**Contents**


Reprinted from: *Applied Sciences* **2020**, *10*, 6709, doi:10.3390/app10196709 . . . . . . . . . . . . . . **131**


### *Editorial* **Application of Wood Composites**

**L'uboš Krišt'ák \* and Roman Réh**

Faculty of Wood Sciences and Technology, Technical University in Zvolen, T.G. Masaryka 24, 96001 Zvolen, Slovakia; roman.reh@tuzvo.sk

**\*** Correspondence: kristak@tuzvo.sk

Wood composites are the key material for a number of structural and non-structural applications for interior and exterior purposes, such as furniture, construction, floorings, windows and doors, etc. They can be successfully produced with predetermined specific properties matching the required end uses. Wood composites ranging from fiberboard to laminated beams need to be better known, and more attention must be paid to their research. Laboratories worldwide do innovative research, and new challenges, approaches, and ideas are continuously increasing, allowing us to mirror an exciting and interesting research future [1–4].

The time when this Special Issue had been continuously compiled was mostly a stressful period for all of us, marked with the widespread outbreak of the COVID-19 pandemic, but we hope all *Applied Sciences* readers are healthy and well. We do see a glimmer of hope to return to normalcy on the horizon.

This Special Issue "Application of Wood Composites" addressed various aspects of these important wood materials' use, e.g., mechanical processing of wood composites including their cutting, milling, or sanding incorporating the current analysis of wood dust or grain size measurements and composition of particles [5–9], scientific views on the influence of various adhesives in the creation process of wood composites and the analysis of their behavior in contact with various wood elements under different conditions [10–14], the analysis of input raw materials forming wood composites, including various wood species, but also non-wood lignocellulosic raw materials and, last but not least, the analysis of bark, which in the recent years has become an important and promising raw material involved in the construction of wood composites [15–19]; the study of the development of the sliding table saw also suitably complements this Special Issue [20].

If we take a closer look at the main topic of this publication, it is clear that wood composite materials are engineered and produced with tailored physical and mechanical properties appropriate for a wide variety of applications, known or not discovered yet. Additionally, indeed, the utilization of wood composites in various areas has increased recently due to their outstanding properties, allowing them to successfully and sustainably replace solid wood and other conventional materials. We have tried to respond to this newly created situation with this publication.

One of the publications was aimed at providing the reader with new information on the recent practices in laser cutting of wood and wood composites, and determining the optimal set of cutting parameters by a method of a low-power CO<sup>2</sup> laser in particular. Three factors were investigated, namely the effect of the laser power, cutting speed, and number of annual rings [7]. Other new insights are emerging in another type of wood processing: Sanding. The research results indicate that the factors determining sanding efficiency are the type of wood, and, secondly, the grit size of sanding belts. Maximum sanding efficiency for the softwood surface of wood composites ranged from 1 to 2 min, while for the hardwood species composites surface, it ranged from 2 to 4.5 min at the start of sanding and then decreased [5]. The accompanying phenomenon of sanding is wood dust that poses a serious threat to the health of workers and employees as well as a significant fire and explosion hazard; it accelerates the wear of machines, worsens the

**Citation:** Krišt'ák, L'.; Réh, R. Application of Wood Composites. *Appl. Sci.* **2021**, *11*, 3479. https:// doi.org/10.3390/app11083479

Received: 3 April 2021 Accepted: 12 April 2021 Published: 13 April 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/).

quality of processing, and requires large financial outlays for its removal [21]. Therefore, the aim was to investigate the extent to which the grit size of sandpaper influences the size of the wood dust particles and the proportion of the finest particles which may constitute the respirable fraction [10]. The grain size measurements of wood dust samples from selected tropical wood species were investigated, as well as the conditions under which wood composites are processed, e.g., impact of thermal modification of wood composites surface saturated by steam and its influence on the particle size distribution of the sawing and milling process [8,9].

It is remarkable how many opportunities will arise when using modern methods and high-quality instruments and equipment for the analysis of wood composites. With a compact Time-of-Flight Secondary Ion Mass analyzer, integrated in a multifunctional focused-ion beam scanning-electron-microscope, it was possible to show that the Ga<sup>+</sup> ion source could be detected and visualized in 3D ion molecular clusters specific to polymeric 4,4′ -diphenyl methane diisocyanate (pMDI) adhesive and wood [10]. The bonding of wood with assembly adhesives is crucial for manufacturing wood composites. Various adhesives in the context of their application to various types of wood must be analyzed for the formation of quality wood composites, e.g., polyvinyl acetate (PVAc), lignin-based formaldehyde-free adhesives (lignosulfonates), or new and improved adhesive mixtures of urea-formaldehyde (UF) resin, e.g., with soy flour [11,12]. It has been shown that the properties of wood composites can be improved by using these new and less used combinations of adhesives. The fabricated wood composites achieved close-to-zero formaldehyde content of 1.1 mg/100 g, i.e., the super E0 emission grade (≤1.5 mg/100 g), which allowed their classification as eco-friendly, low-emission wood-based composites [13,14,22,23].

Without a detailed analysis of input raw materials, it is not possible to form meaningful wood composites. The analysis of input wood raw materials must be carried out no matter what kind of wood composites are produced. The surface roughness constitutes one of the most critical properties of wood veneers for their extended utilization, affecting the bonding ability of the veneers with one another in the manufacturing of wood composites, the finishing, coating and preservation processes, and the appearance and texture of the material surface. The surface roughness was examined by applying a stylus tracing method on typical wood structure areas of each wood species, as well as around the areas of wood defects (knots, decay, annual rings irregularities, etc.), to compare them and assess the impact of the defects on the surface quality of veneers [18]. The production possibilities of oriented strand boards (OSB) in the laboratory from a mixture of softwood species and hardwood species were tested [16]. In our times, it is becoming increasingly important to use secondary products in wood processing or to use less known and less used lignocellulosic materials as sustainable alternatives of wood [24–26], such as the bark of various woody plants which has not been fully utilized yet, or utilization of walnut and hazelnut shells which are agricultural by-products, available in high quantities during the harvest season. Very interesting and promising research results were achieved [17,19]. As invasive alien species are one of the main causes of the loss of biodiversity, and thus of changes in ecosystem services, it is important to find the best possible solution to their usability. Research showed that it will be possible to deal with such a problem as well and the production of wood plastic composites is a viable solution [15].

We would like to thank our Section Managing Editor Dr. Kyle Ke for his professional attitude and assistance with publishing.

It is good that such a book publication was created. The topic "Application of Wood Composites" is still relevant, new possibilities for application of wood composite materials are emerging, and therefore it is understandable that MDPI has already opened access to a new Special Issue "Application of Wood Composites II" within the journal Applied Sciences with the possibility of publishing new high-quality original research articles and reviews on the latest advancements in wood composites materials and their applications.

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

**Acknowledgments:** This publication was supported by the Slovak Research and Development Agency under contract No. APVV-18-0378, APVV-19-0269 and VEGA1/0717/19.

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

#### **References**


### *Article* **Efficiency of Machine Sanding of Wood**

**Maciej Sydor 1,\* , Radosław Mirski <sup>2</sup> , Kinga Stuper-Szablewska <sup>3</sup> and Tomasz Rogozi ´nski <sup>4</sup>**


**Abstract:** We hypothesized that the type of wood, in combination with the grit size of sandpapers, would affect sanding efficiency. Fixed factors were used in the experiment (a belt sander with pressure *p* = 3828 Pa, and a belt speed of *v*s = 14.5 m/s) as well as variable factors (three sand belts (P60, P120, P180), six hardwood species (beech, oak, ash, hornbeam, alder, walnut) and three softwood species (pine, spruce, larch)). The masses of the test samples were measured until they were completely sanded. The sanding efficiency of hardwood species is less variable than for softwood species. Maximum sanding efficiency for the softwood ranged from 1 to 2 min, while for the hardwood species, it ranged from 2 to 4.5 min at the start of sanding and then decreased. The average time for complete sanding of the softwood samples was: 87 s (P60), 150 s (P120), and 188 s (P180). For hardwood, these times were 2.4, 1.5, and 1.8 times longer. The results indicate that the factors determining sanding efficiency are the type of wood, and, secondly, the grit size of sanding belts. In the first phase of blunting with the sanding belts, the sanding processes of hardwood and softwood are significantly different. In the second phase of blunting, sanding belts with higher grit numbers (P120 and P180) behaved similarly while sanding hardwood and softwood.

**Keywords:** softwood; hardwood; sanding; belt sander; sandpaper; abrasion; beech; oak; ash; hornbeam; alder; walnut; pine; spruce; larch

#### **1. Introduction**

Sanding is widely used in the furniture industry. The objectives of sanding may be to achieve the required surface smoothness to be painted, to achieve the required roughness necessary for gluing on the surface, and effective and controlled material removal to obtain the desired shape or dimensional accuracy of the workpiece. When planning a technological sanding process, several key aspects should be considered. Providing appropriate working conditions by reducing the exposure of workers to respirable wood dust in the air is the first important aspect [1–6]. Another group of problems are the economic issues of the used technology; in other words, obtaining high quantitative efficiency and productivity and the expected surface quality and/or accuracy of the shape for the workpieces. These two groups of problems are solved by properly selecting the production equipment, parameters of the abrasive tools and parameters of the sanding process [7,8].

Issues resulting from the specific effect of abrasive grains on wood have been studied both from the point of view of machine tool design [9], abrasive tools (type of sandpaper and its grit size) and the technological parameters used (in particular, the contact pressure and speed of the abrasive belt, the size of the surface to be sanded and the orientation of the wood fibers during sanding [10–16]). The influence of the properties of various species of wood on the effects of sanding were also studied [17–21]. One of the most

**Citation:** Sydor, M.; Mirski, R.; Stuper-Szablewska, K.; Rogozi ´nski, T. Efficiency of Machine Sanding of Wood. *Appl. Sci.* **2021**, *11*, 2860. https://doi.org/10.3390/ app11062860

Academic Editor: Roman Réh

Received: 23 February 2021 Accepted: 19 March 2021 Published: 23 March 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/).

important measures of the efficiency of the sanding process is the mass of material sanded per unit of time. Sanding efficiency decreases during the process due to the blunting of the abrasive belt. Ockajova [21], analyzing the literature, identified three phases of sanding belt blunting: initial sharpness, work sharpness and sanding belt blunting. During the initial phase, there is a very large reduction in efficiency during sanding. The limit between the first and second phase is a stabilization of this reduction in efficiency, at a level of about 45–50% in relation to the initial sanding efficiency. In the second phase, where the wear of the abrasive grit dominates, a further, somewhat slower reduction in sanding efficiency is observed (by about 10–20% in relation to the initial efficiency). Characteristic for the third phase is a rapid decrease in sanding efficiency.

Wieloch and Siklienka [22] investigated the effect of long time sanding on the variation in efficiency for beech wood. The analyzed process lasted 480 min. P40, P80, and P120 abrasive belts were used, and different contact pressures were applied: *p* = 1.0, 1.5, 1.85 and 2.0 N/cm<sup>2</sup> (10,000, 15,000, 18,500 and 20,000 Pa). At a pressure of 10,000 Pa, a rectilinear decrease in sanding efficiency was observed. At a pressure of 18,500 Pa, however, the decrease was "bi-rectilinear": first, the efficiency decreased intensively and, after a certain time, the decrease in sanding efficiency slowed markedly. At higher contact pressures, the sanding performance decreased more rapidly. In a comparative study on sanding oak and beech wood, Ockajova et al. [21] found that the contact pressure that ensures long-term operation of the abrasive belt depends on the direction of sanding and the wood species (the pressure on beech wood may be higher). The species of wood in these studies had a greater influence on sanding belt efficiency than the direction of cutting. The examples described here concern studies using manual sanding belt machines. The operating conditions of these machine tools are relatively high pressure (up to 20,000 Pa) and low belt speed (*v* <sup>s</sup> < 10 m/s). Industrial belt sanding machines operate at higher belt speeds (*v*<sup>s</sup> > 10 m/s) and lower pressure (*p* < 10,000 Pa). An example of the description of such research is the work of Saloni et al. [23], where a comparative study of industrial sanding of pine and maple wood is described. As a result of this study, a positive effect of the contact pressure and sandpaper belt speed on sanding efficiency was found, as well as a higher sanding efficiency of pine wood.

However, there is a lack of comparative studies on the influence of wood type and tool grit size on the variability of efficiency during sanding. Taking this into account, it was decided to verify the hypothesis that the type of wood, in combination with the grit size of sandpapers, affects the sanding efficiency during sanding with parameters typical for industrial applications (*p* < 10,000 Pa and *v*<sup>s</sup> > 10 m/s).

