Fault diagnosis of chemical process based on CNN-BiLSTM-GAM model

Due to the issues of high dimensionality, nonlinearity, and strong correlations inherent in petrochemical data, traditional fault diagnosis models often exhibit limited accuracy when diagnosing faults in chemical processes. In response to this issue, this paper proposes a fault diagnosis model based on a CNN-BiLSTM-GAM framework. The model first utilizes a Convolutional Neural Network (CNN) to extract features and reduce the dimensionality of the collected fault data. Next, these features are input into a Bidirectional Long Short-Term Memory (BiLSTM) network for deep feature extraction. Finally, the deep features are passed through the Global Attention Mechanism (GAM) for feature fusion, and fault diagnosis results are obtained via a Softmax function after the fusion process. The proposed CNN-BiLSTM-GAM model is evaluated through fault diagnosis experiments on the reaction regeneration system of a petrochemical plant, and compared with an unimproved LSTM algorithm. Results demonstrate that the CNN-BiLSTM-GAM model achieves higher fault diagnosis accuracy and robustness, offering significant advantages in real-world chemical process applications.