Prediction of Crimp in Cotton Woven Fabrics by Machine Learning Techniques

Abstract

Textile industry is contributing greatly to the country's economy by having share of 55% in the total exports and by providing directly & indirectly employment to about 35% of the total workforce of Pakistan. Weaving is a major segment of textile industry contributing 7.5% to the global cotton fabric exports. Woven fabrics are produced by the interlacements of warp and weft yarns. These interlacements of yarns causes the yarns to follow a wave like path. Therefore, extra length of yarns is required to get a certain fabric length and width. This waviness of yarns in fabric is expressed in terms of yarn crimp or fabric shrinkage. The amount of shrinkage or crimp depends on the fabric structural parameters like yarn linear density, yarn density in fabric (EPI, PPI), oat length, GSM, reed count and reed denting. The production cost along with physical and mechanical properties of fabric vary with the amount of crimp percentage in warp and weft yarns. Crimp percentage in yarns is analyzed by physically measuring the extra yarn length or by predicting it on the basis of fabric structural parameters. Previously developed methods of crimp analysis can be inaccurate, strenuous, tedious, and involve high computation cost. These issues have been addressed in the proposed framework that uses machine learning algorithms that are random forests, KNN, kernel tricks, adaBoost and neural networks predicting models. in the framework, the trained models were cross validated and have a prediction accuracy (R2) of 0.86 and 0.79 for warp and weft yarn crimp respectively. The nally trained models have also been evaluated on unseen datasets that are from industry and related literature. The proposed framework has prediction accuracy (R2) for warp & weft yarns crimp of 0.99 & 0.81 respectively for the industrial data set and of 0.99 for both warp and weft crimp for the related work dataset. Presently the framework has been trained only for fabrics made up of cotton carded yarns on air-jet loom. This framework can be further enhanced for fabrics made of other type of yarns and looms.