Deresination rate prediction of Masson pine wood based on support vector regression (SVR)

Objective Deresination treatment is an important method for improving the performance of pine wood products. However, traditional methods for detecting deresination rate are both time-consuming and destructive to samples. This paper aims to explore a rapid and non-destructive method for detecting the deresination rate based on changes in wood surface color and using support vector regression (SVR) to construct a deresination rate prediction model.

Method Pinus massoniana wood was treated with ammonia gas and water vapor under high-temperature conditions. The effects of different conditions on wood surface color parameters and deresination rate were analyzed, and their correlations were investigated. Three different kernel functions (polynomial kernel function, Sigmoid kernel function, and radial basis function) were used to construct SVR-based deresination rate prediction models, and the optimal model was selected through comparison.

Result After deresination by ammonia gas-water vapor heat treatment, the surface lightness (L^*) and yellow-blue index (b^*) of Masson pine wood were lower than those of untreated wood, while the red-green index (a^*) was higher than that of untreated wood. With increasing ammonia mass fraction and treatment temperature, L^*, a^*, and b^* showed a gradual downward trend, the total color difference (ΔE^*) increased gradually, and the deresination rate improved accordingly. Under treatment conditions of 180 °C and a higher ammonia mass fraction, ΔE^* reached its maximum value of 58.89, and the deresination rate achieved its highest value of 70.00%. The color parameters showed a local quadratic relationship with deresination rate, with the highest correlation coefficient being 0.713. In SVR model with radial basis function as kernel function, the root mean square errors for predicting lipid content and deresination rate were 0.523 and 4.315, respectively, and the coefficients of determination were 0.847 and 0.823, respectively. This prediction model can be applied to preliminary screening in deresination rate detection.

Conclusion This study successfully constructs a deresination rate prediction model for Masson pine wood based on SVR. The model has certain application value in preliminary screening of deresination rate detection, and can achieve rapid, convenient, and non-destructive detection to some extent. This research provides a new method for improving the efficiency and quality of deresination rate detection in Masson pine wood.