The relationship between color and mechanical properties of heat-treated wood predicted based on support vector machines model

Abstract

Abstract Thermal modification or heat treatment can cause the loss of mechanical property of wood. In this study, Poplar (Populus tomentosa Carr.) and spruce (Picea obies Mast.) were heat treated at 180, 200, and 220 °C for 2–10 h. Changes of color (L*, a* and b*) and mechanical strength including modulus of elasticity (MOE), modulus of rupture (MOR) and shear strength after heat treatment were analyzed. Time-temperature superposition methods were used to quantify color and mechanical strength. The prediction models of MOR, MOE and shear strength were assessed with support vector regression model (SVR) based on color parameters. The trends of color change and mechanical strength after heat treatment were highly consistent. The values of apparent activation energy (E a ) calculated from color parameters (110.6–187.2 kJ/mol) were identical to those from mechanical strengths (103.2–219.2 kJ/mol). Color parameters were used as input variables, and the MOE, MOR, and shear strength were output parameters in the established SVR model. Gaussian radial basis function (RBF) was found to be a kernel function for SRV model. Optimal hyperparameters in SVR model were obtained using cross-validation and grid search. The determination coefficients for MOE, MOR, and shear strength were 0.903, 0.835, and 0.865, respectively for poplar. The high correlation suggested that wood mechanical strength can be predicted non-destructively through measuring color parameters after heat treatment.