Table of Contents Author Guidelines Submit a Manuscript
Journal of Spectroscopy
Volume 2018, Article ID 7423501, 10 pages
https://doi.org/10.1155/2018/7423501
Research Article

Hyperspectral Imaging Surface Analysis for Dried and Thermally Modified Wood: An Exploratory Study

Luleå University of Technology, Wood Science and Engineering, Forskargatan 1, 931 87 Skellefteå, Sweden

Correspondence should be addressed to Olena Myronycheva; es.utl@avehcynorym.anelo

Received 25 May 2018; Revised 13 August 2018; Accepted 16 October 2018; Published 14 November 2018

Guest Editor: Ingunn Burud

Copyright © 2018 Olena Myronycheva et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Naturally seasoned, kiln-dried, and thermally modified wood has been studied by hyperspectral near-infrared imaging between 980 and 2500 nm in order to obtain spatial chemical information. Evince software was used to explore, preprocess, and analyse spectral data from image pixels and link these data to chemical information via spectral wavelength assignment. A PCA model showed that regions with high absorbance were related to extractives with phenolic groups and aliphatic hydrocarbons. The sharp wavelength band at 2135 nm was found by multivariate analysis to be useful for multivariate calibration. This peak represents the largest variation that characterizes the knot area and can be related to areas in wood rich in hydrocarbons and phenol, and it can perhaps be used for future calibration of other wood surfaces. The discriminant analysis of thermally treated wood showed the strongest differentiation between the planed and rip-cut wood surfaces and a fairly clear discrimination between the two thermal processes. The wavelength band at 2100 nm showed the greatest difference and may correspond to stretching of C=O-O of polymeric acetyl groups, but this requires confirmation by chemical analysis.

1. Introduction

Increased automation in wood-manufacturing processes would allow a more custom-oriented approach that encourages companies in the wood industry to move from traditional quality grading with high-speed visual inspection camera systems (VIS) to more sophisticated systems for the grading of sawn and further processed timber. The VIS methods are in some cases not sufficiently precise and efficient because of wood variability and wood-treatment heterogeneity [1].

Near-infrared spectroscopy (NIR) is a well-known technology for the nondestructive analysis of wood. The advantages of the NIR technology are speed and the ability to measure a broad range of the chemical and physical properties, which make the technology promising for industrial use [2]. The main challenge for developing an effective predictive model based on NIR data and for its further use in wood-manufacturing processes is, however, to find the reliable absorption range of interest in a highly heterogeneous material such as wood and to calibrate the model for quantitative process parameters assessment [3]. Portable NIR sensors use point-based measurements that have to be developed experimentally, and they will not give a 2-dimensional spatial visualisation of the variation in NIR absorptions from a surface. This weakness can, at least partly, be reduced by using a more stationary technique like hyperspectral imaging for accurate data collection and model development and relating this information to other sources such as a portable NIR sensor.

Hyperspectral techniques can be used for the development of models that quantify the heterogenic chemical composition of the wood surface. The objective of the current study was to use a hyperspectral imaging technique and follow the recommendations of Sandak et al. [4] to build a classification model for wood surfaces in the NIR spectral region. Hyperspectral imaging provides spatially resolved information of the wood surface in the NIR range, and selected wavelengths can be developing calibration models for specific compounds.

Industrial drying, and especially thermal modification, of wood influences the presence and structure of the wood constituents (cellulose, hemicelluloses, lignin, and extractives), and this may in turn influence the characteristics and consequently the service life of the wood material in different products. There are a number of difficulties in describing the changes that take place when wood is heated. A range of chemical reactions take place simultaneously, and since these include both endothermic and exothermic reactions, the determination of the onset temperatures for the different reactions is nearly impossible. The analysis is further complicated by interactions between reactions in different constituents. This means that the information obtained from an analysis of an isolated derivative of one of the components can be very different from what actually takes place inside wood. There are not only interactions between the components in the wood but also interactions between the wood and the treatment atmosphere [5]. Figure 1 shows a schematic representation of the changes in wood components with increasing temperature under humid conditions.

Figure 1: A schematic representation of changes in wood components with increasing temperature under humid conditions, but without regard to time [5].

