Research On Efficient Experimental Strategies Of Complex Simulation System

There exist multiple and numerous uncertainties in the complex simulation systems, which leads to an increase in the complexity of the simulation experiment. So based on the study of existing simulation methods, the reaseach of this paper is carried out at the aim of reducing the number of experiments and improving the efficiency of simulation experiment.Firstly, starting from the principle of the method, this paper introduced Monte Carlo method. Concentrating on the the idea of reducting variance to improve efficiency, several Monte Carlo variance reduction methods are studied and compared. According to the idea that combination methods are better than a single method in reducing the variance, a new method, called “efficient simulaiton experimental methods based on combining sampling stategies ”, is proposed then. It organically combines three methods including variance minimization, Latin hypercube sampling and antithetic random variable, and is able to effectively reduce the number within given relative error of experiments.Secondly, considering that optimal Latin hypercube sampling methods are widely applied and studied, and they show great effectiveness in improving the efficiency. Several basic optimization criterion and a particular optimization algorithm are studied. Then based on "Replicated optimal Latin hypercube method" this paper proposed "Extented optimal Latin hypercube method based on ML2-ESE". Through the thought of "extension", the proposed method can not only make up for the former one in sample waste and improve efficiency, but also keep the excellent properties of seqentially increasing samples.Finally, a tool of efficent experimental methods for complex simulation system is designed and developed. And a practical simulation example is solved by using this tool, which verifies above two proposed methods in the improvement of efficiency.