An EPS-kNN Fault Propagation Analysis Method for Industrial Processes Based on the Fusion of Knowledge and Data

To reduce the false connections resulting from numerous redundant paths in the fault propagation analysis for industrial processes, an equal probability symbolized k-nearest neighbors (EPS-kNN) fault propagation analysis method based on the fusion of fault data and process knowledge is proposed. Subblock interaction monitoring is introduced to identify potential fault areas and eliminate redundant variables. The complex network model is incorporated into the EPS-kNN-based fault propagation path identification to improve accuracy and interpretability. Different types of faults in the Tennessee Eastman process and Ammonia synthesis process are applied to verify the effectiveness, and the result shows that the proposed method can identify the fault propagation path more effectively and reduce the occurrence of false connections compared with the traditional methods.