基于数据均衡化的船舶涡轮增压系统故障诊断
Fault Diagnosis of Marine Turbocharging System Based on Data Equalization
李星贤 1肖文 2张斌 2龚梅杰 2陈辉1
作者信息
- 1. 武汉理工大学高性能船舶技术教育部重点实验室 武汉 430063;武汉理工大学船海与能源动力工程学院 武汉 430063
- 2. 江南造船集团有限责任公司 上海 201913
- 折叠
摘要
在船舶涡轮增压系统的故障诊断方面,针对正常状态与故障状态数据不平衡的问题,采用基于熵权重的Entropy-Weight SMOTE方法对数据样本进行增强,改善样本的不均衡性,并结合支持向量机(SVM)进行故障诊断.基于已有SMOTE算法与熵理论,以舰船动力系统仿真平台运行数据作为样本集,搭建基于熵理论的Entropy-Weight SMOTE与SVM的涡轮增压系统故障诊断模型;将舰船动力仿真平台数据样本导入模型进行仿真计算,综合各类评价指标,评判该方法的可行性.仿真实验表明在采用Entropy-Weight SMOTE进行样本均衡化后,分类准确度和综合指标(F-Measure)提升了 5.1％和6.5％.结果表明:该方法可以有效提高数据样本不平衡时涡轮增压系统的故障分类效果.
Abstract
In the aspect of fault diagnosis of ship turbocharging system,aiming at the imbalance be-tween normal state and fault state data,Entropy-Weight SMOTE method based on entropy weight was adopted to enhance the data samples and improve the imbalance of the samples.Combined with support vector machine(SVM),the fault was diagnosed.Based on the existing SMOTE algorithm and entropy theory,the fault diagnosis model of turbocharging system based on Entropy-Weight SMOTE and SVM was established with the operation data of ship power system simulation platform as the sample set.The data samples of ship power simulation platform were imported into the model for simulation calculation,and the feasibility of this method was judged by synthesizing various evalu-ation indexes.The simulation results show that the classification accuracy and comprehensive index(F-Measure)are improved by 5.1％and 6.5％after sample equalization with Entropy-Weight SMOTE.The results show that this method can effectively improve the fault classification effect of turbocharging system when the data samples are unbalanced.
关键词
支持向量机/涡轮增压系统/样本均衡化/故障诊断/熵权重Key words
support vector machines/turbocharger unit/sample equalization/fault diagnosis/entropy weight引用本文复制引用
基金项目
国家重点研发计划(2019YFE0104600)
工信部绿色智能内河船舶创新专项()
国家自然科学基金(51909200)
出版年
2024
NETL
NSTL
万方数据