摘要
针对自动机运作时的瞬态冲击、非线性、非平稳信号特征,提出一种基于排列熵和支持向量机对小口径高速自动机进行故障诊断的方法。首先,引入排列熵对信号进行分析,发现排列熵能很好地反映自动机工作状态;其次,将排列熵特征量分别作为概率神经网络PNN和SVM的输入参数以识别自动机故障,结果表明:SVM相比于PNN可以提高分类正确率。同时证明基于排列熵和SVM在自动机故障诊断中的有效性。
Aiming at that automata vibration signals are typical transient impact ingredients, non-linear and non- stationary time series, a fault diagnosis method based on permutation entropy and SVM is put forward. Firstly, the permutation entropy theory is applied to analyze the preprocessed signals, the results show that the permutation entropy (PE) can give a good presentation for automata working condition. Secondly, the permutation entropy characteristic parameters are input to SVM and PNN for fault classification. The results show that SVM relative to PNN can improve the fault diagnosis accuracy. Meanwhile, it proves that the validity of the method of the permutation entropy and SVM for automata fault diagnosis.
出处
《机械设计与研究》
CSCD
北大核心
2015年第5期138-140,共3页
Machine Design And Research
基金
国家自然科学基金资助项目(51175480)