**E**ff**ect of Thermal Treatment of Birch Wood by Saturated Water Vapor on Granulometric Composition of Chips from Sawing and Milling Processes from the Point of View of Its Processing to Composites**

#### **Richard Kminiak 1,\* , Kazimierz A. Orlowski <sup>2</sup> , Ladislav Dzurenda <sup>1</sup> , Daniel Chuchala <sup>2</sup> and Adrián Banski <sup>1</sup>**


Received: 1 October 2020; Accepted: 23 October 2020; Published: 27 October 2020

**Abstract:** The goal of this work is to investigate the impact of thermal modification of birch wood with saturated steam on the particle size distribution of the sawing and milling process. Birch wood (Betula pendula Roth) is an excellent source to produce plywood boards. Wastes from mechanical processing of birch wood are suitable to produce composite materials. Granulometric analyses of chips from sawing processes on the PRW 15M frame saw, as well as on the 5-axis CNC machining centre SCM TECH Z5 and the 5-axis CNC machining centre AX320 Pinnacle, proved that more than 95% of chips are chips of coarse and medium coarse chip fractions with dimensions above 0.125 mm. Depending on the shape, coarse and medium-thick chips belong to the group of fiber chips, the length of which is several times greater than the width and thickness. Fine fractions with dimensions smaller than 125 µm are isometric chips that are approximately the same size in all three dimensions. Thoracic dust fractions below 30 µm were not measured. The performed analyses showed that the heat treatment of birch wood with saturated steam did not affect the grain size of chips formed in sawing and milling processes on CNC machining centre and can be used as a raw material for the production of composite materials. Fabric filters are suitable for separating chips extracted from frame saws, PRW-15M or machining centre. Environmental criteria for the separation of chips from transport air in textile filters are met by filters with a fabric classified in class G4.

**Keywords:** birch wood; chips; wood composites; granulometric composition of sawdust and chips; air handling; ecological filtration

#### **1. Introduction**

Thermal treatment of wood with saturated water vapor, in addition to the targeted physico-mechanical changes of wood, is often used in the production of veneer and plywood, bent furniture, or pressed wood. The mentioned treatment is also frequently accompanied by chemical reactions causes the colour changing of wood [1–7]. While in the past, the colour changes of the darkening of thermally treated wood were used to eliminate undesirable colour differences between light beige and dark kernel, or to remove unwanted colour spots caused by evaporation, browning or molding, in recent times the research focuses on targeted changes in wood colour to more

or less pronounced colour shades, namely wood imitations of domestic trees as exotic trees [8–12]. The influence of the hydro-thermal modification process on wood machinability was also investigated by Sandak et al. [13] with four minor species such as black poplar (*Populus nigra* L.), deodar cedar (*Cedrus deodara* Roxb.), black pine (*Pinus nigra* Arnold.) and alder (*Alnus cordata* Loisel) the sharpness of the tool has a lower importance for the final surface smoothness.

Chips from sawing and milling processes were characterized in the literature as a polydisperse bulk material consisting of coarse and medium coarse fractions with a grain size of 0.5–3.5 mm, while the proportion of fine (dusty) fractions with smaller particle sizes below 300 µm is not excluded [14–30]. Dzurenda and Orlowski [31] examined sawdust of thermally modified ash wood obtained during sawing on a sash gang saw PRW-15M, and they revealed increasing of chip homogeneity in the range of granularity *a* = 250 µm–2.4 mm. A combination of wood chips and pine sawdust (*Pinus Sylvestris* L.) with a percentage of sawdust of up to 50% positively affects the board quality by making its structure more homogeneous [32].

The geometry of wood elements used in composites production is one of the important factors affecting board properties. Therefore, the research of birch wood after its thermal treatment on granulometric composition of wood particles from sawing and milling processes is substantial when we consider how large volumes of birch wood are processed in the sawmill industry and in the formatting of plywood boards around the world (Youngquist [32], Maloney [33], and Istek et al. [34]). According to these studies, bending and modulus of elasticity as well as dimensional stability of composites improve with appropriate elements length used for composite production.

The aim of this work was to determine the effect of thermal treatment-modification of the colour of birch wood (*Betula pendula* Roth) by saturated water vapor upon grain size from cutting processes of thermally modified birch wood on different machine tools as follows: the sash gang saw PRW-15M, the 5-axis CNC machining centre SCM TECH Z5 and AX320 Pinnacle. Furthermore, specification of separation technique requirements for the mentioned processes have to be defined.

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

#### *2.1. Material*

Birch wood (*Betula pendula* Roth) in the form of tangential boards with dimensions: 40 (radial direction) × 80 (tangential direction) × 600 mm (longitudinal direction), in total 180 samples, were divided into 3 groups consisting of 60 samples each. The initial moisture content MC of wet birch was in the range MC = 54.7–58.2%. The samples in the Group 1 were not heat treated. The boards of the second group were heat treated with the MODE I and the blanks of third group were heat treated with the MODE II. Thermal treatment of birch wood with saturated water vapor was carried out in a pressure autoclave APDZ 240 (Himmasch AD, Haskovo, Bulgaria) installed at Sundermann Ltd. in Banska Stiavnica (Slovakia).

#### *2.2. Thermal Treatment*

The process of thermal treatment of birch timber with saturated water vapor (steam) is shown in Figure 1, and the technical parameters of each mode are given in Table 1.

**Figure 1.** Mode of colour modification of birch wood with saturated water steam.

**Table 1.** Regimes for thermal treatment of the birch wood using saturated water vapor.


The temperatures *t*max and *t*min in Figure 1 are the intervals between which saturated water vapor is fed into the autoclave to carry out the technological process. Temperature *t*<sup>4</sup> is a parameter of the saturated water vapor pressure in the autoclave to which the autoclave vapor pressure must be reduced before the pressure equipment is safely opened.

Subsequently, thermally treated and not-thermally treated blanks were dried by a low-temperature regime without changing the colour of the wood to a moisture content MC = 12 ± 0.5% in a conventional hot-air dryer: KC 1/50 (SUSAR s.r.o). After the blanks were dried, a portion of the blanks from each group was used to produce *ST* = 5 mm thick slats on a PRW-15M, a portion of the blanks was machined on the 5 axis machining centre AX320 Pinnacle and a part of the blanks was milled on a machining centre SCM Tech Z5.

#### *2.3. Characteristics of Machine Tools and Milling Cutters*

#### 2.3.1. Narrow-Kerf Frame Sawing Machine (Sash Gang Saw) PRW–15M

The lamellae were made from birch thermally untreated and treated wood on the frame sawing machine PRW–15M with the hybrid dynamically balanced drive system and elliptical tooth trajectory movement [35] at the Department of Manufacturing and Production Engineering (Gda ´nsk University of Technology, PL).

ń − The sawing process for both type of materials was carried out with average cutting speed *v<sup>c</sup>* = 3.69 m·s <sup>−</sup><sup>1</sup> and feed per tooth *f* <sup>z</sup> = 0.14 mm. In the case of the frame sawing process, can be assumed that the value of feed per tooth is equal to the value of average uncut chip thickness, *f* <sup>z</sup> = *h*av = 0.14 mm (Figure 2). The saw blades were sharp, with Stellite tipped teeth. The other basic parameters of frame saw and saw blades are shown in Table 2.

**Figure 2.** Mode sawing kinematics on the sash gang saw: fz–feed per tooth, s–saw blade thickness, AD–area of the cut, P–pitch, Y, Z and YM, ZM–machine coordinate and setting axes, Yf–f-set coordinate axis, Pfe–working plane.


**Table 2.** Machine tool and tool settings for frame sawing process.

#### 2.3.2. Milling Centre AX320 Pinnacle

− − − After the sawing process on the frame machine, the bigger parts of sawed samples were subjected to a milling process on 5 axis milling centre AX320 Pinnacle also located in GUT laboratory. The main parameters of milling process such as cutting speed *v*<sup>c</sup> = 3.69 m·s −1 (rotational speed *n* = 4405 min−<sup>1</sup> ) and average uncut chip thickness *h*avg = 0.14 mm (feed speed *v*<sup>f</sup> = 2848 mm·min−<sup>1</sup> ) were used the same values as for frame sawing process. The end milling cutter with blades from cemented carbide was used during experimental cutting tests. This cutter was manufactured by ASPI company, Suwalki, Poland and the main geometry dimensions of milling cutter (Figure 3) and main parameters of the milling process are shown in Table 3. The work movements of the tool were performed by the machine tool in accordance with the CNC program on the Heidenhain TNC 640 control system. The standard

vice with jaws length 100 mm was used to fixture of samples (Figure 4). The height of the milling samples *H* = 25 mm was cut depth *a*<sup>p</sup> also (Figure 4).

**Figure 3.** Cutting edge geometry of mill cutter used for wood milling process at the AX320 milling centre.


**Table 3.** Main parameters of milling cutters and milling processes.

**Figure 4.** Machine tool AX320 Pinnacle with equipment for machining **Figure 4.** Machine tool AX320 Pinnacle with equipment for machining tests.

#### 2.3.3. Milling Centre SCM Tech Z5

The birch blanks as well as after the heat treatment by the individual modes were milled on a 5 axis CNC machining centre SCM Tech Z5 manufactured by SCM-group, Rimini, Italy. A positive spiral milling cutter manufactured by IGM under the designation IGM 193 was used in the experiment (Figure 5). The working part of spiral cutter IGM 193 was manufactured with high-quality carbide VHM (Integral HM). The base technical data of this cutter are shown in Table 3.

**Figure 5.** Mill cutter used for wood milling process at the SCM Tech Z5 milling centre.

#### *2.4. Granulometric Analysis of Chips*

For granulometric analyses of sawdust and shavings from the sawing and milling process of thermally untreated and treated birch wood, sawdust and chips samples were taken by isokinetic procedure from the exhaust pipe of the individual machine tools: PRW-15M frame saw, AX 320 Pinnacle and SCM Tech Z5 CNC machining centres, according to ISO 9096 [36].

<sup>−</sup> τ A granulometric composition of the chips was evaluated by sifting. For this purpose, it was used a special set of sieves arranged one above the other (mesh size: 2, 1, 0.5, 0.25, 0.125, 0.063, 0.032 mm, and the bottom), the sieves are placed on a vibration stand of the sifting machine Retsch AS 200c (Retsh GmbH, Haan, Germany). The parameters of sifting were as follows: frequency of sifting interruption of 20 s, amplitude of sieves deflection: 2 mm·g −1 , sifting time: τ = 15 min, weighed sample: 50 g. The granulometric composition was obtained by weighing of the portions remaining on the sieves after sifting on an electronic laboratory scale Radwag 510/C/2 (Radwag Balances and Scales, Radom, Poland), weighing to an accuracy of 0.001 g. The sifting was realized with 3 samples for each combination of parameters.

With the purpose of specifying information about the size of the smallest particles of fine fraction of dry chips a microscopic analysis of granules of fraction of dry chips with the size lower than 500 µm was realized. The proposed analysis of chips was carried out by an optical method–analysis of the picture obtained from the microscope Nikon Optiphot–2 with the objective Nikon 4×. Granules of chips were scanned by three low-cost television CCD cameras HITACHI HV-C20 (RGB 752 × 582 pixel), with horizontal resolution 700 TV lines and evaluated by a software LUCIA-G 4.0 (Laboratory Universal Computer Image Analysis), installed on a PC with the processor Pentium 90 (RAM 32 MB) with the graphic card VGA Matrox Magic under the operation system Windows NT 4.0 Workstation. The program of analysis of picture LUCIA-G enables to identify the individual particles of disintegrated wood material, quantitative determination of individual particles situated in the analyzed picture, and basic information such as width and length of particles, and circularity expressing the measure of deviation of projection of a given chip shape from the projection of the shape of a circle ψ according to equation:

$$
\psi = \frac{4\pi \cdot \mathcal{S}}{P\_p} \tag{1}
$$

where: *S*—surface of particle [m<sup>2</sup> ], *P*p—perimeter of particle [m].

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

The results of the sieve analysis of sawdust sucked off from the frame saw PRW-15M and the chips sucked off from the machining centre SCM Tech Z5, and AX320 Pinnacle from birch shavings with and without heat treatment, are shown in Tables 4–6.



**Table 5.** Chips from AX320 Pinnacle, average values with standard deviations.



By comparing the proportions of the individual sawdust fractions of the PRW-15M frame saw (Table 4), it can be stated that there are no differences in the proportions of the sawdust fractions of natural birch wood and thermally treated birch wood. The distribution of chip fractions according to the distribution is symmetric, with the largest fraction in the grain size range of 0.250–0.500 µm.

Analyses of the shavings of individual sawdust fractions indicate that coarse and medium coarse fractions over 0.5 mm in size from the sawing process of both untreated and treated birch wood on the PRW-15M frame saw belong to the category of polydisperse fibrous materials, rod-shaped with significant elongation in one dimension. The agreement between the granulometric composition of the steam-saturated wood vapor at a temperature of *t* = 125–135 ◦C and the granulometric composition of the native birch wood chips predicts that the thermal treatment of the birch wood does not affect the chip formation process. This finding is similar to the results of temperature analysis in the process of drying beech timber by low and high temperature regimes on sawdust grain size presented by Orlowski et al. [37].

Microscopic analysis of the chip shape of the fractions below 500 µm indicates that the chip size of these fractions by their shape belongs to the group of isometric chips. Chips having approximately the same dimension in all three directions. Circularity is in the range Ψ = 0.7–1.0.

The proportion of fractions of chip extracted from the machining centre AX320 Pinnacle (Table 5), is similar to sawdust from the PRW-15M frame saw. Chips from machining centre AX320 Pinnacle did not show any significant effect of birch wood thermal treatment on the particle size except for the increased proportion of fine fraction of thermally treated birch wood, which increased from 2.1% to 4.4% for thermally treated birch wood mode I, and to 5.5% for the thermally treated wood mode II.

From the chip shape aspect, only chips of a coarse fraction above 1 mm are fibrous chips. Medium and fine fraction chips are isometric chips with a circularity value in the interval: Ψ = 0.7–1.0.

The chips formed at the SCM Tech Z5 machining centre (Table 6) differ according to the proportion of the individual fractions compared to the sawdust formed by sawing the birch wood on the PRW 15M sash gang saw. The coarse fraction chips above 2 mm make up to 2/3 of the total chips produced. Similar representation of the fractions of beech, maple, and oak wood chips at the SCM Tech Z5 machining centre at a removal of *a*<sup>e</sup> = 3 mm and a feed speed *v<sup>f</sup>* = 3 m·min−<sup>1</sup> , and at feed speed *v<sup>f</sup>* = 5 m·min−<sup>1</sup> reported by: Kminiak and Banski [38]. With the feed speed decreasing at this machining centre when milling wood as the authors stated [39], the share of the coarse fraction is increasing at the expense of the medium coarse fraction.

