Abstract
The effect of sanding, as the last operation before finishing, on the quality of heat treated wood surfaces has been insufficiently explored and explained. This paper compared the effects of sanding with three commonly used sanding grit sizes P60, P100, and P150 on the surface roughness values of beech (Fagus sylvatica L.) wood. The wood samples were treated by the ThermoWood process at 200 °C for 2.5 h. A large range of standard roughness parameters (Ra, Rq, Rv, Rt, RSm, Rsk, Rk, Rpk, and Rvk) and two waviness parameters (Wa, and Wt) were included in the analysis, as well as environmental scanning electron microscope (ESEM) images of the sanded surfaces. The results showed that the heat treatment slightly increased the surface roughness and decreased the wood surface waviness after sanding. All roughness and waviness parameters increased with increasing sanding mean grit diameters by following a strong linear correlation. The processing roughness was closely approximated by the parameter Rk. For both, treated and untreated beech, sanding had a tendency to obscure (in magnitude and number) wood anatomical details in the measured data. However, the influence of wood anatomy in the valleys domain increased as the grit size became finer.
Download PDF
Full Article
Surface Roughness of Heat Treated and Untreated Beech (Fagus sylvatica L.) Wood after Sanding
Lidia Gurau,a,* Mark Irle,b and Julia Buchner b
The effect of sanding, as the last operation before finishing, on the quality of heat treated wood surfaces has been insufficiently explored and explained. This paper compared the effects of sanding with three commonly used sanding grit sizes P60, P100, and P150 on the surface roughness values of beech (Fagus sylvatica L.) wood. The wood samples were treated by the ThermoWood process at 200 °C for 2.5 h. A large range of standard roughness parameters (Ra, Rq, Rv, Rt, RSm, Rsk, Rk, Rpk, and Rvk) and two waviness parameters (Wa, and Wt) were included in the analysis, as well as environmental scanning electron microscope (ESEM) images of the sanded surfaces. The results showed that the heat treatment slightly increased the surface roughness and decreased the wood surface waviness after sanding. All roughness and waviness parameters increased with increasing sanding mean grit diameters by following a strong linear correlation. The processing roughness was closely approximated by the parameter Rk. For both, treated and untreated beech, sanding had a tendency to obscure (in magnitude and number) wood anatomical details in the measured data. However, the influence of wood anatomy in the valleys domain increased as the grit size became finer.
Keywords: Surface roughness; Heat treated beech; Sanding quality; Roughness parameters; Microscopic images
Contact information: a: Transilvania University of Brasov, Faculty of Wood Engineering, Universitatii Str. 1, 500036 Brasov, Romania; b: Ecole Supérieure du Bois, rue Christian Pauc, 44000 Nantes, France;
*Corresponding author: lidiagurau@unitbv.ro
INTRODUCTION
Heat treatment of wood can considerably improve some wood properties (Tjeerdsma and Militz 2005). Many aspects of heat treatment have been studied including: dimensional stability, wood durability, mechanical properties, equilibrium moisture content, mass loss, wettability, colour change, and chemical modification (Esteves and Pereira 2009). Some studies have examined the impact of heat treatment on machinability and the resulting wood surface quality (de Moura and Brito 2008; Skaljić et al. 2009; Budakci et al. 2011; Tu et al. 2014; Gaff et al. 2015; Kminiak et al. 2015; Kvietková et al. 2015a, Kvietková et al. 2015b; Ispas et al. 2016; Pinkowski et al. 2016; Gurau et al. 2017). Machining of a wood surface is characterized by the surface quality, which is generally analyzed by the surface roughness measurements, resulting from the interaction between the cutting tool and the wood surface.
Surface quality can be measured with either optical, non-contact measuring instruments such as a laser, or by means of a stylus contact method. The latter was found to give more repeatable results and seems more reliable for wood (Gurau et al. 2005) because the laser appears to distort the surface data (Sandak and Tanaka 2003). Accurate measurement of wood surface roughness requires the correct selection of the evaluation length as well as the lateral resolution (Gurau and Irle 2017). The measured surface or surface-profile is filtered in order to leave only the high frequency irregularities that characterize the roughness values. The selection of filter is crucial, because wood roughness is known to be distorted by common Gaussian filters (Krisch and Csiha 1999; Molnar et al. 2017). A robust Gaussian filter applied iteratively to a data set was tested and found useful for wood surfaces because it does not introduce any bias, which can occur around certain anatomical features (Fujiwara et al. 2004; Gurau et al. 2006). The need for a robust filter for analysis of wood surfaces was confirmed and tested by the more recent studies of Piratelli-Filho et al. (2012) and Tan et al. (2012). However, no previous research on the surface roughness of heat-treated wood has used a robust filter. All previous researchers have applied simple Gaussian filters inherent in most measuring instruments on the market. After filtering, roughness parameters can be calculated and these are the basis for comparisons between the quality of processed surfaces or/and wood treatment.
