往复压缩机十字头滑块故障的小波包敏感特征提取研究

由于往复压缩机振动信号具有非平稳和低信噪比特点,利用传统的时域或频域分析方法很难提取到反应压缩机的运行状况有效特征。压缩机发生故障时,信号能量沿频率的分布与正常状态有较大差异,本文利用小波包对非平稳信号的分解和时域重构能力,提出一种基于小波包分析的多频带平均能量特征提取方法;针对各特征对故障的敏感度不同,提出了一种基于欧式距离的特征选择方法,选择的特征能较好地反映压缩机的运行状态,最后通过往复压缩机的实验数据验证了该方法的可行性和有效性。

Sensitive Feature Extraction of Cross-head Slider Faults in Reciprocating Compressors Based on Wavelet Packets

It was difficult to extract the effective features to the operation condition of the compressor by using analysis of the Traditional time domain or frequency domain ,Because the Vibration signal of reciprocating compressor is non-stationary and low SNR,. When the reciprocating compressor faulted, the distribution of signal's energy in the frequency was very different comparing with the normal state; In this paper, we proposed a method to extract the feature of multi band average energy based on wavelet packet analysis; Due sensitivity of each feature was different in response the running situation of compressor, a feature selection method was proposed based on Euclidean distance. The selected feature effectively reflected the running state of compressor. Finally, the feasibility and effectiveness of the proposed method was verified by the experiment of reciprocating compressor.