An artificial neural network model for predicting mechanical strength of bamboo-wood composite based on layups configuration

Abstract

The transportation application of the bamboo–wood composite container flooring (BWCCF) has increased considerably. However, materials would be destroyed in the process of common mechanical evaluation, resulting in a waste of resources. Therefore, this paper aims to design artificial neural network (ANN) models to predict mechanical strength of BWCCF. The modulus of rupture (MOR) and the modulus of elasticity (MOE) of BWCCF were predicted by ANN models based on layups configuration, including directions, densities, and thicknesses of 21-layer BWCCF in each layer. According to results, the mean absolute percentage errors (MAPE) and the correlation coefficient (R) were determined as 16.93% and 0.619 in prediction of MOR, and 10.10% and 0.709 in prediction of MOE, respectively. The results indicated that ANN can be applied to predict mechanical properties of BWCCF.

Full Article

An Artificial Neural Network Model for Predicting Mechanical Strength of Bamboo-wood Composite Based on Layups Configuration

Zihua Su, Zhilin Jiang, Yi Liang, Bingzhen Wang, and Jianping Sun *

The transportation application of the bamboo–wood composite container flooring (BWCCF) has increased considerably. However, materials would be destroyed in the process of common mechanical evaluation, resulting in a waste of resources. Therefore, this paper aims to design artificial neural network (ANN) models to predict mechanical strength of BWCCF. The modulus of rupture (MOR) and the modulus of elasticity (MOE) of BWCCF were predicted by ANN models based on layups configuration, including directions, densities, and thicknesses of 21-layer BWCCF in each layer. According to results, the mean absolute percentage errors (MAPE) and the correlation coefficient (R) were determined as 16.93% and 0.619 in prediction of MOR, and 10.10% and 0.709 in prediction of MOE, respectively. The results indicated that ANN can be applied to predict mechanical properties of BWCCF.

DOI: 10.15376/biores.17.2.3265-3277

Keywords: Artificial neural network; Layups configuration; Bamboo-wood composite; Mechanical strength

Contact information: Materials, School of Resources, Environment and Materials, Guangxi University, Nanning 530004, China; *Corresponding author: jpsun@gxu.edu.cn

INTRODUCTION

Bamboo-wood composite materials, which can be defined as composite plates mainly made of bamboo and wood through processing into the same or different forms of structural units, have excellent physical mechanical properties due to the composite effect of materials (Li et al. 2021). Bamboo-wood composites not only can take advantage of the characteristic of bamboo high strength, but also they can give play to the characteristics of high utilization rate of wood veneer (Chen et al. 2017). Based on the anisotropic mechanical nature of the materials, the physical and mechanical properties of the final products, manufacturing parameters and costs can be optimized through rationally arranging the layout of layers (Abdelhamid et al. 2004). Some theories of wood production are based on the influence of layups configuration with mechanical strength of itself. Examples include the method of classical theory, the Euler beam theory (Abdelhamid et al. 2004), the first-order shear theory, mechanics of composite materials, finite element method (Chen et al. 2016b), mechanics of elasticity, theory of laminated plates, and related knowledge of composite materials (Li-Haiwei 2011). From the perspective of composite science, significant differences exist between the products made of the same specific elements while using different combination approaches (Chen et al. 2016a). According to laminated plate theory, the advantages of each layer can be maximized by placing the materials with high strength and high elastic modulus in the outermost layer and the materials with poor performance in the core layer (Chow et al. 2019). For example, a new composite container flooring was formed via the design of rational structure and the research of hot-pressing process based on this theory (He 2005). Other layer characteristics also affect mechanical properties. Bamboo-wood composite on the mechanical properties were affected the layer structure of composite made (Lee et al. 2012) and the quality of adding different layers of bamboo fibers (Zhai et al. 2011). These findings showed that the mechanical properties of composite can be improved by increasing the proportion of bamboo. Similar results were obtained in the study of composite beams (Zhang et al. 2020) and wood-bamboo composites construction concrete (He et al. 2019). Furthermore, the arrangement direction of the configuration has a great influence on the properties of bamboo-wood composites. The influences of angle 0°, 45°, and 90° on the properties of bamboo-wood composites were studied (Ashaari et al. 2016; Verma et al. 2017). In sum, the lay-up of bamboo-wood composite has an important influence on their mechanical properties.

