An MPS Algorithm Based on Pattern Scale-Down Cluster

Abstract

A key evaluation indicator of multiple-point geostatistics modeling algorithm is ensuring model quality as well as harmonizing the modeling calculation of time-consuming and space-consuming RAM. Due to the inherent flaws of SIMPAT, poor efficiency of similarity match computation between the data event and the whole pattern of train image led to the impracticability of SIMPAT many years after it was proposed. Some improvement following methods based on SIMPAT, such as Filtersim and DisPat, still did not resolve the problem. After studying key points of SIMPAT, this paper proposes PSCSIM algorithm based on a pattern scale-down clustering strategy which uses an interval sampling technique. Unlike SIMPAT, PSCSIM replaces the one-step similarity computation with the two-step similarity computation: firstly, comparing the representative patterns of the pattern cluster to the data event to find the most related pattern cluster and, secondly, matching the similarity of whole patterns in a pattern cluster with the data event to search the target pattern. With the same condition, this paper made a comparison of modeling in two dimensions and three dimensions among with PSCSIM, SIMPAT, Snesim, Filtersim, and DisPat in the end. As a result, PSCSIM greatly improves modeling efficiency on the premise of quality assurance.