System-Level Performance-Based Design and Optimization of Wind Sensitive Structures

Abstract

The ultimate goal of this dissertation is to establish benchmark approaches for designing wind sensitive structures that optimize economic criteria, while rationally meeting society’s need for a truly safe built environment. In the face of worldwide population growth and urbanization, a significant percentage of the population work and live in high-rise buildings, many of which are located in regions that are prone to severe windstorms. Hence, the exposure of these infrastructure components, as well as their occupants to extreme wind hazards, is continuously increasing. This risk has triggered interest for proactive strategies that enhance structural performance such that these buildings can provide suitable shelters during extreme wind events and can be repaired to full functionality shortly after the event at minimal costs.

To achieve the desired high-performance goals at affordable costs, this research is focused on developing new design methods that can automatically identify minimal-investment designs that satisfy targets on measures such as system-level reliability and expected loss. The proposed approach is based on the integration of structural optimization methods with probabilistic performance-based wind engineering frameworks. While the integration concept is a very powerful design tool, a major challenge to its implementation in practice is the high computational demand associated with running stochastic simulations within the optimization loop that involves hundreds of design variables. To overcome this burden, a sequential optimization technique is developed that is based on defining a high-quality optimization sub-problem. The sub-problem allows the simulation to be approximately decoupled from the optimization process and can be efficiently solved through gradient-based optimization algorithms. This approach significantly accelerates the optimization process while utilizing information obtained from a simulation that entails site-specific hazard analysis, data-driven wind load models, dynamic structural analysis, as well as fragility-based damage and loss models. The effectiveness and scalability of the proposed system-level optimal design approach have been demonstrated through various applications, including reliability-based design optimization and performance-based design optimization (PBDO) of large-scale wind-excited building systems under loss constraints.

To extend the developed approach to practical design projects that involve multiple objectives, decision support frameworks are needed that can provide candidate designs and associated trade-off information to decision-makers. This motivation has led to the development of frameworks that can maximize system-level performance while simultaneously minimizing the initial cost. The proposed approach decomposes the multicriteria problem into a series of single-objective design problems, which can be solved through the recently introduced PBDO methods. The performance measure also accounts for the variance of the total loss, providing a means to evaluate the design robustness. Hence, the proposed frameworks can help the stakeholders make an informed decision on an optimal design based on the cost, performance, and robustness of the system.