Process Innovation via Adaptive Multi-exemplar Driven Optimization: Enhancing Engineering Design and Global Optimization Performance

Metaheuristic search algorithms (MSAs) are pivotal in tackling complex optimization problems across various engineering domains. Teaching–learning-based optimization (TLBO), noted for simplicity and rapid convergence, faces inherent challenges such as premature convergence and inadequate exploration-exploitation balance, restricting its practical effectiveness. Addressing these gaps, this paper proposes a novel multi-exemplar driven teaching–learning-based optimization (MEDTLBO) as a process innovation in metaheuristic optimization. MEDTLBO incorporates a sophisticated initialization mechanism integrating dynamic oppositional-based learning and diverse chaotic maps, ensuring robust initial population quality. A distinct adaptive strategy assigns tailored learning phases based on individual solution fitness, balancing exploration and exploitation dynamically. Additionally, novel exemplar construction methods in both teacher and learner phases facilitate personalized learning interactions, significantly enhancing search efficiency and diversity. Extensive experimental evaluations using CEC2014 and CEC2019 benchmark functions and various real-world engineering design problems indicate that MEDTLBO achieves competitive or better performance than the compared algorithms in most tested cases, rather than claiming universal superiority. These findings suggest that MEDTLBO is a promising process innovation to advance practical optimization tasks, promote economic productivity, and improve resource efficiency.