Evaluation And Prediction For Knit Inner Wear Comfort Property Based On Artificial Neural Network

This thesis introduces firstly the principle of procedure when heat and moisture transfer from body to outside through clothes based on clothing comfort definition. From clothing material, the basic construction and property of yarn and fabric influence heat and moisture transferring. How to influence is analyzed.It is very important to research how to evaluate the clothing comfort. Generally evaluation method can include experiment measuring, human body test, and all kinds of data processing. As the technology develop, all kinds of evaluation also improve.According to compare all kinds of evaluation, we introduce a new date processing method-Artificial neural network. This is developed lately and advanced technology with many subject mixed. It can simulate human brain and include a lot of nerve cell. It can solve problem with only a lot of original date and exact model is not needed. And also it can learn itself and adjust continually. At end, it get the essence among the appearances and describe as the inexactly input and output. It features self-adjusted, tolerances, imagination and remembering. Artificial neural network show bigger prediction ability than other traditional data processing method.Inner wear touch human skin directly. People realize the importance gradually. These theses research how to use artificial neural network to predict and evaluate clothing comfort property according to knit inner wear.20 kinds of knit are chosen to test the basic construction and heat-moisture property. They include basic indexes (fabric density, dry and wet weight, thickness, loop length), basic heat-moisture property (heat conductivity, moisture regain) and heat-moisture property (heat retention ratio, sensation of cold, moisture transfer under obvious sweat condition, average wick height, air permeability impedance). We take the basic construction and heat-moisture property as the input parameter and take the heat-moisture as the output parameter, and then a model is established. After training in period of time, then we can use this model to predict and evaluate. Result show that test value match prediction value and artificial neural network can be available to predict and evaluate the clothing heat-moisture comfort.