Artificial neural network modeling for predicting final moisture content of individual Sugi (Cryptomeria japonica) samples during air-drying

Sugi (Cryptomeria japonica D. Don) lumber is known to have a large variability in final moisture content (MCf) and is difficult to dry. This study assessed the capability of artificial neural networks (ANNs) to predict the MCf of individual wood samples. An ANN model was developed based on initial moisture content, basic density, annual ring orientation, annual ring width, heartwood ratio and lightness (L* in the CIE L*a*b* system). The performance of the ANN model was compared with a principal component regression (PCR) model. The ANN model showed good agreement with the experimentally measured MCf with a higher correlation coefficient (r) and a lower root mean square error (RMSE) than the PCR model, demonstrating the importance of nonlinearity of the variables and the higher capability of the ANN model than the PCR model. By adding redness (a*) and yellowness (b*) and drying time to the input variables of ANNs, r and RMSE values were improved to 0.98 and 1.2 % for the training data set, and 0.85 and 2.2 % for the testing data set, respectively. Although the developed ANNs are available under the limited conditions of this study, our results suggest that the ANNs proposed offer reliable models and powerful prediction capability for the MCf, even though wood properties vary considerably and their complex interrelations are not fully elucidated.