A novel modified BSA inspired by species evolution rule and simulated annealing principle for constrained engineering optimization problems

The backtracking search optimization algorithm (BSA) is one of the recently proposed evolutionary algorithms (EAs) for solving numerical optimization problems. In this study, a nature-inspired modified BSA (called SSBSA) is proposed and investigated to improve the exploitation and convergence performance of BSA. Inspired by the species evolution rule and the simulated annealing principle, this paper proposes two modified strategies through introducing a specified retain mechanism and an acceptance probability into BSA. In SSBSA, the specified previous individuals of historical population (oldP) and their corresponding amplitude control factors (F) are retained according to the fitness feedback for the next iteration, and a new adaptive F that could decrease as the number of iterations increases is redesigned by learning the acceptance probability. SSBSA has two main advantages: (1) The way to retain the specified previous information improves BSA's exploitation capability. (2) This new F adaptively controls the diversity of population which makes convergence faster. Simulation experiments are carried on fourteen constrained benchmarks and engineering design problems to test the performance of SSBSA. To fully evaluate the performance of SSBSA, several comparisons between SSBSA and other well-known algorithms are implemented. The experimental results show that SSBSA improves the performance of BSA and its performance is more competitive than that of the other algorithms.