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

Wood from six hardwood species (beech, oak, ash, hornbeam, alder, walnut) and three softwood species (pine, spruce, larch) was tested. The wood material for making test samples was dried in an industrial dryer to a moisture content of 12% and stored in a freezer to preserve its physical properties. Then, the wood specimens with dimensions of 120 × 55 × 20 (length × width × height in millimeters) were obtained from it. Each specimen was measured with a caliper with an accuracy of ±0.2 mm and weighed using a WPS 510/C/2 balance (Radwag, Radom, Poland) with an accuracy of ±0.01 g. These measurements were used to calculate the density of the wood and to determine its initial mass. The calculated volumetric mass densities of the wood materials tested and the numbers of samples in the sample sets for each wood species tested are given in Table 1.


**Table 1.** Characteristics of wood specimens used in sandability tests.

Before sanding, the samples were glued with PVAc glue to raw chipboard spacers with dimensions of 120 × 55 × 16 (length × width × height in millimeters) (the purpose of this procedure was to enable complete sanding of the tested wood). The wood samples were positioned so that they were sanded along the wood fibers. The form of the test samples is shown in Figure 1.

**Figure 1.** Form of research samples.

The samples shown in Figure 1 were conditioned for another 3 months to equalize their moisture content in the whole volume.

Abrasive belts type EKA 2000 F 2000 × 75 (length × width in millimeters) (manufactured by Ekamant, Pozna ´n, Poland) with three different grit sizes (P60, P120, P180) were used for sanding; their specifications are given in Table 2.



A small industrial belt sander, Maktek S (Cormak, Siedlce, Poland), with a horizontal abrasive belt arrangement (Figure 2) was used. The gravitational clamping assembly allowed a constant pressure to be exerted by the abrasive belt on the samples (*p* = 3828 Pa). The speed of the sanding belt was constant and was: *v*<sup>s</sup> = 14.5 m/s.

**Figure 2.** Construction (**a**) and kinematic diagram (**b**) of a laboratory sander.

A separate sanding belt was used for each wood species. Each sample was sanded along the wood fibers in 30 s intervals and after each interval, the sample was weighed using a WPS 510/C/2 laboratory scale (Radwag, Radom, Poland). These steps were repeated many times until the entire wood sample was sanded off from the chipboard spacer. In this way, the following time series were obtained for each tested wood species: time-varying sanding efficiency (1), time-varying wood loss (2), and time to sand each sample, which allowed the calculation of the average sanding time for each series of samples (3).

The sanding efficiency for the intervals was calculated:

$$s\_{\varepsilon} = \frac{\left(m\_1 - m\_2\right)}{A} / t\_{\varepsilon} \left(\frac{\text{g} / \text{cm}^2}{\text{min.}}\right) \tag{1}$$

where: *se*—sanding efficiency, *m*1—wood sample mass at the beginning of each sanding interval (g), *m*2—wood sample mass at the end of each sanding interval (g), *A*—sample sanded area (cm<sup>2</sup> ), and *tc*—sanding cycle time (min.).

The wood loss was calculated relative to the initial sample weight:

$$w\_l = \frac{(m\_2 - m\_1)}{m\_0} \tag{\%}$$

where: *wl*—weight loss, *m*0—starting weight of the wood sample (g).

The mean value from the sample set measured every 30 s was taken as the *s*<sup>e</sup> result, and the mean value of the mass loss measured every 30 s was taken as the *w<sup>l</sup>* result until the last sample in the set was ground.

For the comparison of hardwood and softwood, the mean values of *s<sup>e</sup>* and *w<sup>l</sup>* were additionally calculated for all six hardwoods and three softwood species. The parameter *w<sup>l</sup>* was subjected to mathematical analysis. The determination of functional equations and their similarity analysis was based on multiplicity theory for comparing functions and for narrowing functions [24].

Average time for total sanding of wood in a serie:

$$t\_{AM} = \frac{1}{n} \sum\_{i=1}^{n} t\_i = \frac{t\_1 + t\_2 + \dots + t\_n}{n} \tag{3}$$

where: *tAM*—meantime for sanding a sample from the set, *n*—number of samples in a set, *t*<sup>1</sup> + *t*<sup>2</sup> + · · · + *tn*—sanding times of subsequent samples in a set.

#### **3. Results**

The results in the form of sanding efficiency time series are presented separately for hardwoods and softwood species for all three grades of sandpaper (Figures 3–5).

**Figure 3.** Mean sanding efficiencies of belts with grade P60.

**Figure 4.** Mean sanding efficiencies of belts with grade P120.

**Figure 5.** Mean sanding efficiencies of belts with grade P180.

Maximum sanding efficiency for hardwood species lasts from about 0.5 to 3 min, while for softwood species it lasts from 0.5 to 2 min. It is therefore apparent from Figures 3–5 that the first phase of sanding belt blunting ends quite early. In fact, for all wood species and all abrasive belt grit sizes, it is about 2–3 min after the start of machining when the sanding process moves into the second phase. In the case of softwood sanded with a P60 belt, the sample material finishes just after reaching the beginning of the second blunting phase. And in the case of pine wood sanded with the belt with the coarsest coating, it is not possible to enter the third phase of blunting of the coated abrasive before the sample wood is completely worn out.

In such a situation of the rapid progress of machine sanding at speeds higher than in the case of tests with manual sanders, it was decided to interpret the results of the experiment also by analyzing the wood removal rate for the tested wood species during the sanding. In this way, time series were obtained showing the percentage material loss during sanding (relative to a mean initial sample weight). The means were calculated separately for the sets of samples of each tested wood species. Those time series are represented by three consecutive Figures 6–8 (they show only a mean wood loss in each set of samples; without including the possible loss of a chipboard spacer).

**Figure 6.** Average percentage weight loss of specimens sanded with abrasive belt P60.

**Figure 7.** Average percentage weight loss of specimens sanded with abrasive belt P120.

**Figure 8.** Average percentage weight loss of specimens sanded with abrasive belt P180.

In all the cases, the fastest sanding was performed on the pine samples and the longest on the hornbeam samples. The total loss of mass of the specimens in the case of the P60 abrasive belt occurred in 1.5 min (pine) to 6.5 min (hornbeam). For the P120 belt, it ranged from 2 min (pine) to 7.5 min (hornbeam), and for the P180 belt, it ranged from 2.5 to 8.5 min (beech and hornbeam).

#### **4. Discussion**

In Figures 3–5, different rates of decline in sanding efficiency are observed. It seems that the rapid rate of decline in the sanding efficiency is related to the high initial efficiency (the greater the initial efficiency, the more rapid its reduction). This rapid rate of decline in sanding efficiency was observed with lower density samples, especially softwoods. A possible reason for the rapidly decreasing sanding efficiency (which occurs from 0.5 min to 3 min depending on grit size and species of wood) is that the spaces between the coated abrasive become clogged more quickly by wood dust.

The weight loss of sample sets during sanding is uniform (Figures 6–8), which results from the fact that most of the experiment time takes place in the second blunting phase, for which such a course of the sanding process is characteristic. Additionally, in this way, the occurrence of a short time of the first blunting phase was emphasized. Moreover, in the case of most of the wood species, a third blunting phase occurred at the end of the experiment, when the vast majority of the sample mass had already been sanded. The rate of weight loss of the wood during sanding, as known to date, is generally greater for lower-density wood species; but the differences between the high-density species (oak) and the light softwoods (spruce, visible in the graphs) are slight. In addition, the lighter hardwood species (alder, walnut), in terms of wood removal rate, behave similarly to low-density softwoods.

The conclusions of the scientific works to date indicating the effect of wood density resulted from comparisons of a mostly small number of species. Saloni et al. [23] compared parameters of sanding hard maple (*Acer saccharum*) (hardwood) and eastern white pine (*Pinus strobus*) (softwood). They mentioned wood species as one of the factors influencing the sanding results. The wood removal rate was twice as high for pine than for maple. Ockajova et al. [21] considered only two hardwood species: European beech (*Fagus sylvatica*) and English oak (*Quercus robur*). With a slight difference in density (684 kg/m<sup>3</sup> for beech, 678 kg/m<sup>3</sup> for oak), they found significant differences between the wood removal rates of both species. Miao and Li [25] also studied two hardwood species: Manchurian ash (*Fraxinus mandshurica*), and birch (*Betula* sp.). The density of wood samples in this study was respectively 620 and 470 kg/m<sup>3</sup> . In this case, lower values of wood removal rates in all variants of the study were for the denser and harder Ashwood. Thorpe and Brown [26] used

as many as 21 species (17 hardwoods and 4 softwoods) in the study on dust production during hand sanding. They found that the quantity of wood removed during sanding varied irreversibly with wood density.

The results of these studies link the rate of the wood removed during sanding to the density and directly to the species of wood, regardless of whether it is softwood or hardwood. Therefore, to compare these two different types of wood, the total sanding times for all samples of each species were averaged. The calculated average values of these total sanding times for the six hardwood species and three softwood species separately are shown in Figures 9–11.

**Figure 9.** Comparison of average results for hardwoods (beech, oak, ash, hornbeam, alder, walnut) and softwoods (pine, spruce, larch) species; P60 sandpaper.

**Figure 10.** Comparison of average results for hardwoods (beech, oak, ash, hornbeam, alder) and softwoods (pine, spruce, larch) species; P120 sandpaper.

**Figure 11.** Comparison of average results for hardwoods (beech, oak, ash, hornbeam, alder) and softwoods (pine, spruce, larch) species; P180 sandpaper.

The graphs in Figures 9–11 show that in all tested cases, the average sanding time of the samples of softwood species was lower than that of the hardwood species. The sanding times of the softwood species were: 2.5 min (P60), 5 min (P120), and 6.5 min (P180). For hardwood species, these times were 6.5, 7.5, and 8.5 min, respectively. The spread of sanding efficiency values for softwood is much greater than in hardwood species. This is due to the difference in tribological properties of softwood and hardwood. During sanding, what is important is not only the density but also the specific physical and mechanical properties of individual wood species and their morphologies themselves (hardwood ring-porous, hardwood scattered porous, with resin content or without, chemical composition, etc.). The resin content of the softwood has a great influence on the tribological properties of wood and it may be a cause of quickly blunting of the sanding tool [27].

The mean values of the measurements were statistically analyzed. Confidence intervals were calculated using a t-distribution table (*α* = 0.90). It was found that the measurement uncertainty of the means was always greater for softwood and its maximum value was independent of the sandpaper gradation. This uncertainty for hardwood was, respectively: 12.9% (P60), 13.2% (P120), 12.5% (P180). For softwood, it was: 19.1% (P60), 21.6% (P120) and 29.1% (P180).

The main hypothesis of our study is that hardwood differs from softwood in terms of the efficiency of the sanding process. For this purpose, changes in average weight loss were compared in the function of sanding time (Figures 9–11). The equations for these abrasive belts were found. Then, to adjust and compare the course of the functions, the set theory approach was used to narrow down the functions. In the first stage, for the functions to be considered equal, they must satisfy the first condition of the equality of the functions, which says that the functions *f* <sup>1</sup> (*x*) and *f* <sup>2</sup> (*x*) are equal to each other if, and only if, they have the same domains and for each point of the common domain, they assume these are the same values *f*<sup>1</sup> = *f*<sup>2</sup> ↔ *D<sup>f</sup>* <sup>1</sup> = *D<sup>f</sup>* <sup>2</sup> and for each *x* ∈ *D<sup>f</sup>* <sup>1</sup> = *D<sup>f</sup>* <sup>2</sup> , and *f*<sup>1</sup> = *f*<sup>2</sup> [24].

For all three grit sizes of sanding belts P60, P120, and P180, and for both types of wood (hardwood and softwood), the domains of functions being a square function with the general formula were calculated as *y* = *ax* <sup>2</sup> + *bx* + *c* using the procedure: (1) calculation of the root of a function ∆ (∆ = *b* <sup>2</sup> <sup>−</sup> <sup>4</sup>*ac*); (2) determination of parameters *<sup>p</sup>* <sup>i</sup> *<sup>q</sup>* (*<sup>p</sup>* <sup>=</sup> − √ ∆ 2*a* , *q* = <sup>−</sup><sup>∆</sup> 4*a* ). In the case of sanding belt P60, significant differences were found between the examined functions because their domains assumed values *<sup>D</sup><sup>f</sup>* <sup>1</sup>(−5, +∞) and *<sup>D</sup><sup>f</sup>* <sup>2</sup>(−1, +∞), thus *D<sup>f</sup>* <sup>1</sup> 6= *D<sup>f</sup>* <sup>2</sup> , so the functions are not equal. Then the second condition for the definition of function equality was checked, indicating that for each *x* ∈ *D* : *f*1(*x*) = *f*2(*x*) [28]. These functions do not satisfy the equality condition and are therefore different, which

indicates that hardwood differs from softwood in terms of the efficiency of the sanding process for the grit size P60.