Natural seasoning of sawn timber is the oldest way of processing sawn timber to remove water and increase its durability. Compared to drying of industrial timber in a kiln at a temperature far above 20°C, natural seasoning takes a long time and the chemical composition remains close to the natural composition of the wood [6, 7]. In order to improve the industrial sawmill process, kiln drying was developed in the late 1800s to provide a fast and controlled drying process to reach the target moisture content 6–12% [8, 9]. Below 40°C, it is mainly physical changes that occur in the wood, such as the emission of water and volatile extractives [10]. Industrial kiln drying at a temperature of ca. 70°C is the predominant regime nowadays for Scots pine and Norway spruce [11], and at that temperature, the chemical changes occur mainly in the extractives and hemicelluloses. Wood dried at a temperature higher than 100°C has been shown to have a lower carbohydrate content and lower susceptibility to mould growth than wood dried at lower temperatures [12, 13].

Thermal modification (TM) is a process whereby wood properties are altered by treatment at temperatures in the range160–260°C with essentially no oxygen present. During TM, wood undergoes changes in its chemical and physical properties by degradation and partial removal of polymeric components, leading to a loss of mass of the wood material [14]. This modification of the chemical structure during TM results in a darkening of the wood, and thermally modified timber absorbs less moisture than unmodified timber [15]. The TM process removes extractives from the sawn timber, and this can be an advantage, for example, when TM timber is coated due to less bleed of resin through the coating [5].

The degradation temperatures of cellulose and lignin are considered to be higher than those of hemicelluloses and extractives (Figure 1). The degradation of cellulose and lignin is, however, hard to predict, and some studies show that these components can also start to degrade at lower temperatures such as 110°C where minor changes start in the guaiacyl lignin units [14, 16].

Extractives, both native and those formed by degradation in the TM process, tend to migrate towards the surface of wood during drying and TM [17], and some extractives volatilize already at moderate temperatures [10].

In the present work, the chemical composition of surfaces of naturally seasoned, kiln-dried, and thermally modified sawn timber from two different processes (the ThermoD® and WTT processes) has been studied with NIR (the wavelengths are from 900 to 2500 nm). The naturally seasoned and the kiln-dried timber were processed under atmospheric condition, the ThermoD-modified timber was treated in superheated steam at a temperature of 212°C under atmospheric pressure; in the WTT process, the sawn timber was processed in saturated steam at a temperature of 170°C at a pressure of up to 6 bar. Under such conditions, it was possible to achieve a modification similar to the ThermoD process, but at a considerably lower temperature [18, 19].

2. Materials and Methods

2.1. Wood Material

A total of 40 green (never-dried wood) flat-sawn sideboards, 22 × 100 mm in cross section and with a length of at least 4.5 m, from ten different Scots pine (Pinus sylvetris L.) trees were collected directly after sawing from a sawmill in Northern Sweden. The boards were cut into samples with a length of 1 metre giving four end-matched boards. Samples containing only sapwood were selected for the study, and these boards were divided into two groups for natural seasoning or kiln drying.

The naturally seasoned samples were single-stacked between 25 mm thick stickers and dried indoors at a temperature of 20°C and about 10% relative humidity, without any additional air circulation. The average moisture content after 30 days was 4.2 ± 0.6%. This was a very “soft” drying where the migration of extractives to the wood surfaces during the drying was expected to be low.

The kiln-dried samples were double-stacked between 25 mm thick stickers with the bark-side surfaces of each pair turned outwards in order to achieve a steered flow of water/moisture from the inner part of the boards to the bark-side surfaces. The kiln drying was performed in a small-scale laboratory kiln with air circulation (3 m/s) and a maximum dry-bulb temperature of 75°C, for 45 hours. The average moisture content after drying was 13.9 ± 3.8%. Compared to natural seasoning, this drying process was expected to be harsher and to lead to more extractives and decomposition components on the surfaces after drying.

Planed thermally modified Scots pine (Pinus sylvestris L.) boards were obtained from the ThermoD® and WTT® industrial processes.

