Implementation of an Advanced Hybrid Simulation Technology

The beam-to-column connections in any structural building frame are vital components governing the overall behavior of the structure. The wide range brittle fracture failures of the then popular welded connections in steel moment resisting frames during the 1994 Northridge earthquake made these components the focus of countless research projects since then. Unlike the beams and columns whose behavior is well understood, predicting the actual rigidity of the connections has always been a challenge due to their semi-rigid behavior governed by several factors like slippage, gap openings, friction and high nonlinearity.

The response of structures to seismic loads has traditionally been studied by using either experimental testing or analytical testing. Experimental tests conducted on small to large scale physical specimens representative of full structures have the potential of capturing the ‘exact’ response of the structure. But the large costs associated with the testing equipment and construction of the specimen and test rigs, along with the limitations of the testing facility make this option impractical for studying the response of large structures.

The state-of-the-art six degree of freedom shake table at the Richmond Field Station testing facility of the Pacific Earthquake Engineering Research Center (PEER), University of California, Berkeley

The other relatively inexpensive option of analytical testing utilizes mathematical and/or numerical techniques based on approximations and involves the modeling of the structure in a finite element program. This methodology is capable of providing acceptable results but for a limited family of components that are not ‘numerically hard to model’ i.e. whose behavior is well understood, and the boundary conditions and interactions can be properly defined.

In an attempt to keep the best of both these testing methodologies and to overcome their limitations, an innovative technique called hybrid testing came into practice recently. Hybrid testing involves breaking down of the entire structural frame into two sets. The first set called the analytical module is composed of all the portions of the structure whose behavior is well understood and can be modeled analytically with confidence like the beams and columns. The second set called experimental module consists of all the portions of the structure that are difficult to model due to several factors governing their highly nonlinear response, and that solicit experimental laboratory testing to capture their ‘exact’ response like the beam-to-column connections. Hybrid testing integrates the analytically simulated models of the analytical module with the physical laboratory testing of the experimental module in real-time to give the overall behavior of the entire structure.

Actuators used for implementing hybrid simulations in the laboratory

Hybrid simulation can be considered as a conventional finite element analysis where physical models of the some portions of the structure are embedded in the numerical model. The overall simulation is controlled by a hybrid engine that applies the load history, having the physical module updating the analytical module at every increment, and vice-versa. Apart from giving highly realistic and accurate overall results for large structures while keeping the costs to a sustainable level, real-time hybrid-testing captures the dynamic and strain rate effects unlike quasi-static experimental tests. Also, unlike the shake-table experimental tests, hybrid testing is not limited by the size of the shake table as it engages hydraulic actuators and data acquisition systems to transfer the forces and reactions exchanged between the analytical and the experimental modules.

Pacific Earthquake Engineering Research (PEER) Center headquarters, Berkeley, California

The hybrid simulation technique, however, has two major limitations. Hopefully in a future post, I will talk about how the research at the George E. Brown, Jr. Network for Earthquake Engineering and Simulation (NEES) has tried to overcome the first limitation. Also, I will talk about my own research at the University of Cincinnati where I am developing a model based on phenomenology, artificial intelligence and neural networks to overcome the second limitation of the hybrid simulation technology. The model engages several analytical Abaqus modules with its user-subroutine capability as integral components that are tied by a hybrid engine developed by NEES to the experimental module and other OpenSees analytical modules running on different machines over a network. Attached with this article are a few pictures from the hybrid simulation workshop with emphasis on real-time loading at the headquarters of the Pacific Earthquake Engineering Research (PEER) Center at Berkeley, California that I attended last year.