For the sanding belt P120, it was observed that the first condition of the equality of the function was not satisfied (*D<sup>f</sup>* <sup>1</sup>(−1, +∞) and *<sup>D</sup><sup>f</sup>* <sup>2</sup>(−0, 1, +∞)). However, the second condition assuming that for every *x* ∈ *D* : *f*1(*x*) = *f*2(*x*) was satisfied. Thus, a restriction was applied for functions on the selected set of points belonging to the set A (−89.4; −100) where for every *x* ∈ *A*, the functions are equal : *f*1(*x*) = *f*2(*x*) [29]. Based on the calculations related only to this set of arguments, it was found that both functions are similar to each other in the indicated range, and their domains are the same for this set of arguments. A similar situation was observed for the sanding belt P180. Using the second condition of function equality, set A was determined (−81.7; −100) for which these functions are equal. Based on the mathematical analyses, it was found that the sanding belts in the second stage of blunting behave similarly. The belts of grit number (P60) dull faster than belts with higher grit numbers (P180). At the same time, differences were found between softwood and hardwood in terms of the efficiency of the sanding process.

The graph in Figure 12 shows the average sanding times with belts of different grit numbers in the individual sample sets. The graph also shows the Brinell hardness of individual wood species (the macro-hardness determination method was selected from two common hardness measurement methods [30]).

**Figure 12.** Average sanding time for sample sets sanded with different grit sizes.

The average sanding times for sample sets of two wood species are interesting. In the case of ash, the shortest sanding time was obtained for the P120 belt, while in the case of the other tested materials, the shortest sanding time was usually for the P60 belt (in the case of walnut, these times were more or less equal). This can be explained by the fact that it is the hardest species of wood and in this case, the optimal abrasive belt grit from the point of view of sanding efficiency fell on the belt of medium grain (according to the grain size effect described in the publication Sin et al. 1979 [31] caused by the influence of the elasticity of wood [32]). Another interesting species of wood is spruce. In the case of samples made of this material, the greatest effect of the grit number of the sanding belt on the average time for complete sanding of the sample set was observed. The sanding time with the P120 belt increased by as much as 120% compared to the sanding time with the P60 belt. For other grits, these times were either slightly shorter (by 15% for ash and 5% for walnuts) or greater (from 13 to 42%).

The times of sanding with belts of different grit numbers of the individual sets of samples do not seem to correlate with the Brinell hardness of the tested wood species. For example, walnut, which is twice as hard as alder, shows an average sanding time similar to that of alder. Among the softwood species, average sanding times of pine wood are roughly half of that of spruce, although the hardness of both species is similar. The test results show that hardness is not the only factor affecting sanding efficiency. The influence may be caused by other tribological properties, e.g., the instantaneous coefficient of friction, which depends both on the type of wood, hardness, as well as the roughness and temperature, which are time-varying during sanding and dependent on the grit of the sanding belt [33].

#### **5. Conclusions**

The results of the study on the machine sanding of different wood species with sanding belts of various grit numbers indicate that:


In machine sanding of wood at low pressure and high belt speed, abrasive materials with a low grit number and high sharpness affect hardwood and softwood differently. The sanding efficiency of softwood is considerably higher than hardwood in these conditions. Therefore, sanding parameters (pressure and belt speed) should be set at lower values to avoid excessive sanding or over-sanding (a situation when too much material is sanded).

**Author Contributions:** Conceptualization, T.R. and M.S.; methodology, T.R.; validation, M.S., R.M. and T.R.; formal analysis, K.S.-S.; investigation, T.R.; resources, T.R.; data curation, M.S.; writing original draft preparation, T.R. and M.S.; writing—review and editing, T.R. and M.S.; visualization, M.S.; supervision, T.R.; funding acquisition, R.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Centre for Research and Development, BIOS-TRATEG3/344303/14/NCBR/2018.

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** The authors thank Jacek Sydor for the valuable terminological comments.

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

#### **References**


### *Article* **Influence of Grit Size and Wood Species on the Granularity of Dust Particles during Sanding**

#### **Marta P ˛edzik 1,2 , Kinga Stuper-Szablewska <sup>3</sup> , Maciej Sydor <sup>4</sup> and Tomasz Rogozi ´nski 1,\***


Received: 28 October 2020; Accepted: 16 November 2020; Published: 18 November 2020 -

**Abstract:** Wood dust poses a threat to the health of employees and the risk of explosion and fire, accelerates the wear of machines, worsens the quality of processing, and requires large financial outlays for its removal. The aim of this study was to investigate the extent to which the grit size of sandpaper influences the size of the wood dust particles and the proportion of the finest particles which, when dispersed in the air, may constitute the respirable fraction. Six species of hardwood (beech, oak, ash, hornbeam, alder, and walnut), and three species of softwood (larch, pine, and spruce) were used in the research. While sanding the samples under the established laboratory conditions, the following were measured for two types of sandpapers (grit sizes P60 and P180): mean arithmetic particle size of dust and finest dust particles content (<10 µm). Based on the obtained results, we found that the largest dust particle sizes were obtained for alder, pine, and spruce; the smallest size of dust particles during sanding with both sandpapers was obtained for beech, hornbeam, oak, ash, larch, and walnut. The mean arithmetic particle sizes ranged from 327.98 µm for pine to 104.23 µm for hornbeam. The mean particle size of the dust obtained with P60 granulation paper was 1.4 times larger than that of the dust obtained with P180 granulation sandpaper. The content of the finest dust particles ranged from 0.21% for pine (P60 sandpaper) to 12.58% for beech (P180 sandpaper).The type of wood (hardwood or softwood) has a significant influence on the particle size and the content of the finest dust fraction.

**Keywords:** wood dust; sanding; sandpaper; particle-size distribution

#### **1. Introduction**

Wood dust is a waste generated during mechanical wood processing in wood industry plants. Dust poses a threat to workers health, increases the risk of explosion and fire, accelerates the wear of machines, worsens the quality of processing, and incurs high costs for its removal.

Dust with particles smaller than 10 µm affects the respiratory system, eyes, and skin, causing health effects in the form of irritation, allergies, and diseases [1–6]. The dust toxicity is determined to a large extent by the type of wood raw material, which results from the different contents of the main chemical components, such as cellulose, hemicellulose, and lignin in coniferous and deciduous trees [7,8]. Long-term inhalation of air polluted with wood dust, including the most harmful (i.e., beech and oak wood) may contribute to cancer incidence. As a result, wood dust was classified by the International

Agency for Research on Cancer (IARC) among the most dangerous and carcinogenic materials for humans. According to Directive 2004/37/EC of the European Parliament and of the Council of 29 April 2004 on the protection of workers against the risks related to exposure to carcinogens or mutagens at work; its current limit is 3 mg/m<sup>3</sup> . After 17 January 2023, it will be reduced to only 2 mg/m<sup>3</sup> . To meet such high requirements, it will be necessary to use all available technical means of dust reduction.

The dustiness of the air inside production plants causes a fire and explosion hazard. Dust dispersed in the air can create an explosive mixture and settle on walls, floors, and machines, creating a risk of fire and explosion at the workplace. The inflammability of wood dust favors the spread of fire [9–13].

Extraction devices designed to remove dust from processing areas are never completely effective, especially in relation to the very small-sized dust particles. One way to reduce the amount of fine dust generated is to reduce the thickness of the furniture components and thus the diameters and depths of the holes for connectors, which reduces the volume of wood material cut. Such a procedure is effective in eliminating drilling and milling operations; however, it requires the development of new furniture fasteners [14] and does not reduce the need for sanding operations. Another way is to adjust the processing parameters to the type and properties of the material being processed. Appropriate adjustment of the treatment parameters will result in the formation of a reduced amount of fine dust dispersed in the air, could constitute a potentially dangerous inhaled fraction.

With regards to sawing and milling, the technological parameters of the processing, as well as the type and sharpness of the tools used, are particularly important. Among technological parameters, the feed per tooth, the feed rate, and the thickness of the cut layer have important influences on the amount and size of the created dust particles [15–25].

The largest amount of dust is generated during the sanding of wood; therefore, this technological operation is considered to be the source of the most serious hazards related to wood dust. The amount and size of the dust particles created when sanding wood depends on the type of sanding machine used: wide-belt, narrow-belt, or disc sanders. It is difficult to remove the dust from some special types of sanding machines, such as manual belts, discs, and oscillating sanders [26,27].

Recent studies on the size of the dust particles created during the sanding of wood concerned some of the most commonly used wood species in the industry. These tests were carried out using various types of sanding machines and sandpapers with grain sizes typical for the basic technical requirements of sanding operations. Many times, the sandpaper type studied was limited to paper tapes with the grain size of P80. Due to the method of sieve analysis used in these studies to assess the size of dust particles (the sieve with the smallest mesh of 32 µm), it was not possible to quantify the content of dust particles with the smallest sizes, which would constitute the respirable fraction after dispersion in the air [28–30]. Such particles are present in the wood sanding dust. Their presence was confirmed by spectrometric, optical, and laser methods [27,28]; however, no comprehensive and comparative studies have been performed on the content of wood dust particles that would be small enough to be dispersed in the air as a thoracic fraction [31–33].

To reduce the risk to workers' health, increase work safety, and meet future legal requirements for fine wood dust, we examined the most important factor characterizing the sanding process of wood-grit size. Therefore, the aim of the study was to investigate the extent to which the grit size of the sandpaper affects the size of the dust particles created during sanding different wood species and the proportion of the finest particles, which, when dispersed in the air, may constitute a respirable fraction.

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

#### *2.1. Sanding and Particle Size Analysis*

Three species of softwood and six species of hardwood often used in the wood industry were used in this research (Table 1). The densities of wood species were determined according to the method described in the standard ISO 13061–2:2014.



Sanding was performed using a prototype narrow belt sanding machine designed and made in the laboratory of the Department of Furniture Design (Faculty of Forestry and Wood Technology, Pozna ´n University of Life Sciences PULS, Pozna ´n, Poland). EKA 1000 F sandpaper (Ekamant, Pozna ´n, Poland) in the form of belts with dimensions 1000 × 80 mm was used (Figure 1). The grit sizes of the paper were P60 and P180. A cutting speed of 14.5 m/s and a sanding pressure of 0.65 N/cm<sup>2</sup> were applied.

**Figure 1.** Schematic diagram of sanding.

Particle-size determination and calculation of the content of fine dust particles (the content of particles <10 µm) were carried out according to methods described by [20,22,34,35]. In the sieve analysis, a set of sieves with aperture sizes of 250, 125, and 63 µm was used due to the high level of wood dust fineness. Then, the content of the dust particles <10 µm in the sieve fraction <63 µm was measured using a Analysette 22 MicroTec Plus laser particle sizer (Fritsch, Idar-Oberstein, Germany). Based on the results of sieve analysis, the cumulative particle size distribution Q<sup>3</sup> and the particle mean arithmetic diameter *x* was calculated as follows:

$$\mathcal{Q}\_3 = \sum\_{i=1}^n \overline{q}\_{3,i} \Delta x\_i \tag{1}$$

$$\overline{\mathbf{x}} = \sum\_{i=1}^{n} \mathbf{x}\_i \times q\_{\overline{3}i} \tag{2}$$

where *q*<sup>3</sup> is the particle size distribution by mass, *x* is the mean value of particle size class, and *n* is the number of particle size classes.

#### *2.2. Statistical Analysis*

The results recorded in the course of conducted tests were subjected to statistical analysis with the use of STATISTICA ver. 13.1 (StatSoft, Inc., Tulsa, OK, USA) and Microsoft® Excel 2020, Microsoft 365 (Addinsoft, Inc., Brooklyn, NY, USA) software packages. In order to compare grit sizes of the P60 and P180 sandpaper, multivariate comparison procedure was used, with identical letters denoting a lack of differences at the significance level of *P* = 0.05, lowercase letters denote significant differences between the grit sizes of the sandpaper, and uppercase letters denote significant differences between wood species. Correlation coefficients between the wood density and the mean arithmetic particle size of dust were calculated. The multiplicity factors for the mean arithmetic particle size and for the content of the finest particles obtained from paper with the grit sizes P60 to P180 were calculated. Moreover, a step linear discriminatory analysis (SLDA) and the principal component analysis (PCA) were used to separate groups of analyzed dust in the entire population.

