Research Of Performance Monitoring And Fault Detection Method Based On Data-driven

AbstractWith the rapid development of modern industry,all kinds of equipment are increasing and gradually becoming intelligent.In order to ensure the normal operation of various components of the system,a large number of sensors are used to monitor the status of different equipment.The large amount of process data generated when the system is running can be used to effectively detect the performance,operation status and failure of the system.However,the redundant information existing between the data in the actual project cannot accurately monitor the performance of the system and detect the failure of the system.Therefore,the starting point of this paper is to solve practical problems,and the main research contents are as follows:In order to effectively analyze the influence of disturbances and control actions on performance variables under non-Gaussian data,the monitoring efficiency of non-Gaussian process variables can be improved.A no-Gaussian correlation algorithm based on independent component analysis is proposed to monitor the non-Gaussian process variables and non-Gaussian performance variables.First,non-Gaussian information is extracted from the original data center by independent component analysis(ICA).On this basis,the non-Gaussian information is divided into non-Gaussian performance-related subspace and non-Gaussian process-related subspace by canonical correlation analysis(CCA).The proposed method can effectively analyze the influence of disturbance and control actions on performance variables under non-Gaussian data,and improve the monitoring efficiency of non-Gaussian process variables.Finally,a case study is used to illustrate the applicability and effectiveness of this method.To avoid process and sensor noise affecting the difficulty of slowly changing fault detection in high-speed trains.By introducing the idea of dynamic inner into the framework of multiple statistics,a novel dynamic inner slowness feature analysis(DISFA)is proposed in this article.This method considers both dynamic and static conditions and improves the detection speed of slow-change faults of running gears,and improves the detection rate of slow-change faults.Compared with other traditional methods,this method is more sensitive to slow changes,can effectively analyze fault-related data,and improve the detection efficiency of slow-changing faults.First,the validity of the method proposed in this paper is proved by mathematical deduction,then it is verified by actual operationdevices.By dealing with the non-Gaussian measurement and slow-change faults in running gear systems,this paper presents a fault detection scheme named time-series independent component analysis(Ts ICA),where the time-series characteristic is taken into account.The time-series algorithm can extract slow-change data by controlling the weight and update the data set once after each extraction.Since only the influence of non-Gaussian data is considered,the independent component analysis algorithm is used to detect faults after the data is updated.This method effectively extracts the latent information between the data,and also shortens the time for the fault to be detected for the first time.The feasibility of the proposed scheme is verified by the running department system.