The Development Of Quick Color Matching System For Colored Spun Yarn

The rapid development of colored spun yarn,brought today's textile market another new development space.Colored spun yarn with the new technology of "first dyeing,after spinning",shorten the production process after the enterprise processing,become more convenient,and has less pollution,rich colors and so on.Colored spun yarn has the features of "small batch,more varieties,huge change and change frequently",with the colored spun yarn used widely leading to consume time and labor,waste of raw materials,difficult to determine the color ratios,and in the process increased the time and cost.This paper analyzed two major problems in the process of colored spun yarn,matching consumes time and raw materials and difficult to determine the ratios.First to carry out rapid test spinning,then to optimize the SX01 fiber sliver instrument,form the quick color proofing process.After that setting up a BP neural network color training and forecasting model to establish a method to forecast color ratios.Finally,to develop an excellent and fast color matching system for colored spun yarn.In order to shorten the traditional color ratio process,save fiber raw materials,solve the problem of difficult to determine traditional proofing color ratios,provide staff the initial value ratios,and then fine-tuning,save a lot of time,energies and raw materials.The main contents and conclusions of the paper are as follows:Started with the quick proofing of colored spun yarn,different from the traditional process is that use of the SX01 fiber sliver instrument,which is the key of the quick test spinning.After the spinning process,familiar with the spinning and weaving process,and compared with the traditional carding machine,SX01 fiber sliver instrument with the relatively short mixing time,the less fiber materials,the short production process,the higher degree of accuracy is relatively.In the process of color mixing,there is a difference in weight between the input and output fibers,and a small part of the fibers become fallen fibers.After the single factor test,the three factors of fiber weight,process negative pressure and rotor speed were optimized by orthogonal test,fibers falling rate from 4.32% to 3.28%,the rate of the instrument is decreased.It is necessary to optimize the parameters of the instrument,and then it is better to use in the after process,to develop the quick color matching system for colored spun yarn finally.The establishment and achievement of quick color matching model,focusing on solve the difficult problem of matching color.First designed total of 36 sets of color ratios with three different colors,spinning and weaving based on the quick proofing of colored spun yarn to set up a sample library.After the fabric is ironed,scanned the fabrics,and 5 pieces of the same pixel size image are obtained at different positions in each group colored spun fabric to obtain the RGB eigenvalues in the pixel area of the picture.180 sets of RGB eigenvalues and the corresponding fabric color ratios constitute the neural network training and prediction input and output samples.Of which 150 groups as training samples,30 groups as testing samples,through the establishment of neural network color matching model,to determine the neural network for the input layer is 3,the hidden layer is 10,the output layer of 3,it is a 3-layer structure.Set the network parameters,enter the activation function for the tansig,the output activation function for the purelin,learning rules function for the Traingdx.,the training target for the 0.07,the maximum number of training steps for the 2000.In the MATLAB neural network toolbox training,in 1993 when the training target to reach 0.07,the final output of each test number under the five sub-sample ratios,The ratios of the group mean ratios and the design ratios has some errors,and it is necessary to evaluate the accuracy of the forecasting ratios of the fast color matching model in further.After the establishment and achievement of quick color matching model,get the network output ratios.There is some error between the ratios of network output and the actual ratio of the test samples.The model verification and result analysis are needed,That is,different test numbers to the test sample fabrics and the expected output ratios of the fabrics with the color difference analysis.The conversion of the color model uses the color difference formula of L * a * b * to calculate ?E.Among the six groups of different color ratios,the numbers 32,33,34,35 and 36 were compared with those of the original sample and ?E <1.5,basic no color difference,the number 31 Contrast ?E ? 1.5,there is a perceived color difference exists.By further analyzing the reason,the method of predicting the initial value of the BP neural network has certain practicability because the training sample data is less in the course of the experiment and the final prediction effect of the BP neural network is affected.