Expert system for controlling sinter chemistry based on neural network prediction

A sintering predictive model of chemical composition based on many periods was developed by the BP neural network algorithm with appending momentum and adaptive variable step size linear reinforcement. Using knowledge base that was based on database technology and illation with forward inference, an expert system was designed for controlling sinter chemistry. Since the system was plunged into application, the hit ratio of the predictive model is over 90% steadily, and the acceptance of operation suggestion is 92%. The goal of controlling chemical composition steadily is actualized.