基于小波变换的隐Markov模型离心泵故障诊断方法研究

根据离心泵故障振动信号的特点,本文提出了一种结合小波变换与隐Markov模型(HMM)的离心泵故障诊断方法。小波变换具有多分辨率分析并且在时频两域都具有表征信号局部特征能力的特点,利用Daubechies小波对振动信号进行一维8尺度的小波分解,然后从中提取一维信号的低频系数作为特征向量,将其输入到各个状态HMM进行训练,其中输出概率最大的状态即是离心泵的运行状态,从而实现离心泵的故障诊断。最后通过2BA-6A离心泵试验系统验证了该方法的有效性。

Research on Fault Diagnosis Methods for Centrifugal Pump Based on Wavelet Transform and Hidden Markov Model

According to the characteristic of fault vibration signal, a new method for centrifugal pump based on wavelet transform and hidden Markov model is produced in this paper. The wavelet transform has characteristics of multi-resolution analysis, which can be used in time-frequency localization and to analyze signals. 1-D,8-scale Daubechies wavelet decomposition is utilized in vibration signal analysis,then the low-frequency coefficients of 1-D signal which extracted from that can be utilized as feature vectors of running state of a centrifugal pump to train HMM, fault classification can be made according to the maximum-likelihood probability. The model is tested with the experimental data collected from the 2BA-6A centrifugal pump experimental system and the result demonstrated that the model was effective to classify classical faults.