Motor fault diagnosis method based on non-convex non-smooth optimization and graph model

The invention discloses a motor fault diagnosis method based on non-convex non-smooth optimization and a graph model, and the method comprises the steps: extracting the incidence relation of differenthigh-dimensional signals through the non-convex non-smooth optimization, obtaining the implicit state of the high-dimensional signals and a corresponding inverse covariance matrix by solving an optimization problem, building the graph model corresponding to a motor signal, and finally, comparing the real-time inverse covariance matrix and the historical inverse covariance online to judge whetherthe motor is in a fault, excavating the invisible fault of the motor in the high-dimensional time-varying signal through the history under the condition of no calibration, identifying the motor faultin advance, and therefore, the reliability of motor online fault diagnosis is improved to the maximum extent. 本发明公开了一种基于非凸非光滑优化和图模型的电机故障诊断方法，通过非凸非光滑优化提取不同高维信号的关联关系，通过求解优化问题得到高维信号的隐状态以及对应的逆协方差矩阵，建立电机信号对应的图模型，最终将实时的逆协方差矩阵和历史的逆协方差