摘要
针对传统故障监测与诊断算法在船舶柴油发动机燃油系统应用中精度较低的问题,提出一种基于粒子群优化(Particle Swarm Optimization,PSO)算法优化核主成分分析(Kernel Principal Component Analysis,KPCA)和支持向量机(Support Vector Machine,SVM)的故障监测和诊断新方法。首先采用KPCA提取样本数据中的非线性特征,获取其高维信息,同时在特征空间中构建T 2和SPE统计量,实时监测故障的发生;对于监测到的故障样本,通过KPCA提取其非线性主成分,作为多分类SVM的输入样本进行故障模式识别,采用PSO算法分别对KPCA与多分类SVM的核函数参数、多分类SVM的惩罚因子进行优化,以提高故障监测和诊断模型的精度。船舶燃油系统故障监测和诊断试验结果表明,经过PSO优化后的KPCA-SVM故障监测和诊断模型的精度明显提高,验证了所提方法的优势和有效性。
Aiming at the problem of the low accuracy of traditional fault monitoring and diagnosis algorithms in marine fuel systems,a new fault monitoring and diagnosis method based on Kernel Principal Component Analysis(KPCA)and Support Vector Machine(SVM)optimized by Particle Swarm Optimization(PSO)algorithm was proposed.Firstly,using KPCA to extract the non-linear features in the sample data to obtain its high-dimensional information.At the same time,building T 2 and SPE statistics in the feature space to monitor the occurrence of faults in real time.For the monitored fault samples,extract its non-linear principal component by KPCA,used as the input sample of the multi-class SVM for fault mode recognition.The PSO algorithm was used to optimize the kernel function parameters of the KPCA and multi-class SVM and penalty factors of the multi-class SVM respectively in order to improve the precision of fault monitoring and diagnosis model.The results of the marine fuel system fault monitoring and diagnosis experiment show that the accuracy of KPCA-SVM fault monitoring and diagnosis model optimized by PSO is significantly improved,and the advantages and effectiveness of the proposed method are verified.
作者
张志政
王冬捷
张勇亮
Zhang Zhizheng;Wang Dongjie;Zhang Yongliang(Marine Engineering College,Dalian Maritime University,Dalian 116026,Liaoning,China)
出处
《现代制造工程》
CSCD
北大核心
2020年第9期101-107,共7页
Modern Manufacturing Engineering
基金
工信部项目(工信部装函[2018]473号)。
关键词
核主成分分析
粒子群优化算法
支持向量机
模式识别
故障监测和诊断
kernel principal component analysis
particle swarm optimization algorithm
support vector machine
pattern recognition
fault monitoring and diagnosis