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

The basic results of the particle size analysis came from the sieve analysis. The cumulative distributions (Figure 2) showed that, in general, the dust from sanding operations performed with the use of P180 sandpaper is finer than dust created in sanding with P60 sandpaper.

**Figure 2.** Particle size distributions of wood dust.

To compare the mean arithmetic particle sizes of dust created in sanding with P60 and P180 sandpapers within the entire population, that is, between wood species and within wood species, multivariance analysis was used (Figure 3). On the basis of the obtained results, the smallest size of dust particles created during sanding with both sandpapers was found for beech, hornbeam, oak, ash, larch, and walnut. Significantly larger sizes of dust were obtained for alder, pine, and spruce. By analyzing the ratio of mean arithmetic particle sizes of dust obtained from paper with grit sizes P60 and P180, the multiplicity factors were calculated and we found that the average for the entire population was 1.4, which means that the mean particle size of dust obtained from the paper with grit size P60 was 1.4 times higher than that of dust obtained from paper with the grit size P180. When analyzing the dust

from hardwood and softwood species separately, we found that the multiplicity factors were 1.4 and 1.3, respectively. The lowest multiplication factor was found for walnut wood, at 1.1, and the highest for alder wood, at 1.8.

**Figure 3.** Analysis of the mean arithmetic particle sizes of dust; a, b: the same letters indicate no significant differences at the significance level of 0.05 between sandpaper grit size; A, B: the same letters indicate no significant differences at a significance level of 0.05 between wood species.

While analyzing the content of the finest particles (<10 µm), a multivariate analysis was also performed to compare this content for wood dust created during sanding with papers of grit sizes P60 and P180 for different species of wood and to compare the content of particles <10 µm within species (Figure 4). There were no significant differences between the content of the finest dust particles obtained from sanding with both sandpapers for pine and larch. When analyzing the differences for the species, significantly higher contents of the smallest particles were found for alder, walnut, oak, beech, and hornbeam. By analyzing the ratio of dust particles <10 µm obtained from sandpaper of grit size P60 to the paper of grit size P180, the multiplicity factors were also calculated and an analogous tendency was found for the mean arithmetic particle size. The average multiplicity factor P60/P180 was 1.4 for the entire population of wood species tested.

Then, we analyzed the main components on the basis of the value of the mean arithmetic particles sizes and the content of the finest particles for both grit sizes of sandpaper, and the second predictor, next to the wood species (Factor 1), was indicated as the type (hardwood or softwood) of wood (Factor 2). Supplementing the grouping factors with this factor showed that the obtained results could be divided into two groups, as shown in Figure 5. One region of the loop comprises hardwood and the other, softwood. The full separation was obtained by a factor of two. This clear division was also confirmed by the PCA result showing the projection of the variables on the plane (Figure 6).

**Figure 4.** Analysis of the content of the finest dust particles; a, b: the same letters indicate no significant differences at the significance level of 0.05 between sandpaper grit sizes; A, B: the same letters indicate no significant differences at a significance level of 0.05 between wood species.

**Figure 5.** The spread of the mean arithmetic particle size of dust and the content of the finest particles. (1, 2, 3 are beech; 4, 5, 6 are oak; 7, 8, 9 are hornbeam; 10, 11, 12 are ash; 13, 14, 15 are alder; 16, 17, 18 are walnut; 19, 20, 21 are larch; 22, 23, 24 are pine; and 25, 26, 27 are spruce) based on the multiplicity factor of the mean arithmetic particle size of the dust created in sanding with sandpapers of grit sizes P60 and P180 and on the content of the finest particles. The clusters of results that form separate populations are marked in the loops.

**Figure 6.** The result of principal component analysis (PCA) for the entire population of dust.

The correlation coefficients between the wood density and the mean arithmetic particle size of dust were calculated. Very high values of these coefficients for the tested wood species were found at the level of 0.9588 for grit size P60 and 0.8794 for grit size P180 based on this analysis. A similar relationship was found for particles <10 µm. The correlation coefficient value was 0.9227 for the P60 grit size and 0.8812 for the P180 grit size. This confirms the highly significant relationship between the grit size of sandpaper and the mean arithmetic particle size and the content of the finest particles in wood dust.

The share of beech wood dust particles with a diameter of ≤80 µm in the range of 89.21–96.29% was described by Oˇckajová et al. [18]. The finest dust particles that can be found in a such large proportions of small particles are undesirable in working environments. They can penetrate the alveoli and be the source of serious diseases. A similar relationship was obtained when using P180 paper, where the share of this fraction was also significant.

The wood species also has a significant impact on the sanding belt wear. This is due to the specific, mainly microscopic, structure of the wood, as well as different physical and mechanical properties. Beech wood has a uniform structure in spring–summer rings and scattered rings. Oak wood has significant differences in density between spring and summer rings, as well as a relatively high share of extractive substances, even up to 6.1% [36]. Softwood, such as coniferous species, has relatively long fibers, which result in a fraction with larger particles.

#### **4. Conclusions**

Based on these studies, we concluded that: The largest mean arithmetic dust particle sizes were obtained for alder, pine, and spruce. The smallest mean arithmetic dust particle sizes were obtained for beech, hornbeam, oak, ash, larch, and walnut. The highest content of the finest particles was found for alder, walnut, oak, beech, and hornbeam. The type of wood (hardwood or softwood) has a significant influence on the mean arithmetic dust size and the content of the dust fraction with the size <10 µm. The particle size analyses of dust from the sanding of different wood species showed that the sanding of walnut, oak, beech, and hornbeam can be a source of a considerable amount of very fine dust particles, which can from a respirable fraction when dispersed in the air.

**Author Contributions:** Conceptualization, M.S. and T.R.; Data curation, M.P. and K.S.-S.; Formal analysis, M.S.; Funding acquisition, M.S.; Investigation, M.P. and K.S.-S.; Methodology, M.P. and T.R.; Project administration, M.S. and T.R.; Resources, T.R.; Software, K.S.-S.; Supervision, T.R.; Validation, M.S.; Visualization, K.S.-S.; Writing, original draft, T.R.; Writing, review and editing, M.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** The article processing charge (APC) was financed within the European project POIR.01.02.00-00-00102/17, "The first Polish innovative universal system of furniture fasteners for joining various wood and wood-composite materials in the furniture industry", implemented by Digitouch sp. z o.o. (Suchy Las, Poland). The project is a part of the Polish sectoral programme WoodINN financed by the Polish National Centre for Research and Development (NCRD).

**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* **Optimization of Parameters for the Cutting of Wood-Based Materials by a CO<sup>2</sup> Laser**

**Ivan Kubovský , L'uboš Krišt'ák \* , Juraj Suja, Milada Gajtanska, Rastislav Igaz , Ivan Ružiak and Roman Réh**

Faculty of Wood Sciences and Technology, Technical University in Zvolen, T. G. Masaryka 24, SK-960 01 Zvolen, Slovakia; kubovsky@tuzvo.sk (I.K.); jukas@itenec.sk (J.S.); gajtanska@tuzvo.sk (M.G.); igaz@tuzvo.sk (R.I.); ruziak@tuzvo.sk (I.R.); reh@tuzvo.sk (R.R.)

**\*** Correspondence: kristak@tuzvo.sk

Received: 15 October 2020; Accepted: 12 November 2020; Published: 16 November 2020 -

**Abstract:** This article deals with the laser cutting of wood and wood composites. The laser cutting of wood and wood composites is widely accepted and used by the wood industry (due to its many advantages compared to, e.g., saw cutting). The goal of this research was to optimize the cutting parameters of spruce wood (*Pices abies* L.) by a low-power CO<sup>2</sup> laser. The influence of three factors was investigated, namely, the effect of the laser power (100 and 150 W), cutting speed (3, 6, and 9 mm·s −1 ), and number of annual rings (3–11) on the width of the cutting kerf on the top board, on the width of the cutting kerf on the bottom board, on the ratio of the cutting kerf width on the top and bottom of the board, on the width of the heat-affected area on both sides of the cutting kerf (this applies to the top and bottom of the board), and on the degree of charring. Analysis of variance (ANOVA) and correlation and regression analysis were used for developing a linear regression model without interactions and a quadratic regression model with quadratic interactions. Based on the developed models, the optimization of parameter settings of the investigated process was performed in order to achieve the final kerf quality. The improvement in the quality of the part ranged from 3% to more than 30%. The results were compared with other research dealing with the laser cutting of wood and wood composites.

**Keywords:** laser cutting; wood; wood composites; cutting parameters

#### **1. Introduction**

The cutting of materials is one of the most widely used laser machining processes. The principle consists of the movement of a focused laser beam perpendicular to the plane of the machined surface. The absorption of high energy laser radiation causes the material to melt and subsequently evaporate, creating the desired cut. For this purpose, a powerful CO<sup>2</sup> laser is still very frequently used, producing a beam with a wavelength of 10.6 µm [1]. In the past, CO<sup>2</sup> lasers were by far the most widely used devices for laser cutting. In recent years, the laser with the largest market for industry use has been the fiber laser, with a wavelength of 1064 nm. In industrialized countries, such as Japan, material cutting operations performed by beam radiation are very widespread and have grown significantly in recent years [2,3]. Due to their unique efficiency and accuracy, these technologies have become the center of attraction in various fields of application [4]. The cutting process is fast, non-contact, and highly automated and there is only thermal stress in a very limited area, allowing the cutting of a wide range of materials. Additionally, low operating costs are achieved, despite the high initial investment and the requirement for qualified staff [5].

In the case of wood and wood-based products, laser machining is widely accepted by the wood industry [6–8]. The laser processing of wood was one of the earliest applications of laser use in the 1970s [9,10]. The advantage of the laser system is its ability to cut complicated patterns. The application of lasers in the furniture industry for automated cutting processes us also well-known [11,12]. In addition to cutting, CO<sup>2</sup> lasers are also used for the irradiation or engraving of a wood surface [13–17]. The main advantages of the laser cutting of wood in comparison with conventional cutting methods are the highly precise cut, flexibility to start and finish cutting at any point of the board, narrow kerf width (0.1–0.3 mm compared to saw cut of 3–6 mm), and extremely smooth surfaces [18–23]. Other advantages of laser wood cutting are the absence of tool wear, low noise emission and vibration, low sensitivity to very variable processing properties [23–26], reduced amount of sawdust [27–31], and low mechanical stress in the workpiece [32,33].

Factors influencing the process of cutting wood and wood composites by a laser can be divided into three groups [34–36], namely, the properties of the radiation beam, the properties of the laser device and the characteristics of the cutting process, and the properties of the workpiece, specifically, the effects of the beam power, mode, and polarization as aspects of optics; the location of the focal point; the feed speed; the gas-jet assist system and workpiece thickness; the density; and the moisture content [37–44].

In the case of wood composites, Lum et al. [45] investigated the parameters of the process of MDF board cutting with a beam of radiation generated by a CO<sup>2</sup> laser with a power of 520–530 W in pulse and continuous mode. The optimal setting of the parameters for the MDF board cutting was achieved by the experimental change of the cutting speed, the composition of the auxiliary gas and its pressure, and the change of the position of the radiation beam focusing in the material. The results showed that narrow kerf widths are achievable for MDF laser-cut boards, particularly for pulse mode cutting. Striation patterning, although masked by external charring, is evident, but this is of little significance to the overall quality of cut, as evidenced by the low surface roughness values obtained. Burnout was also minimal, even for angular profile cuts of small internal angles. These results were also in agreement with Powell [46]. Eltawahni et al. [47] investigated a means for selecting the process parameters for the laser cutting of MDF based on the design of experiments. They defined a methodology according to which we can partially evaluate the efficiency and quality of radiation beam cutting by the proportion of the cutting joint width on the cut upper workpiece surface to the cutting joint width on the cut workpiece bottom (the so-called ratio). The focal point position and laser power are the principal factors affecting the ratio. In the case of the upper board's surface kerf, the focal point position has the main role, and in the case of the lower kerf, the laser power and cutting speed have the main effect on the lower kerf width. The roughness of the cut section decreases as the focal point position and laser power increase. The roughness increases as the cutting speed and air pressure increase. Smoother cut sections could be processed, but with an increase in the processing operating cost. Barnekov et al. [48] investigated the effect of different focal point positions on the cut surface quality. They found that a smooth surface of the workpiece middle can be achieved with less charring with the focal point at or slightly above the middle. Most of these results are in harmony with previous published research in the case of wood and wood composites [49–51]. The focal point position was also investigated by Barnekov et al. [40]. In their research, they achieved the best quality of particleboard laser cutting when the beam was focused on the surface of the composite material.

