The objective and reliable assessment of the structural condition of a bridge, as well as the early detection of damage type, location, and cause, represent central aspects of structural health monitoring (SHM). Automated monitoring systems that continuously diagnose the behavior and condition of structures are gaining increasing importance. In particular, non-destructive testing methods based on measurements and computational analyses are highly suitable for SHM. However, comprehensive and validated approaches for the automated long-term monitoring of the global structural condition and for the detection of damage in massive structures remain associated with significant challenges.

The integration of automated monitoring systems enables an objective and continuous assessment of the structural condition. This allows a timely detection and evaluation of damages and continuous documentation of the structure’s aging process. Within the framework of the priority program SPP 100+, the “Nibelungen Bridge” in Worms and the research bridge “openLAB” serve as validation structures for the development of a comprehensive approach to automated condition monitoring of highly stressed concrete structures under real environmental conditions (work phases illustrated in Figure 1). The structural condition is to be described through reliable identification of the location and extent of detected damages. The approach involves damage detection based on nonlinear model adaptation and methods of artificial intelligence (AI).

As part of the nonlinear finite element (FE) analyses, sensitivity studies are conducted to identify the detectable structural and damage parameters. Also, the monitoring concept is prepared for application in automated damage diagnosis. Due to the inverse nature of the model adaptation problem, the uniqueness of possible solutions is to be analyzed and evaluated. For implementation, various AI methods are combined with approaches for uncertainty quantification. The optimization task including nonlinear model updating is highly complex: The models under consideration may exhibit different numbers of degrees of freedom, various loading conditions, and multiple potential damage locations. Based on genetic algorithms (GA), an optimization procedure is to be developed and applied to obtain solutions within the model adaptation process. Since such optimization procedures typically do not yield unique solutions, clustering methods are investigated to improve the reliability of the resulting solution descriptions. Each measurement instance is thereby assigned to a numerical model whose system components reliably represent the structural condition. Damage diagnosis, prognosis, and recommendations for maintenance actions are subsequently derived through comparison of system states at different measurement times and the measured structural responses. The concept for automated damage diagnosis developed during research phases 1 and 2 (see Figure 2) will be validated through controlled damage tests on the research bridge “openLAB”.