The Predication Of Ramie Yarn Qualities

Due to the multi-process of yarn production and complicated physical structure of the yarn, there is a complex nonlinear relationship among the fiber quality, the parameters of the spinning and the yarn quality in the process of spinning. It is difficult to describe the relationship clearly in an ordinary mathematical model. So it is necessary and meaningful to find a more appropriate way to describe the relationship.The topic of this dissertation is that I will do further study on yarn quality prediction for ramie spinning based on the former’s research in yarn quality prediction for cotton and wool spinning.In this paper, the main research contents and conclusions are as follows:Firstly, feature selection method is applied to analyze the influence of fiber properties on yarn quality by use of cotton known data.(1) The feature selection method in this paper include principal component analysis, grey relational analysis, rough set theory and WEKA feature selection.(2) In cotton spinning, the strength, length, micronaire value and length uniformity of fiber are the most important factors to affect the main performance of cotton yarn(yarn strength, yarn unevenness, yarn strength irregularity). So we need to focus on these four fiber properties in cotton assorting in order to improve the yarn quality. The analysis results of two groups known data are basically the same, and also conform to the actual producing, which show that using the selected feature selection methods to analyze the effect of fiber properties on yarn quality is feasible.Secondly, the feature selection methods are applied in ramie spinning, and then combined with back-propagation neural network to predict the ramie yarn quality.(1) In ramie spinning, the fineness, string and breaking strength are the more important factors to influence the main performance of ramie yarn (yarn strength, yarn strength irregularity, yarn unevenness, yarn neps).(2) In this paper, a combination of feature selection method and back-propagation neural network model is applied to predict the ramie yarn quality. The most important performances of ramie fiber and combed sliver are selected by the use of feature selection method as the input variables of back-propagation neural network, this approach can optimize the network and then improve the prediction accuracy. And then the data collected from the factory are used to validate the model. Compared with pure back-propagation neural network, the combination of feature selection method and pure back-propagation neural network are better, the mean relative error between the predicted results and measured results are all less than10%.Thirdly, the calculation of the grey relational analysis is simple, and the result is accurate, so the grey relational analysis combined with back-propagation is applied to set up models of the products’performance in the process of carding, combing, finishing drawing, danny roving and spinning.The grey relational analysis is applied to analyze the effect of former process performance on the back process performance, and then combined with back-propagation neural network to set up the forecast models of the products’performance of the carding, combing, finishing drawing, dandy roving and spinning process. The mean relative error between the predicted results and measured values are all lower than10%, which show that these models are available in practice.