Using Linear Mean and Variance Technique to Evaluate the Patterning in Random Winding Process

In random cone winding system, the patterning concept results in many problems in further processes, e.g. yarn breakage and/or early take off of the package. In current winding machines, there is no system to detect the patterning and consequently the anti-patterning system is working continuously. In this research, an online-computervision system is used to detect the concept of patterning through the linear mean and variance technique. In fact, photographs of bobbin surface are continuously taken with specified time interval. In this way, the mean or variance is calculated from columns within the image matrix, and a vector is obtained. Then again, the variance or the mean of the vector is calculated, thus a number is obtained for each matrix. Therefore, there are four methods including mean of mean (MOM), mean of variance (MOV), variance of mean (VOM) and variance of variance (VOV) to give the index number. Accordingly, a diagram of image number versus index value is plotted for each method, and a peak is perceived in place of patterning defect. By finding this peak, the moment of occurrence of defect is diagnosed, and the “patterning” can be recognized. The results show that among the four methods, the MOV is more accurate. Finally, this outcome is mathematically analyzed.