The Prediction Of Remaining Useful Life And Preventive Maintenance Decision For Mechanical System

In the life cycle of a mechanical system, the reliability assurance of the operation is the key to achieving system design and the goal of manufacturing target. The remaining useful life of mechanical system should be predicted based on real-time monitoring information for reliability and safe operation, and make scientific maintenance decision based on the characteristics of the system. In other words, to realize the implementations of the system failure prediction and health management system is a very important technical approach. Therefore, the study of how to construct the remaining useful life prediction models and scientific maintenance decision method based on the degradation characteristics and monitoring status information of the mechanical system has become an important research issue in the field of mechanical engineering reliability.Aiming at different degradation characteristics of mechanical systems, it has been studied that the remaining useful life prediction model and the maintenance optimization decision method. The numerical calculation and some experimental data were taken as examples to verify the feasibility of proposed strategy. The main research work of this thesis is concluded as follows:Aiming at the sort of mechanical systems which has deficiency of history operational data and is incapable of doing status monitoring, a prediction model of the remaining useful life was construction from known distribution functions of system’s initial life under the circumstance of not knowing the distribution functions of system’s degradation status. And a preventive maintenance decision was proposed with the threshold value based on the remaining life. According to the theory of renewal process, a maintenance decision optimization model that using minimizing long-term average cost rate as target functions was constructed with the optimized variables based on prediction interval and remaining useful life threshold value of the system.Aiming at systems capable of real time monitoring and considering remaining useful life and imperfect maintenance effects which have two conditions of independent and dependent, a real-time remaining useful life prediction model and an imperfect maintenance decision model were established based on stochastic filtering theory. A maintenance strategy of imperfect preventive maintenance and replacement strategy by taking remaining useful life as threshold was proposed, and an optimization model was establish to choose the threshold of preventive maintenance as optimal variable and the minimum long-term average cost rate as object function. Further proposed to consider more after imperfect preventive maintenance can reduce system performance does not meet the reliability requirements need to be replaced in the maintenance strategy, establishes system preventive maintenance threshold, preventive replacement threshold and stop the forecast time threshold as optimization variables, to minimize the long-term average cost rate as the objective function of maintenance optimization decision model.In order to solve the problem which is hard to predict precisely remaining useful life after abrupt change of degradation in the fatigue degradation process of the system, a new model modification method for real-time remaining useful life prediction is put forward. This method detects abrupt change points in the degradation process using received real-time monitoring information, revises status space model to change filtering effects after sudden degradation according to life information provided by abrupt change points, and verifies the prediction model using real time monitoring information from the test bed of fatigue life of gear contact.Considering existent stochastic dependency influences of continuous degradation states between components, a dynamic grouping preventive maintenance model that minimizes the long-term average maintenance cost of the system is construct. A penalty function evaluates the additional cost of shifting the maintenance time. The maintenance strategy is dynamic and adaptive, as the grouping maintenance date and the grouping structure updated based on the real-time prediction of components’ remaining useful life over a long-term period.Considering the influence of the continuous degradation state one-way stochastic dependence, introducing the affected parts monitoring information as covariate, a remaining life prediction model of multiple-component system is established based on stochastic filtering. And a dynamic opportunistic maintenance strategy for multi-component systems is establish according to the real-time remaining useful life predicted of the various components.