Industrial Fault Detection Based on Discriminant Enhanced Stacking Auto-Encoder Model

In the recent years, deep learning has been widely used in process monitoring due to its strong ability extract features. However, with increasing layers of network, compression features by model will lead loss some valuable information and affect model’s performance. To solve this problem, a fault detection method based on discriminant enhanced stacked auto-encoder is proposed. An network structure designed, original data added each hidden layer pre-training problem feature extraction process. Then self-encoding combined spectral regression kernel analysis. The category introduced into optimize enhance discrimination extracted Euclidean distance for From Tennessee Eastman experiment, it can be found that accuracy about 9.4% higher than traditional method.