作者：李鹏, 丁瀛, 黄培炜, 杜艺博

作者单位：华东交通大学机电与车辆工程学院，江西 南昌 330013

关键词：剩余寿命预测;特征优选;长短时记忆网络;NSGA-II

摘要：

研究针对航空发动机剩余寿命预测中的特征选择问题，提出一种基于特征优选的航空发动机剩余寿命预测方法。首先，基于长短时记忆网络建立单序列监测数据预测模型，预测不同工况下发动机测试样本的数据集。其次，基于特征融合与相似性匹配法，计算剩余寿命预测评价指标，并基于非劣分层遗传算法进行特征优选。最后，对基于特征优选的剩余寿命预测效果进行验证，结果表明：1）在同一非劣层中，随着特征数量的增加，剩余寿命评价指标会先变优，再变差。这种现象说明过多的特征数量会影响优化结果，证明特征优选的必要性。2）以不同工况的样本集分别用于特征多目标优选和剩余寿命预测研究，其预测误差均值相差小于16%，表明所提方法的正确性和鲁棒性。

Prediction of remaining useful life for aero engine based on feature optimization
LI Peng, DING Ying, HUANG Peiwei, DU Yibo
School of Mechatronics & Vehicle Engineering, East China Jiaotong University, Nanchang 330013, China
Abstract: Aiming at the problem of feature selection in aeroengine remaining useful life prediction, a method of aeroengine remaining useful life prediction based on feature optimization is proposed. First of all, based on the long-short-term memory network, a single sequence monitoring data prediction model is established to predict the data sets of engine test samples under different operating conditions. Secondly, The index of residual life prediction is calculated, and the feature optimization is carried out based on the non-dominated sorting genetic algorithm II. Finally, the effect of remaining useful life prediction based on feature optimization is verified, and the results show that: 1) In the same non-inferior layer, with the increasing number of features, the evaluation index of residual life first become better and then become worse. This phenomenon shows that the excessive number of features affect the optimization results and proves the necessity of feature optimization. 2) The sample sets of different working conditions are used in feature multi-objective optimization and remaining useful life prediction respectively, and the average difference of prediction error is less than 16%, indicating the correctness and robustness of the proposed method.
Keywords: remaining useful life prediction;feature selection;long short-term memory networks;NSGA-II
2022, 48(11):27-33  收稿日期: 2021-04-06;收到修改稿日期: 2021-10-19
基金项目: 国家自然科学基金（52165016）；江西省教育厅科技项目（GJJ210631）
作者简介: 李鹏（1976-），男，江西南昌市人，教授，硕导，博士，主要研究方向为智能结构与材料、状态监测、最优化方法
参考文献
[1] 孙同敏. 基于DBN-SVM的航空发动机健康状态评估方法[J]. 控制工程, 2021, 28(6): 1163-1170
[2] 李永进. 智能航空发动机性能优化控制技术研究[D]. 南京: 南京航空航天大学, 2018.
[3] WANG T. Trajectory similarity based prediction for remaining useful life estimation[D]. Cincinnati: University of Cincinnati, 2010.
[4] LIU K, HUANG S. Integration of data fusion methodology and degradation modeling process to improve prognostics[J]. IEEE Transactions on Automation Science and Engineering, 2016, 13(1): 344-354
[5] 王金杰, 李炜. 混合互信息和粒子群算法的多目标特征选择方法[J]. 计算机科学与探索, 2020, 14(1): 83-95
[6] RAO H, SHI X, RODRIGUE A K, et al. Feature selection based on artificial bee colony and gradient boosting decision tree[J]. Applied Soft Computing, 2019, 74: 634-642
[7] 巫红霞, 谢强. 基于加权社区检测与增强人工蚁群算法的高维数据特征选择[J]. 计算机应用与软件, 2019, 36(9): 285-292,301
[8] 黄海超, 陈景雅, 孙睿. 基于VMD-LSTM轨道交通客流预测模型[J]. 华东交通大学学报, 2021, 38(1): 95-99
[9] MUHAMMAD M A, MUHAMMAD I B, MUHAMMAD H, et al. Industrial financial forecasting using long short-term memory recurrent neural networks[J]. International Journal of Advanced Computer Science and Applications, 2019: 10
[10] 史业照, 郭斌, 郑永军. 基于LSTM网络的IGBT寿命预测研究[J/OL]. 中国测试, 2020: 1-6[2021-12-10]. http://kns.cnki.net/kcms/detail/51.1714.TB.20200810.1312.006.html.
[11] LI X, DING Q, SUN J Q. Remaining useful life estimation in prognostics using deep convolution neural networks[J]. Reliability Engineering & System Safety, 2017(172): 1-11
[12] 肖剑, 王雨时, 张志彪. 基于Pareto遗传算法的无后坐炮内弹道多目标优化设计[J]. 弹道学报, 2019, 31(3): 53-57
[13] LI P, HUANG L W, PENG J C. Sensor distribution optimization for structural impact monitoring based on NSGA-II and wavelet decomposition[J]. Sensors, 2018, 18(12): 4264
[14] 张继红, 冀伟成. 基于改进遗传算法的光伏系统储能优化配置[J]. 中国测试, 2021, 47(1): 160-168