Due to the chemical changes that wood undergoes during a heat treatment, its density decreases, most mechanical properties are degraded, and its brittleness increases with the deterioration of fracture properties (Esteves and Pereira 2009; Bakar et al. 2013; Schneid et al. 2014; Sandak et al. 2017). Thus, the heat-treated wood is more susceptible to mechanical damage during machining and it sometimes requires an adaptation of the processing parameters.
Studies on the influence of grit size on the surface roughness values of solid wood are numerous (de Moura and Hernandez 2006; Kilic et al. 2006; Ratnasingam 2006; Marthy and Cismaru 2009; Sulaiman et al. 2009; Salca and Hiziroglu 2012; Varasquim et al. 2012; Vitosyte et al. 2012; de Moura Palermo et al. 2014; Miao and Li 2014). Even though a trend of decreasing roughness with increasing grit number is unanimously reported, there has not been a thorough analysis by means of several roughness parameters. Furthermore, only few studies in the literature compared the surface roughness of untreated and heat-treated wood after sanding. De Moura Palermo et al. (2014) found an increase in surface roughness of heat-treated Eucalyptus grandis compared with untreated wood after sanding with P80 and P100 grit sizes. In that study, after high temperature treatment (190 C for 6.5 h), the abrasive grits penetrate the wood more deeply than the untreated wood because of the reduced mechanical strength of the former. A number of studies have shown that heat treatment affects the mechanical properties of wood (Bekhta and Niemz 2003; Boonstra et al. 2007; Windeisen et al. 2009; Calonego et al. 2012). A recent example is that published by Borůvka et al. (2018) where they studied beech and birch wood samples.
Increased surface roughness of heat-treated of Eucalyptus grandis and Pinus caribaea as compared to untreated was also found after sanding by de Moura and Brito (2008) and de Moura et al. (2011). These studies contained a comparison between untreated and heat-treated wood for temperatures ranging from 140 C to 200 C subjected to sanding with following grit size combinations: 60-80; 80-100; and 100-120. The results showed a higher surface roughness in the case of heat-treated wood and this increased with the treatment temperature.
A different result was obtained by Tu et al. (2014) on Eucalyptus urophylla x E. camaldulensis subjected to heat treatments with temperatures ranging from 180 C to 210 C followed by sanding with a sequence 60-120 grit size. The surface roughness evaluated by parameter Ra showed slightly lower values for heat-treated wood, but with no significant differences between treatment temperatures.
The measuring length used in these studies was short, only 15 mm. Longer measuring lengths are preferable to cover the wood variation (Ostman 1983). In addition, the only roughness parameter quoted in much of the literature is Ra, despite the fact that it is known that it does not provide sufficient information about the surface topography, because very different surfaces can have the same Ra (Monetta and Bellucci 2012).
Fewer studies are available in the literature on the study of surface roughness of sanded heat treated wood in comparison with untreated wood. However, in order to further understand the effects of sanding on wood an in-depth analysis is required. This study improves on previous research by using a robust filtering procedure, by increasing the measuring length to cover more wood variation and by adding a large range of roughness parameters in order to give a comprehensive interpretation of data. In addition, this study also includes roughness profiles computed in MathCAD, which allow a visual comparison of the effect of grit size as well as of treatment on wood, supplemented by environmental scanning electronic microscopic (ESEM) images. The present study was based on beech wood (Fagus sylvatica L.) heat-treated by the ThermoWood method at 200 °C for 2.5 h and sanded with various grit sizes sequence that is typical prior to applying a finish, i.e., P60, P100, and P150.
EXPERIMENTAL
Wood specimens of dimension, 400 x 50 x 28 mm, were cut from Fagus sylvatica L. Eighteen samples were kept untreated, while another 18 samples were heat-treated by the ThermoWood method (ThermoWood Handbook 2003) in superheated steam in an industrial-scale TekmaWood kiln, manufactured by TekmaHeat Corporation (Lahti, Finland), according to the schedule presented in Table 1.
Table 1. Heat-treatment Schedule
All specimens were conditioned for 4 weeks at 20 °C and 55% relative humidity prior to sanding. The average moisture content of the samples after conditioning was 3% ± 0.2% for the heat-treated wood and 8% ± 0.5% for the untreated specimens. The mean density of the specimens at the moment of testing was 728 kg/m3 for untreated beech wood and 617 kg/m3 for heat-treated wood.
