave Pasteurization and Sterilization**

#### **Aswathi Soni 1, Mahmoud Al-Sarayreh 1, Marlon M. Reis 1, Jeremy Smith 2, Kris Tong <sup>2</sup> and Gale Brightwell 1,3,\***


Received: 11 June 2020; Accepted: 22 June 2020; Published: 26 June 2020

**Abstract:** The model food in this study known as mashed potato consisted of ribose (1.0%) and lysine (0.5%) to induce browning via Maillard reaction products. Mashed potato was processed by Coaxially Induced Microwave Pasteurization and Sterilization (CiMPAS) regime to generate an F0 of 6–8 min and analysis of the post-processed food was done in two ways, which included by measuring the color changes and using hyperspectral data acquisition. For visualizing the spectra of each tray in comparison with the control sample (raw mashed-potato), the mean spectrum (i.e., mean of region of interest) of each tray, as well as the control sample, was extracted and then fed to the fitted principal component analysis model and the results coincided with those post hoc analysis of the average reflectance values. Despite the presence of a visual difference in browning, the Lightness (L) values were not significantly (*p* < 0.05) different to detect a cold spot among a range of 12 processed samples. At the same time, hyperspectral imaging could identify the colder trays among the 12 samples from one batch of microwave sterilization.

**Keywords:** hyperspectral imaging; cold spots; microwave; sterilization; Maillard reaction

#### **1. Introduction**

Food processing technologies and sterilization regimes work best if they can provide uniform heating across the food being processed. This would not only ensure consistency in delivering quality parameters but would also make sure that each spot gets equal inactivation of any microbial contaminants present to confirm the requisites of food safety. While conventional sterilization involves heating food in cans or retort packages for a set period of time (121 ◦C at the coldest spot for more than 5 min to ensure an F0 leading to 12 D (12 log cycle) reduction of *Clostridium botulinum* spores) under pressure, they have been reported to have effects on the heat sensitive components of foods [1–3]. Microwave pasteurization and sterilization is an emerging thermal technology that combines preheating with hot water and microwave energy (915 MHz) to achieve sterilization in a shorter time as compared to the conventional technologies [4,5]. This technology was initially developed by Washington State University and 915 Labs with the funding provided by the US government and specific food companies [6]. The apparatus used in this study and referred henceforth as the Coaxially induced microwave-pasteurization and sterilization (CiMPAS) system was manufactured by Meyer Burger Germany GmbH (Hohenstein-Ernstthal, Germany) and the industrial microwave parts were

manufactured by MUEGGE GmbH (Reichelsheim, Germany). The regime involves a pre-heat cycle that is employed to bring each sample into a uniform temperature before processing, which is followed by bringing the temperature of the treatment chamber to 121 ◦C while holding for a pre-decided time depending on the F0 targeted. This is followed by a cooling cycle as the last part. F0 is the approximate time of exposure of the sample at 121.1 ◦C and this can be calculated using the actual exposure time at a variable temperature, which is calculated for an ideal microorganism with a temperature coefficient of destruction equal to 10 ◦C [7]. A well-reported challenge with microwave processing of food is non-uniform heating [6,8,9]. This drawback has been reduced to a significant extent with sterilization that combines moist heat with the microwave. However, new methods are continuously sought by researchers to identify any non-uniformity in the processed samples. For example, this includes non-destructive hyperspectral technology, chemical markers, and temperature loggers that could be customized for use inside the food package [10–13]. Non-uniform heating might lead to effects on quality and appearance along with concerns around food safety while minor variation might or might not be a concern to food safety. It is worth detecting the least heated or cooked regions to ensure that these spots receive enough heat to meet commercial sterilization requirements [13]. There are many ways to understand non-uniform heating depending on the food products. This includes the measurement of color change induced by thermal treatment, monitoring the concentration of heat sensitive products, or measuring the inactivation of heat-resistant bacterial spores in food. However, each of these methods have a few limitations and cannot be complete without another confirmatory assay. For example, time temperature metallic sensors are expensive, might slightly interfere with the microwave heating [12], and, depending on size, might not be able to cover all critical spots. Heat-sensitive chemical markers, such as Maillard products, have been used for a long time to indicate heat profiles after processing [12,14]. However, these might have limitations above 100 ◦C to depict any change in color using colorimetric assays due to limited sensitivity. Chemical marker M2 (4-Hydroxy-5-methyl-3(2H)-fura-none) is one of the products of Maillard reaction among the three products that have been identified as chemical markers namely 2,3-Dihydro-3,5-dihydroxy-6-methyl-(4H)-pyran-4-one(referred to as M-1) and 5-Hydroxymethylfurfural (M-3) [15]. Kinetics of M2 has been validated and reported using the mashed potato food model to identify cold and hot spots as a result of non-uniform heating (if any) after Microwave-assisted thermal sterilization [15,16]. This method with a slight modification in the composition of the food model has been used in this study to identify cold spots generated after coaxially-induced microwave pasteurization and sterilization.

The principle of spectroscopy in the visible and near-infrared (Vis-NIR) spectral region is the interaction of electromagnetic radiation with the sample, involving light absorption. Regular reflectance (specular) where the light incident angle with the sample surface is equal to the angle at which it is reflected means little or no interaction with the samples. External diffuse reflectance captures information about the surface of the sample. In addition, light scattering is due to the interaction of light with the sample [17,18]. The detection of the outgoing photons (as a result of scattering inside of the sample) enables us to identify absorbing/scattering as well as the amount reflected (specular and external diffuse). The balance of outgoing photons compared to incident photons is commonly used as a measure of how much was lost by absorption and scattering as well as the amount reflected. These processes are wavelength-dependent, which makes the use of the entire Vis-NIR spectrum a rich source of information about the chemical and structural characteristics of samples. The combination of Vis-NIRS with imaging techniques (hyperspectral imaging—HSI) enables scanning a region of interest (ROI) with a Vis-NIR spectrum being acquired per pixel in that region [17,18].

HSI has the potential to be used to understand the properties of different food products and non-uniform cooking that is no exception [19]. It offers an excellent option as it is a non-invasive technology. This has previously been used to identify contaminants in fruits [20] and damages due to handling in vegetables [21] contamination in poultry carcasses [22] and red-meat quality and safety grading [17,18]. Recently, new HSI sensors, called snapshot HSI sensors, have been introduced with advantages of low-cost systems, high-speed data collection (ability to work at the standard video

rate), and completely portable systems. However, these sensors provide limited spectral features on a short-range of wavelengths. Snapshot HSI systems showed success in many tasks in food processing research such as in meat quality and safety [17,23], and fruit and vegetable classification [24].

In the current study, hyperspectral imaging has been used to identify cold spots in CiMPAS-treated mashed potato and directly compared to the results of color changes due to a Maillard reaction after microwave-induced sterilization. This study is also an attempt to confirm the consistency of this method across three different processing replicates.

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

#### *2.1. Mashed Potato Model Food Preparation*

The mashed potato was prepared, as described by Bornhorst et al., [16] by replacing gellan gum with agar and eliminating the addition of calcium chloride. In short, for every 1000 g of mashed potato, 20 g of agar (Roagar, New Zealand) was added to 830 g of boiling water and mixed using a cake mixer at medium speed for 2 min. Then 150 g of mashed potato flakes were added while mixing to avoid the formation of lumps. The mix was cooled to 60 ◦C, which was followed by the addition of D-ribose (10 g) and lysine (5 g). This was then followed by mixing for another 2 min before being loaded into each of the 12 trays until a weight of 250 g was reached. Trays were then placed in BNB1 pouches (Cryovac, Hamilton, New Zealand) and sealed (230 mBar) in a Multivac C200 vacuum sealer (Multivac NZ Ltd, Auckland, New Zealand) before loading on to the carrier tray in preparation for processing.

