Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill
Graph data and multi-sensor information fusion have been integrated into the abnormal vibration type classification and the identification of the rolling mill for extracting spatial–temporal and robust features. However, most of the existing deep learning (DL) based methods exploit only single senso...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Machines |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-1702/13/1/30 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832588071912079360 |
|---|---|
| author | Rongrong Peng Changfen Gong Shuai Zhao |
| author_facet | Rongrong Peng Changfen Gong Shuai Zhao |
| author_sort | Rongrong Peng |
| collection | DOAJ |
| description | Graph data and multi-sensor information fusion have been integrated into the abnormal vibration type classification and the identification of the rolling mill for extracting spatial–temporal and robust features. However, most of the existing deep learning (DL) based methods exploit only single sensor information and Euclidean space data, which results in incomplete information contained in the features extracted by in-depth networks. To solve this issue, a multi-sensor information fusion with multi-scale adaptive graph convolutional networks (M<sup>2</sup>AGCNs) framework is proposed to model graph data and multi-sensor information fusion in a unified in-depth network and then to achieve abnormal vibration diagnosis. First, convolutional neural networks (CNNs) were adopted for the deeper features of multi-sensor signals. And then, the extracted features were fed into the proposed feature-driven adaptive graph generation network to build graphs to extract spatial–temporal correlation between multi-sensor data. After that, the multi-scale graph convolutional networks (MSGCNs) were employed to aggregate and enrich several different receptive information to further improve valuable features. Finally, the extracted multi-sensor features were integrated into a unified network to achieve the abnormal vibration type classification and identification of the rolling mill. Meanwhile, we performed horizontal, vertical, and coupled abnormal vibration experiments, and then three different types of studies were conducted to illustrate the superiority and usefulness of this method in the paper and the feasibility of rolling mill abnormal vibration diagnosis. It can be seen from the results that the proposed M<sup>2</sup>AGCNs can be able to achieve valuable feature extraction effectively from multi-sensor information and to obtain more excellent behavior of the abnormal vibration diagnosis of the rolling mill in comparison with the mainstream methods. |
| format | Article |
| id | doaj-art-893768c59caa412fa72080a79d0207f3 |
| institution | Kabale University |
| issn | 2075-1702 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Machines |
| spelling | doaj-art-893768c59caa412fa72080a79d0207f32025-01-24T13:39:12ZengMDPI AGMachines2075-17022025-01-011313010.3390/machines13010030Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling MillRongrong Peng0Changfen Gong1Shuai Zhao2School of Mechanical and Vehicle Engineering, Nanchang Institute of Science and Technology, Nanchang 330108, ChinaSchool of Education, Nanchang Institute of Science and Technology, Nanchang 330108, ChinaSchool of Information and Artificial Intelligence, Nanchang Institute of Science and Technology, Nanchang 330108, ChinaGraph data and multi-sensor information fusion have been integrated into the abnormal vibration type classification and the identification of the rolling mill for extracting spatial–temporal and robust features. However, most of the existing deep learning (DL) based methods exploit only single sensor information and Euclidean space data, which results in incomplete information contained in the features extracted by in-depth networks. To solve this issue, a multi-sensor information fusion with multi-scale adaptive graph convolutional networks (M<sup>2</sup>AGCNs) framework is proposed to model graph data and multi-sensor information fusion in a unified in-depth network and then to achieve abnormal vibration diagnosis. First, convolutional neural networks (CNNs) were adopted for the deeper features of multi-sensor signals. And then, the extracted features were fed into the proposed feature-driven adaptive graph generation network to build graphs to extract spatial–temporal correlation between multi-sensor data. After that, the multi-scale graph convolutional networks (MSGCNs) were employed to aggregate and enrich several different receptive information to further improve valuable features. Finally, the extracted multi-sensor features were integrated into a unified network to achieve the abnormal vibration type classification and identification of the rolling mill. Meanwhile, we performed horizontal, vertical, and coupled abnormal vibration experiments, and then three different types of studies were conducted to illustrate the superiority and usefulness of this method in the paper and the feasibility of rolling mill abnormal vibration diagnosis. It can be seen from the results that the proposed M<sup>2</sup>AGCNs can be able to achieve valuable feature extraction effectively from multi-sensor information and to obtain more excellent behavior of the abnormal vibration diagnosis of the rolling mill in comparison with the mainstream methods.https://www.mdpi.com/2075-1702/13/1/30rolling millabnormal vibrationmonitoring and diagnosismulti-sensor information fusionmulti-scale graph convolutional networks |
| spellingShingle | Rongrong Peng Changfen Gong Shuai Zhao Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill Machines rolling mill abnormal vibration monitoring and diagnosis multi-sensor information fusion multi-scale graph convolutional networks |
| title | Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill |
| title_full | Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill |
| title_fullStr | Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill |
| title_full_unstemmed | Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill |
| title_short | Multi-Sensor Information Fusion with Multi-Scale Adaptive Graph Convolutional Networks for Abnormal Vibration Diagnosis of Rolling Mill |
| title_sort | multi sensor information fusion with multi scale adaptive graph convolutional networks for abnormal vibration diagnosis of rolling mill |
| topic | rolling mill abnormal vibration monitoring and diagnosis multi-sensor information fusion multi-scale graph convolutional networks |
| url | https://www.mdpi.com/2075-1702/13/1/30 |
| work_keys_str_mv | AT rongrongpeng multisensorinformationfusionwithmultiscaleadaptivegraphconvolutionalnetworksforabnormalvibrationdiagnosisofrollingmill AT changfengong multisensorinformationfusionwithmultiscaleadaptivegraphconvolutionalnetworksforabnormalvibrationdiagnosisofrollingmill AT shuaizhao multisensorinformationfusionwithmultiscaleadaptivegraphconvolutionalnetworksforabnormalvibrationdiagnosisofrollingmill |