摘要
针对齿轮故障振动信号具有多重分形特征,提出多个无标度区的多重分形理论与神经网络相结合的机械故障诊断方法。该方法采用多重分形理论计算齿轮振动信号的多分形谱和广义分形维数,并将多分形谱能和广义分形维数谱能作为特征量,构成二维特征向量。将该特征向量作为概率神经网络的输入参量,并对采自齿轮故障台的振动信号进行故障分类。实验证明,与单一无标度区多分形谱理论特征提取方法相比较,所提出的方法能更精密刻画振动信号特征,并获得更高的识别率。
The mechanical fault diagnosis method of the multiple scale-free area multifractal theory combined with neural was proposed according to the multifractal characteristics of gear fault vibration signals. Through multifractal theory it calculated the multifractal spectrum and generalized fractal dimensions of the vibration time series whose energy can be used as characteristics,a two-dimensional characteristics matrix,were as the input scale-free area multifractal spectrum theory,and the proposed method was precise to describe characteristics of vibration signal and obtain higher recognition rate.
出处
《机械设计与制造》
北大核心
2016年第1期5-7,11,共4页
Machinery Design & Manufacture
基金
国家自然科学基金项目资助(51105284
51475339)
关键词
多个无标度区
多重分形谱
齿轮故障状态
整体特性
Multiple Scale-Free Area
Multifractal Spectrum
Gear Fault State
Overall Feature
