Modified Backtracking Search Optimization Algorithm Inspired by Simulated Annealing for Constrained Engineering...

Area of Science:

• Computational Intelligence
• Numerical Optimization
• Evolutionary Algorithms

Background:

• The backtracking search optimization algorithm (BSA) excels at global exploration but struggles with local exploitation, hindering convergence speed.
• Existing optimization algorithms face challenges in balancing exploration and exploitation for complex numerical problems.

Purpose of the Study:

• To enhance the convergence speed of the backtracking search optimization algorithm (BSA).
• To address the limitations of BSA's local exploitation capability in numerical optimization.

Main Methods:

• A modified backtracking search optimization algorithm (BSAISA) was developed, integrating principles from simulated annealing.
• The amplitude control factor (F) in BSA was redesigned using the Metropolis criterion, allowing adaptive decrease over iterations without extra parameters.
• A self-adaptive epsilon-constrained method was employed for effective handling of strict constraints.

Main Results:

• The BSAISA demonstrated superior effectiveness compared to the original BSA.
• BSAISA showed competitive performance against other well-known algorithms in terms of convergence speed.
• The modified algorithm successfully solved thirteen constrained benchmark problems and five engineering design problems.

Conclusions:

• The proposed BSAISA effectively improves the convergence speed of BSA for numerical optimization.
• The integration of simulated annealing principles and adaptive constraint handling enhances algorithm performance.
• BSAISA offers a more competitive and efficient alternative for solving constrained numerical optimization tasks.