An Enhanced Randomized Dung Beetle Optimizer for Global Optimization Problems.

Area of Science:

• Optimization Algorithms
• Computational Intelligence
• Metaheuristics

Background:

• The standard Dung Beetle Optimizer (DBO) shows potential for complex optimization but faces challenges with premature convergence and accuracy.
• Existing metaheuristic approaches often struggle to balance global exploration and local exploitation effectively.

Purpose of the Study:

• To introduce the Enhanced Reproductive Dung Beetle Optimizer (ERDBO) to overcome the limitations of the conventional DBO.
• To improve convergence rate, stability, and solution precision in optimization tasks.

Main Methods:

• The ERDBO employs a novel three-stage mechanism: larval growth with experiential learning for diversity, reproduction with parent-offspring verification for exploitation, and predator avoidance with Lévy flight for adaptability.
• Algorithm performance was evaluated using CEC2017 benchmark functions and compared against advanced metaheuristic methods.
• The ERDBO was applied to engineering design problems including tension/compression springs, three-bar trusses, and pressure vessels.

Main Results:

• The ERDBO demonstrated superior performance compared to other advanced algorithms in terms of convergence rate, stability, and solution precision.
• Experimental results on benchmark functions confirmed the ERDBO's effectiveness.
• Successful application to engineering design tasks validated its efficiency and practical applicability.

Conclusions:

• The ERDBO offers a significant advancement over the DBO, effectively mitigating premature convergence and enhancing accuracy.
• The proposed algorithm provides a robust and competitive optimization framework suitable for complex real-world engineering challenges.
• The ERDBO's hybrid approach enhances adaptability and accelerates convergence, making it a valuable tool in computational optimization.