The chip shape of the coarse fraction corresponds to the shape and size of the layer to be cut. Medium and fine fractions of birch wood chips with or without thermal treatment are isometric chips.

A comparison of the dust fraction of the extracted sawdust from the PRW-15M frame sawing machine and the chips extracted from the CNC machining centres suggests that the smallest chips in the extracted bulk wood mass are chips of size *a* = 32 µm (Figure 6). Fractions below 30 µm (thoracic dust or respirable dust) were not detected by measurements.

**Figure 6.** Comparison of sawdust and chips fractions from CNC machining centres (**a**) coarse fraction (**b**) medium coarse fraction (**c**) fine fraction.

μ Criterion for establishing a separation technique requirement for the separation of extracted bulk solids, including dust chips from conveyed air in exhaust systems is the limit of separation of separation technology. By comparing the smallest chips in the conveyed bulk material from individual machines *a* = 32 µm with the separability limit of the separators in Figure 7, the possibility of meeting environmental criteria is demonstrated. Separation limits of selected filters and separators are shown in Table 7.

**Figure 7.** Fractional separation diagram of disintegrated wood mass of individual types of separation technique: A—settling chambers, B—dry mechanical separators, C—fabric filters, D—wet separators, E—electrofiltrators.

**Table 7.** Separation limits for separation technique.


≥ μ Based on data in Table 4 and Figure 6, it follows that for separating chips with dimensions above and ≥32 µm, suitable separation techniques are: wet scrubbers, fabric filters, and electro-scrubbers.

μ For the purpose of separating loose wood from transport air in wood processing plants, the best available technology are filters with G4 fabric having a separation limit value of *a*SL = 10 µm, thus meeting the requirements of EN 779/2012 [40].

Fabric filters are fully sufficient to catch the chips extracted from the sawing and milling process native birch wood, and thermally modified wood with saturated water vapor [38–41].

Analyses of the grain size of chips formed in the processes of sawing dried birch wood on PRW-15M frame saws and milling on CNC machining centers showed that heat treatment of birch wood with saturated steam did not affect the grain size of the chips. Based on this fact, it can be concluded that, just as chips from raw birch wood are used for the production of composite materials, chips from heat treatment birch wood are also a suitable raw material for the production of composite materials. The use of a given chip to produce agglomerated materials prolongs the life cycle of wood and thus contributes to increasing the degree of sustainability.

#### **4. Conclusions**

The following conclusions are drawn from the performed granulometric analyses of the extracted sawdust from the frame saw PRW 15M and the chips of the extracted CNC machining centre in order to evaluate the inlet of the thermal modification of the birch wood by the saturated water steam to modify the wood colour:


μ


**Author Contributions:** Conceptualization, L.D., K.A.O., D.C. and R.K.; methodology, L.D. and D.C.; software, D.C. and R.K.; validation, L.D. and K.A.O.; formal analysis, L.D. and R.K.; investigation, L.D., K.A.O., D.C. and R.K.; resources, L.D., K.A.O., D.C., R.K. and A.B.; data curation, L.D. and K.A.O.; writing—original draft preparation, L.D., K.A.O., D.C., R.K. and A.B.; writing—review and editing, L.D., K.A.O., D.C. and R.K.; visualization, L.D., K.A.O., D.C., R.K. and A.B.; supervision, L.D. and K.A.O.; project administration, L.D., K.A.O. and D.C.; funding acquisition, R.K. and D.C. Please turn to the CRediT taxonomy for the term explanation. Authorship must be limited to those who have contributed substantially to the work reported. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Slovak Research and Development Agency, grant number APVV-17-0456 and by Polish Ministry of Science and Higher Education, decision number 21/E-359/SPUB/SP/2019.

**Acknowledgments:** This experimental research was prepared within the grant project: APVV-17-0456 "Termická modifikácia dreva sýtou vodnou parou za úˇcelom cielenej a stabilnej zmeny farby drevnej hmoty" as the result of work of authors and the considerable assistance of the APVV Agency. The authors gratefully acknowledge the Polish Ministry of Science and Higher Education for funding the maintenance of scientific and research equipment–PRW-15M frame saw (decision no. 21/E-359/SPUB/SP/2019).

**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/).

**Petr Klímek 1,2, Rupert Wimmer 2,3,\* and Peter Meinlschmidt <sup>4</sup>**

3

	- Institute of Wood Technology and Renewable Materials, University of Natural Resources and Life Sciences, 3430 Vienna, Austria

**Abstract:** Cup-plant (*Silphium perfoliatum* L.) stalks were investigated as a potential wood-replacement in particleboards (PBs). Two types of PBs were produced—(1) single-layer and (2) three-layer boards. In the three-layer cup-plant PB, the core layer was made from cup-plant, while the surface layer consisted of spruce particles. The cup-plant as well as spruce control panels were produced with polymeric methylene diphenyl diisocyanate (pMDI) as the adhesive, with the physical and mechanical properties measured to meet class P1 of the European EN 312 standard. For the intrinsic morphology of the particleboards, scanning electron microscopy was applied. Wood-based and cup-plant-based particleboards indicated significant differences in morphology that affect the resulting properties of particleboards. Furthermore, an innovative approach was used in the determination of the pMDI bondline morphology. With a compact Time-of-Flight Secondary Ion Mass analyser, integrated in a multifunctional focused-ion beam scanning-electron-microscope, it was possible to show that the Ga<sup>+</sup> ion source could be detect and visualize in 3D ion molecular clusters specific to pMDI adhesive and wood. Mechanical performance data showed that cup-plant particleboards performed well, even though their properties were below the spruce-made controls. Especially the modulus of rupture (MOR) of the cup-plant PB was lowered by 40%, as compared to the spruce-made control board. Likewise, thickness swelling of cup-plant made boards was higher than the control. Results were linked to the specific porous structure of the cup-plant material. In contrast, it was shown that three-layer cup-plant PB had a higher MOR and also a higher modulus of elasticity, along with lower thickness swelling, compared to its single-layer cup-plant counterpart. The industry relevant finding was that the three-layer PB made from cup-plant stalks fulfilled the EN 312 standard, class P1 (usage in dry conditions). It was shown that raw material mixtures could be useful to improve the mechanical panel performance, also with an altered vertical density profile.

**Keywords:** particleboard; three-layer particleboard; cup plant; TOF-SIMS; biomass; bioresources

#### **1. Introduction**

Wood is the traditional and prime raw material in particleboards production since 1887, and annual production volumes in Europa exceed 30 million m<sup>3</sup> [1]. Considering the high production volumes, declining stocks of natural resources [2], i.e., possible future wood shortage situations, could play important roles. In addition, as the use of potentially contaminated waste wood in particleboards reaches 90% in some European countries, it could create environmental concerns, with a higher request for alternative non-contaminated materials.

Non-wood materials could also be utilized in particleboard (PB) production, which have the advantages of achieving higher resource-effectiveness, at ecologically and economically viable conditions. While agricultural crops are primarily cultivated for food, for

**Citation:** Klímek, P.; Wimmer, R.; Meinlschmidt, P. TOF-SIMS Molecular Imaging and Properties of pMDI-Bonded Particleboards Made from Cup-Plant and Wood. *Appl. Sci.* **2021**, *11*, 1604. https://doi.org/ 10.3390/app11041604

Academic Editor: L'uboš Krišt'ák Received: 6 January 2021 Accepted: 3 February 2021 Published: 10 February 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/).

various chemical products, or for biogas production [3], unutilized plant parts could be potentially processed to PBs. Using agricultural residues for PBs might have economic benefits for manufacturers, as the expenses for residues and wastes might be below market prices for wood [4]. Likewise, the utilization of waste materials in industrial production is also reducing environmental burdens, as residues such as stalks, husks, or straw are often left on the fields, or even burned.

Cup-plants seem to be a reasonable candidate for replacing wood in PBs. Dry mass yield between 11 t/ha and 20 t/ha per harvest is high and might compete with the yield achieved in forests, which can be ~16 t/ha per harvest [5]. Cup-plant (*Silphium perfoliatum* L.) originates in Eastern North America [6], but is now well-established across Central Europe. Although it was grown in gardens as an ornamental plant during the 18th century, today it is widely cultivated for energy production [7]. Cup-plant characteristics, including aspects of cultivation and utilization, including particleboard manufacturing have been demonstrated [8].

Particleboards from rice straw or rice husks [9–12], wheat straw [13], sunflower stalks [14–17], from vine prunings [18], cotton stalks [19], apple and plum orchard prunings [20], or even teal oil camellia [21], were already shown, i.e., produced. Balducci et al. [22] and Dix et al. [23] introduced residues of several Central European agricultural plants as a raw material for low density PBs, and Selinger and Wimmer [24] introduced light-weight sandwich PBs from hemp shives and fibers. It is obvious that agricultural resources could provide materials to replace wood in PBs, even if their property profiles are generally below conventional PBs. As the mechanical properties of PBs made with alternative materials are lowered, the anatomical structure and morphology of the utilized particles are of ultimate importance due to the close connections to the relevant properties [25].

Various microscopic techniques were applied to describe the anatomical and structural composition of wood-based composites. As an example, scanning electron microscopy was used for describing anatomical structures in PBs [26]. However, due to the lignocellulosic nature of the samples, there is insufficient compositional contrast to distinguish clearly between wood fractions, and adhesive bondlines [27]. To this end, elemental mapping by means of electron dispersive X-rays might be a feasible method. However, Electron Dispersive X-rays (EDX) techniques are also limited by their spatial resolution, sensitivity, or the ability to detect and map molecules that are indicative of wood-adhesive bondlines, particularly when wood and the used adhesive are both represented by the same chemical elements, albeit different molecular clusters. Here, the alternative method Timeof-Flight Secondary Ion Mass Spectrometry (TOF-SIMS) is suggested. Lately, TOF-SIMS was integrated into focused-ion-beam scanning-electron-microscope (FIB-SEM) TESCAN instruments [28], which brings additional advantages such as high spatial resolution (>5 nm) to elemental mapping, as well as simultaneous ion and SEM imaging. It was shown that TOF-SIMS (single instrument) can describe the distribution of molecules in biological systems [29] and also in wood tissues [30–33], but the use of TOF-SIMS for mapping the molecular distribution in adhesive-wood phases was not demonstrated so far. Consequently, the following hypotheses are stated:

**Hypothesis 1 (H1).** *Compact time-of-flight ion mass analyser (C-TOF) is capable of detecting elements and molecules specific to pMDI adhesion status.*

**Hypothesis 2 (H2).** *The penetration of the adhesive can be visualized in 3D using TOF-SIMS data.*

**Hypothesis 3 (H3).** *The mechanical properties of PBs made from the cup-plant fulfil industrial standards (EN 312, class 1).*

**Hypothesis 4 (H4).** *A three-layer PB made from cup-plant particles in the core layer, and spruce particles in the surface layer, could show improved properties over a single-layer produced panel.*

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

Single-layer as well as three-layer particleboards (PBs) were produced using cupplant stalk particles (*Silphium perfoliatum* L.). Stalks were 1.8 m long and square-sized with <sup>25</sup> <sup>×</sup> 25 mm<sup>2</sup> . As a control, single-layer and three-layers PBs were also produced using spruce particles (*Picea abies [L.] Karst*). The material was first chipped in a Klöckner chipper 120 × 400 H2W.T (Klöckner Maschinenfabrik, Lauenburg, Germany), at a cutting speed of 725 rpm, and a feeding speed of 1 m/s. The obtained chips approximately sized <sup>20</sup> <sup>×</sup> <sup>10</sup> <sup>×</sup> 5 mm<sup>3</sup> were then milled in a Condux-Werk HS 350 (Condux Maschinenbau GmbH & Co. KG, Hanau—Wolfgang, Germany) hammer mill. Afterwards, the particles were screened in the cascade vertical drum screener Allgaier D7336 (Allgaier-Werke GmbH, Uhingen, Germany). The sieve screens had mesh size openings of 5.0 mm, 3.15 mm, 1.24 mm, and 0.60 mm, respectively. Particles used to manufacture PBs were taken from the sieves with openings between >3.15 mm and <5 mm, which were then manually mixed at a weight ratio of 50:50. Afterwards, the particles were oven-dried at 74 ◦C for 4 days, reaching moisture contents between 5% and 7%. For the three-layer PBs, spruce particles at dimensions <1.24 mm were used for the surface layers, while cup-plant particles formed the core layer (3LCP). Same procedure was done for the three-layer PB, with both the core layer and the surface layers made from spruce (3LSP). For both three-layer PB types, the shelling ratio, which is the ratio of the surface layer thickness to the total thickness of the panel, was set at 0.3. —

All PBs were produced with the target density of 600 kg/m<sup>3</sup> , and a panel thickness of 12 mm, and bonded with polymeric methylene diphenyl diisocyanate (pMDI) resin (Huntsman I-BOND® PM4390, Huntsman GmbH, Hamburg, Germany). Two resin dosages were applied—pMDI was applied in amounts of 4% (MDI4), and 6% (MDI6), respectively. Particles were resinated in a drum blender for 5 min, using a pneumatic spraying nozzle. Consequently, the resonated particles were manually distributed in a wooden forming box (550 <sup>×</sup> 550 mm<sup>2</sup> ), and pre-pressed. The pre-pressed mat was then hot-pressed at 200 ◦C, at 3.2 MPa for 100 s, using a hydraulic Siempelkamp press (Siempelkamp Ma-schinen und Anlagenbau GmbH, Krefeld, Germany). The target thickness of the panels was checked at random positions. In total, one PB per type was manufactured (Figure 1). —

— — — — **Figure 1.** Cross-sectional views of the produced particleboards. 3LCP—three-layer particleboard (PB) with a cup-plant core layer, and spruce surface layers; 3LSP—three-layer spruce PB, 1LCP—single layer cup-plant PB, and 1LSP—single layer spruce PB.

— For scanning electron microscopy (SEM), a Tescan S8000 (Tescan Brno, s.r.o., Brno, Czech Republic) was used to study the surface morphology of the various PB types. Likewise, morphology and interactions between cup-plant particles and wood particles were observed as well. The ultra-high-resolution mode was used, by means of an Everhart Thornley secondary electron detector. Low accelerating voltages (between 500 V—1 kV) were used to avoid surface charging. Sample surface was cut with a sliding microtome [34].

