Machine learning-based approach for modelling elastic modulus of woven fabrics

Abstract

There has been a shift of focus from aesthetic properties to mechanical and functional properties of textiles with the recent developments in technical textiles and wearable technology. Therefore, understanding how various fabric parameters influence the mechanical properties of fabrics is paramount. In applications where compression and stretching of fabrics are important, the elastic modulus is a key fabric property that needed to be controlled precisely. Woven fabrics are capable of providing superior elastic properties, but how various fabric parameters affect elastic modulus is not well understood. In this study, two machine learning techniques were implemented to model the elastic modulus of woven fabrics and were compared with multivariable regressions. The two machine learning techniques used are Artificial Neural Network (ANN) and Random Forest Regression. As input variables; weave factor (numerical representation of weave structure), warp yarn count and pick density were used. Both ANN and Random Forest Regression were able to generate reasonably accurate results with Random Forest Regression been the better of the two methods. Using Random Forest Regression, feature importance of the input variables was obtained, and it proved that the weave structure has a notable impact on the elastic modulus of woven fabrics.