Automated assessment of monitoring data
Table of contents
Project data
Titel | Title TP der TU Dresden im Verbundprojekt IDA-KI: Automatisierte Bewertung der Monitoringdaten von Infrastrukturbauwerken | SP of TU Dresden in the joint research project IDA-KI: Automated assessment of monitoring data for infrastructure constructions Förderer | Funding Bundesministerium für Digitales und Verkehr (BMDV) / mFUND Zeitraum | Period 01/2022 – 12/2024 Verbund- und Teilprojektleiter | Leader of joint and subproject Prof. Dr.-Ing. Steffen Marx Bearbeiter | Contributors Max Herbers, M. Sc., Dipl.-Ing. Bertram Richter, Dipl.-Ing. Jonas Scharf Partner | Partners Institut für Digitales und Autonomes Bauen, TU Hamburg | Marx Krontal Partner GmbH, Weimar | Hentschke Bau GmbH, Bautzen Assoziierte Partner | Associated partners Bundesanstalt für Materialforschung und -prüfung (BAM), Berlin | Autobahn GmbH, Berlin | Bundesanstalt für Straßenwesen (BASt), Bergisch Gladbach |
Report from year book 2022
Big data? Smart data!
Infrastructure constructions are subjected to constant ageing, which is why today their condition is to be inspected manually at regular intervals. A repair action is planned only when damage is detected during the regular inspection of the structure. This reactive approach means that damage often remains undetected for a long time, resulting in increased maintenance costs. In the future, this problem-oriented approach is to be replaced by a data-based, predictive maintenance management system. The basis for a reliable assessment of building condition in near real time is structural monitoring. However, current monitoring applications require time-consuming manual evaluation, and it is difficult to detect ageing of the measurement system or measurement errors.
In the project IDA-KI, fully automated evaluation algorithms for monitoring data are developed with the help of machine learning and integrated into a monitoring concept spanning the entire service life from hour zero. The algorithms could already be successfully trained for a real data set of a bridge. For fault detection and correction, it is checked which sensors are correlating with each other. If one sensor shows strong deviations compared to its “partners”, this is an indication of a measurement error. The analytical redundancy approach eliminates the need for additional redundant sensors or a numerical model.
A demonstration bridge is currently under planning, which will be equipped with sensors, e.g. with fibre optic sensors, during construction. This allows the monitoring concept and the evaluation algorithms to be validated in fast motion. Employing load tests as well as targeted damage to redundant measurement technology, a real database can be obtained for the first time. In the future, it will therefore be possible to distinguish between measurement errors, influences from ageing of the measurement system and structural changes to the structure. Condition indicators should enable the intuitive interpretability of large amounts of data. The demonstrator, which will be erected in Bautzen, will be expanded as a real laboratory and will also serve as a place for scientific exchange in the coal region after the project period.