应用人工神经网络预测转杯纱性能

Application of artificial neural network to predict rotor-spun yarn properties

应用人工神经网络对不同混纺比转杯纱的性能进行预测。研究中使用的原料为棉、聚酯纤维、粘胶纤维和亚麻。这些纤维以不同混坊比进行随机混合,在Uster MDTA 3设备上加工成条子。然后,在一台转杯纺小样机上将这些条子纺制成388个转杯纱纱样。测试了纱线的拉伸性能、条干不匀率、毛羽和捻度。研究中采用基于误差反向传播算法的人工神经网络软件ITANET-3.7预测纱线性能。通过对不同网络参数进行优化得到了最佳的网络结构,还研究了该网络的性能。

This project attempted to apply artificial neural network for prediction of rotor-spun yarn properties from different binary blend ratios. Cotton, polyester, viscose and flax were chosen for this study. These fibers were randomly mixed according to different binary blends and those mixes were converted into slivers using Uster MDTA 3 machine. Afferwards, those slivers were spun on a laboratory based rotor-spinning machine and 388 types of yarns were produced. The tensile, irregularity and hairiness and twist values of these yarns were estimated. In order to predict yarn quality, ITANET-3.7 , which is an artificial neural network software based on back propagation algorithm, was applied in this research work. The best net structure was found by optimizing different net variables and performance of that net was studied.