Prediction For The Tactice Comfort Of Worsted Fabrics Based On Complex Atificial Neural Network

In the choice of clothing,people will have to consider the style,size,color, appearance etc.,while another important factor is comfort,of which tactile comfort is an important part.Consumers rely on the sense of touch to identify the material texture and physical properties and feel the comfort of the cloth at the mean time.Whether we can establish a prediction system of the contact comfort of clothes according to the real wearing feeling of the clothes wearer or buyer is of application significance to the consumers and of instructive significance to the textile and garment making corporations.On the basis of the research and expression of the predecessors,the contact comfort,the relationship between the comfort and the four apperceives,which are the sense of prickle,rigid,warm,cool,rough,and the factors influencing on the comfort had been explained and discussed,the background researches of contact comfort had been indicated,the fundamental significances of the research had further been stabilized and the status quo and problem remain existed in this aspect had been summed up,thus this issue in this present thesis was raised.An improved evaluation psychological yardstick was used and a group of healthy male and female students of 20-30 age-year-old were selected as the conners to obtain the evaluation of the tactile comfort of the fabrics.In view of the correlation of the contact comfort feeling and the handle,as well as the fuzzy understanding of the two feelings,independent evaluation tests of the two feelings were conducted,the relevance and their differences of the two feelings were compared and the relationship and differences between the two feeling and the above four feeling factors concerned were simultaneously discussed.On the above basis,a series of objective tests were conducted,and a forecasting model between the objective and subjective test data was established. In the forecast model,multiple regression methods were used for the first try.Due to the large number of variables and inconspicuous relationship between some objective and subjective variables,three different remove methods,namely forward,backward and stepwise variables-removing methods,were used to build multiple regression models.From the results of the forecast models,it was found out that,firstly,the variables-removing,who realized the above purpose well, could choose out the main influence factors and shied away the dependence properties of the variables and the model was efficient and had a certain fitting and forecasts significance.Secondly,although the models had the above strengths, they hadn't a high fitting and forecasting precision,and had some limitations.For this reasons,in sixth chapters of this article,a gray associated model combined with the RBF artificial neural network is used.Through the GSM method, objective experimental data was processed and selected and then these variables were used to construct a gray complex RBF neural network prediction model. Comparing the result with that of the multiple regression method and the RBFANN model,a stunning high fitting performance and good generalization forecast performance were found.The fitting precision reached 10"4,and the generalized forecast error of the subjective grade related to the feeling was below 15%.The forecast results have a certain degree of reliability and practical value.