Search Algorithm, Based On Space Division

Optimization is one type of application technology based on mathematics and used to acquire the optimized resolutions of diversified engineering problems. People have been widely paying attention to it as an important branch of science from the beginning, and it was rapidly carried out and applied in many engineering fields. But the convergent speed of some problems with large scales and high complexity would greatly slow down because of optimization algorithms' inherent fault. Accordingly the theory research on optimization algorithms has remarkable effects on improving performance, perfecting architecture and spreading applicable fields. The intervention of space partition and space contract theory provides new ideas for optimization algorithms.The evolvement of search optimization algorithms is firstly introduced in this thesis by classifying it into two groups: the global search algorithms and the local search algorithms. Then we introduce one typical global search algorithm-genetic algorithm, and one typical local search algorithm-Tabu search algorithm. A new certain algorithm is also described -interval optimization algorithm.One emphasis of this paper is to present a hybrid search algorithm based on grid partitioning. The ideas about space partition and space contracting are applied in this hybrid algorithm: ascertain the elite individuals by one type of global optimization algorithms during the first step, then rapidly partition space and contract it to multi subspaces, in succession the extremum is converged on in short time by hunting nearby the model's extremum using a sort of local optimization algorithms. The simulation results show that the hybrid algorithm has excellent performance in efficiency, applicable fields, results precision and robusticity, therefore it proves our algorithm can work efficiently.Furthermore, some improvements are made to the interval optimization algorithm in the paper. As for one-dimension optimization problems a new step of interval pruning that includes boundary pruning and inner pruning is combined into the algorithm computation process, and it can rapidly delete the space barring global extremums with high efficiency. A hybrid interval evolutionary algorithm is delivered as for multi-dimension optimization problems. The hybrid algorithm takesstrong points from interval method and evolutionary algorithm to offset their weaknesses, hence the two algorithms are syncretized well. The numerical experiments indicate that both one-dimension and multi-dimension algorithms are reliable and efficient. In the final the summarization is presented. Some problems to be solved and the prospect of optimization algorithms are pointed out.