A novel model-free sampling on probability weighted moments

In order to improve the agreement of the statistical information of a target sample with that of observed original sample both globally and locally, a novel model-free sampling was developed, in which the constraints of the probability weighted moments and the empirical cumulative distribution function (CDF) were satisfied simultaneously. The probability weighted moment constraint guarantees the agreement of the local statistical information and the empirical CDF guarantees that of the global statistical information. Additionally, the selection of the initial target sample was improved for the convergence of the novel model-free sampling. The numerical examples demonstrate the advantages of the novel model-free sampling over the traditional one.