COMPUTATIONAL EXPERIMENTS WHEN STUDYING MATERIALS PROPERTIES INFLUENCED BY "MIXTURE" FACTORS

Abstract

Short information on computational materials science is given, with the methodology of material properties fields, in composition and process coordinates, as the part of it and as the background of the study presented in this paper. One of the main means of the methodology is random scanning the whole and local fields. These tools were developed and used to solve many problems in materials science related to the properties defined by mutually independent factors. The purpose of the study presented in this paper has been to develop the tool for random scanning the fields of properties effected by "mixtures" of q components, linearly related portions of components in rangers from 0 to 1, with their sum equal to 1. In these cases, the factors domain (or subregion of it) presents the simplex. The special designs of experiments to get reduced polynomials describing the fields in simplex coordinates are used. Two procedures for generating any number of uniformly distributed points on the simplex have been developed. These points define the virtual mixtures simulated in computational experiments. The procedures were tested by scanning the fields of two rheological characteristics of lime suspension filled with "short", “medium", and "long" cellulose fibers. Experimental-statistical models in the form of reduced polynomials for effective viscosity at shear rate equal to1 s-1 and for the rate of destruction of liquid structure (parameters of power-law model of flow, K = η1 and m) obtained in previous study are used to determine the levels of these characteristic for each of simulated mixture. Computational experiments were carried out, in which the fields of η1 and m in whole simplex domain and in some of its zones were scanned, allowing the generalizing indices of the fields and different correlations between η1 and m in different zones of mixture triangle to be estimated. The developed tools, the procedures of generating random points, which would define the simulated compositions of the "mixtures", make significant contribution to the progress of the methodology of recipe-technological fields of properties and to computational materials science.