How Effective Are Lignin And Extractive Content-Based Multiple (Non)-Linear Regression And Artificial Neural Network Models In The Estimation Of HHV?

Abstract

Cellulose, lignin and extractive material are mixed in certain proportions by having isolated from lignocellulosic materials, such as Zeyrek stem, hazelnut shell and Scotch pine, respectively. Their higher heating values (HHVs) are determined by using a bomb calorimeter system. Estimated HHVs are calculated by applying these mixture ratios to some Multiple (Non)-Linear Regression (M(N)LR) and Artificial Neural Network (ANN) models from the literature. MLR3 model is developed from the data of this study and this model reveals the highest R2 (0.974), lowest MAPE (0.012) and RMSE (0.278) values. The closest estimation accuracy to the MLR3 model is obtained from MLR2 (R2:0.972, MAPE:0.066 and RMSE:1.714) in the comparative analysis. MNLR and ANN equations containing quadratic terms are found to show deviations up to 132.6% (ANN3). It is attributed to the lower size and poor homogeneity of the individual group of samples from which model equations are developed.