摘要
针对齿轮箱的故障表征不明显且传统分类方法精度低等问题,提出一种基于核主成分分析(KPCA)特征提取和蚁群算法优化极限学习机神经网络(ACA-ELM)分类识别相结合的齿轮箱故障诊断方法。首先,从齿轮箱的原始信号中提取时域与频域特征构成特征矩阵,利用KPCA方法降低维度,剔除冗余信息,提取有效的特征指标;其次,利用蚁群算法(Ant Colony Algorithm,ACA)对极限学习机(Extreme Learning Machine,ELM)的网络初始权值与偏置进行优化,得到最优权值与偏置组合;最后,利用ACA-ELM进行齿轮箱故障诊断实验,同时与ELM、BP、ACA-BP、GA-ELM模型对比。实验结果表明,该方法进行故障诊断的准确率可以达到98.3%,能够有效地进行齿轮箱故障诊断。
Aiming at the problem that the fault representation of gearbox is not obvious and the accuracy of traditional classification methods is not high,a gearbox fault diagnosis method based on kernel principal component analysis(KPCA)feature extraction and ant colony algorithm optimization extreme learning machine neural network(ACA-ELM)classification and recognition is proposed.Firstly,time domain and frequency domain features are extracted from the original signal of gearbox to form feature matrix.The KPCA method is used to reduce dimension,eliminate redundant information and extract effective feature indicators.Secondly,the ant colony algorithm(ACA)is used to optimize the initial network weights and biases of ELM,and the optimal weight and bias combination is obtained.Finally,ACA-ELM is used for gearbox fault diagnosis experiments,and compared with ELM,BP,ACA-BP,GA-ELM models.The experimental results show that the accuracy of fault diagnosis can reach 98.3%and this method can effectively diagnose gearbox fault.
作者
李梦瑶
周强
于忠清
LI Meng-yao;ZHOU Qiang;YU Zhong-qing(School of Data Science and Software Engineering,Qingdao University,Qingdao Shandong 266071,China)
出处
《组合机床与自动化加工技术》
北大核心
2021年第4期87-90,95,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
山东省重点研发计划(重大创新工程)项目(2019JZZY020101)
。
