Predicting bursting strength of single jersey 100% cotton plain knitted fabrics using different machine learning models

Abstract

Bursting strength is an important parameter of knit fabrics. It depends on multiple factors. This study aims to determine the best machine learning model to predict bursting strength. Besides, determining the optimum GSM and yarn count to get best bursting strength using RSM (Response Surface Methodology). We tried 9 machine learning models. Among which, the XGBoost (Extreme Gradient Boosting), Random Forest and ANN models performed best having 99%, 85% and 80% R2 value, respectively. The Pearson correlation shows the most significant factors is stitch density, whereas porosity and GSM have most meaningful correlation according to scatterplot. In the same way, RSM determined the GSM and GSM square are significant for bursting strength having p-value > 0.10 (at 90% confidence level). The RSM also depicted the optimum GSM range is from 150 to 245 and yarn count is 24/1 Ne to 31/1 Ne. Again, the RSM found that the GSM, yarn count, and stitch length can influence the bursting strength significantly.