基于非线性Wiener过程航空发动机性能退化预测

Performance Degradation Prediction of Aero-Engine Based on Nonlinear Wiener Process

针对基于线性随机过程航空发动机性能退化预测精度不高的问题,提出了一种漂移系数为指数形式的非线性Wiener过程发动机性能退化建模的方法,可以预测航空发动机的性能退化。基于直接监测发动机性能退化数据,构建发动机性能退化模型,根据Wiener过程首达阈值时间的数学性质,推导出性能退化的概率分布。通过极大似然估计构建退化模型中未知参数的似然函数,利用遗传算法得到发动机总体模型参数的离线估计值。考虑到不同发动机个体间的差异性,采用贝叶斯公式,结合发动机的实时监测数据与总体模型参数的先验分布对模型中随机参数进行实时更新,从而达到对个体发动机性能退化的实时预测。最后,选择商用航空发动机仿真数据集(C-MAPSS)进行实验,结果表明:针对个体发动机基于非线性随机过程方法,实时更新非线性Wiener方法能够提高航空发动机运行后期性能退化预测的准确性,提供更加可靠的预防性维修决策。

Aiming at the problem of low accuracy of aero-engine performance degradation prediction based on linear stochastic process,a nonlinear Wiener process engine performance degradation modeling method with exponential drift coefficient is proposed,which can predict the performance degradation of aero-engine.Based on the direct monitoring of engine performance degradation data,the engine performance degradation model was constructed,and the probability distribution of remaining life was deduced according to the mathematical properties of the first threshold time of Wiener process.The likelihood function of unknown parameters in the regression model is constructed by maximum likelihood estimation,and the off-line estimation of the overall model parameters of the engine is obtained by genetic algorithm.Considering the differences among different engine individuals,the Bayesian formula is used to update the random parameters in the model in real time by combining the real-time monitoring data of the engine with the prior distribution of the overall model parameters,so as to update the residual life prediction of the individual engine in real time.Finally,the commercial aviation engine simulation data set(C-MAPSS)is selected for experimentation.The results show that:for individual engines based on the nonlinear stochastic process method,real-time updating of the nonlinear Wiener method can improve the accuracy of the performance degradation prediction of the aircraft engine in the later stage of aero-engine operation,and provide more reliable preventive maintenance decisions.