Brötzmann, Jascha ; Thiele, Christian-Dominik ; Rupp, Maximilian Michael ; Xing, Mingchen ; Rüppel, Uwe
Hrsg.: Papadrakakis, M. (2023)
A Machine Learning Leveraged and Minimal Hardware-based Train and Load Recognition System.
COMPDYN 2023. Athen, Griechenland (12.06.2023-14.06.2023)
doi: 10.7712/120123.10478.20409
Konferenzveröffentlichung, Bibliographie
Kurzbeschreibung (Abstract)
In the field of structural health monitoring, it is very important to quantify the impacting loads to better estimate and track the remaining service life of the structure. For railway bridges, it is therefore necessary to identify the passing trains. In current research projects, such as ZEKISS (www.zekiss.de), this is often done using sensors, e.g., acceleration sensors. The collected data is then analysed with respect to the impacting loads and possible changes in the supporting structure. However, the determination of the passing train is complex and the instrumentation of bridges with multiple sensors and a corresponding edge device is very expensive and time consuming. In addition, due to the large number of ageing bridges, an economical solution is needed. For those reasons, it is investigated how passing trains can be identified using minimal hardware at minimal costs. The following concept is proposed and tested. The first step is to classify different types of trains as freight or passenger trains. This can be done using image recognition algorithms. The hardware used should be as simple and inexpensive as possible so that it can be replicated for many bridges. Therefore, a single board computer with a camera module for image acquisition and expandability for other components is used. In addition, the hardware is extended to include lasers or photo sensors to classify train and wagon types using light barriers. This can be done by the length and bogie of each wagon and using databases to match the corresponding train. Distance and time can also be used to calculate the speed of the train, which is useful for dynamic studies of the bridge and calibration of the sensors. With all this together, a train and load recognition system is created.
| Typ des Eintrags: | Konferenzveröffentlichung |
|---|---|
| Erschienen: | 2023 |
| Herausgeber: | Papadrakakis, M. |
| Autor(en): | Brötzmann, Jascha ; Thiele, Christian-Dominik ; Rupp, Maximilian Michael ; Xing, Mingchen ; Rüppel, Uwe |
| Art des Eintrags: | Bibliographie |
| Titel: | A Machine Learning Leveraged and Minimal Hardware-based Train and Load Recognition System |
| Sprache: | Englisch |
| Publikationsjahr: | 2023 |
| Ort: | Athen |
| Verlag: | National Technical University of Athens (NTUA), Institute of Structural Analysis and Antiseismic Research School of Civil Engineering |
| Buchtitel: | COMPDYN 2023: 9th International Conference on Computational Methods in Structural Dynamics and Earthquake Engineering : proceedings ; volume 1 |
| Veranstaltungstitel: | COMPDYN 2023 |
| Veranstaltungsort: | Athen, Griechenland |
| Veranstaltungsdatum: | 12.06.2023-14.06.2023 |
| Auflage: | First edition |
| DOI: | 10.7712/120123.10478.20409 |
| Kurzbeschreibung (Abstract): | In the field of structural health monitoring, it is very important to quantify the impacting loads to better estimate and track the remaining service life of the structure. For railway bridges, it is therefore necessary to identify the passing trains. In current research projects, such as ZEKISS (www.zekiss.de), this is often done using sensors, e.g., acceleration sensors. The collected data is then analysed with respect to the impacting loads and possible changes in the supporting structure. However, the determination of the passing train is complex and the instrumentation of bridges with multiple sensors and a corresponding edge device is very expensive and time consuming. In addition, due to the large number of ageing bridges, an economical solution is needed. For those reasons, it is investigated how passing trains can be identified using minimal hardware at minimal costs. The following concept is proposed and tested. The first step is to classify different types of trains as freight or passenger trains. This can be done using image recognition algorithms. The hardware used should be as simple and inexpensive as possible so that it can be replicated for many bridges. Therefore, a single board computer with a camera module for image acquisition and expandability for other components is used. In addition, the hardware is extended to include lasers or photo sensors to classify train and wagon types using light barriers. This can be done by the length and bogie of each wagon and using databases to match the corresponding train. Distance and time can also be used to calculate the speed of the train, which is useful for dynamic studies of the bridge and calibration of the sensors. With all this together, a train and load recognition system is created. |
| Freie Schlagworte: | minimal hardware; machine learning; predictive maintenance |
| Fachbereich(e)/-gebiet(e): | 13 Fachbereich Bau- und Umweltingenieurwissenschaften 13 Fachbereich Bau- und Umweltingenieurwissenschaften > Institut für Numerische Methoden und Informatik im Bauwesen |
| Hinterlegungsdatum: | 28 Nov 2023 09:13 |
| Letzte Änderung: | 05 Apr 2024 11:32 |
| PPN: | 516940783 |
| Export: | |
| Suche nach Titel in: | TUfind oder in Google |
 |
Frage zum Eintrag |
Optionen (nur für Redakteure)
 |
Redaktionelle Details anzeigen |
Drucken
Impressum