Development of a Data-driven-based Multi-element Hybrid Simulation Technique in Seismic Response Assessment of Steel BRBFs

Hybrid simulation stands as an advanced technique in structural testing, enabling the assessment of seismic response of structures by leveraging the reliability of experimental testing and the efficiency of numerical simulations.
In hybrid simulation, critical elements of the structure which are anticipated to undergo instability or significant nonlinearity are physically tested in the lab, while the remaining portion of the structure, i.e., well-known elements, are computationally analyzed through the finite element methods.
Yet, hybrid simulation outcomes might be prone to bias when dealing with structures containing multiple critical elements like seismic fuses in multi-story buildings or bridge piers. Limitations within the laboratory testing such as lack of equipment and financial constraints often lead to the experimental testing of onl-y one or a few of such critical elements. Moreover, errors raising from the test controller, hydraulic actuators, signal delays, and similar factors can impact the reliability of multi-element hybrid simulation. To address these constraints, this research introduces a framework for the seismic response assessment of multi-storey steel buckling-restrained braced frames (BRBFs). This framework integrates data-driven techniques into conventional hybrid simulation, facilitating multi-element hybrid simulation for structural systems with several potential critical components.
Initially, a multi-storey buckling restrained braced frame is first selected and designed following the guidelines outlined in the Canadian steel design standard. In the BRBF, the first-storey BRB is virtually simulated, representing the test specimen, while other BRBs utilize a data-driven model trained on past experimental data initially and then updated with new incoming data from the tested specimen during excitation. Real-time updates of model parameters during hybrid simulation use data received from the virtual test specimen.
The results confirm that the proposed data-driven-based hybrid simulation technique can offer a viable solution to address the shortcomings of conventional seismic hybrid simulation by taking advantage of model updating and multi-element simulation platforms.