Self-attention convolutional neural network based fault diagnosis algorithm for chemical process

Aiming at the problem that the traditional "feature Classifier" model is difficult to make full use of the dynamic information of chemical process data, a self-attention convolutional neural networks (SA-CNN) fault diagnosis algorithm based on time-series self-attention mechanism is proposed. The algorithm uses the moving window method to divide the data into blocks, and then splices the features of the sub blocks to form new features with local dynamic information; At the same time, self-attention mechanism is introduced to characterize the correlation between samples in the window, and combined with one-dimensional convolutional neural network to enhance the feature extraction ability of the model. This method is applied to the fault diagnosis of Tennessee Eastman (TE) process data, and compared with convolutional neural network (CNN), long-term and short-term memory network (LSTM), artificial neural network (ANN) and support vector machine (SVM). The results show that our proposed algorithm can effectively improve the accuracy of chemical process fault diagnosis.