Research On The Methodology Of Intelligent Quality Prediction And Control For Spinning

High quality and efficiency with cheap price is the competition of the international textile market.Quality is the key to enhance the core competitiveness of enterprises,so that it is seen as the life of textile enterprises.But many textile enterprises for product quality control is mainly based on the experience,such as traditional measuring and recording,checking the site and estimating the occasion and the adjustment of equipment,this situation can't meet the needs of the modern textile processing process.Textile processing is a typical complex and non-stationary product processing,So it is a trend to apply the advanced automation technology,computer technology and intellectual technology in textile production,which will make the textile quality control stable,Intelligent production is gradually becoming the main features of modern textile production.Combining with the actual production situation of the enterprises which have different requirements and production habits,research on the yarn quality intelligent forecast and control technology in the process of production has the important meaning.The main steps of the research in this paper are summarized as follows:1)The analysis with neural network helps to build the model,which is about the complex nonlinear relationship between the process parameters and the textile quality.The model adopts the genetic algorithm to complete the optimal search of the network weights and threshold space.The experiment indicates that the neural network based on genetic algorithms can improve the accuracy and stability of the yarn quality forecasting model,and select the optimized process parameters of spinning effectively.2)A knowledge discovery algorithm based on rough set theory(named RSrule)is proposed for extracting the important knowledge existing in the relationship between the process input parameters and the output product quality.Moreover,construction of the integration of rough set and neural network(KBANN)prediction model of product quality,which provides a reasonable method to determine the initial weights and number of neurons in hidden layer parameters for prediction model,and significantly improve the predictive model of learning efficiency.The proposed hybrid learning model is capable to resolve the problems,such as monitoring,diagnosis,the adjustment of process variable and so on.This research lays an important foundation for fully applications of intelligent hybrid leaning models in textile manufacturing process of quality control.3)Through extracting the time domain features from the original Processing data as the control charts characteristics,on the basis of knowledge reduction of theory in the neighborhood rough sets,the optimal selection of original features is realized;Using GA to self-adapted to the optimal parameters of SVM,which is the recognizer.Use this method to construct a model of abnormal detection in dynamic process and to achieve automatic identification of the pattern of control chart under the complicated condition.According to the change of process which control chart describes in the production,the model makes a determination whether the textile process is in control,and as far as possible adjusted according to the production rule.