The Quality Index Prediction Model Based On Data Dispersion Clustering For Slashing Process

The slashing process is a critical process which raises the loom efficiency, increasesthe economic efficiency. The object of slashing is to prepare the warp which can stand thestresses, strains and abrasive forces that are acting on yarn is impregnated with the pastemainly composed of adhesives and lubricants or softeners to increase the tensile strength,better fiber-lay and resistance to abrasion. The slashing process employs complex physical,chemical and thermal processes. Due to the complexity of its manufacturing process, somekey quality indicators in slashing process (such as moisture regain, size add-on, elongation,etc.) cannot be timely and accurately predicted, and these quality index play an importantrole for sizing product quality (especially the first grade product rate), therefore, timely andaccurate predicting quality indicators is one of the key problems to be solved in the textileindustry and also the hot issues of domestic and foreign scholars.Based a brief introduction on sizing goal and sizing mechanism, the composition andfunctions of the various parts of slashing are analyzed, the key quality index, such as sizeadd-on and moisture regain in slashing process is given and the main factors that affect thesizing quality are analyzed.A fuzzy clustering method based on data dispersion is proposed for solving theproblem that data distribution and regularity is not strong since the variation ofenvironmental conditions, noise interference in textile slashing process. The datadispersion index is introduced for quantitative analyzing the extent on dispersion data setsand a non-Euclidean distance is constructed as the clustering objective function in order toreduce the clustering results impact of noise and isolated data, enhance the robustness ofclustering algorithms. And considering the inter-data separation and the intra-datacompactness, optimal number of clusters is obtained by fuzzy clustering validity functionbased on modified partition coefficient.The quality index prediction model based on RBF Neural Network is used forslashing process in which includes complex physical, chemical, thermodynamic changes and the usual differential equations and the problem is hardly described by empiricalformula. Firstly, RBF neural network center is obtained by proposed fuzzy clusteringalgorithm based on data dispersion, and neural network prediction model based on RBF isestablished. In order to ensure the accuracy of neural network prediction model, thedynamic update mechanism is introduced for training data. The simulation shows that theaccuracy of the established size add-on prediction model is high.