基于GRNN神经网络的面料热阻预测模型研究

Study on thermal resistance prediction model of fabrics based on general regression neural network

热阻是衡量面料热舒适性的一项重要指标,为获得不同环境下面料的热阻值,多采用测试获得。文章通过YG(B)606G型纺织品热阻和湿阻测试仪,对不同面料在不同环境下的热阻进行测试。运用Matlab,基于GRNN(General Regression Neural Network)广义回归神经网络,使用少量输入参数,对不同环境下的热阻值进行预测。与传统的测试相比,GRNN神经网络实验量小,方便快捷、省时省力且预测结果准确性好;与BP(Back Propagation)神经网络相比,GRNN神经网络人为设定量更少,更为客观,预测结果更加准确。经Wilcoxon符号秩检验配对样本检验发现,GRNN神经网络预测值与实际值更加接近,可信度更强。

Thermal resistance is an important measurement index of the thermal comfort of fabrics. The heat resistance value of fabrics in different environments is mostly gained by testing. Thermal resistance values of different fabrics were tested under different environments through YG（ B） 606 G textile thermal resistance and moisture resistance equipment. Thermal resistance values of fabrics under different environments were predicted with a few input parameters by Matlab and GRNN（ General Regression Neural Network）. Compared with traditional test,smaller experiment indexes are needed by using GRNN. At the same time,the method is simpler,more convenient and more accurate. Compared with BP（ Back Propagation） neural network,fewer subjective indexes are needed in GRNN,so the prediction result of the model is more objective and more accurate. Wilcoxon signed rank test paired sample test result indicates that,the predictions of GRNN are more accurate and more reliable.