The ThermoD boards were treated according to the Finish Thermowood© process at 212°C by Heatwood AB in Hudiksvall, Sweden. The process was run as it is described in ThermoWood Handbook [20] for the class ThermoD. The ThermoD process consists of three phases:(1)Temperature increase and high-temperature drying, using heat and steam: the temperature is raised rapidly to 100°C. Thereafter, the temperature is increased steadily to 130°C. The total time is 12 hours.(2)Thermal modification: the kiln is increased up to 212°C. When the target level has been reached, the temperature remains constant for 3 hours. The total time is 12 hours.(3)Cooling and moisture conditioning: the temperature is lowered by using water spray systems. When the temperature has reached 80–90°C, the wood is remoistured to reach the wood moisture content to 4–7%. The total time is 12 hours.

The WTT boards were thermally modified at 170°C according to the Danish WTT process by the Uteträ company in Arvidsjaur, Sweden. During the first hour, the temperature was raised to 40°C and prior vacuum was applied. During the next 45 minutes, the temperature was raised to 60°C and held constant (60°C) for 25 minutes. The kiln was warmed up to 140°C during 4 hours and held constant (140°C) for 30 minutes. The kiln was further warmed up for around 2 hours until the temperature reached 155°C and held constant (155°C) for 15 minutes. Thereafter, the kiln was warmed until 170°C and held constant (170°C) for 30 minutes. During the process, the pressure was applied when the kiln temperature reached 60°C and the pressure was raised steadily together with the temperature up to 6 bar at 170°C. After thermal modification at 170°C, the kiln was cooled down first to 80°C for 8 hours and to 65°C for 2 hours. The pressure was steadily decreased to 0 when the kiln temperature reached 60°C.

After drying and thermal modification, all the samples were sawn to give specimens with dimensions of 15 × 50 × 220 mm (T × W × L), which means that the thermally modified specimens had both planed and sawn surfaces. All the specimens were stored under the same conditions at room temperature for five months prior to the NIR measurements.

2.2. Image Acquisition

A push-broom system equipped with high-speed hyperspectral camera (Specim SWIR 3) in the short-wave infrared (SWIR) 900–2500 nm range with a cryogenically cooled mercury-cadmium-telluride (HgCdTe) detector with 384 spatial pixels and 288 spectral band resolution was used for the NIR measurements (Figure 2). The camera pixel size was 24 × 24 µm, and a 31 mm focal length lens was used. Internal dark and white reference (Spectrolon) standards were used, i.e., the dark adjustment was based on camera dark current and Spectrolon was used as a white reference before each measurement.

Figure 2: Test setup for wood imaging: (a) the spectral camera and (b) the conveyer belt for sample movement.

The samples for imaging were divided in two groups: (1) naturally seasoned and kiln-dried specimens and (2) thermally modified specimens.

Three naturally seasoned and three kiln-dried matched specimens were randomly selected from the sample group, and a cross-cut was prepared from each specimen. A naturally seasoned specimen attacked by stain-fungus growth was added to the dataset. The sample origin is listed in Table 1.

Table 1: The random positions of specimens from naturally seasoned and kiln-dried specimens.

Six of the specimens had fine-sawn surfaces, and two had a rougher surface in order to examine the influence of surface conditions on light reflection. The specimens were placed randomly on the conveyer belt in order to obtain a representative light-depth penetration during imaging. All the measurements were made on sapwood cross-sections, with camera-belt movement in the tangential direction.

For the hyperspectral image analysis, data for the naturally seasoned and kiln-dried specimens were combined with data for samples from the thermal modification processes, ThermoD, and WTT, using a “merge” function in the Evince software version 2.7.9 software, Prediktera, Umea, Sweden [21]. A mean-centring procedure was applied as a general preprocessing technique for all the data sets collected. Results from the merged digital images are presented in Figure 3. The standard RGB images of the dataset in Figure 3 show that the planed surfaces of the ThermoD (Figure 3(b)) and WTT (Figure 3(d)) had a more intense colour than the other surfaces.

Figure 3: Merged digital images of the specimens: (a) air-kiln dried, (b) ThermoD planed surface, (c) ThermoD rip-cut surface, (d) WTT planed surface, and (e) WTT rip-cut surface.
2.3. Image Cleaning and Processing

The complete data matrix from eight specimens of naturally seasoned and kiln-dried groups and four specimens of thermally modified wood contained 1,875,840 observations and 288 variables (spectral bands) as measured parameters. The merged absorbance images were exported and analysed with Evince software version 2.7.9 [21].

