Fault-tolerant sensor networks for wireless structural health monitoring

Wireless structural health monitoring systems, being low-cost, flexible and easy to install, essentially consist of wireless sensor nodes that are connected to a central server. Modern wireless sensor nodes are not only capable of collecting, but also of processing and analyzing sensor data recorded from a structure. However, it is well known that the dependability and the accuracy of wireless monitoring systems are significantly affected by sensor faults of the wireless sensor nodes. Although a broad wealth of fault diagnosis concepts is widely applied in related disciplines, these concepts cannot be adopted to the specific monitoring problems in civil engineering, because civil engineering structures are usually unique, and the sensor nodes are exposed to harsh environmental conditions, being in unattended operation at remote sites with very limited energy and computing resources. This project aims at developing new methods and models enabling fault tolerance in wireless monitoring systems in order to improve the quality of structural health monitoring. Compared to existing approaches towards fault tolerance, the methods and models proposed in this research address three important new concepts: First, fault tolerance is achieved in a fully decentralized fashion; second, different types of sensor faults (Figure 1) are self-diagnosed by the wireless sensor nodes in a resource-efficient way based on embedded software; third, the monitoring quality and the effects of sensor faults are quantitatively assessed through an information entropy approach. Methodologically, partial system models will be embedded into the wireless sensor nodes for decentralized, autonomous fault diagnoses. The partial system models - leading to dynamically coupled, nonlinear global system models - are based on so called online learning neural approximators. The neural approximators, adapting to the actual structural condition of the monitored structure, implement the concept of analytical redundancy. It is expected that the dependability and the accuracy of wireless structural health monitoring can considerably be enhanced, facilitating more precise condition assessments as well as reduced maintenance and operating costs of civil engineering structures.

Figure 1: Classification of sensor fault types.

Project type
Internal research project

Since 2014

Project-related publications (selection)

  • Dragos, K. & Smarsly, K., 2016. Distributed adaptive diagnosis of sensor faults using structural response data. Smart Materials and Structures, 25(10), 105019.

  • Smarsly, K. & Law, K. H., 2014. Decentralized Fault Detection and Isolation in Wireless Structural Health Monitoring Systems using Analytical Redundancy. Advances in Engineering Software, 73(2014), pp. 1-10.

Awards and honors

  • "Best Paper Award" of the Forum Bauinformatik 2015, bestowed by the German Association of Computing in Civil Engineering (GACCE) for the paper "An analytical redundancy approach towards decentralized autonomous fault detection in wireless structural health monitoring" co-authored by Katrin Jahr, Kosmas Dragos, and Eike Tauscher (paper, further publications of the Chair of Computing in Civil Engineering)

  • Third place within the national competition "Built on IT" ("Auf IT gebaut"), awarded by the German Federal Ministry for Economic Affairs and Energy (BMWi) for the research work by Katrin Jahr "Autonomous fault detection in wireless structural health monitoring systems using decentralized neural networks" (based on a M.Sc. thesis supervised by Prof. Smarsly and Mr. Kosmas Dragos) (competition website, report, poster of the M.Sc. thesis)

Professor Dr. Kay Smarsly
Bauhaus University Weimar
Computing in Civil Engineering
Coudraystraße 7, Room 517
99423 Weimar
Email: kay.smarsly[at]uni-weimar.de