Research of foreign fibers in cotton yarn defect model based on regression analysis

Abstract

The detecting of foreign fiber may not be very effective, particularly around the detection zone where many types of foreign fibers may coexist. In order to eliminate the fibers more effectively, a model has been established to detect foreign fiber faults in yarn. Relevant data were collected through investigation of a number of standard samples, with the length and area of foreign fibers as the independent variables, and the number of defects as the dependent variable, which were combined using linear regression theory to establish a regression equation for different fiber defects. The equations to find the regression coefficients, which include the model fitting degree, the Durbin–Watson value, the standard error, and the Cook distance, were rigorously tested, and the regression equation was eventually compiled to produce the yarn faults model. When the fiber detection equipment recognizes fibers with a foreign profile, the calculated profile fiber size is used in a corresponding regression equation which obtains the defect points and compares them with each other, so that foreign fibers which are potentially more dangerous can be identified and preferentially eliminated. In order to verify the model, spinning experiments are performed. The actual defects from the experiment are compared with the predicted theoretical defects from the equation, and the prediction accuracy was found to reach more than 95%, showing that the foreign fiber yarn faults model, which lays a theoretical foundation for foreign fiber detection, is accurate and effective.