Efficient analysis and evolutionary optimization of structural protective systems

Over the past decade, significant effort has been devoted to the development of passive and semiactive control techniques for the protection of structural systems against the adverse effects of earthquakes. In order to extract maximum benefit from these protective systems, there is a need to consider their optimal design. However, recent research efforts in the optimization of passive control systems usually only consider two-dimensional planar frame structures using lumped parameter models. In the first portion of this dissertation, finite element models for classical fluid viscous, solid viscoelastic and fluid viscoelastic dampers are constructed and then incorporated into a computer program 3DPASS for efficient dynamic analysis of three-dimensional structures. In particular, the finite element fluid viscoelastic damper model developed in this research offers an efficient way of incorporating the Maxwell type of energy dissipation devices into the dynamic analysis. Emphasis then shifts to the optimal design problem, here an evolutionary design methodology based on a genetic algorithm (GA) is developed for the configuration optimization of the passive damper systems within three-dimensional irregular building structures.; Semiactive control systems offer the potential for enhanced performance with minimum power requirements. Two different algorithms that can accommodate system uncertainties, namely, the sliding mode control and adaptive control, are implemented for semiactive variable dampers. These algorithms are shown to be robust with respect to structural uncertainties, while the traditional linear quadratic regulator (LQR) control deteriorates quickly. Subsequently, a genetic algorithm is employed for the optimal placement of a limited number of semiactive dampers in multi-degree-of-freedom (MDOF) structures. The binary encoding is partially controlled to ensure non-duplication of dampers at each story. It is shown from this study and the previous one on the optimization of passive control that GA is a very versatile optimization method.; However, the present GA approach is computationally intensive for complex three-dimensional structures. As an initial attempt to improve this situation, the incomplete Cholesky conjugate gradient method (ICCG) is implemented with the development of a procedure to assemble the global matrix into the compressed sparse row (CSR) format. Numerical examples demonstrate the saving in computational time and storage requirement of the ICCG method in comparison with the direct method. A computational framework for the development of multilevel aggregation method for building structures is also established. These efficient solvers may make the evolutionary optimization of passive energy dissipation system for large-scale complex three-dimensional structures a reality in the not-too-distant future.