Darzin International Company Multidisciplinary Engineering Science Open 2 Serial Number 2 2025 01 01 Surrogate-Assisted Global Optimization for Expensive Engineering Systems: From Trust Regions to Bayesian Optimization Surrogate-Assisted Global Optimization for Expensive Engineering Systems: From Trust Regions to Bayesian Optimization 1 11 EN Journal Article 2024 10 22 <p>This review aims to synthesize methodological and conceptual advances in surrogate-assisted global optimization (SAGO) for computationally expensive engineering systems, highlighting the evolution from deterministic trust-region frameworks to probabilistic Bayesian optimization approaches. A qualitative systematic review design was employed using content analysis of peer-reviewed literature. Twenty articles published between 2010 and 2025 were selected through purposive sampling after comprehensive database searches in Scopus, Web of Science, IEEE Xplore, and ScienceDirect. Only studies addressing surrogate-assisted strategies for expensive or multi-fidelity optimization were included. Data collection relied exclusively on literature review, and theoretical saturation was achieved after analyzing 20 studies. The qualitative coding and thematic synthesis were conducted using Nvivo 14 software, following open, axial, and selective coding procedures to extract major conceptual themes related to surrogate frameworks, optimization strategies, and robustness mechanisms. Three overarching themes emerged: (1) Evolution of Surrogate Modeling Frameworks—the transition from polynomial and RBF surrogates to probabilistic Kriging, multi-fidelity, and deep learning-based surrogates such as physics-informed neural networks; (2) Global Optimization Strategies and Trust-Region Adaptation—the convergence of deterministic trust-region methods with Bayesian acquisition-based algorithms that integrate uncertainty-aware exploration and exploitation; and (3) Robustness, Generalization, and Application Integration—the expansion of surrogate-assisted methods into real-world workflows emphasizing uncertainty quantification, transfer learning, and digital twin integration. Together, these themes reveal a paradigm shift toward scalable, adaptive, and hybrid optimization systems that unify physics-based modeling with data-driven intelligence. Surrogate-assisted optimization has evolved from local curve-fitting into a data-efficient, uncertainty-aware framework fundamental to modern engineering design. The field now converges toward hybrid, physics-informed, and AI-integrated paradigms that enable robust, automated decision-making in computationally intensive environments.</p> https://jmesopen.com/index.php/jmesopen/article/download/40/35