Our nation's aging infrastructure presents a significant engineering and public policy challenge, particularly given that projected financial resources will not be adequate to fund the requisite structural system repairs and replacements to stem this infrastructure deterioration. The inevitable result is that buildings, bridges, roads, tunnels and other structural systems will frequently remain in service past their initial design lifetimes and design loads. Consequently, novel cost-effective solutions must be pursued to ensure the structural integrity of such systems, particularly when subjected to natural hazards such as earthquakes and strong winds. Structural systems equipped with smart damping systems, whose properties can be adapted to their changing environments, are one promising solution. This award supports fundamental research to enable design methods to incorporate dissipative and nonlinear nature of controllable dampers yielding smart damper designs with increasing performance at reduced cost. This will benefit the U.S. society by making controllable dampers a realistic approach for addressing the dilemma of aging infrastructure. The research involves disciplines of structural mechanics, control theory and computational science. This multi-disciplinary and multi-institutional project will help broaden participation by students from groups traditionally underrepresented in engineering research and will positively impact engineering education at both the undergraduate and graduate levels.
Hybrid Model Predictive Control can affect a significant advancement in controllable damping capabilities to reduce structural response and improve structural safety. The damper design will incorporate the physical dissipative limitations of the dampers through the use of hybrid system models to capture the on-off switching and to model predictive control for nonlinear or constrained dynamical systems. While the promise of this approach for controllable damping design has been established, several barriers must be overcome before wider implementation is possible. Using a set of bridge and building testbed structural systems and several models of controllable damping devices, this project will investigate learning-based methods and parallel computing techniques. The result will reduce the often prohibitive computational expense of repeatedly solving the high-order mixed integer-quadratic programming problems that arise from application to realistic structure models. Critical numerical and laboratory experiments will validate the efficacy of these nonlinear controllable damping approaches and establish that the resulting designs are robust to uncertainties and errors in the structural model, sensor noise and hardware limitations that challenge implementation.
