A Study on Fault Diagnosis Method of Chemical Process Based on WOA-LSTM Hybrid Mode

In order to solve the problem of dynamic timing modeling and parameter optimization in chemical process fault diagnosis, this paper proposes a long short-term memory neural network (LSTM) diagnosis method based on Whale Optimization Algorithm (WOA) optimization. The adaptive search strategy of WOA optimizes the key parameters of LSTM (number of hidden layer nodes, learning rate, and number of iterations), and constructs a diagnostic model with time series feature learning ability. In the TE process benchmark test, the average diagnostic accuracy of the proposed method reaches 97.40%, which is 3% higher than that of the unoptimized LSTM. Experiments show that WOA-LSTM can effectively capture the dynamic characteristics of chemical processes and provide new ideas for fault diagnosis of complex industrial systems.