An improved method for predicting pulp properties and scheduling the ratio of waste paper

摘要

Focusing on the automatic production scheduling of the ratio of waste paper in a paper mill, the research target was minimizing the purchase cost of waste paper under multiple constraints. Having divided, the field data (mixing ratios of waste paper and pulp properties) into the training and validation data sets, the scheduling ratios of waste paper were optimized. Firstly, from the point of view of the average pulp properties and the variances, the predicted results proved that the BP-NN predicted model accuracies of the pulp properties with the mixing ratio of waste paper were better than those of SVM and GA-SVM. Secondly, the minimization of the purchasing costs of waste paper under some constraints were obtained with the BP-NN predicted model and the non-dominated sorting genetic algorithm (NSGAII), Comparing with the general GA in the previous study, the scheduling results improved the pulp brightness by 9.24%, and reduced the purchasing costs of waste paper by 2.16%.