Font Size: a A A

A Soft-sensing Model For Slashing Quality Indexes Using ANFIS Based On Non-Euclidean Distance Clustering

Posted on:2016-06-11Degree:MasterType:Thesis
Country:ChinaCandidate:X DongFull Text:PDF
GTID:2181330467990225Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
Theslashing process is a key textile production process. Slashing quality indexes directlyinfluent textile cost and the quality of product. The detections of slashing quality indexes arebeneficial to improve the quality of fabric and reduce the textile cost. Size add-on is one of theimportant quality indexes. The stability of size add-on plays an important impact on thequality of fabric and slashing quality. To ensure the quality of textile production, thedetections of slashing quality indexes become a kind of problem to be solved. So it isimportant to build a slashing quality indexes soft-sensing model.Three parts are mainly completed as followed:(1) The development situation domestically and abroad of soft-sensing model aresummaried. The slashing process is described and the key factors of influence on size add-onare pointed out. The relations between the process parameters and size add-on are analyzed bystatistical analysis software SPSSV13.0.(2) For some problems of adaptive neural fuzzy inference system based on uniform meshpartitioning, for example, fuzzy rules are so much, the model’s calculation is complex andanti-interference capability is weak. So the adaptive neural fuzzy inference system based onnon-Euclidean distance clustering is proposed, the non-Euclidean distance clustering methodis used to divide the input space instead of the original division method, and the hybridlearning algorithm is used to estimate the model parameters. The model has many advantages,such as the little fuzzy rules, the short convergence time, the high training precision, the stronganti-interference capability. The simulation results show that the adaptive neural fuzzyinference system based on non-Euclidean distance clustering has much more advantages thanadaptive neural fuzzy inference system based on uniform mesh partitioning under the noiseenvironment, such as the smaller error, the faster convergence speed, the less effect from noiseand the better approximation effect. So the adaptive neural fuzzy inference system based onnon-Euclidean distance clustering is more suitable to deal with the the nonlinear systemmodeling problem.(3) The adaptive neural fuzzy inference system based on non-Euclidean distanceclustering is used to establish size add-on soft-sensing model, the non-Euclidean distance clustering method is used to divide the input space, the number of optimal cluster and numberof cluster center are determined, the hybrid learning algorithm is used to estimate modelparameters. The validity of size add-on soft-sensing model based on adaptive neural fuzzyinference system which based on non-Euclidean distance clustering is proved. The simulationresearch respectively uses adaptive neural fuzzy inference system based on uniform meshpartitioning, BP neural network, RBF neural network to establish size add-on soft-sensingmodel, the performance of models is compared. The simulation results show that size add-onsoft-sensing model based on adaptive neural fuzzy inference system which based onnon-Euclidean distance clustering has the fastest convergence rate and the highest modelprecision, which is more suitable to calculate size add-on.
Keywords/Search Tags:Soft-sensing model, Adaptive neural fuzzy inference system, Non-Euclideandistance clustering, Slashing processes, Size add-on index
PDF Full Text Request
Related items