The multifunctional focused-ion beam SEM, TESCAN LYRA3 (Tescan Brno, s.r.o., Brno, Czech Republic), was used, which also had an integrated compact time of flight secondary ion mass analyser (C-TOF-SIMS) (TofWerk AG, Thun, Switzerland). An area of interest (AOI) of 50 <sup>×</sup> <sup>50</sup> <sup>µ</sup>m<sup>2</sup> was scanned with a focused Ga + ion beam (4092 pA, 30,000 V), while time-of-flight of secondary ions, and their clusters (molecules) were continuously analysed by C-TOF-SIMS. C-TOF was operated at 10 µs dwell time, which provided the mass range of 0–170 *m*/*Q*. In parallel with the AOI scanning using focused ion beam

(FIB), C-TOF was used to record spectra at negative ion polarity and to capture elemental distribution maps at resolution of 1024 × 1024 pixels, by always binning 4 × 4 pixels. Data were derived from 100 scanned frames, which resulted in a crater with 1.7 µm depth in the sample (Figure 2, particularly B and D). With TOF-SIMS, the distribution of typical elements in wood, along with molecules indicating pMDI adhesives, were possible to visualize. Here, the distribution of carbon (C), oxygen (O), hydrogen (H), and hydroxyl (OH) groups were displayed to indicate wood, while CNO, CNH, and CN ion clusters were taken to display the pMDI adhesive distributions. Sample surfaces were coated with platinum in a sputter coater, prior to measurements, which avoided charging of the sample when exposed to the primary ion beam.

–

**Figure 2.** Selected area of interest (AOI) SEM of a spruce particle interphase prior (**A**), and crater-formations due to the TOF-SIMS analysis (**B**). AOI of a cup-plant particle interphase prior (**C**) and crater formations due to the TOF-SIMS analysis (**D**).

−1 Mechanical testing was carried out on a Zwick ® 1474 universal testing machine using the testXpert II software (Zwick GmbH & Co. kg, Ulm, Germany). Three-point bending tests according to EN 310 [35] were employed for the bending properties. Samples sized <sup>12</sup> <sup>×</sup> <sup>50</sup> <sup>×</sup> 290 mm<sup>3</sup> were subjected to a loading rate of 7 mm·min −1 , until failure. Internal bonding (IB) strength following EN 319 [36] was measured with squared samples (50 <sup>×</sup> 50 mm<sup>2</sup> ). Prior to testing the samples were sanded and then glued to the stainlesssteel blocks. The blocks were then positioned in gimbal-mounted holders and pre-loaded with 5 N in tension. Subsequently, a loading rate of 1 mm/min was applied until failure was reached.

Thickness swelling was determined according to EN 317, with conditioned samples sized 12 <sup>×</sup> <sup>50</sup> <sup>×</sup> 50 mm<sup>2</sup> fully immersed in 20 ◦C distilled water. Thickness swelling was determined at two-time intervals, i.e., after 2 and 24 h. After the immersion time had elapsed, the test samples were removed from the water and excess water was removed with a paper cloth. Then, the thickness swelling was measured manually, using a thickness gauge, at the center of the samples. Vertical density profiles (VDP) were measured with the x-ray density analyzer GreCon RG44 (GreCon, Germany). Five samples were measured from each type, with samples sized 12 <sup>×</sup> <sup>50</sup> <sup>×</sup> 50 mm<sup>2</sup> . The obtained data were processed with Statistica v.12 (StatSoft, inc., Tulsa, OK, USA) software. Normality of the data were checked by the Shapiro-Wilk test. The Statistical significance was set at *p* < 0.05 for the analysis of variance (ANOVA), with Scheffé post-hoc tests.

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

#### *3.1. Scanning Electron Microscopy*

SEM images indicate that spruce fines located in the surface layer of a particleboard (PB) have a better adherence with each other than the core-layer particles (Figure 3A). Further, an apparent porosity was seen at the transition of the spruce surface layer migrating into the cup-plant core layer (Figure 3D), a fact that potentially affected the mechanical properties. The anatomical structure of the core layer vs. surface layer was clearly different. While the wood showed a rather regular and compact cellular structure dominated by the tracheids (Figure 3B,C), the cup-plant structure was more diverse, showing annular thickenings (Figure 3E) and pitted perforations (Figure 3F), constituting a wider range of pore sizes, all potentially influencing resulting properties.

**Figure 3.** Cross-sectional SEM images of the spruce (**A**– –**C**) and cup-plant core layer—spruce surface layer ( — **D**– –**F**) PBs.

#### *3.2. TOF-SIMS Analysis*

For the spruce wood (Figure 4) and the cup-plant PB (Figure 5), cross-sections were prepared for the detection of carbon at mass-to-charge *m/Q* of 12, hydrogen at a *m/Q* of 1, oxygen at a *m/Q* of 16, and hydroxyl ion cluster OH at a *m/Q* of 17. Secondary ion molecular clusters typical for pMDI adhesives were also detected. The cyanide ion anion (CN) was identified at the *m/Q* peak of 26, hydrogen isocyanide (CNH) at the *m/Q* peak 27, while cyanate (CNO) was detected at a *m/Q* of 42. Due to the detection of molecules associated with pMDI (CN, CNH, and CNO), the resin distribution within the composite could be displayed. Results showed that in the pMDI-bonded PB, the pMDI-wood bondlines were not spot-like, but appeared rather even and non-regular, with a penetration deep into the wood structure. This finding corresponded to data presented by Mahrdt et al. [37]. We are showing that C-TOF attached to FIB-SEM with Ga + ion source could detect ion molecular clusters specific to pMDI adhesive and wood. Additionally, with the Ga + ion source, it should be possible to detect G-lignin at peak *Q/m* 137 [30], however, the applied C-TOF setup delivered only a low secondary ion signal (although visible), which did not allow an elemental mapping. This could be further elaborated in a future study. The relevance of CN, CNH, CNO being related to the pMDI distribution in the wooden structure was also confirmed by the FIB-SEM image (Figures 4 and 5).

− − − −

particle **Figure 4.** Elemental distribution maps of the selected isotopes and their molecular clusters for a spruce particle. The element exponent indicates a detected mass of ions or their clusters used for displaying the maps. Brighter regions in the FIB SE image refers to pMDI (marked); see Figure 2B for the selected area of interest SEM.

**Figure 5.** Elemental distribution maps of the selected isotopes and their molecular clusters for a cup-plant particle. The element exponent indicates a detected mass of ions or their clusters used for displaying of maps. Brighter regions in the FIB SE image refers to pMDI (marked); see Figure 2D for the selected area of interest SEM.

Particle–particle bondlines for both PB types were visualized by TOF-SIMS. The Q/m 12 peak (carbon) was adopted to represent the genuine wood structure, while for the particle–particle bondline, the molecular cluster CN (Q/m 26) was taken. The results showed that the pMDI adhesive was more dispersed in the cup-plant PB than in the spruce PB (Figures 6 and 7). While the spruce PB showed very narrow particle–particle bondlines, with penetration in non-compressed cell lumina regions (Figure 6), the adhesive in cupplant PBs appeared to be more spread-out (Figure 7). This could be related to the greater and more dispersed porosity present in the cup-plant. An essential and novel outcome of this research was also that individual particle–particle bondlines could be visualized through molecular C-TOF SIMS identification. Here, no additional sample preparation such as staining was required, with the samples not getting modified in any way, as it is the case with other bondline identification methods [27]. With C-TOF, it is possible to visualize bondlines in 3D, different to regular electron or light microscopy imaging. The capability of 3D imaging also approved Hypothesis 2. The obtained approach delivered data for in-depth analysis of bondline mechanics and substrate interaction, since the dataset could be transformed into the finite element model, with the stress and strain distributions of the structural components to be further assessed [38]. – – – – – –

– – **Figure 6.** 3D-visualisation of the spruce particle–particle bondlines, using TOF SIMS data, reconstructed by the ORS software (Object Research Systems, Montreal, QC, Canada); (**A**,**B**) spruce PB with visible particle–particle bondline, (**C**) pMDI bondline imaging (Q/m 26). – –

– – – – **Figure 7.** 3D-visualisation of cup-plant particle–particle bondlines, TOF SIMS data reconstructed by the ORS software (Object Research Systems, QC, Canada); (**A**,**B**) cup-plant PB with visible particle–particle bondline, and (**C**) the pMDI bondline imaging (Q/m 26).

#### *3.3. Mechanical Properties*

–

–

The results showed that the mean MOR values of the cup-plant PBs (3LCP, 1LCP) were below the mean measured for spruce PBs. Further, mean MOR of the three-layer cupplant PB (3LCP) was above the one-layer cup-plant PB (1LCP), although not statistically significant. Additionally, MOR of three-layer cup-plant PB (3LCP) was not different to the one-layer spruce PB (1LSP, *p* > 0.05; Figure 8). The spruce PB types and the three-layer cup-plant PB did meet the EN312 class P1 standard [39], which is for general use in dry

conditions. In contrast, average MOR of the one-layer cup-plant PB did not reach the P1 standard.

— — — — **Figure 8.** Modulus of rupture (MOR) of measured particleboards (PBs). 3LCP—three-layer cup-plant PB, 3LSP—three-layer spruce PB, 1LCP—single-layer cup-plant PB, and 1LSP—single-layer spruce PB (*n* = 10). — — — —

Average MOE measured for both cup-plant PBs types (3LCP, 1LCP) did not differ, (*p* > 0.05), and were also not statistically different from the one-layer spruce PB (1LSP, Figure 9). In addition, MOE of both cup-plant PBs were significantly (*p* < 0.05) lower than the three-layer spruce PB. Even with all seen MOE variability (Figure 9), the cup-plant PBs could be classified as P1 of EN 312, which refers to PBs used in dry conditions for interior fitments, including furniture.

**Figure 9.** Modulus of elasticity (MOE) of measured particleboards (PBs). 3LCP—three-layer cupplant PB, 3LSP—three-layer spruce PB, 1LCP—single-layer cup-plant PB, and 1LSP—single-layer spruce PB (*n* = 10).

The measured MOR and MOE of the one-layer spruce PB, and the three-layer spruce PB were consistent with data reported by Rofii et al. [40]. It was also documented that MOE and MOR of three-layer PBs were above the one-layer PBs [41]. MOE and MOR were both affected by particle alignments, surface layer density, as well as by the nature of the used raw material [42]. As cup-plant particles differ in their anatomical structure [8] and sizes when compared to spruce particles, this has evidently an effect on the properties of the produced PBs. Juliana et al. [43] reported a rise in MOE and MOR of three-layer PB made from Kenaf stalk cores, when the surface layers were made of rubberwood. MOR and MOE of the produced one-layer cup-plant PB showed similar values than PBs made from waste tea leaves [44], bleached straw [13], eggplant stalks [45], or rice straw [9].

— — —

—

Internal bonding values (IB) of the cup-plant PBs (3LCP, 1LCP) were also below the spruce PB types (Figure 10). One-layer cup-plant PB showed an average IB of 0.30 MPa, which was not significantly different from the IB of the three-layer cup-plant PB (0.34 MPa). Overall, the PBs made from cup-plant were again found suitable as general usage panels in dry conditions, as defined in EN 312 P1 [39]. In PBs, the core layers strongly determine the internal bonding, which explains why there are no significant differences for IB between single-layer and three-layer PBs. Rofii et al. [40] reported that surface layer characteristics of PBs have no significant effect on the IB. Additionally, Balducci et al. [22] showed that the surface layer of three-layer Miscanthus PBs had no significant influence on the measured IB. A reduced IB was found for rice husks PBs [46], hazelnut husks [47], or waste tea leaves [36]. An IB lower than 0.2 MPa was measured for rice straw [9], and for waste grass clipping PBs [48].

— — — — **Figure 10.** Internal bonding strength (IB) of the measured particleboards (PBs). 3LCP—three-layer cup-plant PBs, 3LSP—three-layer spruce PB, 1LCP—single-layer cup-plant PB, and 1LSP—singlelayer spruce PB (*n* = 20).

#### *3.4. Thickness Swelling and Water Uptake*

The three-layers PBs made from cup-plant, and the three-layer PB made from spruce wood, both showed similar thickness swelling after 2 h (TS2h, Figure 11). Data also show that the spruce particles in the surface layer did significantly lower TS2h, as it is the case with the cup-plant PBs (3LCP). The TS2h of single-layer cup-plant PB was almost twice the TS2h of the three-layer cup-plant PB. No TS2h difference was found between 3LSP and 3LCP. A lower thickness swelling of three-layer PB than single-layer PB was also measured by [22], where Miscanthus or topinambour stalks were utilized for the core layer in three-layer PB.

— — — — **Figure 11.** Thickness swelling (TS) and water absorption (WA) of produced particleboards (PBs). 3LCP—three-layer cup-plant PB, 3LSP—three-layer spruce PB, 1LCP—single-layer cup-plant PB, and 1LSP—single-layer spruce PB.

It is further shown that TS24h of both cup-plant PBs were significantly higher than the three-layer spruce comparison. It can be noted that in the 3LCP the spruce surface layer has a positive effect on TS24h. In addition, TS24h of the three-layer cup-plant PB was significantly lower (*p* < 0.05) than the TS24h measured for the single-layer cup-plant PB.

Water uptake results after 24 h were different in a way with the three-layer spruce particleboards absorbing the least water, while the single-layer cup-plant had the highest water uptake. Spruce particles as the surface layer is reducing the water-uptake, meaning that these particles seemingly reduce water access. This could be linked to a less accessible pore structure in spruce, compared to cup-plant particles (see Figure 3).

#### *3.5. Vertical Density Profile*

As seen in Figure 12, vertical density profiles of the cup-plant PB and the spruce PB were quite different. Single-layer cup-plant PB showed a flat density profile without distinct surface peaks. With spruce particles present in the surface layers, in an otherwise cup-plant PB, the density profile was also altered. It was found that the core layer density of the three-layer cup-plant PB was higher than that in the surface layers. It was evident that the density profile altered the mechanical performance of the cup-plant PBs. In Wong et al. [49] it was found that the density profile was like the one measured for spruce PBs, which is beneficial to MOE and MOR. Likewise, a flat density profile measured for the single-layer cup-plant PB was commonly connected to reduced bending properties, as shown with own data. The three-layer cup-plant PB had a higher density in the core layer, which at first glance should provide better IB for the three-layer cup-plant PB. Interestingly, this is not shown with our data. It must be noted that commonly not only the core layer, but the transition zone (TZ) between the surface and core layer is more prone to fail during an internal bonding test [50] (Figure 6). As seen in Figure 3, the three-layer cup-plant PB's transition zone had a higher proportion of pores and could be seen as a "weak layer", with a density similar to the core layer of the single-layer cup-plant PB. Thus, internal

bonding of the 3LCP did not improve over the other panel types. The three-layer spruce particleboard had the highest surface layer density, which explains the high MOR and MOE values of this PB type. layer", with a density similar to the

' be seen as a "wea

— — — — — **Figure 12.** Vertical density profile of the produced particleboards (PBs). 3LCP—three layer-cup-plant PB, 3LSP—three-layer spruce PB, 1LCP—single-layer cup-plant PB, and1LSP—single-layer spruce PB; TZ—transition zone between surface and core layer.

