Sensitivity analysis of model parameters in a nonlinear model-updating approach for prestressed concrete bridges

Automated monitoring systems in structural health monitoring (SHM) using finite element models can measure system responses of the structure and provide information about behavior and damage scenarios. To describe the reactions and damages of the physical structure, the parameters of the FE-model have to be adjusted accordingly. A common approach therefore is model-updating, an iterative modification of unknown parameters of a finite element model by adapting the response of the model to the measurement results of the physical bridge. The adapted model can be used for assessments and predictions of the structure and provides accurate structural properties to a digital twin. The proposed monitoring concept includes damage detection, based on nonlinear model-updating using Artificial Intelligence (AI) methods and is applied to a prestressed, reinforced concrete bridge based on the observation of static response. For a practical application in long-term SHM, system properties and load values are identified simultaneously by using optimization methods based on Evolutionary Algorithms (EA). Due to the high dimensionality and complexity of the optimization results, Principal Component Analysis (PCA) including autoencoder networks for dimensionality reduction are required to evaluate and visualize the identified structural properties for their reliability. Based on physical 3D nonlinear finite element sensitivity analysis, the influence of damage scenarios and load situations on the bearing behavior of the bridge will be analyzed. The generated data and the results are presented and serve as the basis for the reliability analysis and evaluating the proposed concept. Furthermore, by comparing the measurements of the parameterized investigated models, conclusions about the behavior of the structure and the sensor placement can be obtained.