#### *2.2. CiMPAS Treatment*

To understand the possibility of detecting a difference in heat achieved at various locations, the apparatus for coaxially-induced microwave pasteurization and sterilization (CiMPAS) technique was used. CiMPAS goes through the three steps in the sterilization protocol, which include preheating, hot water immersion, and microwaving followed by cooling. The CiMPAS tool was initially preheated by running a heating program. Then the hot water vessel was stabilized at 130 ◦C and the warm water vessel was stabilized at 30 ◦C, respectively, with a +/−0.5 ◦C tolerance on both. Packaged products were then loaded into a carrier tray, which was placed in the tool and processed using the sterilization regime. As the first step, the vessel was flooded with warm water at 30 ◦C for preheating the food products for 20–30 min. Following the preheating step, hot water was flushed into the vessel, microwave power was switched on at 12 kW, and the carrier tray was passed through antennae for a set period. This was followed by cooling water (30 ◦C) being flushed into the vessel to cool the product. Processed trays were removed from the carrier tray and placed into the chiller (4 ◦C) overnight before analysis. Samples were collected from three processing runs conducted on three different days separately for colorimetry and HSI analysis. Controls were exposed to similar storage conditions except CiMPAS processing.

#### *2.3. Hyperspectral Imaging System*

A low-cost snapshot hyperspectral imaging system in the reflectance sensing mode was applied for image collection. The snapshot HSI system, as shown in Figure 1, consisted of a sample stage and a snapshot hyperspectral camera (MQ022HG-IM-SM4 × 4-VIS [3,12]), which acquires data of 16 wavelengths in the spectral range of 466–639 nm, an illumination unit with a 70-watt quartz-tungsten-halogen (ASD Illuminator unit) (Malvern Panalytical Ltd, UK), and computer running image acquisition software (HS Imager) (Headwall Photonics, Massachusetts, USA). The distance between the camera and the sample stage was set to 60 cm. This value was empirically assessed and then synchronized with the camera by adjusting the exposure time and the frame rate of the camera to 5.1 ms and 7 images/second, respectively, which resulted in images with a spatial resolution of 0.37 × 0.37 mm/pixel.

**Figure 1.** Demonstration of the implemented snapshot hyperspectral imaging (HSI) system.

The exposure time of the camera was adjusted to prevent the saturation of the sensor. The threshold to prevent the saturation is '511' (i.e., any intensity over 511 is defined as saturated). The value of '511' was estimated by the sensor's manufacture based on size of each pixel in the sensor. The exposure time was adjusted based on the measurement of white tile used as a white reference, which is explained below.

Those configurations were used for collecting the raw HSI images *R*\_0 followed by calibration of the reflectance value as follows:

$$R = \frac{(R\_{\!\!\! -D} - D)}{(W - D)} \times \mathbb{C} \tag{1}$$

where *R* is a calibrated image, *R*\_0 is the raw image irradiance, *D* is dark reference data of the sensor, and W is the white reference data of the light source. The dark reference is a hyperspectral image collected when the camera was closed with its cap. Similarly, the white reference is an image of a standard white tile. These reference measurements are used to reduce the impact of experimental variation in the setup and lighting source. The ratio (*R*\_0−*D*) (*W*−*D*) defines a scale between "zero" corresponding to a dark reference and "one" corresponding to a white reference. The scaler *C* (equal to 511) is used to retrieve the original scale of the HSI sensor as "511" corresponds to a maximum value for non-saturated pixels.

A collection of HSI images, by the implemented system, were acquired for each tray individually in each processing replicate, which resulted in 36 HSI images including 12 images for each replicate. These images were then processed as per Equation (1) for obtaining the reflectance values.

#### *2.4. Image Segmentation*

After obtaining the images in reflectance, image segmentation was performed to extract the (ROI). The segmentation process includes several image processing operations including image thresholding using a specific numerical value (reflectance intensity) at a particular wavelength (536 nm) to remove the pixels that belonged to the background and then a set of image morphological operations for removing the noisy pixels and maintaining the shape of the samples. The sequence of these operations included closing, opening, filling holes, erosion, and dilation. The collected HSI images were masked by these segmentation steps for obtaining HSI images with only return on investments (ROIs) for further analysis (Figure 2).

**Figure 2.** Return on investment (ROI) extraction processes for mashed-potato hyperspectral imaging (HSI) images.

#### *2.5. Colorimetric Analysis*

Mashed potato model food samples (in trays) after CiMPAS processing were used for the colorimetry analysis on the three different layers of the product inside. A meat slicer with a uniform height was used to cut the middle slice and nine spots on each layer (top, middle, and bottom) were mapped in each tray for measuring the change in color. The top and bottom layers were then excluded after the preliminary assay as a heat/color pattern on both the top and bottom layer due to being in contact with the immersion water at 12 ◦C showing uniform browning. The mid-layer, which was more dependent on microwave penetration, was considered to be a better indicator. Each tray was divided into nine different spots as represented by the dots in Figure 3. The change in color of the samples was estimated using L\*a\*b\* (CIELAB) color space using a CR20 colorimeter (Minolta, Osaka, Japan). The aperture of the measurement tool was placed on the samples while taking care to avoid any light penetration through gaps and a cling wrap transparent sheet was used to cover the aperture to avoid liquid penetration into the lens. Calibration was done every time using a white tile before a new sample measured. The lightness (\*L values) using \*L a B space was recorded as previously reported [16,25]. The coldest spot was determined as the location with the highest (significant at 95% confidence level) L value.

**Figure 3.** Tray configuration for colorimetric analysis (**a**), where the numbers indicate their position in the processing tray and the arrow indicates the direction toward the waveguide. Division of each tray into nine regions for measuring L values (**b**).

#### *2.6. Validation: E*ff*ect of Temperature Increase on Browning*

For the kinetic study, Digital High-Temperature Oil Bath (Interlab, Wellington, New Zealand) was set at 121 ◦C. To measure the come-up time, three capsules (Figure 4) with mashed potato and pre-set TMI probes inserted were used.

**Figure 4.** Capsule set up for the oil bath work with the main parts.

Once the come-up time was determined, six capsules filled with mashed potato (15 g) were immersed in the oil bath while making sure there was no dripping or leakage from the capsules. Capsules were removed at an interval of 0, 2, 4, 6, 8, and 10 min. This was followed by immediately cooling in an ice slurry to avoid any further reactions due to residual heat. Three experimental replicates were conducted under the same setup. The change in color for each sample was estimated as described in Section 2.5.

