Research On Reliability Assessment And Remaining Useful Life Prediction Of Mechanical Systems With Performance Degradation

Mechanical equipment(such as: wind power system,aircraft engine and weapon equipment,etc.)has the characteristics of complex structure,function integration,long life and difficult maintenance.The system reliability assessment and remaining life prediction of such equipment can find out the weak links of system working ability in time and formulate scientific and reasonable maintenance strategy,so as to effectively extend the service life and reduce the cost.The performance degradation of mechanical system will gradually increase during operation,and the failure state of parts is no longer a single normal fault,especially in complex conditions,it presents a variety of complex characteristics,such as: multi-fault state,dynamic and uncertainty,etc.at the same time,the system will also be impacted by the external operating environment,and there is competition with the performance degradation failure of the system itself The relationship between them.During the degradation process,the system will produce a large number of data which can represent the degradation information,and the residual life prediction model can be established by using the performance degradation data.Therefore,starting from the performance degradation process of mechanical system in operation,this paper discusses and studies the reliability and residual life prediction index to measure the working capacity of the system.(1)Considering the problems of multi-fault state and difficult to obtain accurate failure rates in the operation of mechanical system,a reliability assessment model of multi-fault state system based on T-S fault tree and interval-valued fuzzy Bayesian network is established.The T-S fault tree is constructed by the failure model of the research object,and it is mapped to the Bayesian network model.Then,the fuzzy set and interval theory are introduced to fuzzify the boundary value of the fuzzy subset of failure rate,and the interval-valued triangle fuzzy polymorphic Bayesian network is constructed,which can express the multi-fault state and deal with uncertain factors of the system.Under the condition of knowing only the root node fault state,the reliability index solution process is obtained.(2)Aiming at the problem that the performance degradation process has the influence of external shocks on its own degradation,it is assumed that the external shock type is the extreme shock model.Combined with the distribution function of extreme shock,the problem that the external shock amplitude exceeds the critical threshold will accelerate the degradation rate of the system itself and change the failure threshold is studied,and the general expression of reliability based on failure correlation is established.The rationality of the model is verified by a specific example.(3)In order to solve the problem of difficult extraction of multi monitoring information in the process of performance degradation,a feature extraction and processing method of mechanical system based on Feature Engineering is proposed.Firstly,the data set or sensor which has rich information and can reflect the degradation trend of system performance is selected.Secondly,the data sets are normalized and processed.Finally,the kernel principal component method is used to reduce and fuse the high-dimensional degradation feature sets to obtain the low-dimensional relative feature sets that can represent the degradation of equipment.(4)Through the fusion of multi-monitoring information,the remaining useful life prediction method of deep learning model is proposed to improve the prediction accuracy of the model for complex systems.Firstly,a bidirectional long-term and short-term neural network model integrating kernel principal component method is established.The lowdimensional feature set is input into the model,and the advantages of using the network model to deal with long-term sequences and the characteristics of bidirectional propagation are used to obtain the mapping relationship between state monitoring data and residual life,and the prediction results are obtained.Finally,the results are verified by aero-engine and compared with other models.