Application of a novel metaheuristic algorithm inspired by Adam gradient descent in distributed permutation flow shop...

Area of Science:

• Optimization Algorithms
• Computational Intelligence
• Mathematical Optimization

Background:

• Swarm intelligence algorithms often exhibit unbalanced exploration and exploitation due to similar designs.
• Mathematical properties combined with stochastic processes can improve optimization.
• Existing metaheuristics struggle with complex continuous optimization and engineering challenges.

Purpose of the Study:

• Introduce the Adam Gradient Descent Optimizer (AGDO), a novel meta-heuristic algorithm.
• Address limitations in exploration-exploitation balance in current optimization techniques.
• Evaluate AGDO's effectiveness on continuous optimization and engineering problems.

Main Methods:

• Developed AGDO inspired by the Adam optimizer, incorporating three key rules: progressive gradient momentum integration, dynamic gradient interaction system, and system optimization operator.
• Assessed AGDO performance against 19 established and new metaheuristics on CEC2017 benchmarks across multiple dimensions (10, 30, 50, 100).
• Evaluated AGDO on six practical engineering challenges, including the Distributed Permutation Flow Shop Scheduling Problem (DPFSP), and compared it with six state-of-the-art algorithms.

Main Results:

• AGDO demonstrated strong performance across dimensions 10, 30, 50, and 100, achieving the highest Wilcoxon rank-sum test scores in three dimensions.
• The algorithm maintained an excellent equilibrium between exploration and exploitation, converged rapidly, and evaded local optima.
• AGDO showed significant effectiveness on complex real-life engineering challenges, particularly excelling in the Distributed Permutation Flow Shop Scheduling Problem (DPFSP).

Conclusions:

• AGDO offers a superior approach to continuous optimization and engineering challenges by effectively balancing exploration and exploitation.
• The algorithm's mathematical foundation and novel operators contribute to its robust performance and ability to escape local optima.
• AGDO represents a significant advancement in meta-heuristic optimization, with practical implications for complex scheduling and engineering tasks.