Research On Properties & Applications Of Modified Soybean Protein Fibres

Soybean protein modified fibre, one of the regenerated plant-protein fibres, can be wet-spun by using the globular protein extracted from the degreased bean dregs with the addition of some functional agents. It is an ideal and prospected textile material. Its handle is as soft as cashmere, its colour is as delicate as silk, its warmth retention is as nice as wool, its comfort is as fine as cotton.As a modern textile material, soybean protein modified fibre's structure and property have not been studied thoroughly. Reports in this field remain gaps abroad and superficial at home. Research of the structure of soybean protein modified fibre was carried out in this paper, which includes the component, the content of amino acids, the crystallinity etc. Properties of the fibre were investigated and analyzed. Series of sportswear were developed.Nowadays, technique is developed towards high-new-technique, high-quality, high-benefit and low-cost. With the development of the computer technology, great changes have taken place in the textile industry. The paper mainly deals with two problems in the soybean protein modified fibre's processing by using the computer technology.The first is the identification technique between soybean protein modified fibre and PVA fibre, because the two are similar in chemical properties and morphological structures. With the longitudinal section images of light microscopy and MATLAB toolboxes, two identification methods, support vector machine and discriminate analysis, are compared. The former is a modern identification process developed on the base of statistical learning theory, which is suitable for problems with few samples, non-linear or high-dimension. The result revealed that the correct rate of support vector machine was 100%, and its prediction process was fast and simple. It is ideal for imitation designs and quality inspections in the enterprises.The second is the prediction of yarn tensile behaviour based on the fiber property and processing parameters by using neural network technique. This was mainly done by the mathematics or experience models in the past. In the last 10 years, neural network has been used to replace the traditional method for the yarn property prediction. When neural network is involved, the experience can be gathered. If enough data are supplied, neural network model can replace the sample processing, which is a process with great consumption of manpower, material resource and time.The most important is that neural network model can solve the non-linear relationships between the fibre property, machine parameters and the yarn properties. Three methods were employed in the research for comparison, they are BP neural network, RBF neural network and the traditional linear multi-regression method. The result indicated that the RBF neural network model was considered to be with the high-precision, stable-structure and sole-result. And it was found that the more the training data, the higher the precision of the model. Spinning parameters were optimized based on the best prediction model, the lowest yarn twist, and definite yarn strength. Satisfied results were achieved and the UI is good.Compared with the traditional method, when using the neural network to control the textile processing, no output of formula or experienced data can be expected. That is the reason it gets the nickname "black box", which means one can only see the results without knowing anything in the procedure. With the self-adaptablity and mistake tolerate ability, the neural network can solve problems difficult to deal with in the traditional way. As the research continues, the neural network may see a broader prospect.