In the case of wood, Tayal et al. investigated the significance of the focus position of the CO<sup>2</sup> laser optical system when hardwood cutting [52]. They concluded that the focal point should be at a location Z/2 below the surface to achieve a maximum average laser power density, where Z is the focal length (depth of focus). They verified these results by experimental observations. Nukman et al. [53] used a CO<sup>2</sup> laser to cut a wide range of Malaysian wood and plywood. The processing variables taken into account were the laser power, focal point position, nozzle size, assist gas pressure, types of assist gas, cutting speed, and delay time. The wood properties observed were the thickness, density, and moisture content of wood. The analyses considered the geometric and dimensional accuracy (straight sideline length, diameter of the circle, kerf width, and percent over cut), material removal rate, and severity of burning. A guideline for cutting a wide range of Malaysian wood has been outlined. The influence of

the laser power, cutting speed, and shield gas on the cut quality in the case of hard and soft timber was confirmed by Khan et al. [6]. McMillin [54] dealt with the moisture content and its influence on the process of pine wood cutting by a beam of radiation generated by a CO<sup>2</sup> laser in his work. He found that the cutting speed of wetter wood is slower compared to drier wood. The reason for this is that the moisture content in wood increases the thermal conductivity, which results in energy loss in the heating zone. Grad and Mozina [41] showed that the CO<sup>2</sup> laser beam is almost completely absorbed by wood. This result has been confirmed by Hattori [42], who claimed that a CO<sup>2</sup> laser is most suitable for wood processing due to the wavelength used. The authors of the research [55,56] proved that a cut section quality with less roughness results from an increasing beam power.

The optimal conditions for laser cutting were determined by several authors using theoretical models. Zhou and Mahdavian [57] explored the possibilities of cutting wood and particleboards with various levels of laser power and different workpiece cutting speeds. They introduced a theoretical model that estimates the depth of the cut in terms of the material properties and cutting speed. The experimental cutting results were compared with theoretical predictions. Two correction parameters were introduced in the analysis to improve the theoretical model. Numerous other theoretical models for laser machining have been developed by other researchers, among them, Moradi et al. [58], Choudhury et al. [59], Yang et al. [60], Elsheikh et al. [61], Alizadeh and Omrani, and others [62–64]. Various modes of heat transfer, phase changes, workpiece motions, and material properties have been taken into account in these analysis.

There are also several simulations and mathematical and statistical models, in addition to theoretical models. Statistical and experimental analyses of the multiple-pass laser cutting of wet and dry pine wood were presented by Castaneda et al. [65]. The parameters investigated were the laser power, traverse speed, focal plane position, gas pressure, number of passes, direction of cut (normal or parallel to the wood´s tracheids), and moisture content. The experimental results were compared against process responses defining the efficiency (i.e., kerf depth and energy consumption) and quality of the cut section (i.e., kerf width, heat-affected zone, edge surface roughness, and perpendicularity). An energy balance-based simple analytical model was developed and validated with experimental results by Prakash Kumar [66]. The optimal properties of the cutting process can also be determined by suitable simulation methods. Jianying and Yun [67] developed a simulation technique with ANSYS software using the finite element method in order to propose the optimal parameters of the wood beam cutting process. Polak et al. [68] used ANSYS and the numerical finite element method to model the radiation beam cutting and drilling process. Yang et al. [69,70] analyzed the influence of the ablative mechanism of wood processed with a nanosecond laser on the cutting quality and established a prediction model through multiple linear regression equations. Hardalov et al. [71] used the finite element method in their work to quantify the temperature gradient on the surface of ceramic material in the process of irradiation with pulsed and continuous beams using FEMLAB software. In the research by Modest [72], a previously developed three-dimensional conduction model for the scribing of a thick solid was extended to predict the transient temperature distribution inside a finite thickness slab that was irradiated by a moving laser source. The governing equations were solved for both constant and variable thermophysical properties using a finite-difference method on a boundary-fitted coordinate system. An evaluation of the radiation beam cutting quality was also addressed by Rogerro et al. [73], whose method is a combination of neural networks and traditional algorithmic techniques. Bianco et al. [74] used COMSOL Multiphysics 3.2 software to solve a transient two- and three-dimensional temperature field irradiated by a beam of radiation. Babiak et al. [75] dealt with the analysis of temperature profiles created by a moving Gaussian energy beam on a wood surface. Yilbas et al. [76] analyzed both mathematical and numerical solutions of material heating by a beam of radiation.

The aim of this research was to find the optimal combination of parameters of the cutting process of an 8 mm thick spruce board with a moisture content of 12% by a CO<sup>2</sup> laser. The required cut quality was determined by the minimum width of the heat-affected zone, the minimum value of the degree of charring, and the ratio of the cutting joint width on the upper and bottom board´s surface. The parameters were the laser power, cutting speed, and number of annual rings intersecting the area of the cutting kerf.

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

The experiments were carried out on Norway spruce (*Picea abies*(L.) H.Karst). Wood was harvested from the Polana region in Slovakia. Cutting kerf was created by the cutting of tangential spruce lumber with dimensions of 8 mm x 100 mm x 1000 mm (tangential x radial x longitudinal), with a relative moisture content *w* = 12 ± 1% and average density ρ = 428.4 ± 27.9 kg·m−<sup>3</sup> . The moisture content of the samples was determined according to ISO 13061-1 [77]. The wood density was determined according to ISO 13061-2 [78]. The CO<sup>2</sup> laser LCS 400-1/W (TST Strojárne Piesok, Piesok, Slovakia) was used for cutting with a wavelength of 10.6 µm operated at a continuous mode output power of 100 and 150 W. Three speeds of cutting were used, with values 3, 6, and 9 mm·s −1 . A power of 150 W and speed of cutting of 9 mm·s <sup>−</sup><sup>1</sup> were used as a reference, since this combination is used in practice due to the speed of production. The focal length was 127 mm (5"), beam diameter was 10 mm, and spot diameter was 0.3 mm. The focal point position of the laser beam was set to 1/2 of the sample thickness (measured from the upper surface of the board). The process gas was supplied via a laval contour nozzle with 0.25 MPa pressured air.

All cuts were made parallel to the wood fibers in the tangential direction (Figure 1). The quality of the cut section of the wood was examined using Digital Microscopy (DM).

**Figure 1.** Cutting scheme.

The Response Surface Method (RSM) was used to prepare the experiment. The influence of three factors was investigated, namely, the effect of the laser power P, cutting speed V, and number of annual rings AR (Table 1) on the width of the cutting kerf WKU (width of the cutting kerf on the upper board´ s surface) and WKD (cutting kerf width on the board bottom), on their ratio WKR (ratio of the cutting kerf width on the upper and bottom board´s surface), on the width of the heat-affected areas on the spruce board surface WHAZx (width of the heat-affected area on both sides of the cutting kerf is equal, and this applies to the upper board´s surface, as well as to the bottom of the board), and on the degree of charring B.



Measuring of the cutting kerf width WKU, WKD, and WHAZx was observed by DM using the K-means clustering segmentation method [79]. The K-means clustering algorithm is an unsupervised algorithm and it clusters or partitions the given data into K-clusters based on the K-centroids. The algorithm is used when we have unlabeled data (without defined categories) and the objective is to minimize the sum of squared distances between all points and the cluster center. In our case, three different color-coded regions were searched in image processing software, namely, the area of the unaffected workpiece, the area of the heat-affected zone, and the area of the cutting kerf. The mean values of the width of the cutting kerf and especially the width of the heat-affected zones were determined from the segmented photos by applying the theorem on the mean value theorem for integrals (Figure 2).

**Figure 2.** Photo used for measuring the width of the cutting kerf and heat-affected zone (**left**), and photo edited by segmentation (**right**) by Digital Microscopy (DM).

The measuring of the degree of charring consisted of an analysis of cutting kerf in a direction perpendicular to the surface and a subsequent analysis of the morphology (texture) of the examined surface. Surface charred areas are amorphous areas of cutting kerf without a visible texture with a damaged anatomical wood structure due to the thermal degradation of wood. The bimodal histogram was modeled as a mixture of two Gaussian density functions. The use of adaptive particle swarm optimization for the suboptimal estimation of the means and variances of these two Gaussian density functions, and then the computation of the optimal threshold value, was straightforward [80]. The result of this analysis was obtained as the degree of charring in %.

The Shapiro–Wilk test was used to test the normality of measured data, followed by Box-Cox nonlinear normalizing transformation. For greater data clarity of the response, variables were transformed into variables that also passed the normality test. Analysis of variance (ANOVA) and correlation and regression analysis were used for developing a linear regression model without interactions and a quadratic regression model with quadratic interactions. Based on the developed models, the optimization of parameter settings of the investigated process was performed in order to achieve the required response.

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

The total number of measurements was 108 in one block, with 864 measured or calculated data. Due to the random factor AR, nine measurements were performed at each monitored cutting speed. The WKR ratio was calculated and the degree of charring was subsequently determined from the measured data. The data were then tested by the Shapiro–Wilk test of good agreement of the measured data, with the normal distribution at a significance level of 0.05. Due to the fact that the measured data did not meet the assumption of a normal distribution, a normal distribution was achieved by a nonlinear Box-Cox transformation (Table 2).


**Table 2.** Box-Cox transformation.

The results of the measured data after the transformation enabled the use of parametric mathematical-statistical methods of experimental evaluation (in the next parts asterisk symbol \* is used for transformed values). Descriptive statistics were subsequently produced for all monitored transformed responses (Table 3).

**Table 3.** Descriptive statistics for transformed values of all parameters.


It follows from the descriptive statistics tables for each monitored transformed response that the cutting kerf width at the board top is larger than the cutting kerf width at the board bottom. The term "upper board´s surface" refers to the side where the primary radiation beam interacts with the material. The values of the variation coefficients are smaller on the upper board´s surface compared to the values of the variation coefficients of the observed responses on the board bottom. This is related to the properties of the wood from the point of view of heat transfer, as well as to the power of the radiation beam required for thermal decomposition of the wood in the cutting process. Larger values of variation coefficients also point to defects in the form of burns on the surface of the board bottom, which are caused by the sudden blowing of burns from the space of the cutting kerf with auxiliary gas (air).

#### *3.1. Analysis of Variance and Regression Analysis*

A three-factor experimental model without interactions was used to primarily determine the influence of the investigated factors AR, P, and V for responses WKU\*, WHAZU\*, WKD\*, WHAZD\*, WKR\*, and B\*, with an evaluation of the variability degree of the investigated response (Table 4). Additional information on the influence of interactions between factors was analyzed by a second-degree polynomial model with first-order interactions (Table 5). The coefficients of the factors of the examined responses are marked in gray, which are significant in terms of impact at the significance level of 0.05. The force measures of the models are also marked in gray, which can be considered as sufficiently explanatory of the spruce board cutting process. The coefficients of the regression function were determined by the least squares method. The significance of the investigated factors was evaluated by analysis of variance (F test, *p* test, R<sup>2</sup> , standard deviation, PACH analysis, and percentage error of the model were performed in all cases), graphically by a Pareto diagram, and by a graphical analysis of their influence (Figure 3).


**Table 4.** Summary of linear models without interactions (e.g., WKU\* = 0.9691 + 0.0031·AR + 0.1762·P − 0.0586·V). Background color was used for statistically significant values for 99% confidence interval.

**Table 5.** Summary of linear models with interactions (e.g., WKU\*= 1.96353 + 0.00695·V <sup>2</sup> + 0.08847·P\*V + 0.00032·AR\*V − 0.00745·AR\*P + 0.00526·AR 2 − 0.25323·V − 0.29207·P − 0.05862·AR). Background color was used for statistically significant values for 99% confidence interval.


**Figure 3.** Influence of the main factors: First line (left) WKU\* and (right) WHAZU\*; second line (left) WKD\* and (right) WHAZD\*; and last line (left) WKR\* and (right) B\*.

