Performance Analysis of Multi-Objective Simulated Annealing Based on Decomposition

Simulated annealing is a metaheuristic that balances exploration and exploitation to solve global optimization problems. However, deal with multi- many-objective problems, this balance needs be improved due diverse factors such as the number of objectives. To issue, work proposes MOSA/D, hybrid framework for multi-objective simulated based on decomposition evolutionary perturbation functions. According literature, strategy allows diversity in population while perturbations add convergence toward Pareto front; however, question should asked: What effect components when included part design? Hence, studies performance MOSA/D considering its implementation two widely used operators: classical genetic operators differential evolution. The proposed algorithms are MOSA/D-CGO, operators, MOSA/D-DE, evolution operators. main contribution analysis using both identifying one most suitable framework. approaches were tested DTLZ three objectives CEC2009 benchmarks two, three, five, ten objectives; considered measured through hypervolume (HV) inverted generational distance (IGD) indicators. results pointed out there promising improvement favor MOSA/D-DE.