Non-destructive testing of materials enables the physical and mechanical properties to be obtained without permanently or noticeably altering the materials (Abdelhamid et al. 2004). Its emphasis depends on whether the material is isotropic or anisotropic (Esteban et al. 2009), but mechanical properties are usually based on mathematical models formed from tests carried out on structural members in order to be reliable and statistically robust (Cavalli et al. 2016). Recently, interest in artificial neural networks (ANNs) has grown (Agha and Alnahhal 2012). ANNs have been applied successfully to several areas including medicine, engineering, geology, and physics, to design solutions for estimation problems, classification, control, etc. (Agha and Alnahhal 2012). ANNs can be used as predictive models because they are capable of modelling complex functions. The ANN can learn and extract knowledge to simulate the functions of human brain, becoming a powerful tool to simulate the complex relationships between variables (Esteban et al. 2009). For this purpose, ANNs have been used for predicting the properties of solid wood (Tiryaki and Aydın, 2014), e.g., simulating and predicting the change of wood moisture content (MC) (Chai et al. 2018), drying process of wood, calculating wood thermal conductivity and compressive strength (Tiryaki and Aydın 2014), predicting the modulus of rupture (MOR) and elastic of modulus (MOE) of lumber heat treatment (Esteban et al. 2009; Tiryaki and Hamzacebi 2014), and calculating heat treatment of wood best bonding strength (Tiryaki et al. 2014). Furthermore, the short span of shear force and MOE of plywood were predicted (Fernández et al. 2012). It has been possible to raise the utilization efficiency of the wood board and shorten the time of the design process (Jun et al. 2003). Wood composites have been studied by ANN, for example, the Buckingham’s p-theorem and the MOR and MOE predicted based on a multiple regression of particleboard (Arabi et al. 2011), studying the effects of board thickness, moisture content and specific gravity of plywood boards (Fernández et al. 2012), forecasting the plywood combined strength (Demirkir et al. 2013), and frequencies prediction of laminated timber plates (Sun et al. 2020).

Bamboo-wood composites have great requirements in various sectors such as cement templates, container flooring (Jiang et al. 2021), concrete formwork, truck platform floor, inner floorboard, and are mostly used as engineering structural parts (Jiang et al. 2002). Bamboo-wood composite structure material is subject to external forces, which requires sufficient strength, stiffness, and durability. For instance, bamboo-wood composite used as the bamboo-wood composite container floor (BWCCF) is the main bearing component for container loads. The layups configuration is complex, and the structure is over 20 layers, usually 21, 22, and 23 layers. There has been a need for non-destructive testing of bamboo-wood composite used as the BWCCF. According to the national standards, the vertical and horizontal MOR and MOE is greater than or equal to 85 and 35 MPa and 10000 and 3500 MPa, respectively (GB/T 19536 2015). Therefore, this study used ANN through MATLAB to design the model of BWCCF, which can predict mechanical properties to obtain more economic and safe results without doing comprehensive tests.

Fig. 1. Flow chart of the experiments

EXPERIMENTAL

Figure 1 shows the flow chart of this experiment. First, the BWCCF was prepared. Next, BWCCF were tested for the MOR and MOE, and the results served as the output of the ANN models. The models were trained, and the generalization performance was verified by the input-output sample sets. Finally, the effective prediction models were obtained.

Materials

Seven commercially produced sheets (Nanning Diwang Village Wood Co.) from China were used, including bamboo curtain, eucalyptus, rubber wood, pine, and other miscellaneous woods, with dimensions of 2400 mm × 1160 mm or 1626 mm × 636 mm. And the thicknesses were 2.2 mm, 2.2 mm, 1.7 mm, 1.2 mm, and 0.5 mm, respectively. The adhesive used was a water-soluble urea formaldehyde (UF) resin (with 48% or 55% solids content produced by Wood Glue Industrial, Guangxi, China).