#### **4. Conclusions**

In this research, we successfully produced single-layer as well as three-layer cup-plant particleboards (PBs). We approved Hypothesis 1, as such that C-TOF and a combination of SEM-FIB with Ga + ion source was able to detect ion molecular clusters specific to pMDI adhesive and wood. The actual setup of the primary beam did not provide enough secondary ion signals to capture G-lignin cluster, although a specific peak at *m/Q* 137 was visible. This needs to be further elaborated. Mechanical and physical properties of PBs were compared with spruce-made particleboards. Hypothesis 2 was also confirmed, as the bondline was visualized in 3D with datasets acquired by C-TOF. Hypothesis 3 was only partly approved, as the three-layer cup-plant PBs fulfilled the requirement for a general usage panel (EN312, P1). However, MOR of the single-layer cup-plant PB needs to be increased to meet EN 312, P1 requirements. Hypothesis 4 was proven with the restriction that the IB of the cup-plant three-layer PB was not statistically different from the cup-plant single-layer PB. Nevertheless, three-layer cup-plant PBs delivered better MOE and MOR values than the single-layer cup-plant PBs. Raw material mixtures with spruce might be useful to raise MOR and MOE values. The density profile of the three-layer cup-plant PB has been altered in a way the core layer had higher densities than the surface layer.

**Author Contributions:** Conceptualization, P.K. and R.W.; methodology, P.K. and P.M..; software, P.K.; validation, P.K., R.W. and P.M.; writing-original draft preparation, P.K. and R.W.; writing-review and editing, R.W. and P.M.; visualization, P.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Tescan Orsay Holding a.s., 62300 Brno, by providing the relevant analytical resources. Further, the Fraunhofer-Institut für Holzforschung—Wilhelm-Klauditz-Institut provided all material characterisations.

—

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

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

#### **References**


### *Article* **Bonding of Selected Hardwoods with PVAc Adhesive**

**Ján Iždinský 1,\* , Ladislav Reinprecht <sup>1</sup> , Ján Sedliaˇcik <sup>2</sup> , Jozef Kúdela <sup>3</sup> and Viera Kuˇcerová 4**


**Abstract:** The bonding of wood with assembly adhesives is crucial for manufacturing wood composites, such as solid wood panels, glulam, furniture parts, and sport and musical instruments. This work investigates 13 hardwoods—bangkirai, beech, black locust, bubinga, ipé, iroko, maçaranduba, meranti, oak, palisander, sapelli, wengé and zebrano—and analyzes the impact of their selected structural and physical characteristics (e.g., the density, cold water extract, pH value, roughness, and wettability) on the adhesion strength with the polyvinyl acetate (PVAc) adhesive Multibond SK8. The adhesion strength of the bonded hardwoods, determined by the standard EN 205, ranged in the dry state from 9.5 MPa to 17.2 MPa, from 0.6 MPa to 2.6 MPa in the wet state, and from 8.5 MPa to 19.2 MPa in the reconditioned state. The adhesion strength in the dry state of the bonded hardwoods was not influenced by their cold water extracts, pH values, or roughness parallel with the grain. On the contrary, the adhesion strength was significantly with positive tendency influenced by their higher densities, lower roughness parameters perpendicular to the grain, and lower water contact angles.

**Keywords:** hardwoods; extractives; pH value; roughness; wettability; PVAc adhesive; adhesion strength

#### **1. Introduction**

The strength and stability of glued joints are the priority properties of all construction and decorative composites based on metals, wood, glass, plastics, and also other traditional and modern materials. This also applies to glued solid wood products for industrial, building, and transport structures, furniture, musical instruments, sports equipment, and other uses. With regards to the glued wood products, not only is the initial strength of glued joints important, but also the stability of the joints during indoor and mainly outdoor exposures, causing one-off or cyclical changes in wood moisture and temperature.

The most essential parameters influencing the overall bonding quality of wood products include the following: (a) the wood's species, density, chemical and anatomical structure, physical and strength characteristics, surface machining determining the surface roughness, grain orientation, moisture content, and pre-treatment with biocides or other additives, (b) the adhesive's chemical structure, weight solid, viscosity, surface tension, and mechanism of hardening, and (c) the bonding technology's pressure, time, and temperature [1–13].

The low density, high permeability, and high surface roughness of the individual wood species are basic factors that play an important role in terms of better adhesive penetration depth, usually in connection with a positive impact on the bonding quality. However, Aicher et al. [8] found that the adhesion strength of bonded wood is not always most prominently connected with its density. When compared, experiments of several

**Citation:** Iždinský, J.; Reinprecht, L.; Sedliaˇcik, J.; Kúdela, J.; Kuˇcerová, V. Bonding of Selected Hardwoods with PVAc Adhesive. *Appl. Sci.* **2021**, *11*, 67. https://dx.doi.org/10.3390/ app11010067

Received: 25 November 2020 Accepted: 18 December 2020 Published: 23 December 2020

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

**Copyright:** © 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 (https://creativecommons.org/ licenses/by/4.0/).

researchers who tested bonded hardwoods with density (ρ) in a range of 300–1000 kg/m<sup>3</sup> showed that the greater values of the adhesion strength were not in all cases found in specimens prepared from denser species (adhesion = 4.095 + 0.014 ρ/MPa/; R<sup>2</sup> = only 0.25), but almost always in those prepared from species characterized by a higher shear strength (adhesion = 0.628 + 0.912 τ/MPa/; R<sup>2</sup> = 0.88). Shida and Hiziroglu [14] examined these tendencies and found out that the adhesion strength of bonded woods was greater in the denser karamatsu species than in the less dense sugi species. Similar tendencies for nine European wood species observed Konnerth [15]. On the contrary, Alamsyah et al. [16] demonstrated a higher adhesion strength in bonded specimens made from the less dense *Paraserianthes falcataria* tropical wood than from the denser *Acacia mangium*.

Water-based adhesives, to which water dispersed adhesives also belong (e.g., polyvinyl acetate (PVAc) adhesives [17]), may reach an adhesion optimum when the water has totally penetrated into the wood substrate [18]. PVAc adhesives are commonly used in the wood industry for general assembly applications, film overlay and high-pressure lamination, edge gluing, wood veneer, and edge bonding [17]. They are safe, non-toxic, non-combustible, easy cleanable, without pollution, cure at room temperature, colorless, transparent and tough after curing, and give a high adhesion strength to bonded wood elements.

Özçifçi and Yapici [19] determined that there was a greater adhesion strength for beech and Scotch pine woods bonded with PVAc adhesive along the tangential direction than the radial one. Burdurlu et al. [20] obtained similar results for Calabrian pine wood bonded with PVAc and polyurethane (PUR) adhesives and recommended performing the bonding process on the tangential surfaces with higher pressures. However, in spite of these results, it is well known that the penetration depth of liquid adhesives within the wood structure is influenced not only by the wood's anatomical direction, density, moisture, and final permeability, but also by the physical and chemical characteristics of the adhesive and the technological bonding conditions, such as pressure, temperature, and time. For example, Sernek et al. [1] found better penetration of the water-based urea-formaldehyde (UF) adhesive into beech wood in the tangential direction at pressure application, while no significant difference between penetrations in the tangential and radial directions occurred when pressure was not applied.

The roughness of wood surfaces depends, first of all, on the wood anatomy and the mode of its machining [19,21–23]. Recognition and quantification of the surface roughness are important from the viewpoint of wood bonding and surface treatment, as the wood surface morphology significantly influences the wood wetting with film-forming materials and the adhesion of these materials to the wood substrate. The circular rotary saw usually causes a higher surface roughness of woods, in comparison with their planning or sanding [22,23]. Shida and Hiziroglu [14] inspected four Japanese wood species—sugi, hinoki, hiba, and karamatsu—and determined that their adhesion strengths with the PVAc adhesive achieved greater values if their surfaces were pre-finished with 80-grit sandpaper, compared to those surfaces pre-finished with finer 120- and 240-grit sizes. Burdurlu et al. [20] found out that a greater roughness of Calabrian pine wood surfaces was caused by their machining in this order, from most to least influential: sawing with a circular ripsaw, sanding, and planning. The shear strengths of specimens bonded with the PVAc adhesive were better for sanded or sawed wood surfaces compared with planned ones. Hiziroglu et al. [5,24] also documented the increased roughness of wood surfaces resulting from using sandpapers with lower grit sizes and the better wood surface adhesion with the PVAc adhesive. However, the experiments of Özçifçi and Yapici [19] showed that higher adhesion with various adhesive types had smoother planed wood surfaces than those prepared by band or circular sawing.

The high polarity and good wettability of wood surfaces is given mainly by the presence of hydroxyl, carbonyl, and carboxyl groups in the lignin-polysaccharide matrix of the cell walls. This results in the formation of strong physical bonds with various polar adhesives. Wood surfaces with higher polarities are more wettable with water-based adhesives [16,25]. The consequence is a higher penetration of the adhesives through the lumens of cell elements on the wood surface. However, the penetration rate can partly be limited by the formation of Van der Waals interactions, dipolar interactions, and hydrogen bonds of polar adhesives with the lignin-polysaccharide matrix of wood cell walls [26].

The wettability of wood, stability of the adhesive systems, and quality of the final adhesion can negatively or positively be influenced by wood extractives and also by preservatives or other excipients added to the wood [27–30]. Polar and nonpolar extractives play a major role in wood bonding processes, as they can contribute to or determine the relevant bonding properties of wood, such as acidity (pH value), wettability (contact angle, surface free energy), or even permeability (clogging of lumens by crystals). Extractives of a high acidity accelerate the curing of acid curing urea-formaldehyde (UF) and melamine-urea-formaldehyde (MUF) resins, decelerate bonding with alkaline hardening phenol-formaldehyde (PF) resins, or degrade PUR adhesives [9,11,31]. Starch and monomeric sugars, which belong to the primary polar water-soluble extractives present in all wood species, have a negative effect on the bonding of wood with cement, MUF, and PF adhesives [11,31]. Secondary extractives specifically occur in various hardwood and softwood species. These extractives, which are typically situated in the heart zones of some European and several tropical hardwood species, contain either various polar polyphenols with a hydrophilic character, such as flavonoids, tannins, sterols, flobafenes, rubrenolide, rubrynolide, and quinones or coumarins, as well as various nonpolar or semi-polar waxes, fats, and oils with a hydrophobic characteristic [32–35]. Studies of tropical woods bonded together with a PVAc adhesive showed that the extractive content of the wood species had an adverse effect on the bonding quality [36]. In addition, extractives have pronounced inhibitory or supportive effects on the wood sorption capacity, which is reflected in the swelling and shrinkage coefficients associated with moisture changes. The consequence is a stress state at the wood–adhesive–wood interface, impairing the adhesion of the glued joint [37]. Generally, the impact of wood extractives manifests itself more significantly when the bond line is exposed to multiple negative factors. For example, the loss of the adhesion strength in regard to polar extractives is more apparent in the wet state of bonded woods [11].

In the case of wood bonding with adhesives, it is necessary to consider wood's rheological properties that might induce additional retardation of the process of wood surface wetting with the gluing substance on its own. The impact of rheological performance can be, to a considerable extent, mitigated with the aid of mechanical and physical forces applied during adhesive application, during the pressing of the wood at bonding, as well as at exposure of the bonded wood to climatic changes.

The issue of the bonding of European and tropical hardwoods for construction and furniture purposes has been addressed by several researchers, evaluating the factors influencing the shear strength and delamination of glued joints in particular [7,8,38–40]. In summary, they proved that several European and tropical hardwood species meet the requirements for the adhesion strength of glued joints set by the relevant standards and are potentially suitable for glued furniture and construction products.

The aim of this work was to analyze the impact of the selected structural and physical characteristics of hardwoods (e.g., the density, cold water extract, pH value, roughness, and wettability) on their adhesion strength with a water-based PVAc adhesive in dry, wet, and reconditioned states.

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


The heart zones or central zones of 10 tropical and 3 European wood species were used for the experiment (Figure 1). Their names are usually defined by the standard EN 13556 [41]: bangkirai (*Shorea obtusa* Wall.; Sh. Spp.), European beech (*Fagus sylvatica* L.), black locust (*Robinia pseudoacacia* L.), bubinga (*Guibourtia demeusii* (Harms) J. Léon.), ipé

(*Tabebuia serratifolia* (Vahl) Nicholson), iroko (*Milicia excelsa* (Welw.) C. C. Berg), maçaranduba (*Manilkara bidentata* A. Chev.), dark red meranti (*Shorea curtisii* Dyer ex. King), European oak (*Quercus robur* L.), Santos palisander (*Machaerium scleroxylon* Tul.), sapelli (*Entandrophragma cylindricum* Sprague), wengé (*Millettia laurentii* De Wild.), and zebrano (*Microberlinia brazzavillensis* A. Chev.). The experimental wood material was obtained from the trading company JAF Holz, Ltd. (Špaˇcince, Slovak Republic) in the form of naturally dried boards, having a moisture content of 13% ± 2.5%. Test samples with dimensions of 80 mm × 20 mm × 5 mm (longitudinal × radial × tangential) were prepared with a circular ripsaw (Freud Pro LP30M 026P) having these parameters: a diameter of 255 mm, a cutting thickness of 2.8 mm, a sawblade body thickness of 1.8 mm, a number of teeth of 40, and a maximum rotation speed of 7800 rpm. The wood samples were of a high quality (i.e., without bio-damages, knots, or other inhomogeneities), and before other technological operations, they were conditioned at a temperature of 20 ◦C ± 2 ◦C and a relative air humidity of 50% ± 5%, achieving an equilibrium moisture content of 8% ± 2%. č

**Figure 1.** Hardwoods used in the experiment, numbered (from 1 to 13) according to the increasing values of the adhesion strength in the dry state demonstrated in Section 3.2: (1) bangkirai, (10) beech, (4) black locust, (13) bubinga, (7) ipé, (2) iroko, (11) maçaranduba, (3) meranti, (8) oak, (12) palisander, (9) sapelli, (5) wengé, and (6) zebrano.

