Research on Data-driven Distributed Process Monitoring for the Entire Process
Due to significant fluctuations in SER statistics and incomplete number of fault detections in existing monitoring methods,this study focuses on data-driven distributed process monitoring for the entire process.This method converts different subunits into an undirected graph model,calculates the similarity level of nodes,decomposes the graph model elements,and analyzes the interrelationships of each sub variable.Then,construct a minimum volume 3D surface sphere containing all training samples on each sub variable,calculate the distance range from the center of the 3D surface sphere,and detect the occurrence of faults.Finally,a Bayesian strategy is adopted to obtain global monitoring results,which are compared with variable control thresholds to determine the operating status and complete the monitoring.The experimental results show that during the sampling process,the SER of the experimental group was in a normal state,and even after introducing faults,it did not exceed the red line.After 12 tests,the number of fault detections obtained was the same as the actual number of faults,achieving a good monitoring effect.
万方数据
维普
