Effect of fibre-quality parameters on pulp properties by using multiple linear regression and artificial neural network

Paper strength is developed through pulp refining. This paper investigates a correlation of pulp fibre properties changes by pulp refining with the development of papermaking properties by multiple linear regression (MLR) analysis and artificial neural network (ANN). A total of 21 observations were made on the samples. The available dataset was divided into two groups: 81% of experimental results were selected as training data and 19% for model verification. A satisfactory level of the correlation between simulation data and experimental data was obtained. The linear regression analysis could be used for predicting tensile index (Adj. R 2 = .85), tear index (Adj. R 2 = .97) and burst index (Adj. R 2 = .91) based on the fibre characteristics such as fine content (%), fibre length (mm), fibre width (μm), degree of external fibrillation (%), curl index, kink index and coarseness (mg/m). However, the tensile index, tear index and burst index could be estimated by ANN more efficiently ( R 2 > .95).