Research on equipment fault prediction expert system based on big data dimension reduction

Abstract

Due to the complex structure and massive volume of large equipment in petrochemical enterprises, it was very difficult to carry out inspection and maintenance work. The traditional expert system used a single knowledge base and reasoning machine, so the processing efficiency and prediction accuracy were low. In this paper, the dimensionality reduction which is used to process a large amount of physical information collected by the sensors, remove redundant components and extract main characteristic parameters. Meanwhile, combined with the powerful pattern recognition and judgment ability of fuzzy neural network, the forward reasoning mechanism and error back propagation keeping continuous training correction until meeting the accuracy requirements, the expert system gave the equipment fault prediction conclusion. Finally, taking main fan oil station circulating oil pump and motor of heavy oil catalytic cracking unit and nine stages of bearing damage as monitoring objects, the fault predictions was implemented by the following steps: data collection, feature extraction, dimension reduction of big data and expert system with the fusion of fuzzy neural network, and it is proved that the expert system with fuzzy neural network based on dimension reduction can greatly improve the convergence speed.