Fault Diagnosis And Performance Prediction Of Complex Dynamic Systems

With the development of social science technology,high technology has been injected into all aspects of human life,which makes the systems more complex,integrated and intelligent in the fields of industry,communications,aviation,etc.At the same time,with the rapid development of these complex engineering systems,development and production costs are getting higher and higher,reliability and safety requirements are also significantly improved.When faults occur in these systems,they can cause serious economic losses and environmental pollution,even endanger personal safety and cause catastrophic damage.Therefore,it is important to perform effective troubleshooting and performance prediction in these systems.In this thesis,two complex dynamic systems of TE process and lithium-ion battery capacity decay process are studied on fault diagnosis and performance prediction:(1)In the fault diagnosis of complex systems,the traditional method has the problem of low diagnostic accuracy.To solve this problem,an optimized probability neural network(PNN)fault diagnosis method is proposed and the improved butterfly algorithm is employed to optimize the PNN network to make it more efficient.Firstly,for the traditional butterfly optimization algorithm,convergence speed always is much low,and it is easy to fall into the local optimal.Hence,the opposite learning strategy is used to adjust the distribution position of the butterfly's initial population,to make them closer to the optimal position of the search space.Secondly,the improved butterfly algorithm is applied to the smoothing factor optimization of PNN,which effectively improves the ability of the neural network to diagnose.Finally,the optimized PNN network is verified in the fault diagnosis of TE process.The simulation results show that,comparing with the traditional fault diagnosis methods,the proposed method has a higher fault detection accuracy,which proves the effectiveness and superiority of the method.(2)In view of the performance of complex systems,the traditional method has the problem of low prediction accuracy,a performance prediction method based on particle filtering algorithm optimized by the improved butterfly algorithm proposed before.Firstly,the proposed improved butterfly algorithm is employed to optimize the resampling process of the particle filter algorithm,which improves its predictive performance,overcomes the particle degradationand increases the particle diversity.Secondly,the improved particle filter algorithm is introduced to predict the remaining useful life(RUL)of a lithium-ion battery system.The experiment results illustrate that compared with the traditional particle filter algorithm,the proposed method can perform a more accurately prediction of the RUL of the lithium-ion battery.