Generalized framework for robust design of tuned mass damper systems

The primary purpose of this contribution is to develop a novel framework for generalized robust design of tuned mass damper (TMD) systems as passive vibration controllers for uncertain structures. This versatile strategy is intended to be free of any restriction on the structure-TMD system configuration, the performance criterion, and the number of uncertain parameters. The main idea pursued is to adopt methods and concepts from the robust control literature, including: (1) the linear fractional transformation (LFT) formulation pertaining to the structured singular value (mu) framework; (2) the concept of weighted multi-input multi-output (MIMO) norms for characterizing performance; and (3) a worst-case performance assessment method to avoid the unacceptable computation burden involved with exhaustive search or Monte Carlo methods in the presence of multiple uncertainties. Based on these, the robust design framework is organized into four steps: (1) modeling and casting the overall dynamics into the proposed LFT framework that isolates the TMD system as the controller, and the uncertainties as a structured perturbation to the nominal dynamics; (2) setting up the optimization problem based on generalized indices of nominal performance, robustness, and worst-case performance; (3) implementing a genetic algorithm (GA) for solution of the optimization problem; and (4) post-processing the results for systematic visualization, validation, and selection of preferred designs. This strategy has been implemented on several illustrative design examples involving a seismically excited multi-story building with different combinations of assumptions on the uncertainty, TMD configuration, excitation scenarios, and performance criteria. The resulting solution sets have been studied through various post-processing methods, including visualization of Pareto fronts, uncertain frequency response plots, time-domain simulations, and random vibration analysis. (C) 2010 Elsevier Ltd. All rights reserved.