Remaining Useful Life Prediction For Aero-engines Based On Stochastic Model

As the complexity degree for aero engines is increasing, It is necessary to implement engine health management(EHM) on purpose of ensuring safety, reliability and reducing the life-cycle-cost for aero engine. One of the key issues in implementing EHM is how to predict the remaining useful life(RUL), while, the accurate RUL prediction can provide significant information for maintenance decision. On account of the shortcomings in the present study, the following contents were proposed:Firstly, Aiming at aero-engines with less failure data, a process monitoring model based on PCA and Kernel PCA is presented. On condition of multivariate sensors data, we define two statistics namely T-square and SPE to achieve the fault detection requirement for aero engines. Afterwards, a fault identification method on the basis of contributions is developed for identifying the faulty variables for aero-engines.Secondly, taking nonlinearity, randomness and measurement errors in consideration simultaneously, a RUL prediction approach is established for aero engine conditioned on nonlinear Wiener process. In addition, in order to predict the RUL on-line, degradation state and model-parameters are estimated jointly integrated EM algorithm with Kalman algorithm, and the RUL probability distribution is updated subsequently.Lastly, upon the RUL prediction, a predictive maintenance optimization model to decide replacing time and spare ordering time jointly for engine fleets is raised. After then the biogeography-based optimization algorithm is designed to solve the optimization model numerically. Calculations show that the mentioned model can reduce the expected maintenance-cost-rate, which satisfied the need of EHM.