Evolutionary Big Data Analytics and Multi-Objective Optimization

Abstract

Optimization is a key concept in solving many industrial problems, often complicated by constraints that increase the complexity of finding feasible solutions. Many practical problems require optimization under constraints, a complex area extensively studied by researchers across various disciplines. This thesis introduces innovative approaches to handle constraints in single and multi-objective optimization tasks using population-based algorithms. The proposed methods focus on dynamically adjusting the boundaries of variables based on constraints, minimizing the generation of infeasible solutions, and guiding the search process toward feasible regions in the solution space.

These methods are applied to benchmarks and real-world optimization challenges with varying objectives and scales, employing state-of-the-art algorithms tailored for single-, multi- and many-objective scenarios. Examples include solving renowned engineering problems like the welded beam design and car side optimization problems using these newly proposed techniques. Moreover, the open-pit mining optimization problem is used as a large-scale optimization problem and is solved with the suggested method. The results demonstrate that the proposed methods outperform traditional techniques, improving the speed of finding feasible solutions and enhancing the objective space across both single- and multi-objective optimization tasks. Significant improvements are observed in handling multiple objectives and constraints, validated through comparative analyses with state-of-the-art algorithms. The findings have broad implications for the field of optimization, offering advancements in computational efficiency and practical applications in structural design, automotive engineering, and mining.