基于LMD和GNN-Adaboost的滚动轴承故障严重程度识别

提出一种基于局部均值分解（Local Mean Decomposition，LMD）和遗传神经网络自适应增强（Genetic Neural Network Adaptive Boosting，GNN-Adaboost）的滚动轴承损伤程度识别方法。通过LMD方法将轴承振动信号分解为若干个瞬时频率有物理意义的乘积函数（Production Function，PF），对能反映信号主要特征的PF提取能量矩，结合原始振动信号的时域特征参数（方差、偏度、峭度），组成故障严重程度识别特征参数矩阵。将基于LMD方法的特征参数矩阵作为GNN-Adaboost方法的输入向量，对不同载荷与转速工况下的轴承进行故障严重程度识别。结果表明，基于LMD和GNN-Adaboost的方法能够有效提高轴承故障严重程度识别准确率，对滚动轴承等关键旋转部件的故障识别与定位具有重要意义。

Rolling Bearing Fault Severity Recognition Based on LMD and GNN-Adaboost

A method for recogniting the damage degree of rolling bearings with local mean decomposition (LMD) and genetic neural network adaptive boosting (GNN-Adaboost) is proposed.The bearing vibration signal is decomposed by the LMD into several production functions (PF) with meaningful instantaneous frequency.The energy moment of PF which can reflect the main characteristics of the signal was extracted,and then the energy moment and the time domain characteristic parameters (variance,skewness,kurtosis) of the original vibration signal were grouped into a characteristic parameter matrix.Then the characteristic parameter matrix based on LMD was used as the input vector of GNN-Adaboost to identify the fault severity of bearings under different load and speed conditions.Compared with others,the method based on LMD and GNN-Adaboost shows that it can effectively improve the accuracy of bearing fault severity recognition,which is of great significance for fault identification and location of key rotating parts such as rolling bearings.