关键词: 电梯曳引电机, 故障诊断, 格拉姆角场差, 卷积神经网络, 麻雀搜索算法, 极限梯度提升树

Abstract: To address the challenges of fault feature extraction and insufficient diagnostic accuracy for elevator traction motors under variable operating conditions and high-noise environments,an intelligent fault diagnosis method is proposed. This method integrates the gramian angular difference field ( GADF) , convolutional neural network ( CNN) , and sparrow search algorithm ( SSA) to optimize the extreme gradient boosting ( XGBoost ) model. GADF was used to transform onedimensional vibration signals into two-dimensional images, preserving temporal dependencies and dynamic evolution characteristics. A lightweight CNN model was constructed to extract deep features. Subsequently, the sparrow search algorithm was employed to globally optimize the five core hyperparameters of XGBoost. Experimental results based on monitoring data from over 20 elevator traction motors demonstrate that the proposed method achieves a diagnostic accuracy of 92. 4%,validating its effectiveness and superiority in fault diagnosis for elevator traction motors.

Key words: elevator traction motor, fault diagnosis, gramian angle field difference, convolutional neural network;sparrow search algorithm;extreme gradient boosting tree