#### *2.7. Data Analysis*

#### 2.7.1. Statistical Analysis of Images

Principal component analysis (PCA) is an established chemometric technique for dimensionality reduction and visualizing HSI data [21]. In the current study, a PCA was used to visualize the variation in the distribution of the heating pattern across the 12 trays, which were processed by one processing run as one replicate. Three processing replicates were considered. By applying PCA to the data, the two-dimensional data was converted into a two-dimensional matrix where each pixel can be converted into one row of reflectance values. This was further used to obtain the Eigenvalues or scores. To obtain a good fit into the PCA model, the average ROI for each image was not suitable as it generated a spectrum for each tray that could not represent the spatial variation. As a solution, super-pixel segmentation approaches have been previously used for converting the image into a set of non-overlapped regions (super-pixels), where pixels in each region share the same spectral and spatial (neighboring) information [26]. In the current study, a simple linear iterative clustering (SLIC) algorithm [26] was selected and adapted for generating the overlapped regions in an HSI image, where the Simple Linear Iterative Clustering (SLIC) algorithm was originally proposed for RGB images. In the SLIC algorithm, super-pixels are generated by clustering pixels based on their spectral similarity and proximity in the 2D plane of the image. The similarity is defined as the Euclidean distance in spectral and spatial domains (spectral response + xy), where the spectral response is the pixel reflectance vector (16 wavelengths) and xy are the coordinates of the pixel in the image plane. The Euclidean distance in a spatial domain is normalized by the maximum distance between any two segments because the

distance is not limited and depends on the image size. The algorithm takes two input parameters, which include a desired number of clusters (superpixels) and compactness parameter to control the shape of resulting clusters. In the clustering processes, the centers of clusters were initialized at regular grid intervals in the image plane. Followed by this, in an iterative process, the cluster centers were updated and the pixels belonging to a specific cluster were defined based on the similarity of cluster centers and pixel values (i.e., reflectance vector and x-y coordinates).

In this study, the SLIC algorithm [26] was used for over-segmenting the extracted ROIs into a set of non-overlapping regions (super-pixels) for each HSI image of the middle layer of the mashed-potato (Figure 5) after CiMPAS processing. The desired number super-pixels and the compactness parameter were set into 50 and 0.4, respectively. The values of these parameters were empirically selected to obtain super-pixels that approximately covered only dark or bright pixels in the image. Importantly, the large super-pixels (i.e., low number of super-pixels) could cover regions with mixing of dark or bright pixels in the super-pixel, which could affect the analysis of identifying cold and hot regions in the tray, while small super-pixels (i.e., high number of super-pixels) could increase the impact of noisy pixels in the analysis. The mean spectrum of each segment was extracted and used for fitting the PCA model and the HSI image of replicate 1 was used to fit the PCA model by using 600 spectra (on average of 50 segments in each tray of the 12 trays) for covering both spatial and spectral information of this replicate. The other trays were used for validating the results of the fitted PCA model.

**Figure 5.** Non-overlapping regions of a mashed-potato HSI image.

#### 2.7.2. Visualization

The impact of heating was investigated using heating map graphics. For generating these heating maps, the fitted PCA model was used to project spectra corresponding to all pixels from each hyperspectral image onto their PCA space. The color scale for this map was defined based on heat treatment. Thus, the PCA scores corresponding to the control samples are defined as the zero (no heat treatment). As a result, it is possible to define a direction of growth on values of scores regarding the application of heat. For example, if the scores in a principal component for the control samples are higher than the scores of samples in the heat-treated samples, it means the direction of heat application corresponds to a decrease of score values and vice versa. To enhance the visualization of the effect of heat treatment, the scores of heat-treated samples were re-scaled between 0 and 100, using a minimax approach, where zero indicates the coldest regions while 100 indicates the hottest regions. In this case, score values that are closest to values from control samples corresponds to 'zero' (lowest amount of heat treatment) and those with score values furthest from values corresponding to control samples are '100' (highest amount of heat treatment).

#### 2.7.3. Statistical Analysis of Colorimetry Results

The significant difference (*p* < 0.05) among the \*L values (Lightness values) was estimated using One-way ANOVA and then post hoc analysis at a 95% confidence level to see which values were different from each other. This method was used to understand the average difference in L values across 12 trays per replicate and to understand the difference between nine different regions that were measured per tray. Three processing replicates were used separately, and three technical replicates were averaged for analysis in each processing replicate.

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

The amount of heat penetration or distribution in food post CiMPAS processing depends on several factors including (but not limited to) dielectric properties of food [27], which could be related to the salt content [28], moisture content [29], and the waveguide distribution inside the treatment chamber [15] along with the temperature of the immersion water and pressure. Browning through a Maillard reaction has been reported to be an indicator of non-uniform heating [11]. However, the drawback remains in the lack of sensitivity using colorimetric analysis, which reduces the limit of detection. To improve the sensitivity, mashed potato samples were simultaneously analyzed by hyperspectral imaging as well as by colorimetry (\*L values). For the analysis, each processing replicate was considered as a separate set of work and, for the data analysis, technical replicates were used. This not only removed unclear normalization of the data but also helped to detect the variation in regions across three replicates.

#### *3.1. Kinetic Assay to Understand the E*ff*ect of Heat on Browning*

The Maillard reaction has been reported as a reaction in which a reducing sugar (ribose) condenses with a compound possessing a free amino group (amino acid) to give a condensation product, which is an N-substituted glucosamine. N-substituted glucosamine further gets rearranged to form the Amadori rearrangement product (ARP) [30]. This reaction depends on the pH. In neutral and acidic pH, it forms a furfural whereas, in alkaline pH along with many other reactive fission compounds, it forms a furanone. This is again followed by a range of chemical reactions, which can include cyclisation, dehydration, retro-aldolisation, rearrangements, isomerization, and further condensations. In the final stage, they form polymers called meladonins that lead to the formation of a brown color. Chemical marker M-2 (4-Hydroxy-5-methyl-3(2H)-furanone) has been reported as an effective tool to monitor heating patterns of foods in microwave sterilization [15,16]. However, to understand the browning at different layers of the mashed potato gel, a firmer gel was formulated by increasing the concentration of agar (2% of total volume). To verify the pattern of browning as an effect of an increase in time of exposure at a pre-defined temperature of 121 ◦C, the kinetic assay was used. The come-up time of mashed potato to match the temperature of the surrounding oil was found to be 4 mins, which was estimated by averaging the time taken by three independent mashed potato samples to reach 121 ◦C (data not shown). There was no significant difference in the \*L, \*a, or \*B values of mashed potato before and after the come-up time. However, the color changed (visually toward browning with an increase in exposure time at 121 ◦C (Figure 6a). The lightness (\*L) values showed a significant (*p* < 0.05) decrease and a linear reduction with time (R2 = 0.95) (Figure 6b). The \*a axis represents the green–red component with green in the negative direction and red in the positive direction. The \*a values showed a significant (*p* < 0.05) increase from 1.5 at time zero to nine after 10 min (Figure 6c). The b\* axis represents the blue-yellow component with blue in the negative direction and yellow in the positive direction. The b values did not show an upward gradient where samples after four, six, eight, and 10 min were not significantly different from each other, which indicated a saturation in values (*p* < 0.05). This indicated that b values would not be an appropriate measure to understand the change in color in mashed potato samples (Figure 6d).

**Figure 6.** Mashed potato samples with M2 after exposure up to 10 min at 121 ◦C. The visual change in color (**a**), change in \*L values (**b**), change in \*a values (**c**), and change in \*B values (**d**). The similar letters in each graph indicate no significant difference (*p* < 0.05).

This finding was in agreement with the previous report by Bornhorst et al. [16] when temperatures up to 100 ◦C were taken into consideration. In the current study, the temperature tested was 121 ◦C and visual dark browning was seen within 10 min.

Browning increased with exposure at 121 ◦C (tested up to 14 min, including come-up) and any difference in heat exposure reflected as a change in color (Figure 6). While these findings indicated that heat has a significant effect on browning, the next step involved the analysis of this effect on heat generated by microwave sterilization.

