Predicting the color index of acrylic fiber using fuzzy-genetic approach

Various methods can be utilized in manufacturing acrylic fibers; one of them is the dry spinning process. There are many parameters in this method and the relations between them are nonlinear, since the complexity of the process is high. In this study, to predict the behavior of the dry spinning process different parameters such as temperature for various sections, time, and material properties were measured. The color index of the manufactured fibers was considered as a quality index. Using statistical methods, the parameters that affect the color index the most were determined. In the next step, in order to reduce effects of noise and complexity of the patterns, the collected data were clustered into subpopulations through Kohonen neural network. Then, adaptive neuro-fuzzy inference system (ANFIS) was used to predict the color index. In order to achieving ANFIS with the highest accuracy, genetic algorithm was employed to determine ANFIS parameters. Moreover, obtained results from ANFIS were compared with the linear regression model and it was found that ANFIS can predict the color index with higher accuracy using clustering.