Application Of The BP ANN In The Spunmelt Nonwoven Product Development

Based on current spunmelt non-woven manufacturing companies'new productdevelopment practice and demand,this paper try to utilize the advantages of BP neuralnetwork in the nonlinear mapping treatment,with the platform of MATLAB neural networktoolbox,to research and construct a BP neural network,then used it for assisting spunmeltnon-woven new product developmentFirstly this paper analyzed and summarized the spunmelt non-woven product keyprocess parameters,product performance index and their inherent relationship:overall theperformance of the spunmelt non-woven can be divided into two categories:the base fabricfeatures and additional functions from post-finishing,there is hard correlation between theadditional functions and post-finishing while the base fabric features are more related to themachine and raw material For medical and hygiene market using there are different testmethod and standard requests for spunmelt non-woven,but the most important are strengthand HSH:the strength describes the fabric'S resistance ability to the external damage,and theHSH reflects fabric'S barrier property．According to the process control practice,select basisweight,AD spunbond beam spinpump speed(in rpm),BC meltblown beam spinpumpspeed(in rpm),calander bonding temperature T and bonding nip pressure P as the 6 keyparametersBased on the analysis of product process parameters and performance,with theMATLAB neural network toolbox platform,this paper constructed the BP neural network,andtrained BP network with the company'S existing product process and performance records,thenetwork learned very well:it came to convergence in the end Base setting for the BP network:three layers,use tansig transferring function for input and hidden layer,use logsig function forhidden and output layer,16 neurons for hidden,use trainlm for training function,usecompany'S existing product records as samples to train the BP network,set training erroradO～magnitude．the network can come to convergence quickly．Subsequently,,product performance was simulated with the trained the BP network；compared with the actual data,predicting error of the product strength and HSH were within5％．it fully meet the practical requests for product development and production；Used thetrained BP network to guide the spunmelt non-woven new products development,it will bringhigher efficiency and lower cost．and ithas great commercial value for business...