Smooth appearance analysis of the finished cotton fabric affected by structural parameters based on Hopfield neural network

Cotton as the most important natural fibre is used to produce apparel in clothing industry. The appearance of the cotton garment is an important factor for consumers. Therefore, in textile industry, the cotton fabric is treated with anti-wrinkle finishing material. In this process, the cross-linking is the main parameter that produces the resiliency in the chemical structure of the fiber and gives the crease retention to the cellulosic fabrics. According to this introduction, is the crosslinking unique parameter in the wrinkle resistance appearance of treated cotton fabric or others important parameters related to the fabric structure lead to the wrinkle resistance? Reply of this question is the first novelty of this article. For this purpose, we intend to use a novel tool in computational intelligence to model the wrinkle appearance of the treated and non-treated cotton fabric due to the structural parameters of the fabric. Therefore, the effect of the fiber diameter, yarn twist, fabric thickness and bending length parameters are evaluated experimentally and theoretically on the wrinkle grade of the fabric. In experiments, the mentioned parameters are measured for various provided fabrics. Then wrinkle degree of the samples is ratted according to the light line method. In theoretical study, a new Hopfield neural network model is presented as one of the novelties of this article. Four mentioned parameters are considered as the inputs and the wrinkle degree is obtained as the output of the developed Hopfield NN model. This model approximated a nonlinear function between the output and all four inputs. Simulation and experimental results show the effect of four mentioned inputs on the wrinkle degree of the non-treated cotton samples. Simulation of the treated wrinkle grade shows that the fiber diameter, yarn twist, fabric thickness and bending length parameters affect the wrinkle grade and the crosslinking is not any effect on cotton wrinkle grade. Modeling results show that the proposed Hopfield neural network has high learning capability, fast convergence and accuracy (greater than 98%) and finally negligible error value (smaller than of 1%), so it can be reasonably used in clothing industry.