It follows from the table of results on the analysis of variance for WKU\* for the transformed values of the cutting kerf width (Table 4) that the influence of the cutting speed factor V and the laser power P is statistically significant. The influence of the AR factor was not confirmed by the mathematical-statistical analysis, which is related to the wood density. The reason for this may be the fact that the thermal conductivity of wood only plays a major role in heat transfer in the phase of penetration of the energy beam into the material, while on the surface, respectively in the phase of energy contact with the board surface, the surface properties of wood are more pronounced, e.g., color, roughness, reflectivity, and the properties of the laboratory environment. The cutting speed has the greatest influence on the value of the width size of the cutting kerf on the upper board´s surface. However, this effect is disproportionate, and the width of the cutting kerf WKU\* decreases with an increasing cutting speed. We can explain this phenomenon by the decreasing amount of energy interacting with the board material per unit time. For this reason, it is necessary to include the interactions between the investigated factors in the model, which correlates with the results of Hernandez [65] in the case of Scots pine, with Ready [81] in the case of Silver fir, and with Liu et al. [82] in case of cherry wood. On the contrary, the width of the cutting kerf increases with an increasing laser power, which was confirmed in the work of Nukman et al. [53] in their research on selected Malaysian wood cutting. If we consider the factor AR, then its effect on WKU\*, although negligible, is directly proportional. In the case of the interaction model of factors for WKU\*, in addition of the factors P and V, the interactions P\*V and V<sup>2</sup> are also statistically significant. This confirmed the assumption of the mutually important influence of V and P on the cutting kerf width, so these factors significantly affect the amount of absorbed and reflected energy in the board structure, as pointed out by Hernandez in his work [65] on pine wood cutting, as well as Barnekov [48]. A statistically significant factor AR in the form of the self-interaction AR<sup>2</sup> already appeared in the interactions model. This is indicated by the fact that the wood density affects the properties of the cutting kerf, which is also stated in the work of Asibu [83].

It follows from the results of the analysis of variance for the transformed values of WHAZU\* that the influence of the cutting speed factor V and the laser power factor P is statistically significant. The influence of the factor AR, which is related to the wood density, was not confirmed by the mathematical-statistical analysis. The significant influence of the speed factor V points to the fact that the cutting speed significantly affects the energy distribution in the cutting process by the radiation beam on the workpiece surface at the point of energy interaction with the material. This is consistent with Hernandez's results [65]. This follows from a detailed analysis showing that the cutting speed V and laser power P have the greatest influence on the width value of the heat-affected area on the upper board´s surface in comparison with the other investigated factors. In the case of the cutting speed V, the effect is directly proportional, and the width of the heat-affected area WHAZU\* increases with an increasing cutting speed. The result obtained is in contrast to the findings published in the work of Asibu [83], Barcikowsky [7], and Lum [45]. The reason for this is that these authors examined the heat-affected area below the surface of cut wood, while the subject of our research was the size of the heat-affected area on the board surface. This reduces the amount of energy in the process of decomposing the material below its surface by increasing the cutting speed, as well as increasing the amount of energy on the board surface. Therefore, it is clear that it increases the area affected by heat due to energy excess on the board surface, which is consistent with the results of Arai and Hayashi, [84,85]. The width of the heat-affected zone increases as the power of the laser P increases, as Barcikowski [7] also pointed out. If we also take into account the factor AR, then its effect on WHAZU\*, although also negligible, is directly proportional. All model quality indicators improved after the inclusion of interactions in the model WHAZU\*, but not so significantly that we can consider this model as credible. However, we can use it to explain the influence of the investigated factors and their interactions on the investigated response, in this case, WHAZU\*. The mutual interaction between the laser power P and the cutting speed V determines the degree and form of energy distribution, as in the case of the WKU\* or WHAZU\*; this causes the wood chemical decomposition and thus determines the width of the heat-affected area. The cutting speed also affects the properties at the same time, and in particular, the distribution of the flowing auxiliary gas on the board surface.

It follows from the results of the analysis of variance for the transformed values of the cutting kerf width on the board bottom of the WKD\* that the influence of the cutting speed factor V, the laser power P, and the factor AR is statistically significant. These results are in harmony with Ready [79]. The reason for this is again that the wood thermal conductivity plays an important role in heat transfer in the phase of penetration of the energy beam into the material, while on the surface, respectively in the phase of energy contact with the board surface, the surface properties of wood are manifested analogously, as in the case of WKU\* and WHAZU\*. It is clear that the cutting speed V increase, as well as the density increase, and respectively the number of annual rings AR, results in a reduction

in the cutting kerf width on the board bottom. The results correspond to the results of Mahdavian and Zhou [57] or Lum and Black [45,49] in a cutting depth analysis. On the contrary, an increase in the power P of the laser results in an increase in the value of the cutting kerf width WKD\* on the board bottom, which is in accordance with the published results, e.g., Eltawahni et al. [47]. It was found, by comparing the effect of the factors for both widths WKU\* and WKD\*, that the power factor P and the speed factor V act in the same direction on the width values of the WKU\* and WKD\*. However, the factor AR, respectively the wood density, acts to increase the cutting kerf width on the board surface, in contrast to the effect on the board bottom, where it reduces this width by its influence. The extent value of the variance explanation only improved insignificantly in the case of the interaction model for WKD\*, so the model without interactions sufficiently and simply describes the influence of the factors P, V, and AR on WKD\*.

The effect of the factor AR could not be confirmed by the results of the variance analysis for WHAZD\*; however, a significant effect was confirmed by the cutting speed factors V and laser power P. It is clear that the value of WHAZD\* increases with an increasing laser power P and, conversely, the value of the width of the heat-affected zone on the surface of the board bottom decreases with an increasing speed V. It follows from the results of a detailed analysis that a significant improvement in all properties of the model will not be achieved, even when considering the interactions between the investigated factors. Considering this, it is necessary to include other factors in the analysis in the case of continuing research, e.g., the auxiliary gas flow rate.

It follows from the results of the variance analysis for the transformed values of the width ratio of the cutting kerf WKR\* that the influence of the cutting speed factor V, the laser power P, and also the factor AR is statistically significant. Analogous to WKD\*, the wood thermal conductivity plays an important role in the material, while the wood surface properties are reflected on the surface. A speed increase, as well as density increase, causes the WKR\* value to decrease, while a power increase causes the WKR\* value to increase. The results are in line with the results obtained for the analysis of WKU\* and WKD\*, as well as the works dealing with the joint assessment of the cutting kerf width on the upper and bottom board´s surface in the case of other softwoods, e.g., those by Ready [81] and Eltawahni et al. [47]. It improved all the indicators in the model in the case of the model with interactions for WKR\* and this was mainly due to the mutual interaction between power P and speed V, while the factor of power P and speed V had the most significant effect on WKR\*.

Analysis of variance did not confirm a statistically significant effect of the investigated factors AR, P, and V in the case of B\*. Nevertheless, it is clear from a physical point of view that the power factor P, together with the cutting speed V, i.e., feed of the energy source, determines the amount of energy absorbed by the wood, and thus the amount of energy needed to create the burns. Peters and Banas [86] and Barnekov et al. [48] also reached analogous conclusions. This is due to a decrease in the reaction time of the radiation beam interaction on the workpiece material. It evaporates the material at the upper board´s surface just below the surface due to the uneven distribution of energy supplied to the board material. Below the area, where the material evaporates, is the area where there is less energy, and thus the wood is decomposed by burning. The interactions between the laser power P and the cutting speed V are statistically significant, highlighted by the interaction V\*V in the case of the interaction model.

#### *3.2. Optimal Parameter Settings of the Investigated Process*

It is desirable to minimize the subsequent post processing due to minimizing the cost of the cutting process by the energy beam. It is possible to achieve cost minimization by optimally setting the parameters of the investigated process, in order to achieve the required values of process responses based on a mathematical-statistical model. The goal is to reach a minimum value in the case of WKU\*, WHAZU\*, WKD\*, WHAZD\*, and B\*, and in the case of WKR\*, the goal is to achieve the value 1 and at the same time to find the optimal values of the factors P, V, and AR. There are six polynomial functions

based on Table 5, five of which need to be minimized and one of which needs to be normalized, i.e., needs to be equal to one: · · · · · · · · − ·

− · · · · − · −

− · − · − · · − ·

· · ·

· · ·

− − · − · − · − · ·

**WKU\*** = 1.96353 + 0.00695·V <sup>2</sup> + 0.08847·P\*V + 0.00032·AR\*V − 0.00745·AR\*P + 0.00526·AR<sup>2</sup> − 0.25323·V − 0.29207·P − 0.05862·AR = **min, WHAZU\*** = 0.23073 − 0.00193·V <sup>2</sup> − 0.02749·P\*V − 0.00108·AR\*V + 0.00257·AR\*P − 0.00142·AR<sup>2</sup> + 0.08067·V + 0.19104·P + 0.02398·AR = **min, WKD\*** = −0.13999 − 0.00489·V <sup>2</sup> − 0.02015·P\*V − 0.00208·AR\*V − 0.05651·AR\*P + 0.00034·AR<sup>2</sup> + 0.03693·V + 0.81739·P + 0.06392·AR = **min, WHAZD\*** = −0.78817 + 0.00099·V <sup>2</sup> + 0.01778·P\*V + 0.00121·AR\*V + 0.02135·AR\*P − 0.00349·AR<sup>2</sup> − 0.05375·V + 0.00574·P + 0.01209·AR = **min, B\*** = 1.04444 + 0.00467·V <sup>2</sup> + 0.06385·P\*V + 0.00172·AR\*V + 0.01333·AR\*P + 0.00107·AR<sup>2</sup> − 0.15057·V − 0.51849·P - 0.03579·AR = **min,** − · · − − · − · − · − · − · · · · <sup>−</sup> .

**WKR\*** = −1.01551 − 0.09964·V <sup>2</sup> − 0.09821·P\*V − 0.00352·AR\*V − 0.05468·AR\*P − 0.00263·AR<sup>2</sup> + 0.23656·V + 1.17681·P + 0.10382·AR = **1.**

The modeling was performed by the method of linear programming, which determined the optimal values of the investigated factors (based on a set of solutions of a system of six linear equations and inequations using the simplex method of solving the general problem of linear programming).

The results of joint optimization of the investigated factor settings reached values of AR = 3, P = 100 W, and v = 9 mm·s −1 . The suitability of the joint optimization reached the value of 0.92. We can conclude that we can approach the optimal parameters of the cutting process of a given board by increasing the cutting speed at the optimal performance based on the analysis of the response area (Figure 4 left). It is possible to achieve the same or a similar result by reducing the power of the radiation beam at the optimum cutting speed of a given board (Figure 4 right).

**Figure 4.** Response areas for joint optimization in the case of fixed power P = 100 W (**left**) and in the case of fixed cutting speed V = 9 mm·s −1 (**right**).

−

−

− The correctness of the optimization was also confirmed by Table 6. From this table, it is clear to see that the chosen parameters of cutting after optimization (AR = 3, P = 100 W, and v = 9 mm·s −1 ) are improved over the reference parameters of cutting (AR = 3, P = 150 W, and v = 9 mm·s −1 ) by 3% to 30%.

−


**Table 6.** A table presenting the parameters at 150 W power (reference) and 100 W power (optimal), for a cutting speed of 9 mm·s <sup>−</sup><sup>1</sup> and AR = 3.

In case we need to optimize only one parameter separately, we can achieve even better results, since, in the case of joint optimization, the values are related to each other. The obtained values of separate optimization for all parameters are given in Table 7. In practice, in the case of laser wood cutting, it is not always necessary to achieve the optimized quality of all parameters at the same time. In some cases, with specific requirements for the cutting zone created in laser cutting, optimization of the cutting zone for only one parameter is needed, e.g., the maximum surface flatness is required for subsequent gluing (WKR = 1); in case of increased requirements for the tensile shear strength of the glue joint, minimization of charring is required (B = min); and in the case of subsequent edging and the requirement of one viewing area, minimization of the HAZ is required (WHAZxU alebo WHAZxD = min). In some cases, the perfect surface is counterproductive (in the case of adhesion of the glue, etc.). Alternatively, there is a requirement to combine the optimization of two parameters, e.g., at subsequent edging and the requirement of two viewing surfaces (WHAZxU and WHAZxD = min). These requirements are applied, especially in the case of wood composites (plywood, particleboard, fiberboard, etc.), in the production of furniture elements. In practice, in most cases, the second optimization is more important due to the above reasons.


**Table 7.** Results of separate optimization of the investigated factor settings.

The obtained results show a high correlation with experimentally measured results, despite the difficulties largely resulting from the anisotropy of spruce wood, and at the same time, they are comparable with the results of authors investigating cutting by a radiation beam generated by a CO<sup>2</sup> laser for other wood species, as well as wood composites [53,65,81–83].

