Prediction Model for Non-woven Blending Ratio by Neural Network

Abstract

Along with the trend of automation in textile industry, an objective, correct and efficient fiber identification method has become an important indicator in having a rapid respondent system that textile industry strives for. By applying stain test and artificial neural network, a prediction model for non-woven blending ratio has been developed in this study. As shown in the study results, the correct classification of the classification model for 10 non-woven fabrics, including C/PET, C/W, C/PAN, C/N, W/PET, W/PAN, WIN, PET/PAN, PET/N and PAN/N, developed by adopting CIE XYZxy values was nothing less than perfection. Then, by further adopting CIE XYZxy color values, 10 prediction models for non-woven blending ratio were established individually. The high correlation coefficient of more than 0.99 between the actual blending ratio and its predicted value of each model clearly indicated the strong fitted ability of all models.