#### *3.2. Identification of Cold Spots Post CiMPAS Processing*

The CiMPAS processing tray consists of 12 slots for trays where each tray containing 250 mg of mashed potato after vacuum sealing. Each tray was first distributed into three layers: top, bottom, and mid-layer. Each layer was assigned nine spots for measurement of the color change. The nine spots on the mid-layer in each tray were measured and the average \*L values (*n* = 9) per tray was used to compare the 12 trays, which could be accommodated in one processing run. Initially, the average L values of nine spots in each tray were compared to the average L values of each of the other 11 trays and control. On the other hand, a comparison by colorimetry to identify browning can give us indicative results for a tray but not for a large sample set. There was no significant difference among the L values of the 12 trays and this was reproducible across three processing replicates (data not shown).

However, when the average \*L values (*n* = 3 processing replicates) of the nine spots were compared to each other in each tray and hot and colder regions were identified based on significantly (*p* < 0.05) lower and higher \*L values, respectively (Figure 7). Though there was a difference in heating patterns, it was not significant enough to understand a colder region among the 12 trays of the processing vessel. However, the colder spots on each tray were identified (Figure 3). These spots spanned both the outer regions of the processing tray where the trays would encounter the surrounding water. Though the water is at 121 ◦C most of the time, there is a preheating and cooling cycle, which holds the water at 60 and 30 ◦C and this would explain why the browning indicated a difference on the outer sides. To increase the sensitivity of the assay and to be able to look at the difference when all the 12 trays were being compared, hyperspectral image analysis was used.

**Figure 7.** Mid-layer analysis of each tray after processing at 121 ◦C. Similar letters indicate no significant difference (*p* < 0.05).

#### *3.3. Hyperspectral Image Analysis*

The average spectrum of regions in the tray results from HSI presented in Figure 8, where a clear difference between the control and heat treat samples is observed. Overall, control samples present a higher reflectance.

**Figure 8.** Average spectra for 12 processed trays and one control of mashed potato.

The scatter plot of scores from both principal component (PC) 1 and PC 2 are presented in Figure 9, where it is shown that PC 1 captures the difference between control and heat-treated samples. In this plot, the control samples present high positive values and heat-treated samples present negative values. Thus, PC 1 was used to develop the heating maps. In this case, scores in PC 1 are close to zero and correspond to areas least affected by heat-treatment, herein defined as cold, while scores in PC 1 that are furthest from zero correspond to areas more affected by heat-treatment, herein defined as hot. The visual representation that consisted of the hyperspectral data converted to a heat map clearly showed a difference in the heating pattern within each tray as well as in the whole processing tray consisting of 12 trays (Figure 9ic–iiic). The high variance described by the PC 1 (99.8%) suggests high correlation among variables and/or high correlation among samples. This indicates a linear variation in the concentration of pigments associated with the changes in color and that the ratio among these pigments did not vary due to variation in the temperature of different regions in the sample.

Hyperspectral imaging was found to be time-efficient in data acquisition as compared to the colorimetry and concurrently for increasing the limit of detection and ruling out wrong negatives due to a huge variation of the colorimetry results (Figure 9i–iii).

Clearly and in agreement with the above statistical analysis, these plots show that, for all replicates, the cold and hot spots are projected on the same location in the PCA space with the approximately same distance from the control sample. For instance, in the first replicate (Figure 9i), tray 3 was subjected to be the coldest region as per the average reflectance (*p* < 0.05) and PCA analysis. Similarly, in replicate 2 and 3, tray 1 and 2 were subjected as the coldest regions, respectively. These trays were close to each other and near the door in the processing set-up, which also indicates that this would be the colder region that accumulates less heat as compared to the regions away from the door.

**Figure 9.** *Cont*.

**Figure 9.** Hyperspectral analysis of processing runs (i, ii, and iii): PCA plot (**a**) average reflectance (**b**) and heat map (**c**) as a comparison of 12 trays (tray number 1 to 12 indicates its location in the processing vessel), where the image c in each figure indicates 'zero' (lowest amount of heat treatment) and '100' (highest amount of heat treatment).

The observations support the potential to use a snapshot HSI imaging system for visualizing the heating processes of the processed food like mashed potatoes, especially where colorimetry (\*L values) might not be sufficiently sensitive. Prediction of the colder regions through hyperspectral imaging would not only indicate regions that could lead to a significant difference in sensory food attributes but

can also be used as a tool to identify critical spots for inoculation of bacterial sterilization indicators, which are required for validation of any thermal process. The results were also in agreement with the previous study by Pan et al. [10], where HSI (400–1050 nm) was investigated as a method to identify non-uniform regions post microwave sterilization and the results were comparable to those obtained using an infrared (IR) thermal imaging technique. However, the food matrix itself can induce noise in the results obtained. Hence, a comparison with control samples of each batch would help rule false results.

#### **4. Conclusions**

Hyperspectral image analysis was successfully able to increase the limit of detection to identify colder regions in the processing tray with 12 trays of mashed potato model food after CiMPAS processing, while colorimetry could not identify these colder regions. This also confirms the use of spectral modelling as a tool for cold spot detection. The results showed consistency in detection when samples from three independent processing runs were analyzed. Hence, the detection of the worst critical points via non-destructive HSI indicates the potential to identify critical colder spots, which forms an essential part of ensuring the consistency of microwave-induced sterilization. It also indicates a potential for research on modelling a wide range of food that cannot be formulated as a model system or spiked with a heat-sensitive biomarker.

**Author Contributions:** A.S., M.M.R., and M.A.-S. conceived and designed the experiments. A.S., K.T., and M.A.-S. performed the experiments. A.S. and M.A.-S. analyzed the data. A.S. wrote the paper and A.S, M.A.-S., M.M.R., G.B., J.S., and K.T. significantly edited and reviewed the manuscript. All authors have read and agreed to the published version of the manuscript.

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

**Acknowledgments:** This research was carried out as part of the Food Industry Enabling Technologies program funded by the New Zealand Ministry of Business, Innovation, and Employment (contract MAUX1402).

**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* **Application of Ohmic–Vacuum Combination Heating for the Processing of Senior-Friendly Food (Multiphase Food): Experimental Studies and Numerical Simulation**

**Sung Yong Joe 1, Jun Hwi So 2, Seon Ho Hwang 2, Byoung-Kwan Cho 1,2, Wang-Hee Lee 1,2, Taiyoung Kang 3,\* and Seung Hyun Lee 1,2,\***

	- +82-42-821-6718 (S.H.L.); Fax: +1-808-956-4024 (T.K.); +82-42-823-6246 (S.H.L.)

**Abstract:** The popularity of senior-friendly food has been increasing as the world enters the age of an aging society. It is required that senior-friendly food products are processed with the new concept of processing techniques that do not destroy the nutritional and sensory values. Ohmic heating can be an alternative to conventional heating methods for processing senior-friendly food with retaining excellent taste and quality because of less destruction of nutrients in the food. In this study, the ohmic–vacuum combination heating system was developed to process a multiphase type of senior-friendly food. Changes in physical and electrical properties of senior-friendly model foods were investigated depending on the experimental conditions such as vacuum pressure intensity and vacuum pretreatment time. Numerical simulations based on the experimental conditions were performed using COMSOL multiphysics. The ohmic–vacuum combination heating method with agitation reduced the heating time of the model food, and non-uniform temperature distribution in model food was successfully resolved due to the effect of vacuum and agitation. Furthermore, the difference was found in the hardness of solid particles depending on the vacuum treatment time and intensity after the heating treatment. The ohmic–vacuum combination heating system appeared effective when applying for the senior-friendly foods in multiphase form. The simulation results matched reasonably well with the experimental data, and the data predicted through simulation could save the cost and time of experimentation.

