摘要
选择齿轮箱作为分析对象,经函数拟合后再通过主成分分析(PCA)提取得到特征参数,构建得到RBFELM模型。根据PCA具备高鲁棒性以及良好稳定性的优势来实现对极限学习机(ELM)进行鲁棒性优化的效果,同时利用该算法具备高计算效率的特性,设计了一种新方法。测试了PCA-RBF-ELM在故障检测方面的性能,相对传统算法实现了检测精度的大幅提升。研究结果表明:PCA与PCA-SVDD只能判断故障个数与准确性,并不能确定产生了哪类具体的故障;PCA/RBF-ELM能够检测故障种类,而缺乏足够的准确率。为判断本文算法可靠性,将其与其他算法进行了对比,得到PCA/RBF-ELM是各算法中具备较高检测率,且能满足快速运算的算法。该研究在齿轮箱早期故障排出方面具有很好的价值,对降低齿轮箱故障隐患具有很好的意义。
The gear box was selected as the analysis object,and the characteristic parameters were extracted by PCA after function fitting,and the RBF-ELM model was constructed.According to the advantages of HIGH robustness and good stability of PCA,a new method was designed to achieve the effect of robust optimization on ELM,and at the same time,the algorithm had high computational efficiency.The performance of PCA-RBF-ELM in fault detection is tested,and the accuracy of junction detection is greatly improved compared with the traditional algorithm.The results show that PCA and PCA-SVDD can only judge the number and accuracy of faults,but can not determine the specific type of faults.PCA/RBF-ELM and PCARBFELM can detect fault types,but lack sufficient accuracy.In order to judge the reliability of the proposed algorithm,PCA/RBF-ELM was compared with other algorithms,and PCA/RBF-ELM was the algorithm with higher detection rate and could meet the requirements of fast operation.This research has a good value in the early fault elimination of gearbox and has a good significance to reduce the hidden trouble of gearbox.
作者
贾林
李峰
高晓伟
JIA Lin;LI Feng;GAO Xiaowei(College of Information Engineering,Jiaozuo Teachers College,Jiaozuo 454150,Henan,China;School of Computer Science and Technology,Henan Polytechnic University,Jiaozuo 454000,Henan,China;Zhengzhou Coal Industry(Group)Gaorui Electric Power Co.,Ltd.,Zhengzhou 450000,Henan,China)
出处
《中国工程机械学报》
北大核心
2025年第5期908-912,共5页
Chinese Journal of Construction Machinery
基金
国家自然科学基金资助项目(50775157)。
关键词
齿轮箱
故障检测
函数型
主成分分析
极限学习机
gear box
fault detection
functional
principal component analysis
extreme learning machine
