Modeling of cotton yarn hairiness using adaptive neuro-fuzzy inference system

This paper reports the modeling of cotton yarn hairiness using adaptive neuro-fuzzy inference system, combining the advantages of both artificial neural network and fuzzy logic. Three cotton fibre properties, namely mean length, short fibre content and maturity, measured by the advanced fibre information system and the yarn linear density (English count, Ne) have been used as the inputs to the model. Two levels of membership function have been considered for each of the four inputs and sixteen fuzzy rules are trained. The developed model predicts the cotton yarn hairiness with average error of around 2% even in the unseen test samples. Trained fuzzy rules give good understanding about the role of various input parameters on the cotton yarn hairiness. Yarn count and cotton fibre mean length are having major role in determining the yarn hairiness. Higher cotton fibre maturity reduces the yarn hairiness.