Prepared Composites

Bamboo curtains were subjected to dipping, the key process for manufacturing bamboo mat, which determines its bonding performance (Huang et al. 2019). The bamboo curtains were dipped for three seconds with UF (solid content of 48 ± 2%, pH of 10 ± 1, and viscosity of 45 to 100 mPa.s). In the dipping process, the MC of bamboo curtain needs to be below 10% so that the resins can get into it. All wood veneers were gummed with UF (solid content of 48 ± 2%, pH of 10 ± 1, and viscosity of 50 to 180 mPa.s). The next step was assembly, which is the most important part. The dipped bamboo curtain and coated wood veneer were assembled according to the design plan for lay-up. Finally, the assembled veneers were compressed with hot-pressing at a pressure of 2.5 to 3.0 MPa and 130 ± 5 °C for 45 min. The commonly used size (length × width × thickness) for the BWCCF was either 2400 mm × 1160 mm × 28 mm or 1626 mm × 636 mm × 28 mm.

Blending Properties Test

The mechanical properties, such as MOE and MOR of the BWCCF were determined according to the procedure specified in the plywood standard of container floors (GB/T 19536-2015) and plywood performance standard of international freight container floors (IICL TB 001-2014). According to these standards, the test requires primary specimens with dimensions of 610 mm × 50 mm × 28 mm for the MOR and MOE experiments. The tests were carried out using the computer software SmartText to control the WDW-100E machine (Beijing TIME High Technology Ltd, Beijing, China), and three points bending tests were carried out to calculate the strength. The final mechanical properties are the average value of the six repetitions of the above measurement for further analysis.

ANN Models Prediction Analysis

Two models were built based on layups configuration and mechanical property parameters of BWCCF and were trained by ANN, as shown in Fig. 2. The parameters of blank (direction, thickness, and density) were input to predict the mechanical properties (MOR and MOE) of the bottom plate.

Fig. 2. The ANN architecture used as the prediction model for (a) MOR and (b) MOE

The concept of ANN is inspired by the biological system of the brain, which comprises many neuro connected through synapses that process information (Watanabe et al. 2012). In this study, the proposed ANN models were designed by software developed using the MATLAB Neural Network Toolbox. The training was carried out by making attempts to establish different ANNs with different network configurations and learning parameters (Tiryaki and Aydın 2014). In this study, the 51 groups of data were divided randomly into two parts during model construction: 41 groups of training set (80% of the total), 10 groups of test data (20%). The input layer data were the 63 blank parameters (21 directions, 21 densities and 21 thicknesses, the average of raw material veneer) of the BWCCF. respectively. The data of the output layer were the mechanical properties (the MOR and MOE). In the first model, the 63 blanks parameters were considered as input variables and the MOR used as output, and the 63-10-1 neurons configuration was employed. In the other model, the MOE was used as output, and the 63-11-1 neurons configuration was designed. There is no scientific and explicit method for determining the number of nodes; the number of nodes in the hidden layer could be determined through experience and multiple training (Tiryaki and Hamzacebi 2014). The input signal is propagated forward to the hidden node. After passing the transfer function, the output information of the hidden node is propagated to the output node, and the output result is given after processing. The prediction performance of each model was evaluated and compared for each case using statistical and graphical comparisons (Demirkir et al. 2013). The mean absolute (MAPE) values were considered as the most important performance criteria (Tiryaki and Aydın 2014). The average absolute error (MAE) describes the dispersion of data samples, and the root mean square error (RMSE) indicates the deviation between the predicted value and the real value. The correlation coefficient (R) is used to describe the degree of linear correlation between predicted and actual values, which can reflect the positive or negative correlation between them. The determination coefficient (R²) indicates the correctness of model fitting. The higher values of R and R² present the greater similarities between the measured and predicted values. MAPE, MAE, and RMSE values are calculated using Eqs. 1 through 3 (Ozsahin and Aydin 2013; Tiryaki and Aydın 2014; Tiryaki et al. 2014),