**Keywords:** senior-friendly food; solid–liquid mixture; ohmic heating; vacuum

#### **1. Introduction**

The world's older population has risen over the past few years. Especially, the number of elderly people in Korea and Japan is set to grow at an unprecedented rate because the fertility rate has been drastically decreased [1–3]. According to the portion of the older population, it can be classified into aging society (7% or more), aged society (14% or more), and super-aged society (20% or more) [4]. Japan already entered the super-aged society in 2006, and the older population is expected to exceed 40% of the total population by 2050 [5,6]. Korea is also becoming an aging society. Since a large portion of the elderly population suffers from eating disorders such as masticatory disorders, swallowing disorders, and digestive disorders, they cannot take the essential nutrition from foods [7]. Eating disorders can inevitably lead to an unbalanced diet, which causes malnutrition in the elderly population [8,9]. Malnutrition leads to a decrease in muscle and blood volume, especially causing a rapid decrease in physical function [10,11]. An overall lack of energy

S.H.; Cho, B.-K.; Lee, W.-H.; Kang, T.; Lee, S.H. Application of Ohmic–Vacuum Combination Heating for the Processing of Senior-Friendly Food (Multiphase Food): Experimental Studies and Numerical Simulation. *Foods* **2021**, *10*, 138. https://doi.org/10.3390/ foods10010138

**Citation:** Joe, S.Y.; So, J.H.; Hwang,

Received: 11 December 2020 Accepted: 9 January 2021 Published: 11 January 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/).

can also lead to a vicious cycle of decreased appetite as vitality and activity decrease [12]. To overcome the nutrient intake problem of the elderly people, the interest in senior-friendly foods that can reflect the physical characteristics of elderly consumers and satisfy various tastes is growing [13]. Senior-friendly foods refer to all kinds of food manufactured and processed for the purpose of substituting general meals or keeping the body healthy for the elderly people suffering from metabolic functions. Senior-friendly foods, which have a strong characteristic of patient food, have been developed to satisfy the demand from elderly consumers with an emphasis on maintaining taste and shape. Therefore, these foods should focus on retaining food properties and nutrients during processing so that elderly consumers can easily take enough nutrients.

Conventional cooking methods require long processing time and have difficulty maintaining food quality such as flavor, aroma, texture and appearance [14,15]. In order to process senior-friendly foods, it is necessary to develop new thermal processing technology that enables proper sterilization (or pasteurization) while preserving nutrients in the food through minimum thermal treatment [16]. Among food thermal treatment methods, ohmic heating (OH) has been widely used for sterilization and pasteurization of heat sensitive food products. Resistance heat can be generated inside the food product during OH by passing alternative current through the food product with close contact to electrodes. Compared with other thermal treatment methods (i.e., microwave heating, infrared heating), the OH of single-phase food (liquid or solid phase food) can provide thermal uniformity enhancement, high heating rate and energy conversion efficiency [17,18]. For high-temperature short-time (HTST) sterilization of multiphase foods containing solid particles, OH allows large solid particles to be simultaneously treated with liquid phase food, which was not possible using conventional heat exchangers [19]. However, the OH heating rate significantly depends on electrical conductivities of foods. Since non-uniform heating of multiphase during OH can occur due to the differences of electrical conductivities between the solid phase and liquid phase [20], it can lead to over or under-treatment of multiphase food, consequently resulting in quality deterioration.

The ohmic–vacuum (OH–VC) combination heating method has not been well investigated [21]. During the OH of liquid phase food, the temperature of liquid continues to rise until it reaches its boiling point [22]. To keep its temperature constant, the power source must be controlled or some cooling medium must be applied during the heating process [23]. However, by combining OH with vacuum, the boiling temperature of liquid can be lowered [24]. When the liquid food reaches the boiling point, its temperature remains constant as long as the vacuum pressure does not change [25]. The uniform temperature distribution of multiphase food during OH–VC combination heating can be obtained by controlling the boiling point of liquid phase food using a vacuum [16]. A vacuum combination heating method can improve energy efficiency and the texture of the processed product [26]. Controlling the exact temperature distribution of multiphase food plays an important role in the OH process; however, it is complicated because of the heat transfer between solid and liquid phases [27,28]. In order to determine the temperature distribution of multiphase food in OH, a number of parameters should be experimentally evaluated and complex numerical methods should be applied [29,30]. Magnetic resonance imaging (MRI) temperature mapping was used to observe the temperature distribution of multiphase food during OH [29,31]. Even though temperature distribution could be observed in real time, the cost of MRI was relatively high and the additional space was required [27]. The use of computational simulation can accurately predict temperature distribution of multiphase food under OH [32]. Understanding the behavior of the OH process is essential to demonstrate the correct reliability of the heating system and to the safety of the process [33]. Numerical modeling provides an insight into the heating behavior of OH [34]. Temporal and spatial temperature distributions of multiphase food during OH can be provided from the reliable numerical model [35].

This study was conducted to develop an OH–VC combination heating system for processing senior-friendly food products consisting of solid particles and liquid, and to determine temperature uniformity of multiphase foods under OH–VC combination heating depending on the presence of agitation. In addition, computational fluid dynamic (CFD) models were developed to validate the multiphase food OH process with agitation. As far as can be determined from the accessible literature, the CFD model for OH of multiphase food with agitation was not attempted and developed.

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

#### *2.1. Preparation of Model Food*

The solution model food (base solution) used in this study was prepared by imitating commercialized senior-friendly foods (2 types of soft diets (A, B), 3 types of liquid foods (C, D, E), and 1 pudding food (F)). The constituents, viscosity, and electrical conductivity of six different types of commercialized senior-friendly foods purchased from the local silver food market were analyzed prior to preparing the base solution. The samples were stored in 4 ◦C refrigeration condition. Samples were taken out at room temperature for 1 h before the experiment, and the viscosity and electrical conductivity of the samples were measured at room temperature (around 21 to 23 ◦C). The viscosity of the samples was measured by using a viscometer (CL-L2, CAS, Inc., Seoul, Korea).

The electrical conductivities of the samples were determined using a custom-made Teflon OH test cell (0.01 m in inner diameter and 0.1 m in height) consisting of food grade stainless steel (SUS 316) electrodes (0.01 m in diameter) that were placed at both ends of OH cell through a pair of spacers. During the electrical conductivity measurement, the sample temperature was measured using a thermocouple (K-type KK-K-30, Omega Engineering Inc., Stamford, CT, USA), which was inserted at the center of the sample through a small hole on the surface of the OH test cell. The electrical conductivities of samples were calculated by the following equation;

$$
\sigma = \frac{I}{V} \cdot \frac{L}{A} \tag{1}
$$

where *σ* is the electrical conductivity (S/m), *I* is the current (A), *V* is the applied voltage (V), *L* is the distance between the electrodes (m), and *A* is the contact area between electrode and sample (m2).

The viscosity of the samples ranged from 20 to 160 cPs above 80% torque. The measured electrical conductivities of the samples as a function of temperature are summarized in Table 1. Based on the measured viscosity and electrical conductivity of samples, the base solution was made by using whole milk powder (Seoul milk, Seoul, Korea) and black bean soup (Chung's Food Co., Ltd., Cheongju, Korea). The mixing ratio of the base solution was 1 (whole milk powder) to 6 (black bean soup). The viscosity of the base solution was 100 cPs at over 80% torque, and the electrical conductivity of the base solution was linearly increased with increasing temperature (Table 1).

**Table 1.** Electrical conductivities of commercialized senior-friendly foods and base solution at different temperatures.


Types of commercialized senior-friendly foods: soft diet (A and B), liquid food (C, D, and E), and pudding (F).

