A Modified Dragonfly Optimization Algorithm for Single- and Multiobjective Problems Using Brownian Motion

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Metaheuristics

Background:

• The dragonfly algorithm (DA) is a recent metaheuristic optimization technique inspired by dragonfly swarm behavior.
• The standard DA utilizes the Levy flight mechanism (LFM) for randomization, which can lead to search space overflow and flight interruptions due to large step sizes.

Purpose of the Study:

• To improve the randomization process of the dragonfly algorithm.
• To address the limitations of the Levy flight mechanism in the standard DA.
• To evaluate the performance of a modified dragonfly algorithm using Brownian motion.

Main Methods:

• The Levy flight mechanism in the dragonfly algorithm was replaced with Brownian motion.
• The modified dragonfly algorithm was tested on 15 single-objective and 6 multi-objective benchmark optimization problems.
• Performance comparison between the original DA and the modified DA was conducted.

Main Results:

• The modified dragonfly algorithm demonstrated significant improvements, achieving up to 90% enhancement in minimum point access compared to the original DA.
• Application to the welded beam design problem resulted in a 20% reduction in optimum cost.
• The modified algorithm showed superior performance in both benchmark testing and a practical engineering design problem.

Conclusions:

• Replacing Levy flight with Brownian motion effectively enhances the randomization stage of the dragonfly algorithm.
• The modified dragonfly algorithm offers improved efficiency and effectiveness for solving complex optimization problems.
• The enhanced DA presents a viable alternative for optimization tasks, including engineering design challenges.