RESULTS AND DISCUSSION

Predicting MOR and MOE by ANN

Two BP artificial neural networks were designed in this study. This approach makes use of error gradient descent algorithm to minimize the mean square error between the output value of network and the actual output value (Yu and Xu 2014). While layups direction, thickness, and density were considered as input variables, mechanical properties of the BWCCF were used as output variables in the models. The relationship between the experimental values and predicted values was obtained using the ANN prediction model is shown in Fig. 3. The solid line shows the linear fit between the predicted values and measured values of MOR and MOE. As the R values approaches 1, prediction accuracy increases (Chai et al. 2018). According to Fig. 3, the R values for training of MOR and MOE were 0.69 and 0.66, respectively; the R values of testing for MOR and MOE were 0.71 and 0.71, respectively. The results obtained are satisfying compared with those obtained by other authors on applying ANNs for modeling to predict a variety of properties of wood and wood-based products. R values obtained in these studies can be summarized as follow: Fernández et al. (2012) predicted MOR and MOE of structural plywood board using an ANN, and the maximum value obtained for the R value was 0.71; Samarasinghe et al. (2007) found a R value as 0.62 in modeling to determine the fracture toughness of wood; Watanabe et al. (2012) predicted a final MC of individual Sugi samples during air drying, where the R value was 0.80.

Fig. 3. Relationship between actual and predicted values of ANN training and testing models for (a, c) grain MOR and (b, d) grain MOE

The verification percentage error of ANN model obtained and the distribution of model error are shown in Fig. 4 and Table 1. As shown in Table 1, the prediction values obtained by ANN were determined with not high percentage errors. The prediction error of MOE was less than MOR, which is consistent with the results of R values in Fig. 3. Overall, the prediction accuracy of each measurement point is moderate and can meet the demand for stratified mechanical property prediction. These levels of errors are satisfactory for predicting MOR and MOE values.

Fig. 4. The error of validation

Table 1. The Percentage Error of Validation

MAPE, MAE, and RMSE values were utilized to evaluate the performance of the neural networks, as shown in Table 2. MAPE values of the MOR and MOE training set were 10.5% and 5.98%, while in validation set the corresponding values were 16.9% and 10.1% respectively. The ANN approach has a sufficient accuracy level in the prediction of MOR and MOE values. Furthermore, the predicted optimum bonding strength of heat-treated woods training and testing MAPE were 1.49% and 3.06%, respectively (Tiryaki et al. 2014). Ozsahin and Aydin (2013) predicted the optimum veneer drying temperature, and the MAPE of training and testing were 1.08% and 2.29%, respectively. In this study, the thickness and density of veneer were difference in the process of prepared composites. Moreover, the detection of mechanical properties was affected by the cracks of BWCCF. The training and testing MAE of MOR were 381 and 15.9, and MOE were 630 and 1170. The training and testing RMSE of MOR were 13.5 and 19.4, MOE were 1006 and 1328, respectively. By other studies, scientists have calculated values of RMSE and MAE obtained of the various wood products. For example, predicting compression strength of heat-treated woods learning, the RMSE of training and testing were 0.432 and 1.082, respectively, and the MAE of training and testing were 0.336 and 0.960 (Tiryaki and Aydın 2014). In addition, by predicting the MOR and MOE of the heat-treated material, the MAE for training and testing of the MOR were 0.654 and 0.922, respectively, and the MAE for training and testing of the MOE were 91.0 and 170, respectively (Tiryaki and Hamzacebi 2014). The values of the MOR and MOE were 60 to 100 MPa and 7000 to 12000 MPa, respectively. In the present study, the results were slightly larger. The discrete nature of the data was the cause of insufficient model training sample and influences the result directly. These levels of errors were satisfactory when using ANN models for predicting MOE and MOR.

Table 2. Collection of ANNs Model Data

The actual and predicted values of MOR and MOE were compared. As shown in Fig. 5(a), the consistent agreement between the predicted and measured values of mechanical strength in training set shows that the decision method was effective. According to Fig. 5(b), the prediction values were closer to the measured values in validation set. The results indicate a consistent agreement between the outcomes of the ANN modeling and the experimental results.

Fig. 5. Comparison of the real and predict values results in (a) training and (b) testing for all data

Relationship between Mechanical Strength and Layups Configuration

The materials of bamboo-wood composite contain both bamboo and wood, and the material properties of bamboo and wood are different. Bamboo is superior to wood in some mechanical properties. Therefore, the proportion and position of bamboo and wood should be considered when making BWCCF. The direction, thickness and density of all raw materials affect the mechanical properties of composite.

