Data driven structural behaviour modelling in civil engineering

The project proposes: (a) development of procedures for Bayesian finite element model updating and uncertainty quantification for large-scale civil structures under service loading, and (b) application of advanced methods of artificial intelligence for analyses of structural health monitoring data in order to build data models. The related overall objectives of the project are: (a) to choose a few representative civil engineering structures, perform in-situ forced vibrations test, and apply structural identification process in the framework of Bayesian inversion in order to improve fidelity of the finite element models, and (b) to undermine the potential application possibilities of advanced methods of artificial intelligence to ameliorate the integration of vibration tests data and structural health monitoring data into the maintenance, and present this for a large highway bridge. The scientific objectives of the project are: (i) to assess the advantages and disadvantages of the uncertainty quantification, sensitivity analysis and Bayesian finite element model updating for large-scale civil engineering structures, (ii) to bring out possibilities of an automated update process of data driven model when receiving the sensors data continuously, (iii) to develop a method to identify systematic modelling error of the finite element model, and (iv) to test a novel idea of implementing a mixture density network for the finite element model update and compare it with the conventional and recent existing techniques. The plan is to study a few large-scale structures in order to assess existing and novel technologies and ideas. The structural health monitoring data will be available for one large highway bridge.