基于排列熵和CHMM的齿轮故障诊断

针对齿轮故障特征提取和状态识别困难的问题,提出一种基于排列熵和连续隐马尔可夫模型(CHMM)的齿轮故障诊断方法。首先对提取的目标齿轮啮合信号作降噪处理,再采用排列熵算法进行分析,提取排列熵均值、方均根、最大值、最小值作为特征量输入到CHMM中训练和识别,通过对比最大对数似然概率值来确定齿轮的故障。最后在变速箱齿轮故障模拟试验台上,对正常、轻微磨损、严重磨损和断齿4种齿轮状态进行试验验证,结果表明该方法能有效地对齿轮故障进行诊断。

Fault Diagnosis of Gear Based on Permutation Entropy and CHMM

Considering the difficulties of fault feature extraction and state recognition, a fault diagnosis method for gear based on permutation entropy and Continuous Hidden Markov Model (CHMM) has been proposed. After the vibration signals of gear were denoised, the noise-reducted signals were analyzed by permutation entropy method to extract the mean, root-mean-square, maximum and minimum values. Then CHMM was adopted to carry out gear diagnosis by training algorithm and comparing the maximum log-likelihood values. Finally the four gear states of normal, mild wear, heavy wear and broken were tested by gearbox simulation experiment. The results show that this method can diagnose the fault of gears effectively.