基于强化健康指数相似度量的航空发动机失效阈值设置

Aero-engine failure threshold setting based on enhanced health index similarity

中文关键词  航空发动机健康监测  健康指数构建  寿命预测  失效阈值设置  退化特征增强

英文关键词  aero-engine health monitoring  health index construction  RUL prediction  failure threshold setting  enhanced degradation features

中文摘要

      针对航空发动机剩余寿命预测中多监测参数对发动机性能退化表征能力不强及失效阈值设置困难问题, 采用强化健康指数相似度量法进行了失效阈值加权设置. 该方法首先利用Box–Cox变换增强发动机监测参量与飞 行循环间的关联性, 再采用降噪自编码器进行退化特征的提取. 为了增强退化特征对发动机剩余使用寿命的表征能 力, 采用多重核典型相关技术构建发动机的统一健康指数(HI). 引入全阶时间幂灰色预测模型对全寿命参考发动 机HI的失效阈值进行预测, 并结合基于KL距离的HI曲线相似度量方法对失效阈值加权融合, 以确定测试发动机的 失效阈值. 试验结果表明, 本文提出的失效阈值设置方法充分考虑了航空发动机性能退化过程的个体差异性与相似 性, 克服了传统失效阈值的主观设置问题, 有效地提高了航空发动机RUL的预测精度, 为实现航空发动机的健康管 理提供一种新思路与手段.

英文摘要

      Aiming to the weak characterization ability of aero-engine monitoring parameters and the difficulty in setting failure threshold in aero-engine residual life prediction, a weighted failure threshold setting method based on similarity measure of enhanced health index was proposed. This method applied the Box–Cox transform to enhance the correlation between aero-engine monitoring parameters and flight cycles, and then used the denoising autoencoder to extract degradation features. In order to enhance the characterization ability of degradation features for the RUL of aero-engine, the multiple kernel canonical correlation was used to construct the health index (HI) of aero-engine. The full order time power grey prediction model was introduced to predict the failure threshold (FT) of referencing aero-engines. Combined with the similarity measure of HI curve based on KL distance, the weighted fusion of reference FTs is used to determine the FT of test aero-engines. The experiment results showed that the FT setting method proposed in this paper fully considered the individual differences and similarities of aero-engine degradation process, effectively overcame the problem of subjectively FT setting, effectively improved the prediction accuracy of aero-engine RUL. It provided a new idea and means for the realization of aero-engine health management.