Development of models for the prediction of quality performance of cotton/elastane core yarn

Core yarn is a type of yarn that has a filament fiber in the center with a different fiber wrapped around it. They have a raising importance in the textile industry. That's why error free production for cotton/elastane core yarns and the design of models that can correctly predict the product quality parameters from fiber quality and spinning parameters are needed more and more. Since the multivariate data size reduces the successful prediction chance, the ways to decrease the dimension of data matrix was investigated. Principal Component Analysis (PCA) and Analysis of Variance (ANOVA) were both used to reduce the dimension of input matrix. Since cotton/elastane yarn quality performance prediction problem is non linear, Artificial Neural Networks (ANN) and Support Vector Machines (SVM) were proposed to predict the quality control parameters of core yarns. Both inputs from PCA, ANOVA and the unreduced input matrix were used to compare the power of each size reduction. The predicting models were designed using the data from a textile plant. Mean Square Error (MSE), Mean Absolute Percentage Error (MAPE) and R squared values were used as performance indicators. Neural Networks with PCA reduced input matrix was shown to have the best MSE and MAPE values. The best model has shown to have over 90% success rate in MAPE for most of the yarn quality characteristics.