Wang Qibing, Chen Luqiang, Xiao Gang, Wang Peng, Gu Yuejiang, Lu Jiawei
College of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou, 310018, China.
Shanghai STEP Electric Corporation, Shanghai, 201801, China.
Sci Rep. 2024 Dec 28;14(1):30713. doi: 10.1038/s41598-024-78784-7.
The rapid development of urbanization has led to a continuous rise in number of elevators. This has led to elevator failures from time to time. At present, although there are some studies on elevator fault diagnosis, they are more or less limited by the lack of data to make the research more superficial. For such complex special equipment as elevator, it is difficult to obtain reliable and sufficient data to train the fault diagnosis model. To address this issue, this paper first establishes a numerical model of vertical vibration for elevators with three degrees of freedom. The obtained motion equations are then used as constraints to acquire simulated vibration data through PINNs. Next, the proposed e-RGCN is employed for elevator fault diagnosis. Finally, experimental validation shows that the fault diagnosis accuracy with the participation of digital twins exceeds 90%, and the accuracy of the proposed model reaches 96.61%, significantly higher than that of other comparative models.
城市化的快速发展导致电梯数量持续增加。这使得电梯故障时有发生。目前，虽然有一些关于电梯故障诊断的研究，但或多或少受到数据缺乏的限制，使得研究较为表面。对于电梯这种复杂的特种设备，很难获得可靠且充分的数据来训练故障诊断模型。为解决这个问题，本文首先建立了具有三个自由度的电梯垂直振动数值模型。然后将得到的运动方程作为约束条件，通过物理信息神经网络（PINNs）获取模拟振动数据。接下来，将所提出的电梯关系图卷积网络（e-RGCN）用于电梯故障诊断。最后，实验验证表明，数字孪生参与下的故障诊断准确率超过90%，所提模型的准确率达到96.61%，显著高于其他对比模型。
