【摘要】 为了提高工业机器人的运行效率，采用数字孪生技术和卷积神经网络算法设计了工业机器人故障的预测方法。首先设计了工业机器人故障预测的基本架构，然后利用数字孪生技术构建了虚实映射的工业机器人的物理模型，并提取工业机器人的运动姿态特征，最后利用卷积神经网络算法构建了工业机器人电机故障的预测模型，通过对故障特征信号特征提取和分类，实现了故障的实时预测。仿真结果表明：通过孪生虚拟模型获取的工业机器人运行状态与实际运行状态高度重合，采用提出的故障预测方法对100组数据处理，得到的正确率、精确率、召回率和F1值4个性能指标分别为0.961 7、0.903 5、0.925 4和0.923 1，均明显高于其他两种对比方法，为工业机器人进行故障诊断和预测提供有力的技术支持。更多还原

【Abstract】 To improve the operation efficiency of industrial robots, a fault prediction method for industrial robots is designed using digital twin technology and convolutional neural network algorithm. Firstly, the basic structure of industrial robot fault prediction is designed. Then, the physical model of industrial robot with virtual reality mapping is constructed using digital twin technology, and the motion posture characteristics of industrial robot are extracted. Finally, the prediction model of industrial robot motor fault is constructed using convolutional neural network algorithm, and the real-time fault prediction is realized through the feature extraction and classification of fault feature signals.The simulation results show that the operation state of the industrial robot obtained by the twin virtual model is highly coincident with the actual operation state, and the proposed fault prediction method is used to process 100 groups of data, and the four performance indicators of accuracy, precision, recall and F1 value are 0.961 7, 0.903 5, 0.925 4 and 0.923 1 respectively, which are significantly higher than the other two comparison methods, providing strong technical support for the fault diagnosis and prediction of the industrial robot.更多还原