Multi-Strategy Improved Aquila Optimizer with Adaptive Exploration and Individual-Level Stagnation Control: A...

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Metaheuristic algorithms are crucial for complex optimization without gradient information.
• Premature convergence and population stagnation limit current metaheuristic algorithm effectiveness.
• Existing methods often struggle with exploration diversity and persistent stagnation.

Purpose of the Study:

• To introduce the Stagnation-Aware Aquila Optimizer (SAAO) to overcome limitations in metaheuristic algorithms.
• To enhance the Aquila Optimizer (AO) with mechanisms addressing stagnation and exploration.
• To improve the robustness and applicability of optimization solvers.

Main Methods:

• Developed SAAO by integrating adaptive exploration probability, individual stagnation counters, and diversity maintenance into the AO framework.
• Incorporated physics-grounded operators from the Animated Oat Optimization (AOO) algorithm.
• Evaluated SAAO against nine state-of-the-art algorithms on CEC2015 and CEC2022 benchmark suites.

Main Results:

• SAAO achieved the best Friedman mean rank on both benchmark suites.
• Demonstrated statistically significant performance advantages over most competitors via Wilcoxon rank-sum tests.
• Achieved competitive results on engineering design problems and 98.23% accuracy in equipment anomaly prediction.

Conclusions:

• SAAO effectively mitigates premature convergence and population stagnation in optimization.
• The hybrid approach offers a robust and computationally tractable solution for diverse optimization challenges.
• SAAO shows strong potential for both theoretical benchmarks and practical, real-world applications.