Fault diagnosis driven by the single signal has been widely used in motor fault diagnosis, but it can not meet the diagnostic requirements of complex motor system. In this study, a motor fault diagnosis method using attention mechanism and improved AdaBoost driven by multi-sensor information is proposed. Firstly, the corresponding frequency domain feature information is obtained by Hilbert transform and Fourier transform in different signals. The improved AdaBoost multi-classification classifier is then used to train signals from different sources and obtain sub classifier results. Finally, a dynamic weight distribution matrix is used to obtain the final diagnosis results with sub classifiers. The proposed method is verified by current, magnetic and vibration signals. The results show that the proposed method dynamically evaluates sensitivity of different detection signals to different faults. Compared with the conventional method, the proposed method can enhance the robustness, generalization ability and accuracy of fault diagnosis.

单信号驱动的故障诊断已被广泛应用于电机故障诊断中，但不能满足复杂电机系统的诊断要求。提出了一种基于注意力机制和多传感器信息驱动的改进型AdaBoost的电机故障诊断方法。首先，通过希尔伯特变换和傅立叶变换在不同信号中获得对应的频域特征信息。改进的AdaBoost多分类分类器然后用于训练来自不同来源的信号并获得子分类器结果。最后，使用动态权重分布矩阵通过子分类器获得最终诊断结果。通过电流，磁和振动信号验证了该方法的有效性。结果表明，该方法可以动态评估不同检测信号对不同故障的敏感性。与常规方法相比，该方法可以提高故障诊断的鲁棒性，泛化能力和准确性。