A method for industrial process fault diagnosis based on stacked denoising autoencoders and outlier detection algorithm.

This paper proposes a fault diagnosis method for industrial processes. For relatively complex processes with abundant data, a stacked denoising autoencoder (SDAE) is utilized to extract data features. Deep features are obtained through pre-training and fine-tuning. Additionally, outlier detection algorithm is employed to diagnose the occurrence of fault patterns. The proposed method is tested using the Tennessee Eastman (TE) chemical process dataset and compared with other fault diagnosis methods in terms of specific fault recognition accuracy. Experimental results demonstrate the correctness and effectiveness of the proposed method in this paper.