Research on Fault Condenser Feature Extraction for Multi-modal Processes
To address the issue of online monitoring and fault feature extraction in multi-modal operation process of condenser equipment,a fault feature extraction method integrating k-means clustering and re-constructed principal component analysis was proposed.First,k-means clustering was used to identify modes.The clustering results show significant differences in squared prediction error(SPE)statistics a-mong different modes,and it is necessary to establish monitoring models for sub-modes.In order to obtain fault features effectively and suppress residual pollution in fault separation process,fault feature vectors were obtained by reconstructing principal component analysis method,and the fault feature extraction for multi-modal process was realized.Results show that fault variables and their feature vectors can be separa-ted and detected by using the reconstructed contribution graph method,and the residual pollution problem can be avoided effectively,and the fault location accuracy is good.
condensermulti-modal processfault feature extractionreconstruction-based principal com-ponent analysisfault isolation
万方数据
维普