Hyperspectral image modelling of the multiple image dataset in Figure 3 was used to determine the useful area of interest to study the influence of the different factors (wood drying and thermal treatment). Principal component analysis (PCA) was used to clean background (conveyer belt) and edge effects by the brushing method [21]. Mean centering and base-line correction were applied to obtain clear classes in the score plots. Eight wavelengths at the beginning of the spectral range were deleted due to the darker colour of the TM samples [22] so that only the data between 980 and 2500 nm were analysed. The final spectral range included measurements of 280 wavelength bands, and this was used in the PCA model. Partial least squares discriminant analysis (PLS-DA) modelling was applied to data for the thermally modified specimens to differentiate between thermally modified processes and surfaces. The final PCA model included only a base-line correction for mean-centred data.

3. Results and Discussion

The merged and “background cleaned” hyperspectral dataset from the wood surface images in Figure 3 was processed by PCA, and the results are presented in Figure 4, which relates spatial data from the hyperspectral images to positions in a score plot with a grouping of the data (Figure 4(a)) and also provides spectral information (Figure 4(b)). In Figure 4(a), five clusters of the absorbance image can be seen (numbered arrows). These clusters are related to particular areas in the wood surfaces. The PCA model used to evaluate differences in the combined hyperspectral image, which made it possible to identify main distinctions among the samples and to see the grouping pattern in the dataset. The PCA model explained the variance of the data where the index R2X_cum = 0.97 (eigenvalue 3.47) explained variance of X matrix that was in our case the wavelength numbers. After cleaning (background and edge effect removal), the PCA model was improved (R2X_cum = 0.99461, eigenvalue 1.3426) with four components. The first and second principal components (PC) of the PCA model explained 94% of the total sum of squares (%SS) and 4.34% SS of the data variability, respectively.

Figure 4: (a) Score plot after background image cleaning, and (b) average spectrum of image related to the surfaces of the naturally seasoned and kiln-dried specimens shown in Figure 3(a) (numbers relate to description below and to Table 2).
Table 2: Band assignment from average spectrum of naturally seasoned and kiln-dried wood.

Six zones can be identified in the score plot in Figure 4(a) related to the specimens and surface characteristics in the RGB image of the surfaces presented in Figure 3:(1)Colour difference in knots due to the presence and conditions of extractives and a higher density than the surroundings. The area indicated by a circle in the score plot is related to the knots and latewood “bow” close to the knot on the upper right-hand side of the surfaces of ThermoD-modified wood (Figure 3(b)).(2)The rip-cut surface of specimens from ThermoD and WTT processes represented by the area indicated by the yellow arrow.(3)Almost all rip-cut surfaces of the WTT process and unexpectedly the areas in air-dried samples represented by the area indicated by the violet arrow.(4)Naturally seasoned and kiln-dried specimens represented by the area indicated by the red arrow.(5)The planed surface of the WTT modification specimen represented by the area indicated by the green arrow.(6)Planed surfaces of thermally modified specimens (thermoD and WTT processes) represented by the area indicated by the blue arrow.

The spectra of pixels in the image related to naturally seasoned and kiln-dried specimens (Figure 3(a)) show spectra from TM and air/kiln-dry for comparison and presented as an average single spectrum in Figure 4(b), indicating that there is no spectral difference.

The spectral band assignment of the NIR spectrum in Figure 4(b) was compared to the spectra of wood and wood components provided by Schwanninger [23]. The peaks in the average spectrum in Figure 4(b) are related to pixels in images of naturally seasoned and kiln-dried samples and described in Table 2.

PC2 of the area related to the naturally seasoned and kiln-dried specimens indicated by the red arrow in Figure 4 is presented as a 2D image in Figure 5. It was confirmed that there was no difference between the naturally seasoned and the kiln-dried wooden surfaces.

Figure 5: Contour 2D image of PC2 related to the area representing data from naturally seasoned and kiln-dried wooden surfaces indicated by red arrows in Figure 4(a).