#### **4. Conclusions**

The goal of our research was to find the optimal combination of parameters of the cutting process with a beam of a CO<sup>2</sup> laser energy of a 8 mm thick spruce board with a moisture content of 12% and thus to achieve the required quality of cutting. The targets of past research were mainly equivalent homogeneous wood material. The target of this research was highly anisotropic material. Therefore, parameter AR was considered and discussed, in addition to standard parameters, such as the effect of the cutting speed V and beam power output P, and the influence of these parameters on responses was also analyzed too, including the cutting kerf width on the upper board´s surface WKU, cutting kerf width on the board bottom WKD, the width of the heat-affected area on the upper board's surface WHAZU and the board bottom WHAZD, the ratio of the cutting kerf width on the board bottom to the width on the upper board´s surface WKR, and the degree of char B. Using a mathematical-statistical model, we achieved the optimal parameters (AR = 3, 100 W, and 9 mm·s −1 ). These parameters of cutting are much better than the non-optimized control parameters (AR = 3, 150 W, and 9 mm·s −1 ). The improvement in the quality ranged from 3% to more than 30%. The results exhibit a high correlation with experimental measurements and they are also comparable to the results of research on the cutting of more homogeneous wood and wood composite materials cut with a CO<sup>2</sup> laser.

Research has further shown that even inhomogeneous materials such as spruce wood can be cut with a low power laser. However, in order to make this process more predictable, further research should focus on assessing the impact of other factors, such as the spruce moisture content and composition and flow rate of the auxiliary gas, and extending the ranges of factors considered, especially the beam power output and the cutting speed. We can use the method of measuring the cutting kerf width in other areas, where we require a fast and inexpensive method to measure small lengths, which are determined by the resolution of the scanning device. At the same time, this method is suitable for the proposal of electronic equipment that would perform the measurements of small two-dimensional objects. The used method of char measuring can be applied in the mass identification of errors or characteristic properties of the surface of the scanned workpiece by division into characteristic areas.

As mentioned in the introduction, from a comprehensive point of view, the cutting process of wood and wood-based products with a CO<sup>2</sup> laser can be evaluated as a fast, highly-automated, and workpiece-friendly approach that is suitable for cutting homogeneous and inhomogeneous wood and wood composite materials. Other benefits include the highly precise cut, the narrow kerf width, the smooth surface, no tool wear, and the reduced amount of sawdust.

Innovative solutions are essential for sustaining the market position in the competitive business environment. This clearly supports the position of the wood-working industry among the sectors completely fulfilling the requirements of green business products and principles of sustainable development [87–91].

**Author Contributions:** Conceptualization, I.K. and R.I.; methodology, M.G.; formal analysis, L'.K. and J.S.; investigation, J.S. and M.G.; resources, L'.K.; data curation, I.R.; writing—original draft preparation, L'.K. and R.R.; writing—review and editing, L'.K. and R.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Slovak Research and Development Agency under contract no. APVV-18-0378, APVV-19-0269, and VEGA 1/0717/19.

**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* **Tropical Wood Dusts—Granulometry, Morfology and Ignition Temperature**

#### **Miroslava Vandliˇcková, Iveta Marková \* , Linda Makovická Osvaldová , Stanislava Gašpercová, Jozef Svetlík and Jozef Vraniak**

Department of Fire Engineering, Faculty of Security Engineering, University of Žilina, Univerzitná 8215/1, 010 26 Žilina, Slovakia; miroslava.vandlickova@fbi.uniza.sk (M.V.); linda.makovicka@fbi.uniza.sk (L.M.O.); stanislava.gaspercova@fbi.uniza.sk (S.G.); jozef.svetlik@fbi.uniza.sk (J.S.); vraniak.jojo@gmail.com (J.V.)

**\*** Correspondence: iveta.markova@fbi.uniza.sk; Tel.: +421-41-513-6799

Received: 26 September 2020; Accepted: 26 October 2020; Published: 28 October 2020

**Abstract:** The article considers the granulometric analysis of selected samples of tropical wood dust from cumaru (*Dipteryx odorata*), padauk (*Pterocarpus soyauxii*), ebony (*Diospyros crassiflora*), and marblewood (*Marmaroxylon racemosum*) using a Makita 9556CR 1400 W grinder and K36 sandpaper, for the purpose of selecting the percentages of the various fractions (<63; 63; 71; 200; 315; 500 µm) of wood dust samples. Tropical wood dust samples were made using a hand orbital sander Makita 9556CR 1400 W, and sized using the automatic mesh vibratory sieve machine Retsch AS 200 control. Most dust particles (between 50–79%) from all wood samples were under 100 µm in size. This higher percentage is associated with the risk of inhaling the dust, causing damage to the respiratory system, and the risk of a dust-air explosive mixture. Results of granulometric fractions contribution of tropical woods sanding dust were similar. Ignition temperature was changed by particle sizes, and decreased with a decrease in particle sizes. We found that marblewood has the highest minimum ignition temperature (400–420 ◦C), and padauk has the lowest (370–390 ◦C).

**Keywords:** tropical wood dust; granulometric sieve analysis; morphology shape of particles; temperature of ignition

#### **1. Introduction**

During wood processing, dust is created as a by-product [1–7], and plays a negative role in assessing the risk of fire [8] or explosion [9–14]. Wood fust also poses a significant risk to the health of the human body [15,16].

The damaging effect of wood dust is determined according to the particle size. Larger size fractions tend to settle [17,18], whereas if the particle is smaller (such as below 100 µm), the dust becomes airborne. In the production process, dust is formed that contain particles of various sizes [5]. A granulometric analysis determines the degree of crushing of the base material—which is one of the characteristic abilities of form airborne dust mixture [8].

Reinprecht et al. [19] classify tropical woods as wood that demonstrates significant resistance to biological agents and machine wear and tear, together with solid dimensional stability and pretty aesthetics. These types of wood are commonly used for exterior constructions, tiles, garden furniture, or special plywood [20–22]. The expansion of their processing brings the creation of their wood dust.

Reinprecht et al. [23] made a detail analysis of seven types of tropical woods: Kusia (*Nauclea diderichii* Merill), bangkirai (*Shorea obtusa* Wall; Sh. spp.), massaranduba (*Manilkara bidentata* A. Chev.; M. spp.), jatobá (*Hymenaea courbaril* L.), ipé (*Tebebuia serratifolia* Nichols.; T. spp.), cumaru (*Dipteryx odorata* (Aubl.) Wild.)—996 kg·m−<sup>3</sup> , and cumaru rosa (*Dipteryx magnifica* (Ducke))—1014 kg·m−<sup>3</sup> . The samples were studies during different weather conditions, using a 36-mount in the exterior, and results showed the lowest lightening cumaru samples than others.

Tropical woods have natural durability [24]. There are resistant to decay fungi [19], insects, and dimensional changes. This is due to extractants that have a biocidal effect, such as coumarins, flavonoids, and tannins, and a hydrophobic effect, such as fats, oils, and waxes [25,26]. Giraldo et al. [27] declare that morphology, together with differences in the inorganic constituents, significantly affects the combustion process of wood.

The major parameter to assess the risk of airborne dust ignition is ignition temperature. Ignition temperature is closely surveyed using standardized equipment [28], where the airborne tropical wood dust is loaded heat. The ignition temperature (SIT) is the lowest temperature at which, under the defined test conditions, ignition occurs by heating, without the presence of any additional flame source [29].

This study aims to examine and seek comparative similarities between the granulometric structure of wood sanding dust from cumaru (*Dipteryx odorata*), padouk (*Pterocarpus soyauxii*), ebony (*Diospyros crassiflora*), and marblewood (*Marmaroxylon racemosum*). The prepared tropical dust was analyzed for the purpose of identifying its morphological structure, and determining the given physical properties (average dust moisture and bulk density). We have focused on microfractions of tropical wood dust (size of particles ≤100 µm), and focused the minimum particle size (<100 µm) required to cause ignition within airborne tropical dusts.

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

#### *2.1. Samples—Tropical Woods*

Four samples of wood dust from foreign wood species were used for this study. Samples of tropical wood were selected while considering their use in the production of floor coverings, furniture, and interior decorative items (Table 1).


**Table 1.** Samples used in the experiment.

<sup>1</sup> Association Technique Internationale des Bois Tropicaux (ATIBT) in France.

Cumaru (*Dipteryx odorata*), together with abiurana (*Pouteria guianensis*), garapeira (*Apuleia molaris*), jequitiba (*Cariniana* sp.), Cedro (*Cedrela odorata*), angelim (*Parkia pendula*), angelim pedra (*Hymenolobium excelsum*), and cerejeira (*Amburana acreana*) belongs to the group of Amazonian woody plants [30,31]. The cumaru tree is very dense (950–1000 kg·m−<sup>3</sup> ), tough, highly durable, and resistant to cracking when exposed to sunlight. Therefore, it is suitable for solid flooring, stair treads, furniture, and pool decks [32]. Moreover, it is frequently found in the states of Acre, Amapá, Amazonas, Pará, Rondônia, and Mato Grosso, as well as in neighboring countries like Guyana, Venezuela, Colombia, Bolivia, Peru, and Suriname [32].

Padauk (*Pterocarpus soyauxii*) is moderately heavy, strong, and stiff, with exceptional stability. It is a popular hardwood among hobbyist woodworkers because of its unique color and low cost. It has a unique reddish-orange coloration, and the wood is sometimes referred to by the name 'Vermillion'. It is commonly used in flooring, musical instruments/objects, tool handles, furniture, other small particular wood objects, and as veneer [33].

Ebony woods have many common names, such as Gabon Ebony, African Ebony, Nigerian Ebony, and Cameroon Ebony, and originate from the equator part of Western Africa. Heartwood (955 kg·m−<sup>3</sup> ) is

usually jet-black, with little to no variation or visible grain. Occasionally, dark brown or grayish-brown streaks may be present. Ebony's common uses small items, such as piano keys, musical instrument parts, pool cues, carvings, and other small specialty items [34]. Ebony is highly valued in the Hindu religion as a building material [35], base material [36,37], and for use in other items [38].

Marblewood, also known as Angelim Rajado, is distributed from Northeastern South America. This heartwood (1005 kg·m−<sup>3</sup> ) is yellow to golden brown, with irregular brown, purple, or black streaks [39]. It is commonly used for tuned instruments/objects, flooring, carpentry, sliced veneer, and delicate furniture. Vivek et al. [40] introduced the possibility of using marblewood dust as a partial replacement for cement and sand in concrete.

The basic samples (Figure 1a) of cumaru (152 × 38 × 38 mm), Padouk (131 × 50 × 20 mm), ebony (142 × 36 × 30 mm), and marblewood (120 × 35 × 35 mm) were made by a private wood company in Žilina (Slovakia) by a wood cutting saw (CNC Panel Saw Machine, Shandong, China) (Table 1). The moisture content of the basic samples was approximately 8–10%.

**Figure 1.** Experimental samples (**a**) in the form of plates; (**b**) in the form of prepared wood dust. The legend: (1) Cumaru (*Dipteryx odorata*); (2) African Parakeet (*Pterocarpus soyauxii*); (3) African Ebony (*Diospyros crassiflora*), and (4) Marblewood (*Marmaroxylon racemosum*).

#### *2.2. Preparation of Tropical Wood Dust Samples*

Technical equipment was used to prepare the homogeneous dust particles dust samples (Figure 1b) was Makita 9556CR 1400 W disc sander (Makita Numazu Corp., Branesti Ilfov, Romania) and K36

sandpaper (Topex, Kinekus, Žilina, Slovakia). The grinding was carried out by the Experimental Laboratory at the Faculty of Security Engineering, University of Žilina (Slovakia). Tropical wood dust samples were prepared by a specialist in grinding. We aimed to ensure the grinding process was as close to reality as it is possible (in terms of pressure of the grinding surface of the component, grinding speed, and grinding direction (cross)). The prepared dust particles were amassed in the hopper with a hermetically sealed glass container, because it needed to stop the dust from absorbing any moisture. After the samples were ground three times, the hopper always was cleaned (in between samples). Overall, 300 g of dust was amassed from each board (three boards collection), and served for the granulometric examination. Detailed information about the preparation of dust samples can be found in Reference [29].

#### *2.3. Experimental Methods Utilised for Characteristics of Tropical Wood Dusts and Sieve Analysis*

The moisture of the tropical wood samples determined according to Reference [41], and the bulk density determined according to Reference [42], were selected characteristics of samples before sieve analysis (Table 2). The Retsch AS 200 sieve shaker vibration machine (Retsch AS 200 control, Retsch GmbH, Haan, Germany), using the seven fraction sizes (500, 315, 200, 100, 71, 63, and <63 µm), was used for sieve analysis [43]. The sieves, in addition to the tropical wood dust, weighed 30 g (using laboratory scales with precision readings of 0.001 g). The measurement procedures were conducted five times lasting 10 min. The wood dust moisture testing was carried out according to Reference [44]—in a heated oven, at a temperature of 103 ± 2 ◦C for 24 h (Table 2).

