Relationship Between Fiber Properties And Rotor Spun Yarn Strength

The relationship between fiber properties,and yarn properties has been the focus of research,and considerable success has been achieved.Many mathematical models have been used to understand and predict the complex relationships between fiber parameters and yarn characteristics,and substantial research has been done to determine methods of predicting yarn properties.But the regression approach has been used more intensively for prediction of the yarn strength with an assumption of linearity.Recently,some studies have showed that relationship between yarn strength and fiber properties is nonlinear.In this study we present a comparison study of three models for predicting the strength of rotor spun cotton yarns from fiber properties.The adaptive neuro-fuzzy inference system(ANFIS)and Support Vector Machines(SVM),generally called Artificial Intelligent Techniques,that are capable of mapping non linear relations and Multiple Linear Regression models are used to predict the rotor spun yarn strength. HVI(high volume instrument)and Uster AFIS(advanced fiber information system) fiber test results are used to train and test the three models.Three important stages have been involved in this study.In the first stage,we used the ANFIS method to predict rotor spun yarn strength form fiber properties;that is,we identify the relationship between yarn strength and fiber properties.The impact of each fiber property on the rotor spun yarn strength has been analysed.The graphs illustrating the relationship between yarn strength and one of the fiber properties,with all the other properties held constant have been plotted. An examination of each graph revealed the nonlinear relationship between rotor spun yarn strength and all fiber properties.The direction of each fiber property influence on yarn strength has been also expected.Increasing fiber strength,upper half mean length,length uniformity and yarn count has a positive impact whereas increasing micronaire,yellowness and short fiber content has a negative impact on rotor spun yarn strength.Impacts of fiber properties are non-linear.The study showed also how to control the yarn quality using the knowledge on fiber properties through the yarn strength leaned surfaces on fiber properties.In the second stage,we used Support Vector Machine(SVM)model to predict the rotor spun yarn strength.Using the SVM results,we have demonstrated the relative importance of each fiber property on rotor spun yarn strength.It was confirmed that all fiber properties play a role in performance of rotor spun yarn strength.The results show that the rotor spun yarn strength is influenced to a greater or lesser degree by the fiber properties.Finally,the predictive performances of the two models are estimated and compared to those from classical linear regression method.The comparison showed that the results provided by ANFIS are better than those provided by both regression and SVMs methods.The comparison between ANFIS and SVM with conventional methods indicated that these new approaches worked better in prediction of rotor yarn strength and provided a good understanding of the nonlinear relationship between fiber properties and rotor spun yarn strength.