Together, PC1 and PC2 describe 98.3% of the data variation. The Contour 2D image and the loading plot for PC2 are shown in Figure 6. The PC2 is dominated by the variation in the 1972 nm and 2135 nm wavelengths regions. The peak at 1972 nm is linked to changes in moisture content in the wood surface. The peak at 2135 nm corresponds to Car-H stretching and C=C stretching in lignin and extractives and to C-H stretching and C=O stretching of acetyl groups in hemicelluloses [23]. These peaks can be used for multivariate calibration of the difference between naturally seasoned/kiln-dried wood and thermally modified wood. The red areas in the PC2 score image shown by a red circle in Figure 4(a) are related to the small red areas in Figure 5 due to presence of knots.

Figure 6: The loading plot for PC2.

The average spectrum of the red area of the ThermoD surfaces in Figure 6(a) is presented in Figure 7. The detailed interpretation of the spectral signal recommended by Schwanninger has been combined with suggestions from Wokman and Weyer [24] and is described in Table 3.

Figure 7: The average spectrum for the red area (knots) on the ThermoD surface in Figure 6 (numbers related to Table 3).
Table 3: Band assignment for the average spectrum of naturally seasoned and kiln-dried wood.

The average spectrum of the surface contaminated by fungi is presented as a green line in Figure 8. The spectrum of the fungal-attacked surface does not differ greatly from the average spectrum of the naturally seasoned and kiln-dried samples (Figure 8).

Figure 8: The average spectrum of naturally seasoned and kiln-dried wood surfaces (red line) and the average spectrum of surfaces contaminated by fungi (green line).

The regions where the greatest difference in the minimal preprocessed spectra was observed was the 1100–1200 nm region and at wavelengths longer than 1650 nm (Figures 7 and 8). The absorption in the regions around 1695 and 1744 nm is greater in Figure 7 than in Figure 8 and may be related to extractives.

In a previous study, it was shown that low-molecular sugars and fatty/resin acid compounds migrate to naturally seasoned and kiln-dried surfaces [25]. It is known that chemical compounds are degraded and that new compounds such as extractives/phenols are formed during thermal modification [26]. In the present study, it was not possible to distinguish the signals of fungal-attacked wood from those of the nonattacked wood, due to an inability to identify the fungal species that contaminated the material and to assign bands with the polysaccharide matrix.

A discriminant analysis of the four different surfaces of the thermally modified specimens (b, c, d, and e in Figure 3) was performed, and the result is shown in Figure 9. The greatest differentiation was found between the planed and rip-cut surfaces, but a rather clear discrimination was also found between the two thermal modification processes. The score plot of PC1 and PC4 shows a clear differentiation between the surfaces (Figure 9(a)), and the model coefficients for PC1 and PC4 indicate the wavelengths that are strongest for the discrimination but they do not give the answer without further analysis. The model showed high R (explained variation) and Q (predicted variation) values (0.99 and 0.77, respectively), which indicate a clear difference in the NIR spectra for PC4 in Figure 9. The wavelength band at 2100 nm (arrow in Figure 9(b)) is especially interesting and may be related to the C=O-O of polymeric stretching [23], which may be related to a hydrolysis of acetyl groups to volatile carboxylic acids [27].

Figure 9: (a) Score plot of PC4 versus PC1 for the two thermal modification processes and (b) loadings for PC4 in PLC-DA model. The arrow shows the greatest difference in the spectral dataset.

The studied surfaces had a heterogeneous chemical nature due to tree growth factors and sawing patterns and because of drying and thermal modification. In addition to the wood diversity and processing factors, biological contaminants from the surroundings probably also influenced the chemical properties of the wood surface, but this influence is suggested to have a minor influence on the results of this study.

4. Conclusions

The use of hyperspectral imaging and multivariate analysis makes it possible to distinguish (a) Scots pine wood surfaces from two different thermal modification processes, (b) a rip-cut surface from a planed surface, and (c) surfaces of wood thermally modified from surfaces of naturally seasoned and kiln-dried wood. It was not however possible to distinguish between naturally seasoned and kiln-dried surfaces.

Regions with a high absorbance were related to wood components such as lignin, cellulose, and hemicellulose as well as to extractives with phenolic groups and aliphatic hydrocarbons. The band difference related to the peak at 2148 nm obtained from a PCA model was found to be useful for multivariate calibration between naturally seasoned and kiln-dried wood and thermally modified wood.

