Abstract
In response to challenges in the design process of concrete arch pre-supporting systems (CAPS) for urban subway stations, we propose a novel approach that integrates robust multi-objective decision-making, ensemble learning, and clustering algorithms. This integrated framework aims to achieve an optimal design by considering both design objectives and geotechnical uncertainties and achieving a balance between safety and cost. While Multi-Objective Optimization (MOO) is widely used to balance multiple objectives, traditional methods like Genetic Algorithms often neglect uncertainties in variables. Most prior studies have concentrated on finding optimal solutions without considering variable uncertainties or their correlations. Our approach considers soil uncertainties, a critical factor often overlooked in real-world projects, and strives to understand how variability in soil parameters affects support system performance. In fact, this study introduces a comprehensive framework for managing the design of support systems for subway stations built using the CAPS method, under conditions of uncertainty in soil parameters. The proposed framework is designed to provide decision-makers with optimal support system parameters while also assessing the robustness of the system against uncertainties in soil. The framework combines various techniques such MOO, artificial neural networks, robust decision-making, Cholesky decomposition, and Bayesian learning. With this framework, designers can select solutions that align with their specific criteria while minimizing the impact of soil uncertainties. To improve this framework, future research can incorporate additional robustness evaluation metrics and consider uncertainties in other soil parameters. This comprehensive approach has significant potential for evaluating design alternatives through a range of multi-criteria decision-making methods.