Pork sirloin purchased from a local butcher shop was used as solid particles, and were cut into cubes (2 × 2 × 2 cm). The solid cubes were added into the base solution to produce model senior-friendly food (mixture food). The solid fraction in the total volume of mixture food was 10/100 g. The amount of model food for each experiment was 800 g.

#### *2.2. Ohmic–Vacuum Combination Heating System*

The ohmic–vacuum combination heating system is shown in Figure 1. The system consisted of a vacuum chamber and pump, ohmic chamber, a pair of curved rectangular electrodes, overhead stirrer, and ohmic power supply. The vacuum chamber could maintain a vacuum gauge pressure of up to 0.1 bar. The cylindrical ohmic chamber (outer diameter of 12.7 cm, inner diameter of 10.1 cm, and height of 18 cm) was made of Ultem (PEI) to prevent distortion of the chamber by vacuum pressure and heat. Curved rectangular electrodes with circumference of 9.8 cm, thickness of 0.2 cm, and height of 10 cm were fabricated by cutting sanitary stainless steel (SUS 304 pipe), and installed parallel to both sides of the chamber. Ohmic power supply was designed and custom built by using the IGBT module (404GB12E4s, SKYPER 42R driver and board, Semikron, Inc., Hudson, NH, USA). The power supply was able to generate pulsed alternating current with a frequency range between 1 Hz and 20 kHz, on/off duty cycle from 0.2 to 0.8, and maximum current of 30 A at 380 Vrms.

The ohmic chamber was placed at the bottom center of the vacuum chamber. To measure temperature during ohmic–vacuum combination heating, 2 k-type thermocouples were installed at the middle and bottom part of the ohmic chamber through a small hole of the vacuum chamber lid. The curved rectangular electrodes and ohmic power supply were connected similarly to thermocouple installation. An anchor blade-stirring bar was inserted at the center of the ohmic chamber through a vacuum stirring seal installed at the center of the vacuum chamber lid. The stirring bar was connected with an overhead stirrer (ms5020D, Misung Scientific, Seoul, Korea). The vacuum stirring seal was effective in preventing air inflow during ohmic–vacuum combination heating. All data points of temperature, voltage, and current were monitored and recorded by using a differential probe (PR-60, BK Precision, Yorba Linda, CA, USA), wideband current monitor (169820, Pearson Electronics, Palo Alto, CA, USA), oscilloscope (DPO 4034, Tektronix, Beaverton, OR, USA), and data acquisition unit (DAQ) (39704A, Agilent, Palo Alto, CA, USA).

**Figure 1.** Ohmic–vacuum combination heating system: (**a**) schematic diagram of system, (**b**) ohmic heating power supply configuration.

#### *2.3. Measurement of Hardness of Solid Particles*

Texture Profile Analysis (TPA) was conducted using a texture analyzer (TA.XT Plus, Texture Technologies, Scarsdale, NY, USA) to evaluate the change in hardness of solid particles in the model food after OH–VC combination heating depending on vacuum pressure and vacuum pretreatment time. TPA mimics the chewing effect of putting food in

the mouth and chewing it with teeth. It is a method of analyzing the physical properties of food by applying compression force twice consecutively (Figure 2). The hardness of solid particles was measured using a 30 mm diameter circular probe. Solid particles and liquid particles were immediately separated after the OH experiment, and then the hardness of solid particles was analyzed.

**Figure 2.** Diagram of Texture Profile Analysis (TPA) hardness measurement.

#### *2.4. Experimental Design*

The boiling point of the base solution during OH was investigated depending on the change in vacuum pressure strength and the presence of agitation. The vacuum pressure was controlled within the range of 1.01325 bar (1 atm) to 0.1 bar. The OH experiment was immediately stopped when the liquid boiled at different vacuum intensity levels by applying a voltage of 70 V, duty cycle of 50%, and frequency of 15 kHz. Thermocouples were installed in the middle and bottom of the heating chamber to monitor temperature values of the base solution.

The temperature uniformity of the model food containing pork cubes and base solution during OH was evaluated depending on the presence of agitation. In addition, the effect of vacuum pretreatment on change in temperature of the model food was investigated. The model food was pretreated for 5, 10, and 10 min at different vacuum intensity levels (vacuum gauge pressure: 0.8, 0.5, 0.2 bar, and atmospheric pressure) under agitation before applying voltage to the heat model food. Then, 100 V at frequency of 15 kHz with duty cycle of 50% was applied for the OH of the model food, while maintaining the same vacuum pressure intensity as the vacuum pretreatment condition. In this study, this heating method was named as the ohmic and vacuum ("OH–VC") combination heating with agitation. The temperature of the base solution was measured through the thermocouples (K-type KK-K-30, Omega Engineering Inc., Stamford, CT, USA) installed in the middle and bottom of the heating chamber. The temperature values of pork particles were measured by inserting thermocouples to the core of the particles as soon as the heating process was finished. The OH–VC combination heating experiment of the model food was stopped when the middle or bottom temperature of the base solution reached 90 ◦C. As soon as OH experiments were completed, solid particles and base solution were immediately separated from the model food to evaluate the effect of vacuum pretreatment conditions on change in hardness of solid particles.

Furthermore, the effect of vacuum pretreatment on change in the electrical conductivities of the model food was investigated. After the model food was pretreated by different vacuum pressure intensity levels for different times, solid particles and base solutions were separated and then both electrical conductivities were measured by using the aforementioned method in Section 2.1. The overall experimental protocol is illustrated in Figure 3.

**Figure 3.** A schematic of overall experimental protocol.

#### *2.5. Mathematical Modeling*

The mathematical modeling was developed to understand and predict the heat transfer and heat distribution in multiphase food during ohmic heating combined with agitation by using COMSOL Multiphysics software (COMSOL 5.5, COMSOL, Inc., Palo Alto, CA, USA) including AC/DC, Heat Transfer, and CFD modules.

#### 2.5.1. Governing Equation for Electromagnetic Heat Generation

The electric field distribution in the ohmic heater was determined by using the Laplace equation [18];

$$\nabla \cdot (\sigma \nabla V) = 0 \tag{2}$$

where *V* is the voltage (V), ∇ is the gradient, *σ* is electrical conductivity (S/m).

Since electrical conductivities of foods are a function of temperature, temperature and electrical conductivity can be expressed in a linear relationship [19];

$$
\sigma\_i(T) = \sigma\_0(1 + mT) \tag{3}
$$

where *σ*<sup>0</sup> is the reference value (S/m), *m* is the temperature coefficient, *T* is the temperature (K).

The heat source of the ohmic heating simulation was heat generation inside the food by electric current. Internal energy generation during the OH of multiphase food by conduction can be determined by the following equation [31];

$$
\rho \mathbb{C}\_p \frac{\partial T}{\partial t} = \nabla k \nabla T + Q\_{\mathbb{S}^{\text{cn}}} \tag{4}
$$

where *ρ* is the sample density (kg/m3), *Cp* is the specific heat (kJ/kgK), *T* is the temperature within the sample (K), *t* is the heating time (s), *k* is the thermal conductivity (W/m ◦C).

