摘要
针对含有尖脉冲的齿轮箱振动信号故障特征难以提取且样本较少的问题,提出了一种基于α稳定分布和支持向量机故障诊断的新方法。先设计齿轮箱故障测试方案,获取齿轮箱振动信号;然后提取齿轮箱振动信号的α稳定分布参数,用它作为故障类型的特征样本,并结合决策树和投票法构造多分类支持向量机齿轮箱故障决策系统。该方法较好地解决了小样本学习问题,避免了人工神经网络进行诊断时的过学习、收敛速度慢等缺点。实际齿轮箱故障诊断实验结果表明所提方法有效。
A novel method was proposed based on a-stable distribution parameters and support vector machine about the gearbox fault diagnosis.Firstly,the experiment project on fault diagnosis was designed.Then,the vibration signals of the gearbox were tested,and a-stable distribution parameters of the signals were extracted, which contained the running information.Combining with the basic thought of voting method and decision tree, a special decision-structure of MSVM was designed.Moreover,the decision-structure solved the small sample learning problems well and overcame the shortcoming of over-fitting,longtime training of ANN in fault diagnosis. A practical experimental result demonstrates that the presented intelligent diagnosis method is effective.
出处
《测控技术》
CSCD
北大核心
2012年第8期23-26,30,共5页
Measurement & Control Technology
基金
江西省教育厅科技资助项目(CJJ11244
GJJ11245)
关键词
α稳定分布参数
支持向量机
齿轮箱
故障诊断
α-stable distribution parameters
support vector machine（SVM）
gearbox
fault diagnosis
作者简介
余香梅（1976-），女，江西赣州人，副教授，主要研究方向为机械制造、智能测试技术
