摘要
针对目前数据驱动的故障诊断方法在船舶柴油机应用中存在故障识别率不高的问题,提出一种基于麻雀搜索算法(Sparrow Search Algorithm,SSA)优化栈式自编码器(Stacked Auto-Encoder,SAE)的诊断方法,实现高精度故障诊断。利用SAE的重构误差作为状态监测量,实时监测柴油机故障的发生。将监测到的异常样本输入SAE进行数据分类,实现对故障类型的精确识别。针对SAE在故障类型识别中超参数设置过多、依赖人工经验的问题,采用SSA对SAE多个超参数进行联合寻优,提高故障识别率和稳定性。基于AVL BOOST船舶柴油机仿真数据的试验表明:所提出SSA-SAE诊断方法的故障识别率为96.71%,比SAE、支持向量机(Support Vector Machine,SVM)和极限学习机(Extreme Learning Machine,ELM)具有更高的故障识别率和更优的泛化能力。
A diesel engine fault diagnosis method based on SAE(Stacked Auto-Encoder)with SSA(Sparrow Search Algorithm)is developed to improve the fault identification rate.The engine fault is detected in real time by monitoring reconstruction error of the auto-encoder.The abnormal engine condition samples are sent to the auto-encoder and categorized and identified.Setting of hyper parameters in the auto-encoder are usually done according to experience,therefore,more or less subjective.In order to overcome this problem,a sparrow search algorithm is introduced for optimizing the hyper parameters.Tests are performed with the simulation data produced by AVL BOOST.The achieved fault identification rate is as high as 96.71%,much higher than that with SAE alone or support vector machine.
作者
刘鑫龙
曾鸿
董建伟
和泰山
刘利源
LIU Xinlong;ZENG Hong;DONG Jianwei;HETaishan;LIU Liyuan(College of Marine Engineering,Dalian Maritime University,Dalian 116026,China;China Shipbuilding Industry Corporation Diesel Engine Co.,LTD.,Qingdao 266520,China)
出处
《中国航海》
CSCD
北大核心
2022年第4期45-51,57,共8页
Navigation of China
基金
工业和信息化部高技术船舶科研项目(CJ02N20)。
作者简介
刘鑫龙(1996-),男,山东枣庄人,硕士研究生,研究方向为轮机自动化与智能化。E-mail:liuxinlong1996@163.com;通信作者:曾鸿(1981-),男,福建福鼎人,副教授,博士,研究方向为轮机自动化与智能化。E-mail:zenghong@dlmu.edu.cn。