PLS-DA of thermally modified wood showed the greatest differentiation between a planed wood surface and a rip-cut surface and a rather clear discrimination between the two thermal processes. The wavelength band at 2100 nm revealed the greatest distinction and may correspond to C=O-O of polymeric stretching of acetyl groups.

The hyperspectral imaging in combination with simply operated software opens great opportunities for differently processed wooden surfaces. For the development of proper calibration and prediction models, however, the main wood components with described variability have to be chemically analysed in order to move from the laboratory scale to practical implementation.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that they have no conflicts of interest regarding the publication of this article.

Acknowledgments

The authors wish to thank NORRA Timber, Skellefteå, Sweden; HeatWood AB, Hudiksvall, Sweden; and Carl-Johan Stenvall, Arvidsjaur, Sweden, for providing the test material. Support from the European Regional Development Fund, the County Administration of Västerbotten, the Municipality of Skellefteå, and the Wood Centre North is greatly appreciated and acknowledged. In addition, the authors wish to express their gratitude for the support of COST Action FP1407. The authors thank Prof. Paul Geladi, SLU, Umeå, Sweden, for experimental assistance and fruitful discussions about the hyperspectral imaging technique. This work was supported by the Swedish Research Council for the Environment Agricultural Sciences and Spatial Planning (FORMAS) (Project 942-2016-64, 2016, and Project 419, 2017).

