基于AR模型和谱熵的自适应小波包络检测

针对传统故障诊断的包络问题，提出了一种基于自回归（auto regressive,AR）模型和谱熵的自适应复解析小波包络检测方法。通过AR模型从数据内在规律性上剔除机械振动信号中可线性预测的平稳成分，提取共振衰减的非平稳成分，在不同频带下进行复解析小波包络，结合谱熵在频域内与通带滤波的相关性选定最佳包络。仿真和试验数据分析结果表明，该方法能有效地提取故障特征频率，较传统方法自适应性更强，鲁棒性更高，包络效果更好，在工程应用中具有良好的前景。

Adaptive Wavelet Envelope Detection Based on AR Model and Spectral Entropy

For the envelope problems of traditional fault diagnosis, a method of adaptive complex analytic wavelet envelope detection was proposed based on AR model and spectral entropy herein. The method eliminated the stationary components for linear prediction from the mechanical vibration signals by AR model, and extracted the non-stationary components of resonance damping. The generated signals were enveloped by complex analytic wavelet in different frequency bands, the best envelope was selected based on the correlation between the spectral entropy and the band-pass filter in the frequency domain. This method owns higher adaptivity, better robustness and envelope effectiveness than that of the traditional one. Thus it has favorable prospect in engineering applications.