The heat generation (*Qgen*) during ohmic heating is proportional to the electrical conductivity of food and the square of the voltage gradient [18];

$$Q\_{\mathcal{S}^{\rm ren}} = \left. \sigma \right| \nabla V \Big|^2 \tag{5}$$

#### 2.5.2. Governing Equation for Turbulent Flow

The flow of the model food during OH was caused by agitation. The flow by agitation complicates the determination of the temperature distribution in multiphase food. Therefore, turbulent flow analysis was added to identify the exact temperature distribution inside the food. The flow of model food during OH was determined by the *k* − *ε* turbulence equation and turbulent velocity (*uT*) is significantly affected by turbulent kinematic energy (*k*) and the turbulent dissipation rate (*ε*) [36];

$$
\mu\_T = \rho \mathbb{C}\_{\mu} \frac{k^2}{\varepsilon} \tag{6}
$$

where *uT* is the turbulent velocity (m/s), *ρ* is the density (kg/m3), *C<sup>μ</sup>* is the constant model parameter, *k* is the turbulent kinetic energy (m2/s2), and *ε* is the turbulent dissipation rate (m2/s3).

Turbulent kinetic energy (*k*) can be calculated by following the transportation equation [36];

$$
\rho \frac{\partial k}{\partial t} + \rho u \cdot \nabla k = \nabla \left[ \left( \mu + \frac{\mu\_T}{\sigma\_k} \right) \nabla k \right] + P\_k - \rho \varepsilon \tag{7}
$$

$$P\_k = \mu\_T \left[ \nabla u : \left( \nabla u + \left( \nabla u \right)^T \right) - \frac{2}{3} (\nabla \cdot u)^2 \right] - \frac{2}{3} \rho k \nabla \cdot u \tag{8}$$

where *u* is the mean velocity (m/s), *μ* is the dynamic viscosity (Pa·s), *μ<sup>k</sup>* and *σ<sup>k</sup>* are the constant model parameters, and *Pk* is the production term.

An additional transportation equation is necessary for the calculation of and turbulent dissipation rate (*ε*) [36];

$$
\rho \frac{\partial \varepsilon}{\partial t} + \rho u \cdot \nabla \varepsilon = \nabla \cdot \left[ \left( \mu + \frac{\mu\_T}{\sigma\_\varepsilon} \right) \nabla \varepsilon \right] + \mathbb{C}\_{\varepsilon 1} \frac{\varepsilon}{k} P\_k - \mathbb{C}\_{\varepsilon 2} \rho \frac{\varepsilon^2}{k} \tag{9}
$$

where *Cε*<sup>1</sup> and *Cε*<sup>2</sup> are the constant model parameters.

#### 2.5.3. Mathematical Modeling Setup

Mathematical modeling proceeded in the following order: (1) creation of geometry for modeling, (2) initial and boundary condition assignment, (3) mesh generation and optimization, (4) solver selection, (5) tolerance and time step setting, and (6) built-in convergence solution.

The boundary conditions of the heat transfer equation assumed that all samples were thermally insulated, and the initial temperature values of the entire samples (solid particles and base solution) were set to 303.15 K. The thermal properties of model food were calculated based on experimental data. The specific heat, density, and thermal conductivity of the base solution and solid (pork) particles were 3.9 and 2.71 kJ/kg·K, 1030 and

1099.7 kg/m3, and 0.5948 and 0.21 W/m·K, respectively. The electrical conductivity was increased with an increase in temperature. The electrical conductivities of the base solution and solid particles as a function of temperature were set to 0.01035 × (*T* − 273.15) + 0.61915 S/m and 0.0171 × (*T* − 273.15) + 0.5813 S/m, respectively. In order to improve the mesh quality, the computational domain was discretized with tetra meshes. The mesh geometry consisted of 11,515 tetrahedrons, 1328 boundary elements, and 392 edge elements. The direct linear system solver (PARDISO) was used to increase the convergence rate. The relative tolerance and absolute tolerance used in PARDISO was 0.05.

The geometry used in the simulation is shown in Figure 4. A pair of electrodes were placed on both sides of a cylinder with a diameter of 101 mm and a height of 110 mm. The shaded parts in blue (Figure 4b) are the electrodes and the applied voltage for simulation was 100 V. A stir head shape was employed to mimic the stirring activity and the applied rotation speed was 60 rpm. A spherical particle with a diameter of 10 mm was added to evaluate the temperature difference between solid and liquid. The particles were located at six points to determine the temperature difference depending on the location.

**Figure 4.** Geometry used in the ohmic heating simulation: (**a**) grid mesh geometry, (**b**) electrode layout, (**c**) top view, and (**d**) front view.

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

*3.1. The Effect of Vacuum Pretreatment on Change in Electrical Conductivities of Pork Particle and Base Solution*

After pork particles and the base solution were pretreated at different vacuum pressure intensity levels for different times under agitation, their electrical conductivities were measured (Table 2). The electrical conductivities of solid particles and base solution without vacuum pretreatment were compared to those with vacuum treatment. Regardless of vacuum pretreatment, the electrical conductivities of both samples were linearly increased with an increase in temperature. The vacuum pretreatment caused a slight increase

in the electrical conductivities of both samples. The electrical conductivities tended to decrease with increasing vacuum pretreatment time; however, the difference resulting from pretreatment times was not significant. In vacuum pretreatment processing of multiphase food, the enhanced osmotic pressure by vacuum pretreatment caused the change in the electrical conductivity of solid particles by increasing the solute or water absorption of solid particles from the solution [37]. The vacuum pretreatment used in this study led to the change in the electrical conductivities of both samples by affecting the mutual movement of electrolytes or moisture between the solid and liquid.


**Table 2.** The measured electrical conductivities of pork particle and base solution after vacuum pretreatment.

#### *3.2. Change of Boiling Point of Base Solution Depending on Vacuum Intensity*

The base solution boiled at 97 ◦C under atmospheric pressure, and the boiling point decreased by approximately 3 ◦C as the vacuum gauge pressure decreased until 0.4 bar at a decrement interval of 0.1 bar, as shown in Figure 5. The boiling point rapidly decreased below 0.3 bar, and the boiling point at 0.1 bar was 45 ◦C, which was more than 50 ◦C below the boiling point under atmospheric pressure. It was clearly observed that the boiling point was significantly dependent on vacuum intensity.

**Figure 5.** Change of boiling point of base solution under different vacuum intensities.

#### *3.3. The Effect of Agitation on Temperature Uniformity of Base Solution and Model Food*

Figure 6 shows the temperature difference of the base solution at different locations in the heating chamber during OH depending on the presence of agitation. The initial temperature of the base solution was approximately 30 ◦C regardless of location. When the agitation was not applied to OH, the temperature difference of the base solution at different locations significantly increased after 40 s (Figure 6a). The temperature at the middle of the heating chamber showed a rapid increase rate than the bottom. Although OH is known as the effective food thermal treatment to achieve temperature uniformity of single phase food, temperature non-uniformity was found in this study.

As shown in Figure 6b, the temperature uniformity of the base solution regardless of location was achieved by agitation. The time required for the temperature of the middle or bottom to reach 90 ◦C was 144 s. The temperature difference between the middle and the bottom of the heating chamber was less than 1.4 ◦C during the entire heating process, and showed almost the same increase rate of temperature at all locations in the heating chamber. The forced convection by agitation was effective at achieving thermal uniformity of the base solution during OH.

**Figure 6.** Temperature difference between middle and bottom of base solution: (**a**) without agitation, (**b**) with agitation.

When the model food was treated by OH without agitation at atmospheric pressure, the temperatures of the base solution and solid particles were 90 ± 1.13 ◦C and

73.49 ± 0.38 ◦C, resulting in significant temperature difference. However, in the case of OH with agitation, the temperature difference between solid particles (90.2 ± 0.93 ◦C) and the base solution (87.29 ± 0.06 ◦C) was less than 3 ◦C.

When multiphase food was treated by OH, the temperature difference between solid particles and liquid could be caused by limited convection. An excessive heat treatment is required to increase the temperature of solid particles; however, it will cause over-heating that deteriorates food quality [38]. In this study, it was found that forced convection by agitation was effective in resolving non-uniform temperature distribution in multiphase food.