References

  1. K. Candelier, M.-F. Thevenon, A. Petrissans, S. Dumarcay, P. Gerardin, and M. Petrissans, “Control of wood thermal treatment and its effects on decay resistance: a review,” Annals of Forest Science, vol. 73, no. 3, pp. 571–583, 2016. View at Publisher · View at Google Scholar · View at Scopus
  2. S. Tsuchikawa, “A review of recent near infrared research for wood and paper,” Applied Spectroscopy Reviews, vol. 42, no. 1, pp. 43–71, 2007. View at Publisher · View at Google Scholar · View at Scopus
  3. B. G. Osborne, T. Fearn, P. T. Hindle, and B. G. Osborne, Practical NIR Spectroscopy with Applications in Food and Beverage Analysis, Longman Scientific and Technical, Wiley, New York, NY, USA, 2nd edition, 1993.
  4. A. Sandak, J. Sandak, and R. Meder, “Assessing trees, wood and derived products with NIR spectroscopy: hints and tips,” Journal of Near Infrared Spectroscopy, vol. 24, no. 6, pp. 485–505, 2016. View at Publisher · View at Google Scholar · View at Scopus
  5. P. Navi and D. Sandberg, Thermo-Hydro-Mechanical Processing of Wood, Presses Polytechniques et Universitaires Romandes, Lausanne, Switzerland, 2012.
  6. W. E. Hillis, “High temperature and chemical effects on wood stability,” Wood Science and Technology, vol. 18, no. 4, pp. 281–93, 1984. View at Publisher · View at Google Scholar · View at Scopus
  7. Forest Products Laboratory, “Air drying of lumber,” General Technical Report FPL-GTR-117, U.S. Department of Agriculture, Forest Service, Forest Products Laboratory, Madison, WI, USA, 1999.
  8. N. Terziev, “Industrial kiln drying and its effect on microstructure, impregnation and properties of Scots pine timber impregnated for above ground use. Part 1. Effects of initial, final dryings and preservative on impregnation and timber quality,” Holzforschung, vol. 56, no. 4, pp. 428–433, 2005. View at Publisher · View at Google Scholar · View at Scopus
  9. D. Sandberg, “Primary wood processing in Swedish sawmill 1850-2010,” in Sawn Timber for Prosperity. The Swedish Sawmill Industry History 1850-2010, R. Pettersson, Ed., pp. 326–363, The Royal Swedish Academy of Agriculture and Forestry, Stockholm, Sweden, 2015. View at Google Scholar
  10. A.-M. Manninen, P. Pasanen, and J. Holopainen, “Comparing the VOC emissions between air-dried and heat-treated Scots pine wood,” Atmospheric Environment, vol. 36, no. 11, pp. 1763–68, 2002. View at Publisher · View at Google Scholar · View at Scopus
  11. J. Tomminen and M. Nuorteva, “Pinewood nematode, Bursaphelenchus xylophilus in commercial sawn wood and its control by kiln-heating,” Scandinavian Journal of Forest Research, vol. 7, no. 1–4, pp. 113–120, 1992. View at Publisher · View at Google Scholar
  12. M. Sehlstedt-Persson, “High-temperature drying of Scots pine. A comparison between HT- and LT-drying,” Holz als Roh-und Werkstoff, vol. 53, no. 1, pp. 95–99, 1995. View at Publisher · View at Google Scholar · View at Scopus
  13. F. Shafizadesh and P. Chin, “Thermal deterioration of wood,” ACS Symposium Series 43, American Chemical Society, Washington, DC, USA, 1977. View at Google Scholar
  14. C. Hill, Wood Modification: Chemical, Thermal and Other Processes. Wiley Series in Renewable Resources, Wiley, Chichester, UK, 2006.
  15. P. M. Kekkonen, A. Ylisassi, and V.-V. Telkki, “Absorption of water in thermally modified pine wood as studied by nuclear magnetic resonance,” Journal of Physical Chemistry C, vol. 118, no. 4, pp. 2146–2153, 2014. View at Publisher · View at Google Scholar · View at Scopus
  16. D. Fengel and G. Wegener, Wood: Chemistry, Ultrastructure, Reactions, W. de Gruyter, New York, NY, USA, 1984.
  17. M. Nuopponen, H. Wikberg, T. Vuorinen, S. Maunu, S. Jämsä, and P. Viitaniemi, “Heat-treated softwood exposed to weathering,” Journal of Applied Polymer Science, vol. 91, no. 4, pp. 2128–2134, 2004. View at Publisher · View at Google Scholar · View at Scopus
  18. O. Dagbro, P. Torniainen, and T. Morén, “Thermal modification of birch using saturated and superheated steam,” in Proceedings of Nordic-Baltic Network in Wood Materials Science and Engineering, Oslo, Norway, October 2011.
  19. O. Dagbro, P. Torniainen, O. Karlsson, and T. Morén, “Colour responses from wood, thermally modified in superheated steam and pressurized steam atmospheres,” Wood Material Science and Engineering, vol. 5, no. 3-4, pp. 211–219, 2010. View at Publisher · View at Google Scholar · View at Scopus
  20. Finnish Thermowood Association, ThermoWood® Handbook, ThermoWood, Teuva, Finland, 2003, http://www.thermowood.fi.
  21. Prediktera, Evince–Tutorials and Documentation, 2018, Prediktera, Umeå, Sweden, 2018.
  22. H. Bächle, B. Zimmer, E. Windeisen, and G. Wegener, “Evaluation of thermally modified beech and spruce wood and their properties by FT-NIR spectroscopy,” Wood Science and Technology, vol. 44, no. 3, pp. 421–433, 2010. View at Publisher · View at Google Scholar · View at Scopus
  23. M. Schwanninger, J. C. Rodrigues, and FacklerK, “A review of band assignments in near infrared spectra of wood and wood components,” Journal of Near Infrared Spectroscopy, vol. 19, no. 5, pp. 287–308, 2011. View at Publisher · View at Google Scholar · View at Scopus
  24. J. Workman and L. Weyer, Practical Guide and Spectral Atlas for Interpretive Near-Infrared Spectroscopy, CRC Press, London, UK, 2nd edition, 2012.
  25. O. Karlsson, O. Myronycheva, M. Sehlstedt-Persson, M. Öhman, and D. Sandberg, “Multivariate modeling of mould growth in relation to extractives in dried Scots pine sapwood,” in Proceedings of IRG47 Annual Meeting, Ghent, Belgium, December 2017.
  26. P. Niemz, T. Hofmann, and T. Rétfalvi, “Investigation of chemical changes in the structure of thermally modified wood,” Maderas. Ciencia y Tecnología, vol. 12, no. 2, pp. 69–78, 2010. View at Publisher · View at Google Scholar · View at Scopus
  27. B. Sundqvist, O. Karlsson, and U. Westermark, “Determination of formic-acid and acetic acid concentrations formed during hydrothermal treatment of birch wood and its relation to colour, strength and hardness,” Wood Science and Technology, vol. 40, no. 7, pp. 549–561, 2006. View at Publisher · View at Google Scholar · View at Scopus