基于MIC特征提取与BO-CatBoost的航空发动机RUL预测

针对航空发动机传感器监测的退化参数提取困难,易受噪声干扰及发动机剩余使用寿命预测精度不足等问题,利用最大信息系数、贝叶斯优化算法和类别特征梯度提升算法,提出了一种新的发动机剩余使用寿命预测模型。首先,为有效解决特征提取不足的问题,对采集的传感器历史监测特征进行最大信息系数相关性计算,提取出对发动机寿命运行周期影响较大的关键退化特征。其次,为解决剩余使用寿命预测中的梯度偏差及预测偏移问题,使用基于贝叶斯优化的类别特征梯度提升方法对航空发动机进行剩余使用寿命预测。最后,在美国航空航天局提供的商用模块化航空推进系统仿真数据集上进行实验,结果表明所提预测方法的性能较好,验证了该方法的有效性。

A Remaining Useful Life Prediction of Aero Engines Based on MIC Feature Extraction and BO-CatBoost

Aimed at the problems that the degradation parameters monitored by aviation engine sensors are difficult to be extracted, is subject to noise interference, and accuracy is insufficient in predicting engine remaining useful Life (RUL), a new remaining useful life prediction model is proposed by utilizing the maximal information coefficient (MIC), bayesian optimization (BO) algorithm, and categorical boosting(CatBoost) algorithm. Firstly, to effectively address the issue of inadequate feature extraction, the collected historical monitoring features of sensors are subjected to maximal information coefficient correlation calculation to extract key degradation features from the significant impact on the engine''s operational lifespan.Secondly, to address the gradient bias and prediction offset issues in remaining useful life prediction,the categorical boosting algorithm method based on bayesian optimization is employed to predict the remaining useful life of aero engines. Finally, experiments are conducted on the commercial modular aero-propulsion system simulation (C-MAPSS) dataset provided by national aeronautics and space administration (NASA). The results show that the proposed prediction method has a good performance, and is valid.