Anagrafe della ricerca

Performance-Based Design (PBD) extends the design objectives of structures beyond load-bearing capacity and requires simulators capable of predicting the behaviour of both structural and non-structural elements. However, blind-prediction contests reveal that ab initio structural simulators fail to provide accurate predictions. Hence, calibration of structural simulators against experiments plays a central role in making PBD applicable. Hybrid simulation (HS) is a testing method that combines Physical and Numerical Substructures (PS and NS) conceived to respond to this need. The PS is tested in the laboratory while the NS is numerically simulated. Accordingly, a controller mimics the interaction between PS and NS. The aim is to measure low-quantity-high-value data to calibrate a structural simulator of the PS. However, the definitions of prototype structure and loading excitation are not optimized to maximize the gain of information to support the subsequent model calibration phase. The Hy-Learn project makes a leap in the current practice of model calibration by linking HS and Optimal Experimental Design. The goal is to minimize the cost of attaining a target predictive capability of the calibrated structural simulator. The applicant´s expertise covers computational modelling and calibration. The collaboration with the Earthquake Engineering & Dynamics lab of the Politecnico di Torino will provide knowledge transfer on Experimental Dynamics and Machine Learning. The cooperation established with international institutions, e.g., Aarhus University and ETH Zurich, will provide knowledge transfer on control theory in HS. Finally, the research projects activated at the Politecnico di Torino, including the ones the applicant is currently involved in, can exploit the potential of Hy-Learn for mitigating risks of the built heritage (existing buildings, infrastructure, etc.) and of the environment (NaTech accidents, etc.) against natural hazards.