基于双谱识别和人工免疫网络的智能故障检测

针对故障诊断中人为评估振动谱图而导致诊断结果不稳定的情况,提出基于振动谱图模式识别的故障诊断方法,利用Hilbert包络分析和双谱分析的组合方法来提取振动信号的故障频率特征,进而采用双谱图的灰度共生矩阵（GLCM）及其特征统计量来表征故障特征.改进了人工免疫网络（AIN）分类算法,将特征统计量作为抗原,通过对抗原的学习训练,形成记忆抗体集;通过判断待检验抗原与记忆抗体的匹配程度,实现故障分类识别.滚动轴承故障诊断实践证明,人工免疫网络分类方法具有良好的适应性,取得了较BP神经网络更好的检测准确率.

Intelligent fault diagnosis methods based on bispectrum recognition and artificial immune network

 A diagnosis method based on recognition of vibration spectra was developed aiming at the situation that manual observation of vibration spectra would lead to instability in the diagnosis. A combined method of Hilbert analysis and bispectrum analysis was proposed to extract the frequency characteristics from vibration signs. Then gray level co-occurrence matrix (GLCM) and its characteristic statistics generated from the bispectrum spectrum were selected to denote fault features. Furthermore, the artificial immune network (AIN) classification algorithm was improved by training the characteristic statistics as antigen and forming the memory antibody set. Fault classification was achieved through calculating the matching degree between test antigen and memory antibody set. Practice of the rolling bearing fault diagnosis shows that the AIN classification method has good adaptability, and achieved better detection accuracy than BP neural network．