关键词: 复杂工业过程, 故障预测, 随机森林, 减法平均, 双向长短期记忆神经网络

Abstract: Due to the time-varying characteristics caused by the drift of working conditions in complex industrial processes,in order to solve the problem of low fault prediction accuracy,a hybrid fault prediction model based on SABO-RCA-BiLSTM was constructed. Firstly,the random forest algorithm was used to analyze the importance of the features and perform data filtering to reduce data redundancy and retain key features. Then,the Convolutional Neural Network (CNN) was introduced to solve the problem that the Bidirectional Long Short-Term Memory (BiLSTM) neural network cannot capture spatial features when facing multi-dimensional feature input,and the attention mechanism was added to assign different attention weights to each part of the input feature sequence to enhance the attention to the key information. Finally,the Subtractive Average-Based Optimization (SABO) algorithm was used to optimize the model parameters to further improve the fault prediction performance. The model was verified on the Tennessee-Eastman (TE) process. The results show that in the face of two different types of faults,the average absolute error of the optimized model is reduced by 32 % and 30 % respectively compared with which before optimization. Compared with CA-BiGRU,CA-BiLSTM,MVMD-CA-BiLSTM and SAC-BiLSTM models,the fault prediction accuracy determination coefficient is increased by 23.70 % at most,which effectively solves the problem of low fault prediction accurary in complex industrial processes.

Key words: complex industrial processes, fault prediction, random forest, subtraction average, bidirectional long short-term memory neural network