Design of an Expert System for Pulping Based on Fuzzy BP Network

Pulping process is an important link in paper industry. To control the pulping time and get homogenized and well-proportioned pulp, pulper was improved by an expert system based on BP network and fuzzy controller. Pulp consistency and variation concentration are the inputs and pulping time is the output of the system. By empirical data, membership functions of consistency, variation concentration and pulping time were set up and sixteen fuzzy control rules were made. Combined with BP network, the number of input neurons was fourteen and the number of output neurons was eight. All control rules were training samples and simulation was completed by MATLAB. The results show that comparing with traditional fuzzy controller, capacity of the expert system is improved because of the memory and association of BP network. So the expert system can get knowledge directly from experience and reason automatically, without time-consuming searching in tables. The expert system has been used for about one year in Sichuan Hongyuan Group and the capacity can be used as needed.