The temperature values of the solid particles and base solution under OH–VC combination heating with agitation are summarized in Table 3. The temperature of solid particles was not significantly affected by the vacuum pretreatment time at constant vacuum pressure; however, the effect of vacuum pressure intensity was dominant in the temperature change of solid particles. The boiling point of the model food was lowered by the effect of OH–VC combination with agitation. The solid particles were only heated to the boiling point temperature of the base solution, and the final heating temperature of the base solution at vacuum gauge pressures of 0.8, 0.5, and 0.2 bar regardless of vacuum pretreatment time was 88.99 ± 0.42, 77.89 ± 0.69, and 57.45 ± 0.26 ◦C, respectively. The temperature difference between solid particles and the base solution under all conditions of OH–VC combination with agitation was within 3 ◦C. Since vacuum pretreatment expelled the air inside the solid and made the electrical conductivity of the solution and solid almost similar, it was possible to minimize the temperature difference between solid particles and base solution [39]. The OH–VC combination heating with agitation was effective at improving the temperature uniformity of multiphase food and preventing excessive thermal treatment.

**Table 3.** Temperature values of model food under different ohmic heating–vacuum (OH–VC) combination heating with agitation.


(A) is the base solution, (B) is solid particles.

#### *3.4. Variation of Particle Hardness*

The average hardness of the solid particles treated by OH with/without agitation was 215.7 ± 6.4 and 119.8 ± 10.1 N/m2, respectively. Since the solid particles in the model food were under-processed by OH without agitation, a relatively low hardness value was measured.

Figure 7 indicates the hardness change of solid particles in the model food treated by different OH–VC combination heating conditions with agitation. An increase in vacuum intensity had a great effect on the change in hardness of solid particles. This result is consistent with the previous studies showing that the firmness of papaya treated by vacuum the firmness was reduced compared to the atmospheric pressure treatment [40–44]. However, vacuum pretreatment time was not significant to change the hardness of solid particles. The lowest hardness of solid particles was observed at a vacuum gauge pressure of 0.2 bar and the hardness range was between 170 and 180 N/m<sup>2</sup> under all pretreatment time conditions. Older people suffering from eating disorders prefer soft foods over tough or hard foods. Therefore, the texture of solid particles is a major factor to be considered in the processing of senior-friendly food consisting of solid and liquid phase food. The OH–

VC combination heating made the texture of solids softer than individual OH treatment. In addition, this combination heating was suitable for the processing of the multiphase form of senior-friendly food.

**Figure 7.** Change in hardness of solid particles under different OH–VC combination heating conditions with agitation: vacuum pretreatment time for (**a**) 5 min, (**b**) 10 min, and (**c**) 15 min at different vacuum intensity levels.

#### *3.5. Simulation Verification*

#### 3.5.1. The Simulated Electric Field Strength Distribution in Ohmic Chamber

The electric field distribution inside the OH chamber filled with model food was simulated using an AC/DC module in COMSOL Multiphysics (COMSOL 5.5, COMSOL, Inc., Palo Alto, CA, USA), as shown in Figure 8. The applied voltage to the simulation was 100 V. The electric field overshoot (approximately 3.4 kV/m) was determined at both edges of the curved rectangular electrodes. The current density was higher in the areas of both edges than in other parts. The electric field strength range in the center of the OH chamber was estimated between 1100 and 1200 V/m. In addition, the relatively low electric field strength range (740 to 870 V/m) was observed at the area between the edges of the electrodes. The electric field strength of the solution surrounding the solid particle was observed to be slightly higher than that of other parts, which seems to be a phenomenon that occurs when an electrical current passes through the material composed of different phases having different physical and electrical properties. Moreover, this phenomenon could result in the electric field interruption and non-uniform temperature distribution in the model food.

**Figure 8.** Simulated (**a**) electric potential and (**b**) electric field distributions.

3.5.2. Temperature Distribution of Model Food under OH Depending on Presence of Agitation

Figure 9 shows the simulated heating pattern of model food under OH without agitation. The black arrows represent the total heat flux including conduction and convective heat flow. The rapid increase of temperature near both edges of the electrodes significantly affected the temperature distribution of model food inside the chamber. As the electric overshoot near both edges of the electrodes was estimated in the simulation of electric field distribution, it was predicted that the area near the edges of the electrodes was heated faster than other areas. In the top view of the simulation, the total heat flux was diffused from both edges of the electrodes. However, as shown in the front view, the total heat flux spreads outward from the center. Therefore, when the agitation was not applied to the OH of the model food, the generated internal heat from OH could not be uniformly distributed in the model food because of the limited natural convection. The temperature difference between the center and the outside was about 20 ◦C due to the non-uniform heating pattern. Even though the electrical conductivity of the solid particles was lower than that of the base solution, the heating rate of the solid particles was sharper than that of the base solution in the simulation. Since the locations of solid particles were close to the center of the chamber, the generated heat from OH affected the increase in temperature of the solid particles. The temperature difference between solid particles and base solution increased as heating time increased. The temperature range of the pork particles obtained from the experimental data of OH without agitation was from 60 to 81 ◦C. The simulation results showed that the temperature of the particle located in the center was about 83 ◦C and the outside was about 64 ◦C, which was very similar to the experimental data. The temperature difference between solid particles and base solution varied depending on the location of the solid particles and the difference range was from a minimum of 10 ◦C to a maximum of 30 ◦C.

Figure 10 shows the simulation results of the heating pattern of the model food under OH with agitation. The direction of the total heat flux was the same as the agitation direction. The exacerbated non-uniform temperature distribution between solid particles and base solution was minimized by the effect of forced convection. Regardless of the solid particle locations, solid particles were equally heated and had a similar heating rate. The temperature difference between solid particles and base solution was around 5 ◦C at all heating times and locations; however, the heating rates of solid particles and base solution showed a similar tendency. The simulated temperature values for solid particles and base solution were in good agreement with the experimental data and the maximum prediction error was about 3 ◦C. In this study, heating patterns of model food under OH with agitation could be effectively predicted through the simulation.

**Figure 9.** Simulated heating pattern of model food under OH without agitation.

**Figure 10.** Simulated heating pattern of model food under OH with agitation.

#### **4. Conclusions**

The effects of vacuum and agitation on thermal uniformity of the model food under OH was evaluated in this study. By combining vacuum and agitation in OH of multiphase food, the boiling point of the base solution was lowered and thermal uniformity of model food was improved with softening of solid particles. In addition, the excessive heat treatment inside multiphase food was prevented. The simulation models for the OH of the model food with/without agitation were in good agreement with experimental data. In the simulation for the OH of the model food without agitation, solid particles were heated more rapidly than the base solution. The temperature difference between solid particles, depending on the locations in OH chamber (center and outside), increased with an increase in heating time. The simulation for the OH of the model food with agitation showed thermal uniformity between solid particles and base solution with a maximum difference within 5 ◦C. The developed OH–VC combination heating with agitation has great potential to process senior-friendly foods with improvement of the texture of solid phase food and enhanced thermal uniformity.

**Author Contributions:** Conceptualization and editing, S.H.L., T.K.; experimental work, simulation, and writing, S.Y.J.; experimental work, J.H.S., S.H.H.; simulation work, W.-H.L.; editing, B.-K.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Science, ICT and Future Planning (NRF-2018R1C1B6006014).

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data presented in this study are available in the article.

**Acknowledgments:** This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF).

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

#### **References**

