基于MRSSD与MSDE的滑动轴承故障诊断方法

Fault Diagnosis Method for Sliding Bearings Based on MRSSD and MSDE

针对往复压缩机轴承振动信号强烈的非平稳、非线性的特点,提出了一种基于多重共振稀疏分解(MRSSD)与多尺度符号动力学熵(MSDE)相结合的往复压缩机滑动轴承故障诊断方法。首先,设置高、低品质因子的取值范围,求出能表示故障冲击成分的低品质因子值,对信号进行共振稀疏分解,形成高、低共振分量;然后,根据高共振分量的峭度值评定分解结果,峭度值小于设定阈值时改变高品质因子值,继续对低共振分量进行共振稀疏分解,峭度值大于设定阈值时终止分解;最后,计算最终所得低共振分量的多尺度符号动力学熵,构造故障特征向量,并利用支持向量机进行故障特征识别。试验结果表明,该方法可以逐步降低干扰成分的影响,有效诊断往复压缩机滑动轴承故障,与基于遗传算法优化品质因子的共振稀疏分解和多尺度排列熵(MPE)相结合的方法相比,故障识别率显著提高。

Aimed at strong non-stationary and non-linear vibration signal of reciprocating compressor bearings,a fault diagnosis method for the bearings is proposed based on multiple resonance sparse signal decomposition(MRSSD)and multi-scale symbolic dynamic entropy(MSDE).Firstly,the value range of high and low quality factors is set to find out low quality factor values that can represent fault impact components,and the resonance sparse decomposition of signal is carried out to obtain high and low resonance components.Then,the decomposition result is judged according to kurtosis value of high resonance component.If the kurtosis value is less than given threshold value,the high quality factor values will be changed,and the resonance sparse decomposition of low resonance components is continued.If the kurtosis value is larger than given threshold value,the decomposition is terminated.Finally,the multi-scale symbolic dynamic entropy of final low resonance components is calculated,the fault feature vector is constructed,and the fault feature is identified by support vector machine.The experimental results show that the method gradually reduces influence of interference components and effectively diagnoses fault of sliding bearings for reciprocating compressor,and significantly improves fault identification rate compared with optimized quality factor resonance sparse decomposition method and multi-scale entropy(MPE)feature extraction method based on genetic algorithm.