Fragility-based framework for optimal damper placement in low-rise moment-frame buildings using machine learning and genetic algorithm

Abstract

Commonly, seismic fragility is used as an assessment tool rather than a design constraint for existing building structures. In addition, the degree of seismic fragility safety margin after retrofitting existing structures is usually unknown during the early retrofit design stage. The current study proposes a framework that allows the designer to assign the preferred seismic fragility margin as a design variable before retrofit of a structure using soft computing techniques. In addition, it provides the design damping ratio to achieve the optimal damper capacity distribution required for retrofit. To this end, a machine learning model (MLM) is used to construct safety margin curve (SMC) graphs for the structural system considering different limit states to obtain the design damping ratio. This MLM is trained with a wide range of structural systems subjected to various levels of ground motion intensities for predicting seismic responses of the system. The main role of the MLM and SMC is to provide a reasonable approximate range for damper capacity in each story to reduce the computational time required for the optimization process. After that, a genetic algorithm (GA) optimization process is implemented to obtain the optimum capacities of the energy dissipation devices (EDDs) required for retrofitting the structure using the design damping ratio obtained from the SMC. Displacement-dependent EDDs are used in the current study because of their effectiveness in dissipating seismic energy. The EDDs are arranged such that the seismic demands are controlled to achieve the required basic seismic safety objective maintaining the optimal EDD capacity for each story. The reliability and robustness of the framework are investigated using various case-study examples and validated by comparing the results with those obtained from nonlinear time history (NLTH) and nonlinear static methods. The proposed framework is validated for low-rise moment-resisting frame buildings that can be represented using a single degree of freedom (SDOF) system. The framework is expected to enable researchers and engineers to effectively implement the required level of seismic safety in retrofitted structures with less computational cost and allows the designer to select the most suitable retrofit scheme considering multiple limit state criteria checks.