As shown in Table 3, the following conclusions can be drawn: 1) All the bamboo curtain was longitudinal, indicating that the curtain in the process of making the bottom plate is suitable for grain arrangement, which could contribute maximum mechanical properties. This result is in agreement with most of the bamboo-wood composite literature. (Xiao et al. 2014; Chen et al. 2016a,b). 2) The direction of group 10 was the same as group 11, but one species layer was different. Therefore, the MOR and MOE of group 10 were both lower. This result shows that the mechanical properties of the BWCCF can be improved by adding bamboo veneer, which is in agreement with literature (Chen et al. 2016a; {Yubing, 2019 #56}Zheng 2016). 3) The groups of 4-9, 11-12, 13-14 had the same layups configuration, but their mechanical parameters were different. The differences can be attributed to the influence of mechanical parameters, such as the thickness, density, and the presence of cracks in BWCCF.

Table 3. Layups Configuration and Mechanical Strength of BWCCF

Notes: Within the table, the symbol │means along the grain of wood veneer, whereas ─ means the horizontal grain wood veneer and┆is arranged grain bamboo curtain.

The optimal layup configuration is represented by group 14 in Table 3. Compared with other types of layups, the optimal layup had the following characteristics: 1) The fifth layer (define layer bottom is the first layer) tended to fracture, and this layer was a longitudinal bamboo curtain which has better mechanical properties than wood. 2) No transverse arrangement of veneer was used on the outermost three layers. Moreover, the thickness of veneer was symmetrically distributed in the center.

CONCLUSIONS

1. This study developed artificial neural network (ANN) models to predict modulus of rupture (MOR) and modulus of elasticity (MOE) values of bamboo-wood container composite flooring (BWCCF) based on layups configuration, including directions, densities, and thicknesses of 21-layer BWCCF in each layer. According to obtained data, the following was concluded. In the established ANN models, the MAPE and R fitting parameters for an ANN model for predicting MOR were 16.9% and 0.619, and that of an ANN model for predicting MOE were 10.1% and 0.709, respectively. Considering the complexity of BWCCF, this demonstrates that the ANNs are appropriate for modeling the relations between layups configuration and mechanical properties of BWCCF.
2. In order to further realize the nondestructive testing of BWCCF, it is necessary to continuously explore new testing methods to improve the accuracy of the model. Based on the present study, there are several aspects that need to be improved: 1) Increase the number of training samples to reduce the contingency of data and improve the generality of data; 2) Explore a more concise way to represent the structural parameters of the layups. The input parameters for this study were 63, which leads to a complex training process. Therefore, the simpler digital forms are needed to reflect the thickness, density, and direction.

ACKNOWLEDGMENTS

This work was financially supported by the National Natural Science Foundation of China (31660174).

REFERENCES CITED

Agha, S. R., and Alnahhal, M. J. (2012). “Neural network and multiple linear regression to predict school children dimensions for ergonomic school furniture design,” Appl. Ergon. 43(6), 979-84. DOI: 10.1016/j.apergo.2012.01.007

Arabi, M., Faezipour, M., Layeghi, M., Enayati, A., and Zahed, R. (2011). “Prediction of bending strength and stiffness strength of particleboard based on structural parameters by Buckingham’s p-theorem,” Journal of the Indian Academy of Wood Science 7(1-2), 65-70. DOI: 10.1007/s13196-011-0013-8

Ashaari, Z., Lee, S. H., and Zahali, M. R. (2016). “Performance of compreg laminated bamboo/wood hybrid using phenolic-resin-treated strips as core layer,” European Journal of Wood and Wood Products 74(4), 621-624. DOI: 10.1007/s00107-016-1027-0

Cavalli, A., Bevilacqua, L., Capecchi, G., Cibecchini, D., Fioravanti, M., Goli, G., Togni, M., and Uzielli, L. (2016). “MOE and MOR assessment of in service and dismantled old structural timber,” Engineering Structures 125, 294-299. DOI: 10.1016/j.engstruct.2016.06.054

