A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Cross-device fault diagnosis method based on graph convolution and multi-sensor fusion
ObjectiveFor mechanical equipment in actual production, it is difficult or impossible to obtain a large amount of labeled data, resulting in low accuracy of traditional fault diagnosis methods. To address this problem, a cross-device fault diagnosis method based on graph convolution and multi-sensor fusion, convolutional domain graph convolution network (CDGCN) , was proposed. This method can model class labels, domain labels, and data feature structures.MethodsFirstly, a convolutional neural network was used to extract features from the input signal. Then, the feature structure relationship of the sample was mined through the graph generation layer to construct an instance graph. The instance graph was modeled using a graph convolutional neural network, and a multi-sensor high-level feature fusion method was proposed to perform multi-sensor information fusion. Finally, domain adaptation was achieved by using distribution difference metrics, classifiers, and domain discriminators.ResultsThe proposed method can capture domain-invariant features and discriminant features, and ultimately achieve cross-device fault diagnosis. Migration experiments on two datasets show that the proposed CDGCN not only achieves the best performance among the compared methods, but also extracts transferable features for cross-device domain adaptation.
Cross-device fault diagnosis method based on graph convolution and multi-sensor fusion
ObjectiveFor mechanical equipment in actual production, it is difficult or impossible to obtain a large amount of labeled data, resulting in low accuracy of traditional fault diagnosis methods. To address this problem, a cross-device fault diagnosis method based on graph convolution and multi-sensor fusion, convolutional domain graph convolution network (CDGCN) , was proposed. This method can model class labels, domain labels, and data feature structures.MethodsFirstly, a convolutional neural network was used to extract features from the input signal. Then, the feature structure relationship of the sample was mined through the graph generation layer to construct an instance graph. The instance graph was modeled using a graph convolutional neural network, and a multi-sensor high-level feature fusion method was proposed to perform multi-sensor information fusion. Finally, domain adaptation was achieved by using distribution difference metrics, classifiers, and domain discriminators.ResultsThe proposed method can capture domain-invariant features and discriminant features, and ultimately achieve cross-device fault diagnosis. Migration experiments on two datasets show that the proposed CDGCN not only achieves the best performance among the compared methods, but also extracts transferable features for cross-device domain adaptation.
Cross-device fault diagnosis method based on graph convolution and multi-sensor fusion
SUN Yuanshuai (author) / KONG Fanqin (author) / NIE Xiaoyin (author) / XIE Gang (author)
2024
Article (Journal)
Electronic Resource
Unknown
Metadata by DOAJ is licensed under CC BY-SA 1.0
ROTATING MACHINERY FAULT DIAGNOSIS BASED ON TWO-DIMENSIONAL CONVOLUTION NEURAL NETWORK
| DOAJ | 2020
| DOAJ | 2022