基于IPSO-SVM的动态汽车衡故障诊断方法研究  

IPSO-SVM based fault diagnosis method for dynamic truck scale

作　　者：黄庆程 HUANG Qingcheng(Fujian Provincial Transportation Research Institute Co.,Ltd.,Fuzhou 350004,China)

机构地区：[1]福建省交通科研院有限公司,福建福州350004

出　　处：《机电工程》2024年第12期2310-2319,共10页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金资助项目(52205558);福建省交通运输科技项目(202217)。

摘　　要：针对服役状态下,不易对轴重式动态汽车衡的灵敏度漂移等故障进行在线检测这一问题,提出了一种特征降维下结合莱维飞行改进粒子群算法优化支持向量机(IPSO-SVM)模型,以及信号特征提取与降维的动态汽车衡故障诊断方法。首先,提取了输出信号的时域与频域特征,利用核主成分分析(KPCA),将非线性映射函数输入空间变换到高维空间,实现对特征向量的降维与筛选目的;然后,利用了莱维飞行改进粒子群优化算法(PSO)的寻优能力,并采用改进后的算法对支持向量机(SVM)进行了优化,得到了最优的参数组合,以此构建了全局最优的IPSO-SVM诊断模型;最后,采用建立的诊断模型,对不同车重、不同车速、不同轴型载荷工况下的动态汽车衡进行了故障诊断验证。研究结果表明:采用该动态汽车衡故障诊断方法,其诊断准确率可达98%,证实了引入莱维飞行后的改进粒子群算法可显著改进优化的效率和效果。相比现有诊断方法,IPSO-SVM诊断模型可有效解决PSO算法易陷入局部最优解的问题,准确率得到了较大提升,可实现对汽车衡系统动态故障工况下的全类型高精度诊断。Aiming at the issue of difficulty in online detection of sensitivity drift and other faults in axle-load dynamic truck scales in service state,a fault diagnosis method of dynamic truck scale based on Levy flight improved particle swarm optimization support vector machine(IPSO-SVM)model combined with signal feature extraction and dimension reduction under feature dimension reduction was proposed.Firstly,the time-domain and frequency-domain features of the output signal were extracted,and kernel principal component analysis(KPCA)was used to transform the input space into the high-dimensional space through the nonlinear mapping function to realize the dimensionality reduction and screening of the feature vectors.Then,the optimization ability of Levy s improved particle swarm optimization algorithm(PSO)was adopted,and the improved algorithm was used to optimize the support vector machine(SVM)to obtain the optimal parameter combination,so as to construct the global optimal IPSO-SVM diagnosis model.Finally,the established diagnostic model was used to diagnose and verify the fault diagnosis of the dynamic truck scale under different vehicle weights,different speeds,and different axle load conditions.The experimental results show that the diagnostic accuracy of the prototype can reach 98%using the dynamic truck scale fault diagnosis method,it confirms that the improved particle swarm optimization algorithm after the introduction of Levy flight significantly improves the optimization efficiency and effect.Comparing with the existing diagnostic methods,the IPSO-SVM diagnostic model can effectively solve the problem of PSO algorithm easily getting stuck in local optimal solutions,thus greatly improving its accuracy,and can realize the high-precision diagnosis of all types under the dynamic fault conditions of the truck scale system.

关 键 词：质量计量仪器 故障诊断模型 莱维飞行 信号特征提取 信号特征降维 支持向量机 改进粒子群算法优化支持向量机 核主成分分析 

分 类 号：TH715.1[机械工程—测试计量技术及仪器] U492.321[机械工程—仪器科学与技术]