基于局部敏感哈希 双层随机森林的燃气轮机剩余使用寿命预测

针对燃气轮机剩余使用寿命预测中监测数据信息利用不够充分、退化过程难以表示、预测精度低等问题,提出了一种基于局部敏感哈希-双层随机森林的燃气轮机剩余使用寿命预测方法.该方法首先使用主成分分析充分利用多维监测数据信息构建健康因子来表示退化过程,并通过滑动截取的方法在健康因子曲线上获取训练数据;然后,通过p-stable分布局部敏感哈希相似性搜索算法匹配和需要预测的样本最相似的样本;进而,使用双层随机森林对剩余使用寿命进行回归预测.利用C-MAPSS数据集验证了该方法的有效性和准确性,研究结果可为其他非线性退化系统剩余使用寿命预测提供一定的参考.

Gas turbine remaining useful life prognosis based on local sensitive hash -double-layer random forest

Aiming at the problems of insufficient utilization of monitoring data information in the prediction of the remaining useful life of gas turbines, the difficulty of representing the degradation process, and the low prediction accuracy, this paper proposes a method for predicting the remaining useful life of gas turbines based on Local Sensitive Hash-double-layer random forest. First the health factors is constructed by used the principal component analysis to make full use of multi-dimensional monitoring data ,which represents the degradation process, and then the training data are extracted from the health factor curve; then the sample which is most similar with the sample that needs to be predicted is obtained through the p-stable distribution local sensitive hash similarity search algorithm, and the two-layer random forests are used to make a regression prediction on the RUL. Finally, the method is validated on the C-MAPSS data set, and the results show the effectiveness and accuracy of the method, which can provide a certain reference for the RUL prediction of other nonlinear degradation systems.