*ρ* 2.1.2. Characteristics of Hardwoods: Density, Cold Water Extract, pH Value, Roughness, and Wettability

The density *ρ* of hardwoods was determined in accordance with the standard EN 323 [42].

The cold water extract from hardwoods was obtained in accordance with the standard ASTM D1110 [43], followed by measurement of the pH values of these extracts with a pH meter 7110 (WTW, Wellheim, Germany).

*θ* The wettability of the hardwood surfaces was associated with determining the contact angle with a redistilled water drop with a volume of 0.0018 mL up to its complete soaking into the wood substrate, using a goniometer Krüss DSA30 Standard (Krüss, Hamburg, Germany). The course of the water drop profile evolving parallel to the wood grain, from first contact up to the complete soaking, was recorded with a camera, its scanning frequency set in accordance with the wetting interval. The initial contact angle *θ<sup>0</sup>* was evaluated at the beginning of the wetting process, meaning at the moment of first contact between the water drop and the wood substrate. The drop's contact angle, from the moment of reversion from the acceding contact angle into the receding one, was considered the equilibrium contact angle *θ<sup>e</sup>* . From the values of the contact angles *θ<sup>0</sup>* and *θ<sup>e</sup>* , the abstract contact angle *θw*, corresponding to an ideal smooth surface, was calculated by the method of Liptáková and Kúdela [44].

Three parameters of wood roughness, the *Ra (*arithmetic mean deviation), *Rz (*arithmetic mean of the heights and depressions of the profile at the basic length), and R*Sm* (mean dis-

tance between the valleys), were inspected on the radial surfaces of samples parallel with and perpendicular to the grain, using the Surfcom 130A surface roughness measuring instrument (Carl Zeiss, Jena, Germany) in accordance with the standard EN ISO 4287 [45]. A total measured length for one replicate was 12.5 mm, and a basic length for one analysis was 2.5 mm.

*θ θ*

#### *2.2. Adhesive*

*θ θ*

A bonding of hardwoods was performed with the water resistant, one-component crosslinking polyvinyl acetate (PVAc) adhesive Multibond SK8 (Franklin International, Columbus, Ohio USA). This adhesive is characterized by the following basic technical parameters: weight solids 48.7–52.3%, pH 2.4–3.5, viscosity approximately 4000 mPa.s, and specific gravidity 1.1 g cm−<sup>3</sup> . −

#### *2.3. Wood Bonding*

For the adhesion strength, each individual specimen was prepared by bonding two samples (80 mm × 20 mm × 5 mm) of the same wood species. The PVAc adhesive was applied on the contacting surfaces of both wood pieces in an amount of 120 g ± 10 g per square meter. The bonding process was performed in the press for 60 min at a pressure of 1.2 MPa and a temperature of 20 ◦C ± 2 ◦C. The conditioning of the bonded hardwood specimens lasted 7 days at a temperature of 20 ◦C ± 2 ◦C and a relative air humidity of 50% ± 5%.

#### *2.4. Adhesion Strength: Tensile Shear Strength*

The adhesion strength test—the tensile shear strength of lap joints—of bonded hardwoods was performed by the standard EN 205 [46] in the dry state, wet state, and reconditioned state and evaluated in accordance with the criteria of the standard EN 204 [47] for water-resistant adhesives belonging to the D3 class (Figure 2).

#### *2.5. Statistical Analyses*

The statistical software STATISTICA 12 was used to analyze the gathered data. Descriptive statistics deals with the basic statistical characteristics of studied properties: arithmetic mean and standard deviation. Differences of the adhesion strength in the dry state of the bonded hardwoods were analyzed by the Duncan test. The simple linear correlation analyses together with the coefficient of determination R<sup>2</sup> and the significance level parameter *p* were used as method of inductive statistics to evaluate the measured data.

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

#### *3.1. Density, Cold Water Extract, pH Value, Roughness, and Wettability of Hardwoods*

The selected structural and physical characteristics of 13 hardwoods, theoretically important from the point of view of their bonding with adhesives, are present in Tables 1–3.

The density *ρ* of 13 hardwood species, determined at a moisture content of 8% ± 2%, ranged from 636 kg/m<sup>3</sup> for meranti to 1105 kg/m<sup>3</sup> for maçaranduba (Table 1). The cold water extract of the hardwoods, obtained from their sawdust by the standard ASTM D1110 [43], ranged from lower values of 0.9–1.6% for zebrano, wengé, meranti, and sapelli to higher values of 4.2–4.6% for palisander, oak, and black locust (Table 1). The pH of individual hardwood species ranged from a neutral acidic value of 5.8 for palisander and beech to more acidic values in the scope of 3.4–3.9 for meranti, bangkirai, oak, and ipé (Table 1).

The type and amount of extractives in wood varies from species to species, and for the same wood species this can also be influenced by the geographical origin, climate conditions, tree age, and part of the tree from which a sample originates [48]. Several studies have been carried out on the extractives of tropical woods. For example, Wanschura et al. [49] described the benefits of extractives present in tropical woods for their surface treatments. Kilic and Niemz [50], in the structures of 12 tropical wood species, found very low amounts of lipophilics (0.05–0.38 mg/g); the constituent consisted mainly of fatty acids, while the hydrophilics were composed of phenolic acids, flavonoids, sterols, stilbenes, and a lignan. Jankowska et al. [30] determined that there were large differences in quantity of the hot water soluble extractives in the European and tropical hardwood species, which have been researched by us as well (e.g., 2.92% in light red meranti, 4.24% in beech, 4.33% in wengé, 6.07% in sapelli, 6.47% in iroko, 11.21% in oak, and 12.63% in ipé). The values of the cold water extracts of the same species determined in our experiment were evidently lower, from 1.53% in wengé to 4.33% in oak (Table 1). This difference was probably caused by the fact that hot water dissolves not only polar extractives, which also easily dissolve in cold water (e.g., tannins, gums, sugars, and coloring matter), but also starches. The value of the cold water extract for meranti wood was comparable with that found by Yamamoto and Hong [51]; however, those for the wengé and zebrano woods were smaller compared with the values obtained by [50].

Similar pH values for some of the same tropical hardwood species (Table 1) were reported by Yamamoto and Hong [51], Torelli and Cufar [ ˇ 52], and Ikenyiri et al. [53].


**Table 1.** Densities, cold water extracts and pH values of the hardwoods.

Notes: Mean values of density are from six measurements, and the cold water extract and pH value are from three measurements. <sup>1</sup> By Wagenführ [55].

The roughness parameters *Ra* and *Rz* determined for 13 hardwoods exhibited, on average, 48.2% and 79.1% higher values, respectively, measured perpendicularly to the grains (on average: *Ra* = 10.4 µm, *Rz* = 85.6 µm) than those determined parallel with the grains (on average: *Ra* = 7.0 µm, *Rz* = 47.8 µm). Generally, these represented differences from 8% to 114%, according to the relevant wood species (Table 2). However, the parameter *RSm* was approximately comparable in both measured directions of the wood (Table 2).

**Table 2.** Roughness parameters *Ra*, *Rz*, and *RSm* of the hardwood surfaces.


Note: Mean values are determined from 90 values (15 different measuring spots on 6 replicates). Standard deviations are in parentheses.

The surfaces of the densest maçaranduba wood were characterized by the lowest roughness parameters *Ra* and *Rz*. On the contrary, the highest roughness parameters were exhibited by the least dense wood, the meranti wood, but at the same time by the bangkirai and oak woods also, which had a medium density (Tables 1 and 2). In these circumstances, an unexpectedly higher surface roughness of the bangkirai and oak, belonging to the ring-porous wood species [55], can be explained by their naturally more porous morphological structure.

The water's potential for surface wetting of the tested hardwood species was assessed based on the water contact angles *θ0, θ<sup>e</sup>* , and *θ<sup>w</sup>* (Table 3). Iroko had the highest initial and abstract contact angles (*θ<sup>0</sup>* = 112.5◦ , *θ<sup>w</sup>* = 78.5◦ ), and black locust had the lowest equilibrium and abstract contact angles (*θ<sup>e</sup>* = 19.7◦ , *θ<sup>w</sup>* = 25.4◦ ), while bubinga had the lowest initial contact angle (*θ<sup>0</sup>* = 54.8◦ ) (Table 3).

According to Liptáková et al. [56], the values of the contact angles *θ<sup>0</sup>* and *θ<sup>e</sup>* depend on the wood surface roughness and chemistry. These authors supposed that the calculated contact angle value *θw*, valid for an ideal smooth surface, depends exclusively on the wood chemistry. Consequently, the different values of the contact angle *θ<sup>w</sup>* in Table 3 indicate species-related differences in the wood surface chemistry. This concerns not only the different types and contents of extractives, but also the chemical composition of the lignin and polysaccharides in tested hardwoods.

The wettability of wood is a substantial parameter which gives basic information on the interaction between the solid wood surface and liquids, such as adhesives and paints (e.g., how easily and efficiently the liquids spread over a solid surface) [57,58]. The smaller water contact angles, which usually result from the rougher and more polar characteristics of wood surfaces, indicate deeper penetration of the water-based adhesive into the wood structure [16,59].

When comparing the roughness parameters and the wettability parameters of 13 hardwoods (Tables 2 and 3), it is evident that there was not always a more apparent connection between these two groups of parameters. For example, the maçaranduba and meranti

woods, characterized by totally different values for the density (Table 1) and the parameters of roughness (Table 2), had comparable values for the water contact angles (Table 3).


**Table 3.** Wettability of hardwood surfaces, characterized by the contact angles.

Notes: Mean values are from 30 measurements (5 different measuring spots on 6 replicates). Standard deviations are in parentheses.

#### *3.2. Adhesion of Hardwoods with PVAc Adhesive*

The values of the adhesion strength (the tensile shear strength of the lap joints) of hardwood specimens bonded with the PVAc Multibond SK8 adhesive are shown in Table 4 and Figure 3.

**Table 4.** The adhesion strength (tensile shear strength of lap joints) of hardwood specimens bonded with the PVAc Multibond SK8 adhesive, determined in the dry state, wet state, and reconditioned state by the standard EN 204 [47].


Note: Mean values are from six values. Standard deviations are in parentheses. The Duncan test, using the indexes (a, b, c, and d), identified the significance level of the higher adhesion strength of bonded wood species in relation to the reference bonded bangkirai wood having the lowest adhesion strength in the dry state (e.g., a: very significantly higher, >99.9%; b: significantly higher, >99%; c: less significantly higher, >95%; and d: insignificantly higher, <95%).

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 9 of 15

c: less significantly higher, >95%; and d: insignificantly higher, <95%).

bonded wood species in relation to the reference bonded bangkirai wood having the lowest adhesion strength in the dry state (e.g., a: very significantly higher, >99.9%; b: significantly higher, >99%;

**Figure 3.** The growth tendency of the adhesion strength from bangkirai (No. 1) to bubinga (No. 13) determined in the dry state of bonded hardwoods (a) unequally changed for the individual wood species only in the wet state (b), based on the evidently decreased parameter R<sup>2</sup> and reduced significance *p* (from 99.9% to 99%) of the linear correlation Adhesion = a + b × Number of Hardwood, while this tendency returned in the reconditioned state (c). **Figure 3.** The growth tendency of the adhesion strength from bangkirai (No. 1) to bubinga (No. 13) determined in the dry state of bonded hardwoods (**a**) unequally changed for the individual wood species only in the wet state (**b**), based on the evidently decreased parameter R<sup>2</sup> and reduced significance *p* (from 99.9% to 99%) of the linear correlation Adhesion = a + b × Number of Hardwood, while this tendency returned in the reconditioned state (**c**).

The Multibond SK8 adhesive was as a good glue type for the bonding of several hardwoods exposed in dry and water-soaked conditions. This type of PVAc adhesive, for all 13 bonded hardwoods, usually secured the minimum adhesion strengths required by the standard EN 204 [47] (i.e., on average, (a) in the dry state of 13.6 MPa, from 9.5 MPa for bangkirai to 17.2 MPa for bubinga (required minimum = 10 MPa), (b) in the wet state of 1.8 MPa, from 0.6 MPa for maçaranduba to 2.6 MPa for beech (required minimum = 2 MPa), and (c) in the reconditioned state of 13.0 MPa, from 8.5 MPa for maçaranduba to 19.2 MPa for bubinga (required minimum = 8 MPa)) (Table 4, Figure 3).

In the summary evaluation, using a linear correlation Adhesion = a + b × Number of Hardwood (for the numbering of hardwoods, see Figure 1), the adhesion strength of 13 bonded hardwoods determined in the dry state (Figure 3a) unequally decreased in the wet state due to 4 days of soaking in water (Figure 3b). This was based on the coefficient of determination R<sup>2</sup> declining from 0.6 to 0.06, and on the change of the significance level parameter *p* from 0.000 to 0.021. However, due to 7 days of reconditioning of the wet samples in a dry environment, the adhesion strength recovered quite equally to the initial values found in the dry state, which was in the linear correlation confirmed by the increasing of R<sup>2</sup> to 0.21 and with *p* equal to 0.000 (Figure 3c).

Partially worsened results were achieved in the wet state of the bonded hardwoods, when the adhesion strength decreased in more cases under the criteria value of 2 MPa (i.e., to 0.6 MPa–2.6 MPa) (Table 4, Figure 3), which could be explained, among other factors, with a different penetration of the PVAc adhesive into the individual hardwood species. For example, the lowest adhesion of 0.6 MPa was determined for the maçaranduba wood, the densest species (Table 1) having the lowest roughness (Table 3). On the contrary, the second-highest adhesion of 2.3 MPa was determined for the meranti wood, the least dense species (Table 1) having the highest roughness (Table 3). The highest adhesion of 2.6 MPa was determined for the beech wood (i.e., the species characterized by very good permeability [60,61]). Generally, different microstructures of the wood–adhesive–wood interfaces, when a probable better penetration of adhesives into less dense and more porous woods, as well as into more permeable woods, could be connected with a higher and more water-stable mechanical adhesion of adhesives with wood surfaces.

Results achieved in this work for the bonded beech wood were in accordance with the work of He and Chiozza [62]. For this wood species, bonded it at 23 ◦C with the PVAC glue Vinavil 2259 L, they determined by the standard EN 205 [46] the adhesion strength in the dry state 15.8 MPa and in the wet state 2.7 MPa.

