Prediction of the color change of heat-treated wood during artificial weathering by artificial neural network

The purpose of this study was to predict the color change of heat-treated wood during artificial weathering by an artificial neural network (ANN) model. Chemical component analysis was used to analyze the origin of color change of the heat-treated wood. The network included an input layer consisting of three input nodes, namely, the weathering exposure time, heat treatment temperature, and heat-treated wood species, a hidden layer using six neurons and an output layer consisting of one output node, namely heat-treated wood color. A hyperbolic tangent sigmoid transfer function was used in the hidden layer, and the training algorithm was the Levenberg–Marquardt backpropagation. According to the results, the mean absolute percentage errors (MAPE) were 8.17, 9.70, and 9.85% for the prediction of color change (ΔE) for training, validation and testing data sets, respectively. Determination coefficients (R²) above 0.92 were obtained with the proposed ANN model for all data sets. These results showed that the ANN model can be successfully used for predicting the color change of heat-treated wood during artificial weathering. FTIR results showed that the color change of heat-treated wood during artificial weathering is due to the change in the chemical composition, especially the photodegradation of lignin and wood extractives.