All specimens were then processed by sanding along the grain by using a wide belt sander with aluminium oxide belts, P60 grit size, at a contact pressure of 0.0055 N/m2 and at a feed speed of 4.5 m/min. The depth of sanding was 0.3 mm/pass, and there were three passes for each specimen in order to make sure that irregularities from planing were completely removed. After this first sanding with grit size P60, 6 treated and 6 untreated specimens were kept for measurements and the others were further sanded with P100 by three passes though the machine. Again, 6 treated and 6 untreated specimens were retained and the others were sanded with P150 grit size in three passes.
According to Kantay and Űnsal (2002) and Budakçi et al. (2013), the surface roughness measurements have to be performed on the same type of surface (radial or tangential) and preferably not combined. Therefore, all sanded surfaces were tangential to reduce the influence of wood variability on the measured results.
The roughness measurements were achieved by using a MarSurf XT20 instrument manufactured by MAHR Gottingen GMBH (Göttingen, Germany), with a MFW 250 scanning head that has a vertical range of 500 m. The stylus used had a 2 m tip radius and 90 tip angle. The specimens were measured at a lateral resolution of 5 m recommended in Gurau et al. (2013), at a speed of 0.5 mm/s, and using a low scanning force of 0.7 mN.
Each specimen was measured three times using 42 mm traces across the grain (across the sanding direction) that were randomly positioned down the length of the specimen. Consequently, 18 profiles were obtained for each grit size. This length should be long enough to cover wood growth variability (earlywood and latewood areas) as well as to detect some longer wavelength components in the profile as waviness (Gurau et al. 2012). According to de Moura (2006), the values of roughness measured across the grain are usually higher than those measured along the grain.
The instrument had MARWIN XR20 software (MAHR, Göttingen, Germany) installed in it for processing the measured data.
The sequence of operations for an individual profile followed a standard procedure and began with removing the form error by best fitting a polynomial regression to the dataset so as to generate a primary profile containing waviness and roughness.
Roughness profiles were obtained by filtering each primary profile by using a robust filter RGRF (Robust Gaussian Regression Filter) described in ISO 16610-31 (2010). The cut-off used was 2.5 mm as recommended for wood in previous work of Gurau et al. (2006) and used by other researchers as well (Unsal and Ayrilmis 2005; Sevim Korkut et al. 2008; Korkut et al. 2009; Škaljić et al. 2009; de Moura et al. 2011; Salca et al. 2017).
A range of roughness parameters were calculated from each roughness profile including: Ra, Rq, Rsk, Rv, Rt, and RSm from ISO 4287 (1998); and Rk, Rpk, and Rvk from ISO 13565-2 (1996). The sum of parameters Rk+Rpk+Rvk was used by Magross (2015) in a study on surface roughness of sanded wood and was also included in this study for comparison. Waviness parameters Wa and Wt, from ISO 4287 (1998) measured the longer wavelength components in the profiles.
Wood anatomy is known to bias not only the filtering process, but also the evaluation of the processing roughness parameters of wood, especially when the magnitude of inherent wood irregularities is greater than that caused by processing alone. However, when anatomy is not removed from the measured profile, Rk parameter was found to be a good approximation of the processing roughness (Westkämper and Riegel 1993; Gurau 2004; Sharif and Tan 2011). The parameter Rk measures the core roughness of a profile, and it is sensitive to wood processing and surface heat treatment. It is expected that in the case of sanding, the highest concentration of data points will correspond to the marks caused by the mean grit diameters occurring with the highest frequency. However, grit particles come as a range of values, where minimum and maximum grit diameters will have a low frequency. The grit particles that are larger than the mean value can be expected to create valleys that are deeper than the processing roughness.
Mean parameters Ra (the arithmetical mean deviation of the assessed profile) and Rq (the root mean square deviation of the profile) are common roughness indicators, but alone, they do not provide sufficient information about wood surface topography. The value of Rsk (skewness of the profile) is strongly influenced by the presence of isolated peaks or valleys. Surfaces with a negative skewness have fairly deep valleys below a smooth plateau. The RSm parameter (the mean width of the profile elements) is used in this study to characterize the effect of the sanding marks as far as their width is concerned. The Rt is the total height of the profile calculated as the sum of the maximum profile peak height (Rp) and the largest absolute value profile valley depth (Rv). Therefore, Rv is sensitive to the presence of large diameter vessels and deep processing marks caused by oversized grits or accidental damage.
The parameter Rk (the core roughness depth) is the depth of the core profile and is least influenced by wood anatomy because high peaks and low valleys are excluded from its calculation. This parameter should best indicate the core roughness caused by sanding (Gurau 2004; Magross 2015).
The Rpk parameter (the reduced peak height) is expected to be sensitive to fuzziness (fibres pulled out during sanding). The Rvk (the reduced valley depths) for beech, is typically associated with wood anatomical valleys, but can also be influenced, as Rv or Rt, by isolated high grit penetration in case of rough sanding.
