基于DBN的不均衡样本驱动民航发动机故障诊断  

Fault diagnosis of civil aero-engine driven by unbalanced samples based on DBN

作　　者：钟诗胜[1,2] 李旭 张永健 ZHONG Shisheng;LI Xu;ZHANG Yongjian(School of Mechatronics Engineering,Harbin Institute of Technology,Harbin 150001,China;School of Naval Architecture and Ocean Engineering,Harbin Institute of Technology(Weihai),Weihai Shandong 264200,China)

机构地区：[1]哈尔滨工业大学机电工程学院,哈尔滨150001 [2]哈尔滨工业大学(威海)船舶与海洋工程学院,山东威海264200

出　　处：《航空动力学报》2019年第3期708-716,共9页Journal of Aerospace Power

基　　金：国家自然基金重点项目(U1533202);民航科技项目(MHRD20150104)

摘　　要：在结合深度置信网络(DBN)、采样与集成技术的基础上,提出了基于不均衡样本驱动的民航发动机故障诊断模型。该模型通过分析民航发动机历史飞行数据,利用DBN提取性能参数中的内部特征,利用采样技术将不均衡样本均衡化,采用集成技术进行故障分类。将该模型应用到CFM56-7B系列发动机历史飞行数据,实验结果表明:与常用故障诊断方法相比,该模型的准确率高达0.996,AUC值高达0.948,可以有效处理民航发动机样本高维、不均衡问题。Through combination of deep belief network(DBN),sampling and integration technology,a fault diagnosis model of civil aero-engine based on unbalanced sample driving was proposed.By analyzing the historical flight data of civil aero-engines,the model used DBN to extract the internal features of the performance parameters,then used the sampling technology to equalize the unbalanced samples,and finally adopted integrated technology for fault classification.The model was applied to historical flight data of CFM56-7B series engines.Compared with common fault diagnosis methods,the experimental results showed that the model had higher accuracy of 0.996 and AUC value of 0.948,and can effectively deal with high-dimensional and unbalanced problems of civil aero-engine samples.

关 键 词：民航发动机 故障诊断 不均衡样本 深度置信网络 Adaboost.M1算法 

分 类 号：V263.6[航空宇航科学与技术—航空宇航制造工程]