摘要
针对旋转机械进行故障诊断时,由于邻近机械的干扰,往往无法得到真实的的故障信息以及诊断速度慢的问题,本文提出了一种基于独立分量分析(Independent Component A-nalysis,ICA)和概率神经网络(Probabilistic Neural Network,PNN))的故障诊断方法,采用快速独立分量分析(FastICA)进行特征提取,PNN实现状态识别.通过仿真与实验加以证明,并与经典的前向多层神经网络(BP网络)的故障分类进行对比,结果表明PNN的准确率可以达到100%,而BP网络只有95%,同时PNN所需的时间只有BP的1/3.
In fault diagnosis of machine, the observation signals always include signals from other machines,therefore it is essential to separate the individual signals from mixed signals. In view of the above issues, this paper proposes a new fault diagnosis method based on Independent Component Analysis（ICA） and probabilistic neural network（PNN） . In this method the FastlCA algorithm is used to feature extraction and the PNN is used to status recognition. Compared with the multilayer feedforward neural network method （BP network）, the result showed that the new method is more efficient and more accurate.
出处
《西安工业大学学报》
CAS
2009年第5期490-494,共5页
Journal of Xi’an Technological University