#### *3.3. Connections between Bonding and Selected Characteristics of Hardwoods*

The adhesion strengths of 13 bonded hardwoods, valued in the dry state, were positively influenced—on the 99% or 95% significance level (*p* < 0.01; *p* < 0.05)—by their higher densities *ρ*, lower roughness parameters *Ra* and *Rz* perpendicular to the grain, and lower water contact angles *θ*0, *θ*e, and *θ*w, as was documented by the linear correlations (Adhesion = a + b × Property of wood) (Table 5 and Figure 4a,d–f).

On the contrary, the adhesion strengths of the bonded hardwoods were not significantly influenced by the cold water extract, pH value, or roughness parameters *Ra*, *Rz*, and *RSm* parallel with the grain (*p* > 0.5) (Table 5 and Figure 4b,c).

Generally, the joint quality of bonded timbers depends on several surface characteristics of the wood, including its wetting capacity. For example, beech and bubinga woods were characterized with good wettability values (Table 3), and this property resulted in a positive impact on their adhesion strength with a PVAc adhesive in the dry state (Table 4 and Figures 3a and 4e,f), as well as in the wet state and reconditioned state (Table 4). However, the wettability may not always be a presuming factor. For example, a good bonding quality in the dry state was also determined for ipé, iroko, sapelli, and wengé woods, whose surfaces had evidently higher water contact angles.


**Table 5.** Linear correlation analyses between the hardwood characteristics or properties and the adhesion strength, determined for the wood–PVAc adhesive–wood interface in the dry state.

Note: N is the number of samples. However, several times more measurements were performed and by linear correlations analyzed for the roughness parameters and the contact angles, as for each individual sample, the roughness parameters were determined on 15 spots and the contact angles on 5 spots (see Tables 2 and 3). *θ<sup>w</sup>* 78 0.13 −3.40 0.001 15.82–0.043 × *θ<sup>w</sup>* Note: N is the number of samples. However, several times more measurements were performed and by linear correlations analyzed for the roughness parameters and the contact angles, as for each individual sample, the roughness parameters were determined on

15 spots and the contact angles on 5 spots (see Tables 2 and 3).

50 60 70 80 90 100 110 120 **Contact angle θ<sup>0</sup>**

**(e)**

**Adhesion**

 **[MPa]**

**[°]**

**Adhesion = 18.31 - 0.05 \* θ<sup>0</sup>**

**R 2 = 0.14** 

**(f)**

**Adhesion**

 **[MPa]**

> 20 30 40 50 60 70 80 **Contact angle θ<sup>w</sup> [°]**

**R 2 = 0.29** 

**Adhesion = 17.31 - 0.08 \* θ<sup>w</sup>**

**Adhesion**

 **[MPa]**

**(c)**

**(a)**

**Adhesion**

 **[MPa]**

*Appl. Sci.* **2021**, *11*, x FOR PEER REVIEW 11 of 15

*RSm* 78 0.039 1.75 0.085 12.43 + 0.002 × *RSm*

*Ra* 78 0.071 −2.41 0.018 14.56–0.087 × *Ra Rz* 78 0.081 −2.59 0.011 15.05–0.016 × *Rz RSm* 78 0.006 −0.68 0.498 14.03–0.001 × *RSm*

*θ*<sup>0</sup> 78 0.11 −3.03 0.003 17.52–0.044 × *θ<sup>0</sup> θ<sup>e</sup>* 78 0.10 −2.94 0.004 15.62–0.051 × *θ<sup>e</sup> θ<sup>w</sup>* 78 0.13 −3.40 0.001 15.82–0.043 × *θ<sup>w</sup>* Note: N is the number of samples. However, several times more measurements were performed and by linear correlations analyzed for the roughness parameters and the contact angles, as for each individual sample, the roughness parameters were determined on

**(d)**

**Adhesion**

 **[MPa]**

**(b)**

**Adhesion**

 **[MPa]**

Roughness perpendicular to grain (μm)

15 spots and the contact angles on 5 spots (see Tables 2 and 3).

600 700 800 900 1000 1100 1200 **Density ρ [kg/m<sup>3</sup>**

3.0 3.5 4.0 4.5 5.0 5.5 6.0 **pH-value**

**R 2 = 0.09** 

**R 2 = 0.14** 

**]**

**Adhesion = 9.72 + 0.83 \* pH**

**Adhesion = 8.37 + 0.007 \* ρ**

Contact angle (°)

**Figure 4.** Expression of the linear tendency changes of the mean adhesion strengths of 13 bonded hardwoods, determined in the dry state, in relation to the mean properties of these wood species – (**a**) density, (**b**) cold-water extract, (**c**) pH value, (**d**) roughness parameter *Ra* perpendicular to the grain, (**e**) initial contact angle, (**f**) abstract contact angle. Note: In Table 5, statistical analyses carried out from all individual measurements are present.

#### **4. Conclusions**

The adhesion strengths of 13 hardwoods bonded with the PVAc adhesive Multibond SK8 met, in most cases, the requirements of the standard EN 204 [47] (i.e., it ranged from 9.5 MPa to 17.2 MPa (limit 10 MPa) in the dry state, from 0.6 MPa to 2.6 MPa (limit 2 MPa) in the wet state, and from 8.5 MPa to 19.2 MPa (limit 8 MPa) in the reconditioned state).

0 1 2 3 4 5 **Cold-water extract [%]**

2 4 6 8 10 12 14 16 18 20 **Roughness** *Ra* **perpendicular to grain [μm]**

**R 2 = 0.41** 

**Adhesion = 17.28 - 0.35 \*** *Ra* **perpendicular**

**R <sup>2</sup>= 0.01** 

**Adhesion = 13.10 + 0.184 \* extract**

The water contact angles, expressing the wettability of wood surfaces, were not clearly affected by the other measured structural and physical characteristics of the hardwoods. For example, for the most dense wood, maçaranduba wood (1105 kg/m<sup>3</sup> ), in comparison with the less dense meranti wood (636 kg/m<sup>3</sup> ), was determined to have a 136% greater cold water extract, 36% higher pH value, 63.3% lower roughness parameter *Ra* parallel with the grain, and 71.6% lower roughness parameter *Ra* perpendicular to the grain. However, the initial contact angles of these two tropical woods *θ<sup>0</sup>* were essentially the same, being 94.0◦ and 92.0◦ , respectively. This means that the wettability of wood surfaces with water, which usually dominantly affects the adhesion strength between wood surfaces and waterbased adhesives or coatings, can simultaneously be dependent on the combination of several other structural parameters of the wood, including its specific molecular structure. This should be analyzed in any following experiments.

The linear correlations indicated that the mutual relations between the adhesion strength values in the dry state of 13 hardwoods bonded with a PVAc adhesive and the cold water extract, pH value, or roughness parameters parallel with the grain were not statistically significant. On the contrary, the adhesion strength values showed negative tendencies on the 95% significance level, influenced by the increased roughness parameters *Ra* and *Rz* perpendicular to the grain, and on the 99% significance level, influenced by the increased water contact angles *θ0*, *θ<sup>e</sup>* , and *θ<sup>w</sup>* of the individual hardwoods.

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

**Funding:** This work was supported by the Slovak Research and Development Agency under the contracts no. APVV-17-0583 and no. APVV-16-0177, and the VEGA project 1/0729/18.

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Acknowledgments:** The authors would like to thank the Slovak Research and Development Agency under contracts no. APVV-17-0583 and no. APVV-16-0177, and also to the VEGA project 1/0729/18 for funding and financial support. This publication is also the result of the following project implementation: Progressive research of performance properties of wood-based materials and products (LignoPro), ITMS 313011T720 supported by the Operational Programme Integrated Infrastructure (OPII) funded by the ERDF.

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

#### **References**


### *Article* **Prediction of Mechanical Performance of Acetylated MDF at Di**ff**erent Humid Conditions**

#### **Sheikh Ali Ahmed 1,\* , Stergios Adamopoulos 2,\* , Junqiu Li <sup>1</sup> and Janka Kovacikova <sup>3</sup>**


Received: 19 November 2020; Accepted: 2 December 2020; Published: 4 December 2020

**Abstract:** Change of relative humidity (RH) in surrounding environment can greatly affect the physical and mechanical properties of wood-based panels. Commercially produced acetylated medium density fiberboard (MDF), Medite Tricoya®, was used in this study to predict strength and stiffness under varying humid conditions by separating samples in parallel (//) and perpendicular (⊥) to the sanding directions. Thickness swelling, static moduli of elasticity (MOEstat) and rupture (MORstat), and internal bond (IB) strength were measured at three different humid conditions, i.e., dry (35% RH) and wet (85% RH). Internal bond (IB) strength was also measured after accelerated aging test. A resonance method was used to determine dynamic modulus of elasticity (MOEdyn) at the aforementioned humid conditions. Linear regression and finite element (FE) analyses were used to predict the MDF's static bending behavior. Results showed that dimensional stability, MOEstat, MORstat and IB strength decreased significantly with an increase in RH. No reduction of IB strength was observed after 426 h of accelerated aging test. A multiple regression model was established using MOEdyn and RH values to predict MOEstat and MORstat. In both directions (// and ⊥), highly significant relationships were observed. The predicted and the measured values of MOEstat and MORstat were satisfactorily related to each other, which indicated that the developed model can be effectively used for evaluating the strength and stiffness of Medite Tricoya® MDF samples at any humid condition. Percent errors of two different simulation techniques (standard and extended FE method) showed highly efficient way of simulating the MDF structures with low fidelity.

**Keywords:** acetylation; wood fiber; strength; stiffness; internal bonding strength; thickness swelling; regression; finite element analysis

#### **1. Introduction**

Medium density fiberboard (MDF) is manufactured with wood fibers bonded with water resistant adhesives such as phenol formaldehyde, urea formaldehyde, isocyanate resin, etc. [1,2]. MDF is primarily used in furniture, as a building material and for laminate flooring, since it has good strength and stiffness and it is easy to process. Compared to plywood, MDF panels generally swell more and may not be recovered after drying due to the inherent hygroscopicity of the wood fibers, the residual stresses formed in the fiber mat during hot pressing and some loss of the glue bonds [3]. As a result, when the MDF panel is exposed to any form of water, its constituent wood fibers swell and some of that residual stress is released, resulting in an increase of thickness of the panel. Thickness swelling

markedly weakens the product [4], and the mechanical properties that are most directly affected are shear strength and moduli of elasticity and rupture [5]. Exposure to outdoor environment with varying climate conditions can result in dimensional changes and strength loss, and thus MDF panels are generally not recommended for exterior applications. In terms of computational modelling, these phenomena are considered to be the rheological behavior of the material [6].

Several studies have been carried out using various adhesive systems [7–9], post treatments [10], heat treated fibers [11], alternative fibers [12] and recycled adhesives [13] to improve strength and water resistance of MDF panels. In addition, chemical modification was also used to improve the material properties, such as moisture-related properties, durability and weathering resistance [14–16]. The most described chemical modification method for improved dimensional stability of wood particles and fibers to produce panel products is acetylation with acetic anhydride [17,18]. In acetylation process, acetic anhydride substitutes the hydroxyl (-OH) groups in the wood cells with acetyl groups, resulting in decreased hygroscopicity and increased dimensional stability; while reductions in some mechanical properties also occur depending on the extent of temperature and time of the modification reaction [18–20]. Several explanations of the strength loss were given, as for example the type of the adhesive [21], bondability [22], press pressure [23], etc. Higher bondability is required if the panels are used under severe conditions for a long time. Thus, the effect of weathering on the dimensional changes and mechanical properties of acetylated MDF would be beneficial for predicting its long-term service behavior. Mechanical strength and stiffness along and perpendicular to the sanding direction should be also available to achieve better and proper assembly of acetylated MDF panels in outdoor applications.

Fiberboards can be exposed to a range of environmental conditions during service life. Moisture content of MDF, similar to other lignocellulosic materials, changes with the change of surrounding humidity and temperature. Therefore, it is important to know the relationship between moisture content and strength properties of MDF if used as a structural member subjected to these environmental conditions. Use of non-destructive acoustic testing could provide a rapid and reliable measurement of strength properties of MDF panels. Usually, acoustic testing is carried out by using time-of-flight (TOF) and resonance methods [24]. TOF methods use propagation time of a pulse of ultrasound or a stress wave across the material. On the other hand, resonance methods use the free vibration frequency of the material under forced harmonic vibration. Resonance methods provide more information on the elastic properties of materials and are thus considered more reliable than the TOF methods [25]. Previous studies showed a very good to strong relation between dynamic bending properties measured by acoustic tools with the static bending properties of wood panels [26,27]. However, values of dynamic bending properties vary depending on the method used and, most importantly, on the moisture content levels of wood panels. Prediction of static bending properties of acetylated MDF using acoustic techniques under different humid conditions is lacking. Considering that acetylated MDF is more hydrophobic than conventional MDF, fewer internal bond failures and associated changes of internal structure should be expected by repeated swelling and shrinkage. That in turn should lead to more stable static bending behavior. Nevertheless, establishing the relationship between acoustic and static bending properties of acetylated MDF at different humid conditions would ensure reliable and safe predictions of their performance for intended end uses.

Another available method to predict and analyze material behavior of MDF is creating macro scale finite element (FE) models of MDF board's structure. Following classic design procedures for macroscale modelling, the material characteristics obtained from the experiments presented in this paper were used as input values to define the material, and the geometry and boundary conditions of the experiment set up were imitated in the FE model. Here, two analysis approaches that are implemented in a SIMULIA™ Abaqus/CEA (Systèmes®) were used to imitate a static three-point bending test. The first approach was a standard quasi-static stress/displacement procedure to control the time incrementation, and the approach is named T1 in this work [28]. Additionally, a second and more advanced technique, the extended finite element method (XFEM) [29], was used and is named T2. This technique allows us to model discontinuities as an enriched feature and it is an extension to the conventional finite element method [28]. Both techniques are classified as macroscale techniques that are nowadays typically used while designing structures [30,31]. The outcome of both analyses are displacements, reaction forces and maximum principal stresses of the studied FE models. These models should be later optimized to achieve higher fidelity, meaning to create more sophisticated models accounting for the environmental loads as well as mechanical loads, for example, multiscale models [32].

