摘要
为提高滚动轴承故障诊断的准确性,提出基于信息增益比的奇异谱分析(IGRSSA)与改进粒子群算法优化支持向量机(IPSO-SVM)的诊断模型。首先,引入信息增益比实现信号自适应重构;其次,采用动态惯性权重和梯度信息对粒子群算法进行改进并用于优化支持向量机;然后,用IGRSSA对滚动轴承外圈故障、钢球故障和正常3种状态的振动信号进行降噪并提取时域特征值,使用平均影响值(MIV)筛选出最优特征参量作为后续故障信号特征数据集;最后,将BP神经网络、RBF神经网络、交叉验证优化的SVM、遗传算法优化的SVM和粒子群优化的SVM作为对比算法用于轴承故障诊断。30次有放回的随机抽样诊断结果表明,IPSO-SVM的平均诊断准确率达到97.72%,波动性和收敛误差均优于其他方法。
In order to improve accuracy of fault diagnosis for rolling bearings,a diagnosis model is proposed based on singular spectrum analysis(SSA)of information gain ratio(IGRSSA)and optimized support vector machine(SVM)by improved particle swarm optimization(IPSO-SVM).Firstly,the information gain ratio is introduced to realize adaptive signal reconstruction.Secondly,the dynamic inertia weight and gradient information are introduced to improve PSO(IPSO),which use to optimize SVM.Then,IGRSSA is used to denoise vibration signals of outer ring fault,steel ball fault and normal state of rolling bearings,and the time-domain feature values are extracted.MIV method is used to select optimal characteristic parameters as data set of subsequent fault signals.Finally,BP neural network,RBF neural network,SVM optimized by cross validation,SVM optimized by genetic algorithm and SVM optimized by PSO are used as comparison algorithms for fault diagnosis of the bearings.The results of 30 times random sampling diagnoses with playback show that the average diagnostic accuracy of IPSO-SVM reaches 97.72%,and the fluctuation and convergence errors are better than other methods.
作者
黄磊
马圣
HUANG Lei;MA Sheng(College of Aeronautical Engineering,Jiangsu Aviation Technical College,Zhenjiang 212134,China;Technology Research and Development Department,Chengdu CoreMii Technology Co.,Ltd.,Chengdu 610213,China)
出处
《轴承》
北大核心
2021年第10期60-66,共7页
Bearing
基金
2021年度院级课题资助项目(JATC21010104)。
关键词
滚动轴承
故障诊断
机器学习
支持向量机
谱分析
粒子群优化算法
主成分分析法
rolling bearing
fault diagnosis
machine learning
support vector machine
spectrum analysis
particle swarm optimization algorithm
principal component analysis
