天然气压气机械设备故障信号时频分析特征提纯

天然气压气站压气机械设备工作环境恶劣，故障多发，而机械设备的故障状态下的振动噪声信号是研究机械设备故障诊断的有效载体。为有效进行故障诊断，需要对机械设备的故障信号进行提纯分析和特征提取等研究。提出基于时频分析技术的WVD和Hough变换结合的方法对低信噪比的故障信号进行提纯算法。首先分析了非平稳信号处理技术的基本原理，对原始信号的滤波和检测、频谱分析等处理，分析了时频分析特性和对应的条件，提出采用时间均值、频率均值、时间散布和频率散布四个特征值作为时频分析的特征提取量。仿真实验以某天然气压气站某型压缩机故障振动下采集信号样本为研究对象，进行提纯滤波和特征提取仿真，仿真实验得到降噪滤波和WVD及Hough变换算法下的检测结果，表明在强干扰背景SNR为-8dB下，有卓越的滤波降噪和检测性能，特征提纯检测性能相比传统算法提高18％以上。为天然气压缩机故障诊断奠定了可靠的理论基础。

Time Frequency Analysis and Feature Extraction of Vibration Signal for Natural Gas Machinery Equipment in Fault Status

The working environment of machineries in natural gas compressor station is complicated with high frequency of fault hap- pening. The vibration and radiated noised signal of the machinery in fault status is the good research object in fault diagnosis. In order to reach the effective fault diagnosis, the fault vibration signal needs to be filtered and extracted purely. A new signal extraction algorithm for the low SNR fault signal is proposed based on the Wigner Ville distribution（WVD） and Hough transformation theory with the time frequency analysis technology. The signal processing principle for the unstable signal is analyzed firstly, and the filter, detection and spectrum analysis are processed for the original signal. The property and the corresponding conditions are presented. The four features such as time mean val- ue, frequency mean value, and time scattering and frequency scattering value are proposed as the extracted feature. The vibration signal of faulting compressor are sampled and collected in real. The noise reduction and filtering result and the detection result are get based on the new method. Simulation result shows that the performance of filtering and detection is predominant when the SNR is as low as -8dB. The detection property is improved by 18% compared with the traditional algorithm. It provides the reliable base for the fault diagnosis of corn pressor in theory and practice.