Fault Diagnosis Of Electroslag Metallurgical Process Based On CHMM And DKPCA

Electroslag furnace metallurgy is an important mode of production of special steel,and its products are often of high value.Once the process of a significant failure,it will cause huge economic losses or even security incidents.In this paper,the characteristics of non-linearity,time series correlation and batch data unequal length exist in the data of electroslag metallurgical process,and the dynamic kernel principal component analysis method is applied to the feature extraction of electroslag metallurgical process data Based on the continuous hidden Markov model for fault detection and fault identification,some new process monitoring and fault diagnosis methods are proposed,including:(1)Based on the non-linearity of the data of electroslag metallurgical process and the unequal length of batch data,the DKPCA method is used to extract the characteristic data of the process data and convert the original data into continuous process data so that the continuous process data can be directly used by CHMM Troubleshooting.(2)A fault detection method based on CHMM and DKPCA for electroslag metallurgical process is proposed.Firstly,the DKPCA algorithm is used to extract the characteristics of the electroslag metallurgical process data.The dynamic data tracing of the extracted principal component sequence is carried out by using the VMW technique,and the real-time threshold of the online fault detection is calculated.Then,the real-time statistics And the size of the real-time threshold to determine whether the system failure,in order to achieve based on CHMM and DKPCA electroslag metallurgical process real-time online fault detection.Finally,the process data of the electroslag metallurgical process simulation system are established to simulate the method.It shows that this method can accurately detect the various faults in the process of electroslag metallurgy.(3)A fault identification method of electroslag metallurgical process based on CHMM and DKPCA is proposed.Firstly,the DKPCA algorithm is used to extract the characteristic data of the electroslag metallurgical process data,and then the dynamic data is tracked by the VMW technique.The CHMM model of the fault is trained and all the fault cases are trained.Calculate the degree of matching between the test data and each failure model to determine which type of fault occurred in the data to be measured.Finally,the simulation data of the electroslag metallurgical process based on CHMM and DKPCA are simulated by the simulation platform.The effectiveness of the algorithm is verified.