Yet, there is not enough systematic information on the dimensional stability and mechanical properties of acetylated MDF panels under different humid conditions. Establishing correlations between elastic properties measured nondestructively with bending strength and stiffness of acetylated MDF would lead to a quick and reliable means of assessment of the safety margins for different applications. Thus, this study was focused on elucidating the dimensional and static bending properties of acetylated MDF and on evaluating possibilities to predict its bending behavior from acoustic data by using standard statistical and multiscale prediction modelling methods. Dry (35% RH), standard (65% RH) and wet (85% RH) climatic conditions were considered to represent different moisture content situations as well as accelerated weathering.

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

#### *2.1. MDF Panels*

Commercially produced Medite Tricoya® MDF panels with dimensions of 300 × 210 × 18 mm<sup>3</sup> (length × width × thickness) were used in this study. Formaldehyde free glue is used for the acetylated softwood fibres during the production of Tricoya ® panels. Two different sample sets were prepared, i.e., along (parallel samples symbolized by //) and across (perpendicular samples symbolized by ⊥) the sanding direction. Working samples were prepared according to Table 1 and were stored in three different climatic conditions, i.e., dry (20 ◦C, 35% RH), standard (20 ◦C, 65% RH) and wet (20 ◦C, 85% RH). Five replicates of MDF samples (two different directions and three climatic conditions depending on measured properties) were produced in a total of 80 samples, which were tested for different properties according to EN standards (Table 1).


**Table 1.** Dimensions (length × width × thickness) of samples used for measuring different physical and mechanical properties of Medite Tricoya® medium density fiberboard (MDF) samples.

Moisture content and density were measured on samples after conditioning in each climatic condition. Samples were considered to be acclimatized when the differences were smaller than the 0.1% mass of the sample between two weightings within 24 h.

#### *2.2. Experimental*

#### 2.2.1. Dimensional Changes

The relative changes in length and thickness of the samples were determined in between two equilibrium conditions. The increases in length and thickness due to swelling were measured from 65% to 85% RH in adsorption (first regime), while the reductions in length and thickness due to shrinkage

were measured from 65% to 35% RH in desorption (second regime), according to the standard EN 318 [35]. The samples were exposed to different RH levels until acclimatized at two regimes. The first regime consisted of dimensional changes among consecutive RHs 35%, 65% and 85% at 20 ◦C constant temperature, whilst the second regime consisted of consecutive RHs in the reverse order, i.e., 85%, 65% and 35%, at 20 ◦C constant temperature.

Relative expansion and contraction sample's length were calculated using the formulae below:

$$
\delta l\_{65,85} \text{ (mm/m)} = 1000 \times (l\_{85} - l\_{65}) l\_{65} \tag{1}
$$

$$
\delta l\_{65,35} \text{ (mm/m)} = 1000 \times (l\_{35} - l\_{65}) l\_{65} \tag{2}
$$

where δ*l*65,85 (mm/m) is the relative increase in length due to swelling of sample's length after RH change from 65% to 85%, based on the length *l* (mm) measured at 65% RH and 85% RH; δ*l*65,35 (mm/m) is the relative reduction in thickness due to the shrinkage sample's length after RH change from 65% to 35%, based on the length *l* (mm) measured at 65% RH and 35% RH.

Similar to the calculations of relative change in the sample's length, thickness swelling and shrinkage properties were calculated as follows:

$$(\delta t\_{65,85} \text{ (\%)} = 100 \times (t\_{85} - t\_{65}) / t\_{65} \tag{3}$$

$$(\delta t\_{65,35} \text{ (\%)} = 100 \times (t\_{35} - t\_{65}) / t\_{65} \tag{4}$$

where δ*t*65,85 (%) is the relative increase in sample´s thickness due to swelling after RH change from 65% to 85%, based on the thickness *t* (mm) measured at 65% RH and 85% RH; δ*t*65,35 (%) is the reduction of the sample's thickness due to shrinkage after RH change from 65% to 35%, based on the thickness *t* (mm) measured at 65% RH and 35% RH.

#### 2.2.2. Thickness Swelling

In this test, conditioned samples at 20 ◦C and 65% RH were placed in swelling testers (IMAL SW 200, San Damaso, Italy) having water pH of 7 ± 1, and the temperature was controlled to 20 ± 1 ◦C. Samples were immersed about 25 mm in water and were separated from each other and from the sides of the water bath. After immersion in water for 24 h, the thickness of each test piece was measured by a digital caliper nearest to 0.01 mm. Thickness swelling, *G<sup>t</sup>* (%), was calculated based on the initial thickness *t<sup>1</sup>* (mm) before and final thickness *t*<sup>2</sup> (mm) after soaking in water.

$$G\_l = 100 \times (t\_2 - t\_1) \sharp\_1 \tag{5}$$

#### 2.2.3. Accelerated Aging Test

All the edges of the samples were coated with silicone resin, conditioned at 20 ◦C and 65% RH and were placed in an QUV Accelerated Weathering Tester, QUV/spray (Q-Lab Co., Westlake, NJ, USA). This QUV with AUTOCAL system facilitates testing the external performance of products on their weather ability, light stability or corrosion resistance by simulating sunlight, rain and dew. The test was continued for 426 h and each complete cycle was equal to one week (168 h) following the sequence of condensation at 45 ◦C for under 24 h, a repeat of UV-radiation 60 ◦C (0.89 W/m<sup>2</sup> /nm) at a wavelength of 340 nm for under 2.5 h and water spray (6–7 L/min) for under 0.5 h. One complete cycle was equal to one week (168 h). Commercial MDF of similar thickness (18 mm) intended for indoor use was used for comparison.

#### 2.2.4. Non-Destructive Testing

An acoustic resonance method was used for measuring the dynamic modulus of elasticity (MOEdyn). In this method, a data acquisition logger (PicoScope 4224, Cambridgeshire, UK) connected with the software BING®, version 9.7.2 (Beam Identification by Non-destructive Grading by CIRAD- French Agricultural Research Centre for International Development, Montpellier, France) that controls, processes data and delivers results. A free-free flexural vibration test set-up was used, and more details about this method can be found in [39]. In flexural vibration, the first four modes of vibration were measured and used for determining the dynamic transversal modulus of elasticity, which represents stiffness under bending stress, i.e., MOEdyn. The test was repeated four times for every sample, two times in each side, and the average was calculated.

#### 2.2.5. Static Bending Test

A three-point bending test was performed to determine the static modulus of elasticity (MOEstat) and modulus of rupture (MORstat) of the MDF samples following the standard EN 310 [38]. A universal testing machine (Instron 4466, Buckinghamshire, UK) with 10 kN load capacity was used. A static bending test was performed on the same sample used for measuring MOEdyn in nondestructive testing. Uniaxial load was applied on the flat side of the samples. The load was constant (10 mm/m) so that the maximum load was reached within 60 ± 30 s. An increment of load and deflection between 10% and 40% of maximum load was considered for measuring the MOEstat (MPa).

$$MOE\_{\rm stat} \text{ (MPa)} = \langle l^3(F\_2 - F\_1) \rangle \langle 4bt^3(a\_2 - a\_1) \rangle \tag{6}$$

where *l* is the span length (mm), *b* is the width of sample (mm), *t* is the thickness of sample (mm), *F<sup>1</sup>* and *F*<sup>2</sup> are the increment of load at 10% and 40% of maximum load, and *a*<sup>1</sup> and *a*<sup>2</sup> are the corresponding deflection at the mid-length of the test pieces due to the load *F*<sup>1</sup> and *F*2, respectively.

Bending strength, MORstat (MPa), of the test sample was calculated from the maximum load, *Fmax* (N), using the equation:

$$\text{MOR}\_{\text{stat}} \text{ (MPa)} = \text{(3F}\_{\text{max}} \text{I)} \text{(2bt}^2) \tag{7}$$

#### 2.2.6. Internal Bond Test

Internal bond (IB) or tensile strength perpendicular to the plane of panels was measured following the standard EN 319 [37]. Samples were effectively bonded with a hot-melt glue, and tensile load was applied until rupture using an Instron 4466 universal testing machine (Buckinghamshire, England) with 10 kN load capacity. A loading speed of 8 mm/min was maintained so that the maximum load is reached within 60 ± 30 s. In addition to the conditioned samples at dry (20 ◦C, 35% RH), standard (20 ◦C, 65% RH) and wet (20 ◦C, 85% RH) conditions, IB strength was also measured on samples after accelerated aging. IB or tensile strength (MPa) perpendicular to the plane of MDF test pieces was calculated by following the formula:

$$\text{IB\\_strength (MPa)} = \text{F}\_{\text{max}}/(ab) \tag{8}$$

where *Fmax* is the breaking load (N), and *a* and *b* are the width (mm) and length (mm) of the test pieces, respectively.

#### 2.2.7. Finite Element Analysis

Geometries of finite element (FE) models were identical to the samples manufactured for three-point bending tests and solved as a 3D problem; thus, full models were considered. Specifically, the beam´s part was 300 mm in length, 36 mm in width with a thickness of 18 mm, and the cylindrical pins' parts were 36 mm long with a diameter of 30 mm and placed 26 mm from both ends on a lower side of the beam part and in the middle of the beam´s span on an upper side of the beam part. The beam part was created as a deformable body and pins as discrete rigid bodies. The boundary conditions (BCs) and load were applied to the reference points (RP) created on pins' circular areas. Six degrees of freedom (df) were fixed in both bottom pins´ reference points for applying BCs and five df were set on the upper pin RP that also serves for applying a load. Namely, the following BCs were assigned: all six df are

fastened, thus 1, 2, 3, 4, 5, 6 = 0 for pins at the ends on the lower side of the beam part; for the pin in the middle of the span on the upper side of the beam part movement in the *z*-axis direction was enabled, only five df are fastened allowing movement in direction 3, thus 1, 2, 4, 5, 6 = 0. The global coordinate system and df notations and plus directions are shown in Figure 1. The load applied to the beam was a concentrated nodal force of value 1200 N, thus, larger than a mean value of the breaking load Fmax measured in the experiment (Figure 1).

**Figure 1.** Illustration of parts and boundary conditions assigned in the finite element (FE) model.

3D rigid elements *R3D4* (4-node bilinear quadrilateral) and *R3D3* (3-node triangular facet) were assigned to mesh pins. To create a mesh for the beam, *C3D8R* (8-node linear brick) solid elements with reduced integration were defined.

Normal hard-contact, frictionless tangential behavior, and a general standard surface-to-surface contact, were prescribed to characterize the interaction between pin parts and the beam part. The mechanical properties of the elastic–plastic isotropic material of the studied MDFs had been taken from the experimental data considering the RH of 65% for the standard board perpendicular to bending: thus. Young´s modulus of 2574 MPa, Poisson´s ratio 0.25, and density of 746.2 kg/m<sup>3</sup> .

Additionally, to specify a damage initiation, damage criteria for the XFEM model were prescribed to define the constitutive response for cohesive elements: namely, the maximum principal stress damage criterion (MAXPS) of value 34.47 MPa, taken from experiment measures. This means the damage criterion is using traction separation laws and is established as:

$$\mathbf{f} = \{ \langle \sigma\_{\text{max}} \rangle / \sigma^0\_{\text{max}} \} \tag{9}$$

where σ 0 max represents the maximum allowable principal stress, and the symbol h i represents the Macaulay bracket with the usual interpretation. The Macaulay brackets are here to represent that a purely compressive stress does not initiate damage; rather, damage is initiated when the maximum nominal stress ratio reaches a value of one (f = 1) [28]. Additionally, there was no initial crack assigned to the model. Instead, the XFEM crack domain was designated on the whole beam part. For describing a crack geometry and the crack´s growth motion in 3D space, two level sets for a crack is assumed. First set is the crack surface Φ while second, Ψ, is made so the intersection of two level sets that gives the crack front. The nodal value of the function Φ is the signed distance of the node from the crack face and of the function Ψ is the signed distance of the node from an almost-orthogonal surface passing through the crack front [28].

#### 2.2.8. Statistical Analysis

To determine any statistically significant differences between // and ⊥ samples, mean values were compared by a two-tailed group t-test at 0.05 significance level. In addition, to define relationships between measured parameters (MOEstat and MORstat, MORstat and IB), linear regression analyses were performed.

Prediction models for MOEstat and MORstat by using MOEdyn and RH as input variables were built from multiple regression analysis using the following equation:

$$\text{MOE}\_{\text{slal}} \text{ or } \text{MOR}\_{\text{slal}} \text{ (predicted)} = \text{MOE}\_{\text{dyn}} \times \text{b}\_{\text{MOE}\text{dyn}} + \text{RH} \times \text{b}\_{\text{RH}} + \text{C} \tag{10}$$

where b is coefficient and C is the intercept.

All regressions (linear, multiple) were performed at 95% confidence level using the Microsoft Excel 365 program (Microsoft, Redmond, WA, USA). In addition, the ANOVA is used to check the adequacy of the regression model developed.

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

#### *3.1. Physical Properties*

Moisture content, density and thickness swelling of commercially produced Medite Tricoya® MDF samples at three different RH levels are presented in Table 2. As expected, the EMC and density of samples increased with an increase in RH levels. Medite Tricoya® samples had 48% lower equilibrium moisture content when compared with commercial indoor MDF samples of similar thickness (Li et al., unpublished data) at 85% RH. Similar results can be found in a previous work [17]. As the acetylation process is a single site reaction, which means that one acetyl group is attached to one hydroxyl group resulting in a reduction of moisture absorption sites in wood polymers [40], acetylated MDF panels have low EMC even at high humidity level. Only 7.1% thickness swelling (after soaking in water for 24 h) was observed, meaning that acetylated panels do not swell severely when they are exposed to the water. These results showed that acetylation plays a significant role in the thickness swelling of the Medite Tricoya® samples. However, the extent of the reduced thickness swelling of acetylated MDF depends on the weight percent gain levels by acetylation process [2].

**Table 2.** Physical properties of Medite Tricoya® MDF samples at different RH levels. Values in parenthesis are the standard deviations.


\* Thickness swelling is measured after immersion in water for 24 h at 20 ◦C.

Concerning the results of dimensional changes, the acetylation caused the MDF panels to absorb little moisture during the conditioning and are thus dimensionally stable (Table 3). The linear expansion of sample length and thickness swelling values obtained in adsorption conditions (relative humidity change from 65% to 85%) were higher than those values obtained in desorption conditions (relative humidity change from 65% to 35%). The amount of water held by wood fibers at a given temperature and RH depends on the direction from which equilibrium is approached. The moisture adsorbed at high relative humidity exposure is not entirely released when re-drying by lowering the relative humidity levels, and this phenomenon was well observed for wood-based panels in other studies [41,42]. As a result, acetylation had a major effect on the dimensional stability of Medite Tricoya® samples. In addition, the density of panels can also adversely affect the dimensional stability [41]. Average linear expansion and retraction of Medite Tricoya® samples were, respectively, 23% and 57% lower than

those of standard samples (Li et al., unpublished data) and for thickness changes, those differences were 67% and 45% (comparison was done from Li et al., unpublished data). However, no significant differences of relative shrinkage or swelling in thickness and length were observed in the two principle directions (// and ⊥ direction of sanding). An exception was seen for the relative swelling in length when RH increased from 65% to 85% where swelling was significantly higher in perpendicular samples. However, the reason for that is not quite clear.

