| Industrial production is rapidly moving towards integration,a trend that is directly driving changes in industrial intelligence,digitization and automation.The more complex industrial production equipment becomes,the more its safety becomes a concern for government and society.The overhaul and maintenance of equipment has become an inevitable prerequisite for industrial production and manufacturing.In such a general environment,dynamic system failure prediction and lifetime prediction are also developing,becoming an important means to improve the efficiency and quality of industrial production and to guarantee the right of industrial production at all times,with important research significance and practical significance.This paper reviewed the current state of the art in failure prediction techniques.It is found that for complex equipment,accurate mathematical models are often difficult to construct due to their relatively complicated failure mechanisms and failure modes.At the same time,the working process of dynamic systems generates a large amount of process data,which is huge in volume and has strong dynamic characteristics,and the correlation changes between variables affect the accuracy of data analysis,which in turn affects the accuracy of fault detection and prediction.To address these two problems,this paper proposes a fault prediction method based on the dynamic intrinsic principal component analysis method and a remaining life estimation method based on the Markov chain model for traction motor temperature data.Thus,the research work is as follows:(1)The dynamic system work operation process,generating a large number of process data,this data volume is huge,has a strong dynamic characteristic,the correlation between variables change affect the accuracy of data analysis.In turn,it affects the accuracy of fault detection and fault prediction.The traditional principal component analysis method cannot extract the dynamic characteristics of process data.In this paper,we proposed a dynamic system fault prediction based on dynamic intrinsic principal component analysis for this data characteristic.The main processes are model building,fault detection,fault reconstruction and estimation,and fault magnitude prediction.Dynamic Inner Principal Component Analysis(Di PCA)is applied to model the data,extract the dynamic latent variables within the data,and partition the data into dynamic and static spaces.Then,the fault data are reconstructed and the fault magnitude is estimated,and the estimated fault magnitude is predicted by combining the vector autoregressive model(VAR),and the feasibility and effectiveness of the method are verified by using the ideal data(TE process)and the actual temperature data generated during the operation of the train traction motor.(2)Remaining life prediction is a continuation of fault prediction,and this paper used the Markov chain model to estimate the remaining life of the train traction motor.First,the state characteristics of the traction motor are analyzed to conform to the Markov chain model,the state space of the motor operation is partitioned,and the state transition probability is calculated.Second,the failure time of the traction motor is calculated by combining the failure curve(bath tub principle),and finally the remaining effective life of the motor is estimated.The analysis results show that the effective life of the motor is advanced by 10 h.Finally,the overall work of this dissertation is summarized,the shortcomings of the study are analyzed and the future research directions are also presented. |
