The invention discloses an elevator fault diagnosis method based on a graph neural network. Aiming at elevator faults, firstly, common fault types of an elevator are subjected to mechanism analysis, and main elevator characteristic quantities changing correspondingly with the faults are obtained; the method comprises the following steps: collecting change conditions of main characteristic quantities through various sensors, and carrying out EMD and wavelet noise reduction processing on characteristic quantity data to obtain a sample set; constructing a sample graph through cosine similarity and a K nearest neighbor algorithm; constructing a graph convolutional neural network model, and inputting the adjacent matrix and the node feature matrix of the sample graph and training data into the model for training; and in the trained graph neural network model, data collected in real time are input to achieve fault diagnosis of the elevator. According to the method, on the basis of mechanism analysis on the failure mode of the elevator, valuable information in multi-source heterogeneous data such as sound, electricity and vibration of the elevator is extracted, so that the model is more interpretable, and a better diagnosis effect can be obtained in fault diagnosis.
本发明公开了一种基于图神经网络的电梯故障诊断方法。本发明针对电梯故障,首先对电梯的常见故障类型进行了机理分析,得到了与故障相应变化的电梯主要特征量;通过各类传感器采集主要特征量的变化情况,将特征量数据进行EMD和小波降噪处理后得到样本集;通过余弦相似度和K近邻算法构造样本图;构造图卷积神经网络模型,将样本图的邻接矩阵和节点特征矩阵以及训练数据输入模型进行训练;在训练好的图神经网络模型中,输入实时采集到的数据实现电梯的故障诊断。本发明基于对电梯失效模式的机理分析,并提取了电梯声、电、振动等多源异构数据中的有价值信息,不仅使得该模型更具有可解释性,也可以在故障诊断中获得更好的诊断效果。
Elevator fault diagnosis method based on graph neural network
一种基于图神经网络的电梯故障诊断方法
2023-02-24
Patent
Elektronische Ressource
Chinesisch
Elevator fault diagnosis device and elevator fault diagnosis method
Europäisches Patentamt | 2022
|Elevator fault diagnosis system using recurrent neural network model
Europäisches Patentamt | 2023
|Europäisches Patentamt | 2023
|Europäisches Patentamt | 2021
|