Chai, H., Chen, X., Cai, Y., and Zhao, J. (2018). “Artificial neural network modeling for predicting wood moisture content in high frequency vacuum drying process,” Forests 10(1). 10. DOI: 10.3390/f10010016

Chen, F., Deng, J., Li, X., Wang, G., Smith, L. M., and Shi, S. Q. (2016a). “Effect of laminated structure design on the mechanical properties of bamboo-wood hybrid laminated veneer lumber,” European Journal of Wood and Wood Products 75(3), 439-448. DOI: 10.1007/s00107-016-1080-8

Chen, F., Wang, G., Li, X., Smith, L. M., and Shi, S. Q. (2016b). “Laminated structure design of wood—Bamboo hybrid laminated composite using finite element simulations,” Journal of Reinforced Plastics and Composites 35(22), 1661-1670. DOI: 10.1177/0731684416663852

Chen, F. M., Deng, J. C., Li, X. J., Wang, G., Smith, L. M., and Shi, S. Q. (2017). “Effect of laminated structure design on the mechanical properties of bamboo-wood hybrid laminated veneer lumber,” European Journal of Wood and Wood Products 75(3), 439-448. DOI: 10.1007/s00107-016-1080-8

Chow, A., Ramage, M. H., and Shah, D. U. (2019). “Optimising ply orientation in structural laminated bamboo,” Construction and Building Materials 212, 541-548. DOI: 10.1016/j.conbuildmat.2019.04.025

Demirkir, C., Özsahin, Ş., Aydin, I., and Colakoglu, G. (2013). “Optimization of some panel manufacturing parameters for the best bonding strength of plywood,” International Journal of Adhesion and Adhesives 46, 14-20. DOI: 10.1016/j.ijadhadh.2013.05.007

Esteban, L. G., Fernández, F. G., and de Palacios, P. (2009). “MOE prediction in Abies pinsapo Boiss. timber: Application of an artificial neural network using non-destructive testing,” Computers and Structures 87(21-22), 1360-1365. DOI: 10.1016/j.compstruc.2009.08.010

Fernández, F. G., de Palacios, P., Esteban, L.G., Garcia-Iruela, A., Rodrigo, B.G., and Menasalvas, E. (2012). “Prediction of MOR and MOE of structural plywood board using an artificial neural network and comparison with a multivariate regression model,” Composites Part B: Engineering 43(8), 3528-3533. DOI: 10.1016/j.compositesb.2011.11.054

GB/T 19536 (2015). “Plywood for container flooring,” Standardization Administration of China, Beijing.

He, W. (2005). Study on Floor Structure and Gluing Technology of Bamboo-wood Composite Container, Nanjing Forestry University. (in Chinese)

He, R., Qiu, J., He, H., and Luo, B. (2019). “Research status and progress trend of wood fungal staining,” Journal of Forestry Engineering 4(3), 19-24.

Huang, Y. X., Ji, Y. H., and Yu, W. J. (2019). “Development of bamboo scrimber: A literature review,” Journal of Wood Science 65(1). DOI: 10.1186/s10086-019-1806-4

IICL TB 001-2014 (2014). “Performance standard for new and unused structural container floor panels, International freight containers,” Institute of International Container Lessors, Washington, DC.

Jiang, Z., Liang, Y., Su, Z., Chen, A., and Sun, J. (2021). “Nondestructive testing of mechanical properties of bamboo–wood composite container floor by image processing,” Forests 12(11). DOI: 10.3390/f12111535

Jiang, Z. H., Wang, G., Fei, B. F., and Yu, W. J. (2002). “The research and development on bamboo/wood composite materials,” Forest Research 15(6), 712-718. (in Chinese)

Jun, C., Yi-zhuo, Z., and Qi, Y. (2003). “Combining the genetic algorithms with artificial neural networks for optimization of board allocating,” Journal of Forestry Research 14(1), 87-88.