**Table 2.** Basic physical parameters of samples used in the experiment.


#### *2.4. Experimental Methods for Shape of Wood Dust Particles*

The size, shape, and form of wood dust particles were studied by microscopic analysis, through a wide-field microscopy system (Nikon Eclipse Ni (Nikon Corp., Tokyo, Japan)) with a Nikon DS-Fi2 camera (Nikon Instruments Inc., Melville, NY, USA) [45]. This microscope possesses two c-mount camera ports, and an electric XY stage. This is the normal configuration for the bright field observation of this microscope.

A Nikon DS-Fi2 full-HD color camera (Nikon Instruments Inc., Melville, NY, USA), that uses one port, 2560 × 1920 pixels, was also used. This camera was used to establish the image by connecting it to the computer using a USB cable. Microscopic analyses of tropical wood dusts were performed using 100 µm fraction precision.

The Institute of Research in Banská Bystrica, Slovakia, has a Nikon SMZ 1270 stereomicroscope (Nikon Corp., Beijing, China), which was used to analyze the shape of the dust particles. Equipment for the research of dust particles shape has a range with a 12.7:1 zoom head, and a magnification range of 0.63× to 8×. The samples of the 500 µm, 315 µm, 200 µm, and 100 µm fractions were measured in this stereomicroscope.

#### *2.5. Experimentation So as to Determine the Ignition Temperature Measurement of Airborne Dusts*

Specific test equipment (the detail of which can be found in Reference [29]) (VVUÚ, a.s., Ostrava, Czech Republic), with automatic weighing machines (Steinberg Systems, Łód´z, Poland), an air compressor HL 100 ZU EINHELL (Einhell Corp., Landau an der Isar, Germany), and ALMEMO equipment (Ahlborn, Berlin, Germany) was used to the measurement of minimum ignition temperatures of airborne dusts (Figure 1a) according to the standard instructions [28]. The details of this are described in Vandliˇcková et al. [29].

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

The average bulk density (kg·m−<sup>3</sup> ) and dust moisture (%) were physical parameters connected with the preparation of dust samples (Table 2). Poorter et al. [46] consider of the density of tropical woods and the influence of climatic conditions on the growth and quality of wood mass. Tropical tree cumaro is considered a hardwood [47–50]. Soriano et al. [51] determined the density of cumoro in the range of 1060–1070 kg·m−<sup>3</sup> . Marblewood is also ranked among the hardwoods. Everything about working with marblewood revolves around its incredible hardness and density [52].

Ebony is different—it has a lower density (960 kg·m−<sup>3</sup> ), and the highest value of average bulk density (206.5 kg·m−<sup>3</sup> ).

King et al. [53] performed research on the growth of tropical trees in the Amazon and investigated their physical properties. The research samples included padauk trees, namely, *Dipterocarpus cornutus Dyer.* With a density of (680 ± 0.013) kg·m−<sup>3</sup> and *Dipterocarpus globosus Vesque* with a density of (690 ± 0.017) kg·m−<sup>3</sup> .

#### *3.1. Results of Particle Sizes (Fractions) of Dust Samples*

Sanding dust particulates are minute and of complex composition, and are more hazardous to humans than the cutting dust particulates [54]. Wood dust particles (Figure 2) are normally generated in different sizes [12,55]. The percentages of particle sizes start from the size of 500 µm (Table 3). Larger fractions occurred only at a minimal rate (at 1%). Oˇckajová and Marková [56]; Oˇckajová et al. [5,57] presented the same result, which was presented for the chosen domestic tree samples.

**Figure 2.** Continuous cumulative curve of cumaru, padauk, ebony, and marblewood dust.


**Table 3.** Determination of % particle number fractions of the tropical wood dusts.

The dust particles prepared from African Padauk had similar composition % particle number, as well as spruce, oak, and beechwood [29]. č

Oˇckajová et al. [5] studied the size of fraction particles and share of wood dust particles (beech, oak, and spruce). Their solution formulated that the percentage share of dust particles is also very different, depending on the kind of this tree species. Our results showed similar size of particles (Table 3). Oˇckajová et al. [6] found a connection between the shares and densities of the wood dust. Their analyses regarding the wood dust in the sanding process showed that more dust is produced as the density of wood increases [54]. č

The results of dust sieving are presented by cumulative curves (Figure 2). The results are presented in passed weight percent of the individual fractions collected on a sieve with the appropriate mesh size. The results of the sieve analysis should be presented according to the appropriate standards [58,59]. Prepared continuous cumulative curves are presented in Figure 2 with complete mesh size.

#### *3.2. Shape Analysis of Dust Particles*

The characterization of wood dust samples from a morphological (Figures 3–6), and a dimensional point of view, yields information that can help epidemiologists and toxicologists to understand the causes of respiratory illnesses [60].

**Figure 3.** Cumaru dust particles and their shape.

**Figure 4.** Ebony dust particles and their shape.

**Figure 5.** Padauk dust particles and their shape.

**Figure 6.** Marblewood dust particles and their shape.

μ μ μ The particles in our sample show a range of shapes and sizes of particles [61]. The size 100 µm is the boundary value of a particle size, where dust is expected to becomes airborne and is potentially explosive [62,63]. Within the basic range of size particles, dust can be classified as coarse (particle diameter of >100 µm) or fine (particle diameter of <100 µm) [64].

μ The anatomical structure is preserved in analyzed particles [65]. When magnified, the fractions of 500, 315, 200, and 100 µm ( Figures 3–6) appeared differently—they had their own specific shapes. Figure 3 shows the fractions of cumaru, and Figure 6 fractions of marblewood.

Selected tropical woods produce dust particles within the whole spectrum. The differences in morphology are shown in Figures 3–6. Cumaru and marblewood are hardwoods, with a density of 1000 gm-m-3, and it is possible to state the similarity of the formed particles. The relationship between the particle shape and the initiation temperature can be shown in marblewood dust, which has the highest ignition temperature. Ebony offers a limited or bounded particle shape, and its ignition temperature is lower.

#### *3.3. Microscopic Dust Analysis*

μ μ The particles of tropical dust samples (Figure 7) have the size of <100 µm. From the safety and occupational hygiene perspective, particles below 100 µm are the most dangerous in the working environment [2]. The scans clearly show that the fibers of the dust particles maintain their anatomical structure with the fibrous character of particles [66]. The two-dimensional pictures of particles can be used for determining the smallest particle sizes.

μ **Figure 7.** Light microscopic images of tropical dust fibers with 100×, 200×, and 400× zoom. Legend: Blue line presents a size of 100 microns (µm) in a 2D layout.

Picture of wood samples, after sanding, taken by an electron microscope, show different and complex shapes of particles of wood dust (Figure 7). As described in the Particle Atlas [67], the diverse geometric expression could be observed, such as cylinders, cones, rectangular prisms, and spheres [60].

Mazzoli et al. [60] provided microscope analyses of dusts from two hardwoods (sessile oak, oak-tree), two tropical hardwoods (padouk, iroko), and three softwoods (pine, spruce, and larch). These were obtained using a grinding machine with 360-grit sanding paper.

μ μ μ Gómez Yepes and Cremades [61] analyzed the particle characteristics in Quindio (size distributions, aerodynamic equivalent diameter (*Da*), elemental composition, and shape factors), and particles were then characterized via scanning electron microscopy (*SEM*) in conjunction with energy dispersive *X-ray* analysis. Results from their analysis of particulate matter showed that the cone-shaped particle ranged from 2.09 to 48.79 µm *Da*; the rectangular prism-shaped particle from 2.47 to 72.9 µm *Da*; the cylindrically-shaped particle from 2.5 to 48.79 µm *Da*; and the spherically-shaped particle from 2.61 to 51.93 µm *Da*.

μ μ Oak dust provides a similar comparison with marblewood dust, as oak is the hardest Slovak wood. In all investigated oak fractions, the isometric shape of particles with sharp edges and rounded corners, which is more typical for fraction <100 µm [3].

#### *3.4. Ignition Temperature of Airborne Dust*

Vandliˇckova et al. [29] studied the processes of ignition within airborne wood dusts. They [29] studied the specificities of the dust ignition process in different stages. The first stage is the beginning of an explosion: The dust was sprinkled into a ceramic tube furnace, with dust carbon residue. In the second stage, the unburned wood dust was ignited with the most incredibly intense flame. Subsequently, in the third state, flame slowly diminished with side effects.

The ebony dust of 500 µm, 315 µm, and 71 µm fractions were obtained the maximum flame (Figure 8). It was noted that when using larger particles, the flame had a lower intensity.

**Figure 8.** The flame on ignition (conditions: 0.2 g of dust, air pressure of 30 kPa, the temperature of the ceramic tube furnace 500 ◦C). Images were prepared using a Basler a602fc-2 high-speed camera (Basler AG, Ahrensburg, Germany). Legend: (**a**) Point of ignition for fraction 500 µm of ebony dust; (**b**) point of ignition for fraction 315 µm of ebony dust; (**c**) point of ignition for fraction 71 µm of ebony dust.

These results determined that the ignition temperature of airborne dust (Figure 9) showed a reduction in values relating to the smaller size fraction [68].

The minimum ignition temperature of cumaru for the 500 µm fraction was 410 ◦C, and this temperature decreased with changing particle size.

Particles of 100 µm or less had the most significant effect on the change in the minimum temperature. The minimum ignition temperature of the ebony dust particles 500 µm, and 200 µm was identical (400 ◦C). The 100 µm particle fraction had a minimum ignition temperature of 380 ◦C. The last three size fractions had minimum ignition temperatures, due to strong dispersion in the heating furnace. An assumption was made applied to a larger dispersion of particles in a space for larger particles (of 500 µm), compared to 100 µm with the same weight of the batch.

Marblewood had the highest tested minimum ignition temperatures of all samples. The minimum ignition temperatures started from 400 ◦C at a particle size of 500 µm to 100 µm. Subsequently, the minimum ignition temperatures decreased with the particle size change to the level of 400 ◦C, at a particle size of 63 µm.

Flame propagation behaviors and temperature characteristics of four types of biomass (poplar, pine, peanut, and corn sawdust particles) with two different particle sizes (50–70 µm and 100–200 µm) distributions were studied experimentally by Jiang et al. [69]. The average flame propagation velocity and the amplitude of the velocity fluctuation are functions of the mass density of the biomass particles and depend on the particle size distributions [69].

Tropical woods came to a global market for their use in various products. Information about tropical wood fire parameters is poor. Fire parameters are not described in Safety Data Sheets [70,71]. Carrasco et al. [72] studied the heat transfer in Brazilian woods, using a sample that was thermally loaded to examine the potential for fire. Their results showed that there are curves of temperature on the time experiments that create the thickness *f* chair layer corresponding with Carrasco's numerical models.

**Figure 9.** Ascertaining of the ignition temperature fractions of the tropical wood dusts.

#### **4. Conclusions**

From the perspective of this study, we can claim that we have obtained original results within the field of morphology of dust particles regarding the wood sanding process of tropical four species, as well as results concerning the determination of minimal ignition temperature airborne dust.

Our results show that fraction sizes of tropical wood dust are an important factor for fire ignition. However, the shape and morphology of tropical wood dust particles may also exert influence ignition process. All these aspects deserve to be studied further.

Marblewood has the highest minimum ignition temperature. Thermal stability can be found in its hardness and density.

Cumaru has specific behavior. It is a hardwood with a density comparable to marblewood, but the minimum initiation temperature decreases significantly with decreasing particle size (fraction <63 µm has a temperature of 370 ◦C).

The monitored dust particles retain their anatomical structure; the shapes of individual samples are different. These results offer different particle shapes such as: Rectangular prism-shaped particle, cylindrically-shaped particle and spherically-shaped particle.

**Author Contributions:** Conceptualization, M.V. and I.M.; methodology, investigation and resources S.G., J.S. and J.V.; writing—original draft preparation, I.M.; writing—review and editing V.M., L.M.O. and I.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This article was supported by the Cultural and Educational Grant Agency of the Ministry of Education, Science, Research and sport of the Slovak Republic on the basis of the project KEGA 0014UKF-4/2020 Innovative learning e-modules for safety in dual education.

**Acknowledgments:** This article was supported by the Project KEGA 0014UKF-4/2020 Innovative learning e-modules for safety in dual education.

**Conflicts of Interest:** The founding sponsors had no role in the design of the study, in the collection, analyses, or interpretation of data; in the writing of the manuscript and in the decision to publish the results.

#### **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*
