Using a Multivariate Gaussian Mixture Model With Expectation Maximization to Identify Characteristic Bursting Strength in Woven Hemp Fabrics

Abstract

This paper proposes to identify the strength characteristics for a particular woven hemp fabric from a collection of data representing the strengths derived from bursting strength testing based on moisture content. The Ball Bursting Strength Test, D3787 and ASTM 6797, define the size of puncture tool and the speed of force application for the bursting test procedure. The bursting strength test is a method of defining the strength of a woven fabric in two directions simultaneously given a single force perpendicular to the fabric surface. Plotting the resultant bursting force against the apparent modulus of elasticity for each sample, sets the variance in strength for the elastic range against the variance for the elastic range. The amount of variance of any particular data point from an overall group mean will help identify its association with a group of data points all belonging to a common family of test samples. Recognizing that a particular data point is likely to belong to a group of data points and less likely to belong to another group of data points given the parameters of variance and mean for any group of points, is the function of the Gaussian Mixture Model with Expectation Maximization (GMMEM).