The parameter Wa is similar to Ra, and Wt is similar to Rt, but they apply to the waviness profile, which excludes the shorter wavelength irregularities as roughness.
For each treatment and roughness parameter, a mean value and the standard deviation were calculated.
Individual roughness profiles taken across the grain were computed in MathCAD 2000 Professional (MathSoft Inc., Cambridge, MA, USA) in order to visualize the results. The core roughness (processing roughness) was separated from the other surface irregularities by upper and lower thresholds using a method described in Gurau et al. (2005) to allow visual comparisons between sanding with various grit sizes and wood treatment.
ANOVA and Duncan’s multiple range tests were performed to test statistical significant differences between datasets (treatments and processing types).
The ESEM images were taken with a Quanta 250 made by FEI (Hillsboro, OR, USA). The photos were taken in Lovac mode at a pressure of 90 Pa. The surface relief was more easily observed if the specimen was tilted by 30°.
RESULTS AND DISCUSSION
From Fig. 1, it is difficult to notice any major differences in surface quality between untreated and heat-treated beech wood. However, it was observed that sanding traces were clearly visible as horizontal bands, and their depth and width were gradually reduced with the grit number. For all grit numbers, for both heat treated (HT) and untreated wood (UT), vessels were clearly visible, being uncovered by the grits ploughing into wood. The presence of vessels with simple pits was detailed by higher magnifications (400x) as shown in Fig. 2.
Fig. 1. Comparative ESEM images, with magnification 100x, of untreated (UT – a,c,e) and heat treated (HT – b,d,f) beech (Fagus sylvatica L.) sanded with grit sizes P60 (a,b); P100 (c,d); and P150 (e,f)
Fig. 2. Comparative ESEM images, with magnification 400x, of untreated (UT – a,b,c) and heat treated (HT – d,e,f) beech (Fagus sylvatica L.) sanded with grit sizes P60 (a,d); P100 (b,e); and P150 (c,f)
Table 2 shows that heat treatment at 200 °C for 2.5 h caused an increase in surface roughness after sanding for all the three tested grit sizes. These differences were significant for all roughness parameters and grit sizes with the exception of the parameters, which were sensitive to isolated features, Rv, Rt, and Rpk for grit sizes P100 and P150 and Rvk for all grit sizes.
The surface roughness of HT wood was approximately 2% to 11% higher than that for UT wood for all the roughness parameters except Rt for surfaces sanded with P100. These results were in agreement with Budakçi et al. (2013), who proposed that chemical changes caused by the heat treatment, especially those associated with hemicellulose changes, cause an increase in surface roughness.
Certainly it is well accepted that heat treatment changes the mechanical properties of wood and, in particular, making it more brittle. It is logical to conclude, therefore, that the mechanical response of heat treated wood to the action of a cutting tool will be different to that of untreated wood. Visual comparisons of the photographs in Figs. 1 and 2 do not reveal any dramatic differences that might explain the different roughness parameters given in Table 2.
Heat treatment reduces waviness such that Wa was smaller by 14% to 28%, and Wt by 10% to 14% than the values observed for the untreated wood. The differences were significant as tested with Duncan multiple range test (Table 3). Heat treatment had a marked effect on the mechanical properties of wood and, therefore, its response to mechanical solicitation, such as sanding, could be affected. This may explain the reduced surface waviness observed here and in previous studies on planed heat treated beech (Ispas et al. 2016).
Table 2. Mean Values of Roughness Parameters and Standard Deviations for Treated (HT) and Untreated (UT) Beech Wood Subsequently Sanded with Three Different Grit Sizes
Note: Groups with the same letters in columns indicate that there was no statistical difference (p < 0.05) between the samples according to Duncan’s multiple range test. The meaning for stdev- standard deviation; signif- significance.
Table 3. Ratios of Mean Values of Roughness Parameters, Waviness Parameters, and Standard Deviations for Treated (HT) and Untreated (UT) Beech Wood Prepared by Sanding with Three Different Grit Sizes
Letters refer to the statistical significance in Duncan’s test. The meaning for stdev- standard deviation; signif- significance.
The profiles in Figs. 3 to 8 show clear differences between grit sizes of the same treatment, but the difference between treatments was hard to observe. The processing roughness, highlighted in blue, was assimilated to the mean depth of penetration of the sanding grits, which is characterized by the profile data points detected within the upper and lower threshold and is measured by the distance between the two thresholds. It is best approximated by the roughness parameter Rk, which was greater for the treated wood by absolute values between 1.5 m (for P150) to 2.2 m (for P60), which were too small to be detected by eye. Therefore, a combination of ESEM images, roughness profiles, and roughness parameters was better for comparison rather than using one or the other.