基于性能非线性退化统计模型的发动机剩余寿命预测

Prediction of Engine Residual Life Based on Performance Nonlinear Degradation Statistical Modeling

针对航空发动机在性能退化过程中普遍存在的非线性和不确定性问题,提出一种基于非线性退化数据的统计模型和剩余寿命预测方法。通过对发动机性能真实退化轨道的分析,采用统计回归的建模方法建立发动机退化轨道模型,利用发动机的历史数据,通过最小二乘估计求解模型中的未知参数;根据贝叶斯准则,以发动机实时监测数据与参数的先验分布对模型中的参数进行实时更新,以发动机性能退化量首次达到红线值作为失效依据,采用蒙特卡洛仿真的方法得到发动机剩余寿命分布,实现了对个体发动机剩余寿命的预测;通过试验数据进行发动机剩余寿命的预测,验证了该方法的准确性。结果表明:根据发动机退化数据结合退化模型得到的个体发动机剩余寿命实时预测值末端均方根误差为0.02588,可以辅助指导维修决策。

Aiming at the nonlinear and uncertain problems commonly existing in the performance degradation process of aeroengines,a statistical model and residual life prediction method was proposed based on nonlinear degradation data. Through the analysis of the real degradation trajectory of the engine performance,the statistical regression modeling method was used to establish the engine degradation trajectory model,and the historical monitoring data of the engine were used to solve unknown parameters of the model through least square estimation. According to Bayes criteria,the model parameters were updated in real time with the prior distribution of engine real time monitoring data and parameters. Taking the first time that the engine performance degradation reached the red line value as the failure criteria,the engine residual life distribution was obtained by Monte Carlo simulation,and the prediction of the individual engine residual life was realized. The accuracy of this method was verified by predicting the engine residual life through the test data. The results show that the root mean square error of the real-time prediction value of individual engine residual life obtained from the engine degradation data combined with the degradation model is 0.02588,which can assist in guiding maintenance decisions.