Research On Electromagnetism-like Mechanism Algorithm

In recent years, a large number of optimization problems in practical engineering are becoming more and more complicated. However, it is difficult to solve the complex global optimization problems that are non-differentiable and non-convex with the traditional deterministic algorithms because such algorithms have rather high requirements for the optimization functions. So some heuristic optimization algorithms have got researched and developed greatly due to many significant advantages, such as good versatility, great intelligence and high efficiency. Electromagnetism-like mechanism (EM) algorithm, a new heuristic for global optimization, is inspired by the attraction-repulsion mechanism in an electromagnetic field. The EM algorithm has many advantages, such as simple optimization mechanism, few requirements for resource and high convergence rate. But for some complicated problems, the EM algorithm still has some disadvantages, such as low convergence rate, large possibility of falling into premature convergence, and so on.The optimization principle and mechanism of EM algorithm are studied deeply in this paper. An improved Electromagnetism-like mechanism(IEM) algorithm is proposed here to eliminate the shortcomings of the original EM algorithm, such as non-uniform initial population, low search efficiency, pathological feature in total force formula and poor adaptability in the movement of particles. The uniform design method is used to generate the initial population of the new algorithm, which makes the initial particle s be distributed in the feasible region more evenly. By combining the advantages of chaos search and pattern search, the new algorithm improves the local search algorithm with the hybrid chaotic pattern search method, which can not only avoid the algorithm falling into a local optimum prematurely, but also improve the accuracy of the solution to a certain extent. Besides, by normalizing the objective function value and introducing a correction factor into the total force formula, the power formula is simplified and the pathological feature in the total force formula is eliminated, which accelerate the convergence speed of the algorithm. Finally, an adaptive formula of movement is designed, and niche technology is introduced into the process of particles’movement, which can make particles move to the global optimal solutions more rapidly.Experimental results show that the improved algorithm has better optimizing performance. The new algorithm can improve not only the accuracy of solutions, but also convergence efficiency, which is successful in solving general functions, complex functions, Dixon and Szego functions and high-dimensional functions.In further research, the EM algorithm would be further improved, and the improved algorithms would be used to solve the optimization problems in actual projects, which are mainly constrained optimization problems, discrete optimization problems and multi-objective optimization problems.