An Improved Model-Free Sampling Technique Based on Bootstrap Methods
The distribution and parameters of the random variables is an important part of conventional reliability analysis methods, such as Monte Carlo method, which should be known fist before using these methods, but it is often hard or impossible to obtain. Model-free sampling technique puts forward a method to get the distribution of the random variables, but the accuracy of the extended sample generated by it is not enough. This paper presented an improved model-free sampling technique, which is based on Bootstrap methods, to increase the accuracy of the extended sample and decrease the iteration times. In this improved model-free sampling technique, the method of the selection of initial sample points and the generation of iterative sample is improved. Meanwhile, a center distance criterion, which considers the local characteristics of the extended sample, is added to the generating criterion of dissimilarity measure. The effectiveness of this improved method is illustrated through some numerical examples.