一种基于频域信息的机械系统异常振动识别方法

摘要

In this paper,a novel detection method for abnormal mechanical system vibration in the fre⁃quency-domain is presented. Firstly,the eigenvector is specified as a frequency sequence in a specified range. Next,the Mahalanobis distance is employed to assess the degree of variance between the eigenvec⁃tor and the reference data set,and a corresponding threshold can be obtained using Monte Carlo method. Then,the fault detection is achieved by comparing the measured signal with the threshold value. Finally, the validity of this approach is demonstrated by testing against a marine motor.%　　提出了一种基于频域信息的机械系统异常振动识别方法。该方法将振动特征向量选定为特定频率范围内的频率序列，构建马氏距离表征待评估数据与参考特征向量集之间的差异程度，然后利用蒙特卡洛法设定极限阈值，将待测信号与设定限值进行比较，从而实现对故障状态的探测。通过一个船用电机的试验，验证了所提方法的正确性。