Background and motivation
Civil infrastructure is subject to continuous aging and increasing environmental impacts, causing deterioration and damage. Structural health monitoring (SHM) is deployed to advance predictive maintenance and condition assessment that helps identify deterioration at damage at early stages, supporting life-cycle management based on continuously recorded sensor data. However, common SHM applications require time-consuming post-processing tasks usually conducted manually. Furthermore, sensor faults, miscalibrations, and measuring errors usually remain undetected, which may render SHM applications error-prone and inefficient.
Project goals and scientific approach
This project aims to design fully automated analysis algorithms based on AI and IoT concepts, for detecting deterioration and damage of infrastructure. The analysis algorithms, to be executed by intelligent sensor systems, facilitate infrastructure condition assessment, taking into account sensor faults and sensor miscalibrations. The algorithms will be validated on a real bridge, which will specifically be built for this project, serving as a unique long-term large-scale test structure, which will intentionally be deteriorated and damaged for validating the novel sensor systems (Figure 1).
Objectives of this subproject and expected results
This subproject will specifically focus on developing ML algorithms for autonomous sensor fault diagnosis in structural health monitoring systems. The ML algorithms will be implemented using regression and classification techniques. Individual IoT-enabled sensors will self-detect sensor faults and sensor miscalibrations. Following the swarm intelligence paradigm, the sensors will act as intelligent entities, distributed-collaboratively communicating with each other and with human individuals, able to detect complex patterns in the sensor data. Structural data, inspection data, and system data will continuously be updated in real-time within a digital-twin representation. As a result, the decentralized data analysis algorithms and the fault diagnosis approach proposed in this project will advance digital, predictive maintenance. It is expected to render structural health monitoring more efficient, thus infrastructure maintenance and operation more sustainable.
- Dresden University of Technology, Germany
- Marx Krontal Partner, MKP GmbH, Weimar, Germany
- Hentschke Bau GmbH, Bautzen, Germany
- Autobahn GmbH
- Federal Institute for Materials Research and Testing (BAM)
- Federal Highway Research Institut (BASt)
Professor Dr. Kay Smarsly
Hamburg University of Technology
Institute of Digital and Autonomous Construction