Lee, C. H., Chung, M. J., Lin, C. H., and Yang, T. H. (2012). “Effects of layered structure on the physical and mechanical properties of laminated moso bamboo (Phyllosachys edulis) flooring,” Construction and Building Materials 28(1), 31-35. DOI: 10.1016/j.conbuildmat.2011.08.038

Li, H. W. (2011). “Prediction and text research of elastic mechanical properties of sandwich structure composite material,” Vol. Master of Engineering, School of material science and engineering. (in Chinese)

Li, H., Wang, B. J., Wang, L. B., Wei, P. X., Wei, Y., and Wang, P. Z. (2021). “Characterizing engineering performance of bamboo-wood composite cross-laminated timber made from bamboo mat-curtain panel and hem-fir lumber,” Composite Structures 266. DOI: 10.1016/j.compstruct.2021.113785

Ozsahin, S., and Aydin, I. (2013). “Prediction of the optimum veneer drying temperature for good bonding in plywood manufacturing by means of artificial neural network,” Wood Science and Technology 48(1), 59-70. DOI: 10.1007/s00226-013-0583-2

Samarasinghe, S., Kulasiri, D., and Jamieson, T. (2007). “Neural networks for predicting fracture toughness of individual wood samples,” Silva Fennica 41, 105-122. DOI: 10.14214/sf.309

Sun, J. P., Niederwestberg, J., Cheng, F.C., and Chui, Y. (2020). “Frequencies prediction of laminated timber plates using ANN approach,” Journal of Renewable Materials 8(3), 319-326. DOI: 10.32604/jrm.2020.08696

Tiryaki, S., and Aydın, A. (2014). “An artificial neural network model for predicting compression strength of heat treated woods and comparison with a multiple linear regression model,” Construction and Building Materials 62, 102-108. DOI: 10.1016/j.conbuildmat.2014.03.041

Tiryaki, S., and Hamzacebi, C. (2014). “Predicting modulus of rupture (MOR) and modulus of elasticity (MOE) of heat treated woods by artificial neural networks,” Measurement 49, 266-274. DOI: 10.1016/j.measurement.2013.12.004

Tiryaki, S., Özşahin, Ş., Yıldırım, İ. (2014). “Comparison of artificial neural network and multiple linear regression models to predict optimum bonding strength of heat treated woods,” International Journal of Adhesion and Adhesives 55, 29-36. DOI: 10.1016/j.ijadhadh.2014.07.005

Verma, C. S., Purohit, R., Rana, R. S., and Mohit, H. (2017). “Mechanical properties of bamboo laminates with other composites,” Materials Today-Proceedings 4(2), 3380-3386. DOI: 10.1016/j.matpr.2017.02.226

Watanabe, K., Matsushita, Y., Kobayashi, I., and Kuroda, N. (2012). “Artificial neural network modeling for predicting final moisture content of individual Sugi (Cryptomeria japonica) samples during air-drying,” Journal of Wood Science 59(2), 112-118. DOI: 10.1007/s10086-012-1314-2

Xiao, S. L., Lin, H., Shi, S. Q., and Cai, L. P. (2014). “Optimum processing parameters for wood-bamboo hybrid composite sleepers,” Journal of Reinforced Plastics and Composites 33(21), 2010-2018. DOI: 10.1177/0731684414553281

Yu, F., and Xu, X. (2014). “A short-term load forecasting model of natural gas based on optimized genetic algorithm and improved BP neural network,” Applied Energy 134, 102-113. DOI: 10.1016/j.apenergy.2014.07.104

Zhang, Q., Yang, Y., Fu, Y., and Sun, J. (2020). “Dyeing process and mechanism of eucalyptus veneer with Dalbergia bariensis heartwood pigment as dye,” Journal of Beijing Forestry University 42(3), 151-159.

Zheng, Z. F. (2016). “Production technology of bamboo fiber/wood composite container flooring,” China Wood-Based Panels 23(05), 15-17. (in Chinese)

Zhai, Z. Z., Chen, Y. H., Zhao, B., and Yang, H. M. (2011). “New structural wood-bamboo composite panels for container flooring,” China Wood Industry 25(5), 44-46. DOI: 10.19455/j.mcgy. (in Chinese)

Article submitted: December 29, 2021; Peer review completed: March 12, 2022; Revised version received and accepted: April 6, 2022; Published: April 25, 2022.

DOI: 10.15376/biores.17.2.3265-3277