**Table 3.** Relative changes in thickness and length of Medite Tricoya® MDF sample at different relative humidity (RH) levels. Values in parenthesis are the standard deviations.


//: parallel sample to the sanding direction; ⊥: perpendicular sample to the sanding direction; δ65,35: samples conditioned from 65% to 35% RH; δ65,85: samples conditioned from 65% to 85% RH. \* Significant at the 0.05 level as determined by two-sample t-test. NS, non-significant.

#### *3.2. Mechanical Properties*

Table 4 shows MOEstat, MORstat and IB properties of parallel and perpendicular samples from Medite Tricoya® MDF conditioned at dry, standard and wet conditions. With the increase of RH from 35 to 65%, the average MOEstat and MORstat reduction of Medite Tricoya® samples was 8% and 10%, respectively. When the RH was further raised from 65% to 85%, reduction of those properties was respectively 37% and 20%. Higher humidity environment had a detrimental effect on the MOEstat and MORstat and IB strength values for both sample types.

**Table 4.** Static bending properties and internal bond (IB) strength of Medite Tricoya® MDF samples. Values in parentheses are the standard deviations.


//: parallel sample to the sanding direction; ⊥: perpendicular sample to the sanding direction. \* Significant at the 0.05 level as determine by two-sample T test. NS, non-significant.

Mechanical properties like MOEstat and MORstat measure the elastic behavior and resistance to bending, respectively, and are important properties when MDF is placed under load. Aforementioned properties determine largely the applicability of MDF as a structural component in furniture or other constructions. Those properties also depend on the sample properties, i.e., sanding direction (parallel or perpendicular), density, moisture content and type of MDF [2,5,43]. Dimensional stability is a good indication of the acetylation effect. However, moisture content had a great influence on the strength properties of both types of MDF samples. Significant reductions in MOEstat, MORstat and IB strength values were observed with an increase in RH (Table 4). These findings are in agreement with the results found in previous studies [2,43]. The decrease in such properties at increasing RH level can be attributed to the separation of fibers resulting from the thickness swell of the panel materials [44]. Differences in the static bending properties were also found between parallel and perpendicular samples, and especially at the dry condition these differences were statistically significant. In general, MOEstat and MORstat values were higher in parallel samples compared to perpendicular samples (see Table 4). Lower bending strength and stiffness values of perpendicular samples in MDF and other wood-based panels have been reported previously [26].

Figure 2 shows the linear regressions that determine how well the MORstat are related with MOEstat values in parallel and perpendicular samples at different humid conditions. Statistical analysis showed strong positive and significant relationships (*p* < 0.05) for both sample directions (parallel and perpendicular) at varying humidity levels. Strong correlation between bending strength and modulus of elasticity is known for wood-based panels [45]. When linear regression analyses were performed by separating humid condition, very poor and insignificant relationships were observed. In the parallel samples, the coefficients of determination (R<sup>2</sup> ) were 0.08 and 0.39 at 65% and 85% RH, respectively; whilst in the perpendicular samples, those values were, respectively, 0.20 and 0.25. However, in a dry condition (35% RH), significant relationships (*p* < 0.05) were observed, and R<sup>2</sup> values were 0.85 and 0.88 in parallel and perpendicular samples, respectively. Previous findings also showed higher correlation coefficients at lower moisture content levels for wood-based panels [46,47].

**Figure 2.** Linear relationship between MOEstat and MORstat of the Medite Tricoya® medium density fiberboard (MDF) samples in two different directions conditioned at three different climatic conditions.

A positive and significant relationship (*p* < 0.05) between MORstat and IB was also observed (Figure 3). As shown in Table 4, Medite Tricoya® samples had the highest IB strength at 35% RH. With the increase of RH from 35 to 65%, the average IB strength reduction of Medite Tricoya® samples was 20%. When the RH was further raised from 65% to 85%, reduction of IB strength was 27%.

y = 0.0353x - 0.3681 R² = 0.6753 0.9 1.2 IB (MPa) UV radiation causes photochemical degradation mainly in lignin polymers of the cell walls. UV light in combination with water plays a major role in such type of weathering. When lignin is degraded, water washes away degraded products and subsequently loosens the surface fibers to erode. However, acetylation reduces the loss of surface lignin and thus the erosion caused by accelerated weathering [48]. When compared with commercial MDF samples (indoor use), Medite Tricoya® samples performed much better in terms of the extent of being weathered by moisture and UV light. Thickness swelling

25 30 35 40

35% RH 65% RH 85% RH

MORstat (MPa)

0.3

0.6

of the test pieces from indoor MDF was almost twice as much as Medite Tricoya® (Figure 4). After the accelerated aging test, fibers in the indoor samples were separated and were possible to shred by finger rub. That implied that no residual strength was left. Medite Tricoya® samples showed better resistance against thickness swelling. This could be attributed to the fiber–fiber bonding efficacy of the resin to retain bonding in a very hydrophobic fiber network. After accelerated aging and conditioning (20 ◦C, 65% RH), the IB strength of the Medite Tricoya® samples was found to be 0.79 ± 0.06 MPa. This result showed that Medite Tricoya® samples made from acetylated wood fibers were able to retain the initial IB strength of non-weathered samples. Previous results also showed that a higher residual strength was observed in acetylated MDF after a cyclic test [19]. Such retention of strength can also be attributed to other parameters than the acetylation of fibers, such as the adhesive used [49]. 25 1500 2100 2700 3300 MOEstat (MPa)

1500 2100 2700 3300

Perpendicular

Parallel 85% RH

35% RH 65% RH

y = 0.0085x + 12.458 R² = 0.9195

y = 0.0074x + 14.779 R² = 0.9134

25

40

30

MORstat (MPa)

35

30

MORstat (MPa)

35

40

**Figure 3.** Linear relationship between MORstat and internal bond (IB) strength of the Medite Tricoya® MDF samples conditioned at three different humidity.

**Figure 4.** Medite Tricoya® and commercial standard MDF (for interior use) samples after accelerated aging test. Note the double thickness in the commercial MDF sample.

Acoustic resonance measurements were used to determine the MOEdyn (Figure 5). These values represent the mean stiffness, whilst MOEstat represents the local stiffness of the samples at the highly stressed areas of a specific test set-up [50]. MOEdyn was found to differ significantly between // and ⊥ samples only for the dry condition, i.e., 35% RH. As expected, the increase of RH from dry/standard (35 and 65% RH) to wet (85% RH) conditions resulted in a considerable decrease in MOEdyn in both directions.Resonance frequency of MDF samples decreased with the increase in moisture content and thus affected the MOEdyn. This is because at a dryer state, molecular chains in the amorphous regions of the wood cell wall are distorted with the presence of microvoids between the molecular chains, resulting in lower internal friction, resulting in higher MOEdyn. On the contrary, when moisture content increases, water molecules are embedded in the microvoids and rearrange the distorted molecular chains in the amorphous region. If the moisture content increases further, water acts as a plasticizer and decreases the cohesive forces between molecules, resulting in a higher internal friction and leading to the decrease in MOEdyn [51]. With the increase of RH from 35 to 65%, the average MOEdyn reduction of Medite Tricoya® samples was negligible, i.e., 1%. When the RH was further raised from 65% to 85%, the reduction of MOEdyn was 26%. In addition, MOEdyn values in the perpendicular samples were found lower than the corresponding values in the parallel samples. However, those differences were found to be statistically insignificant. A previous study by Han et al. [26] showed that reduction of stress wave velocity along the perpendicular direction compared to the parallel direction for wood-based panels (plywood, oriented strand board and particleboard) depends on the anisotropic properties

of the products. Smaller differences between the two directions in MDF samples implies a uniform product. However, as expected, MOEdyn values were higher than the static values approximately by 40%. Wood-based panels contain a significant number of glued interfaces, and such a difference is generally explained by the different rates of stresses applied in the dynamic and static tests [52]. The differences between the MOEdyn and MOEstat values were higher at a higher RH. It was 33%, 37% and 49% higher at 35%, 65% and 85% RH conditions, respectively. This is because moisture affects the stress-wave properties for wood-based panels, as they can swell considerably during moisture uptake. The swelling often leads to bond failures and to changes of their internal structure, and as a result, to a decrease of stress wave velocity with the increase in panel moisture content [26].

**Figure 5.** Dynamic modulus of elasticity (MOEdyn) of Medite Tricoya® MDF samples at different relative humidity levels measured by the acoustic resonance method. Error bars represent 95% confidence intervals for the means. \* Significant differences between parallel and perpendicular samples as determined by two-sample t-test at the 0.05 level. NS, non-significant.

Acoustic tools have been used quite successfully to predict MOEstat and MORstat [26,27,53]. Table 5 shows the multiple regression model summary results. Overall, linear and significant relationships (*p* < 0.05) of MOEstat and MORstat with MOEdyn were observed at different humid conditions. It was noted that R<sup>2</sup> values were slightly higher between MOEdyn and MOEstat than those between MOEdyn and MORstat.


**Table 5.** Model summary of regression statistics for MOEstat and MORstat prediction.

\* Significant at the 0.05 level.

The comparison between the experimental and predicted values is shown in Figure 6. The result indicates that the predicted values are very close to the experimental values. F-values for all models (MOEstat and MORstat in parallel and perpendicular samples) were highly significant at the 0.05 level (Table 5).

Statistical analysis showed that MOEdyn and RH can be used as predictors of MOEstat and MORstat of MDF samples. The results also indicated that MOEstat and MORstat have a positive relationship with MOEdyn.

**Figure 6.** Estimated and predicted values of MOEstat and MORstat for Medite Tricoya® MDF samples conditioned at three different climatic conditions.

#### *3.3. Finite Element Analysis*

Magnitudes of reaction forces and the deflections under these reactions forces in the z-axis direction, as well as values of maximal principal stress for T1, T2 procedures and experimental values, are presented in Table 6. Reaction forces are equal to half of maximal loading force.


**Table 6.** Magnitudes of displacements, reaction forces, and maximal principal stress for first approach (T1) and second approach (T2) finite element (FE) procedure, and experimental values.

Finally, it is necessary to add that the values for the T2 model show the maximal values at the time point when the jump in the damage dissipation energy occurs (Figure 7): thus, in this particular case, at the time increment 0.4623 of 1 with index number 30. At this time, the increment of the crack growth is initiated and the crack propagates [28], i.e., we are at beginning of a fracture mechanism.

The crack occurrence is visualized in Figure 8 using the PHILSM function, i.e., a singled distance function that describes the crack surface [28].

When we compare results obtained from the two computational models using different simulations techniques T1 (quasi-static stress/displacement) and T2 (extended finite element) with the experimental values, we see percental errors of 8.6% and 14.3%. As expected, these percent errors are relatively high. This demonstrates that the used computational methods do not provide reliable results, and thus, can be considered to have a low fidelity. Despite this fact, these techniques are today used as the main techniques in standard design procedures because they are highly efficient [29,31]. To be sure, the errors can be minimized by optimizing the input parameters that imitate the real nature of the studied MDF material; however, that requires more necessary testing, like, for example, creep testing and fracture testing, to obtain those data. This means spending more time to set up experiments, more wasted material for creating samples, as well as time to collect and analyze data sets.

240

**Figure 7.** Damage dissipation energy for the extended finite element method model (T2).

**Figure 8.** Singled distance function (PHILSM) for T2 model.

A better solution here is to use the help of micromechanical modelling approaches to specify material behaviour on a constituent level that is summarized in [32,54,55], and that will provide more reliable results compared with experimental ones and reduce the need for new sets of tests. More precisely, the employment of concurrent material and structural design applying multiscale models comes in since viscoelastic creep damage models and rheological behavior of MDF material are considered [56]. Here, we mean the time-dependent mechanical repose of MDF on a constant load considering the influence material´s density as well as fiber orientation, and the existence of voids and interfaces in the material microstructure. More importantly, the macroscale computational techniques used in this work do not take into consideration the very significant dependence of mechanical properties on the moisture and temperature. This, as demonstrated in this work, should be included in the computations to precisely predict mechanical material characteristics such as strength, toughness and elasticity as well as to predict the damage mechanism of the panel product.

#### **4. Conclusions**

The results obtained in this study showed that, due to the lower hygroscopicity of Medite Tricoya® samples, they absorbed less moisture and became more dimensionally stable even at the highest humidity condition. This MDF type can also retain its IB strength after an accelerated aging test. However, IB strength, MOEstat and MORstat were reduced from dry to humid conditions. At the highest humid condition (85% RH), strength and stiffness values did not differ significantly between parallel and perpendicular samples. In addition, multiple regression models were developed from MOEdyn and RH to predict the strength and stiffness of Medite Tricoya® MDF. In both parallel and perpendicular directions, highly significant relationships were observed. Developed models could predict the MOEstat and MORstat values of Medite Tricoya® MDF samples at any humid conditions, which produced an excellent fit to the measured values. This experimental outcome could ensure reliable and safe predictions of Medite's Tricoya® MDF strength and stiffness properties for intended end uses.

This study also showed that employing macroscale computational modelling approaches that are nowadays broadly used in engineering practice, such as quasi-static stress/displacement (T1) and extended finite element (T2) techniques, are not sufficient to obtain reliable results for MDF. These modelling methods are highly efficient and fast, but low in fidelity. Therefore, in future work, it is necessary to employ more precise multiscale models to secure more efficient material and structure design approaches. Additionally, multiscale models will reduce the necessity to set up new testing whenever we need to change the components proportion or component material in composite material.

**Author Contributions:** Conceptualization, S.A.A. and S.A.; methodology, S.A.A. and S.A.; software, S.A.A.; validation, S.A.A.; formal analysis, S.A.A. and J.K.; investigation, J.L.; data curation, J.L.; writing—original draft preparation, S.A.A. and J.K.; writing—review and editing, S.A.; visualization, S.A.A. and J.K.; supervision, S.A. All authors have read and agreed to the published version of the manuscript.

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

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

#### **References**


**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*
