Optimisation of Cotton Fibre Blends using AI Machine Learning Techniques

Fibre blend should be composed regarding the requirements and allowable price of the textile end product. Using the appropriate raw material and optimised fibre blends we can influence the mechanical properties and regularity of a yarn as well as significantly reduce the number of yarn faults. The contribution presents a study of the influence of quality characteristics of cotton fibres and constructional parameters of a yarn on the most important properties of cotton yarn. The achieved results have been used for determination of optimised cotton fibre blends regarding the quality and price of a cotton yarn. A complex procedure of cotton fibre blend determination significantly depends on suitable models for prediction of properties of resulting cotton yarns, in-depth knowledge of characteristics of cotton fibres and consideration of parameters of a production process. The results of a theoretical part of the research have been summed up in a model of cotton yarn engineering, which main components are regression model for prediction of properties of cotton ring and rotor yarns, and model for optimisation of cotton fibre blends regarding the quality and price. The regression prediction model has been designed using one of the popular artificial intelligence methods: the machine learning from examples. The prediction accuracy for forecasting the breaking tenacity, breaking extension, and regularity of cotton yarns, as well as amount of yarn faults, was significantly improved. The obtained regression trees served as a basis for realisation of a model for optimisation of cotton fibre blends regarding the quality and price of resulted yarn. Special linear programming techniques, supplemented by specific spinning technology constraints have been used for this purpose. In order to enable the comparison between the predicted and measured properties of investigated properties of cotton yarns, a new method of analytical evaluation and graphical representation of a so-called Yarn “Total Quality Index – TQI” was developed. The graphical representation has a form of a control diagram and because of its clearness provides a potential for a referential document of a modern spinning mill. Furthermore, it can be successfully used for establishing the indubitable dialogue between a spinning mill and its customers. TQI was developed based on referential information of textile fibres and yarns, included into Uster Statistics. Key-Words: cotton fibre, cotton yarn, fibre blends, artificial intelligence, machine learning