A HYBRID META-HEURISTIC ALGORITHM FOR OPTIMUM DESIGN SKELETAL STRUCTURES

This study proposes a new hybrid algorithm for optimum design of skeletal structures using a search method inspired by such meta-heuristic algorithms as GA, CSS, and PSO. In the proposed method, the exploration space is searched via moving points. Each design in the design space corresponds to a moving point in the exploration space. Collectively, these moving points form a population. By moving within the exploration space, these moving points create an evolutionary process for successive populations while moving towards the optimal point. The movement and displacement of the moving points in the exploration space is consistently based on the factors and characteristics of the previous population. To this end, similar to other meta-heuristic algorithms, the first population is created randomly. Then, the positions of the points in the next population are determined based on the geometric center of the previous population, the geometric center of the selected points, and the positions of

the selected points in the previous population. In this way, the points form a new population by moving within the design space towards the collective center of the points, the collective center of the selected points, and the respective positions of the selected points in the previous population. The average quality of the present population points, the average quality of the set of the selected points, and the quality of each selected point affect the displacement of the moving points. Other significant factors affecting the formation of points in a new population include variation as well as displacement of individual points in the previous population, acting independent parameters in specifying a new position for each point. To evaluate the efficiency of the proposed algorithm, we used a number of the benchmark examples. To this end, we plotted the optimization process convergence diagram for each example to study the method used in the proposed algorithm for obtaining the optimum point. On the other hand, we determined the average number of successive runs obtained for the proposed algorithm for each example. Our results showed that the best and the average run convergence trends calculated for different examples were in good agreement, which is a sufficient proof that the proposed algorithm possesses the required efficiency in obtaining the optimum point.