基于模糊神经网络的面料成形性预测

Prediction of fabric formability based on fuzzy neural network

根据面料FAST伸长及弯曲性能指标与成形性之间的关系,应用基于输出空间的监督模糊C均值聚类算法确定神经网络的模糊规则数,提出一种基于模糊神经网络的面料斜向成形性客观评价系统。将试样从经向到纬向沿每10°角(含45°)进行剪裁后测试,最后分别用66组和33组数据进行神经网络的训练和仿真。实验结果证明,系统可以根据织物的不同结构与物理性能快速准确地给出该织物斜向成形性评价指标,同时最大相对误差绝对值未超过7.7%。

Based on the relationship between FAST extension,bending properties,and fabric formability,this paper applies FCM clustering to determine the number of fuzzy rule,and proposes an objective evaluation system of fabric formability by using fuzzy neural network.Fabric samples are cut every 10°（45°included） from warp-wise to weft-wise,and 66 and 33 groups of data obtained are used for neural network learning and stimulation respectively.The experimental results demonstrate that the proposed system can be used as a prediction system for fabric formability according to its weave structure and physical properties.It gives evaluation indexes quickly and accurately,with absolute value of the maximum relative error less than 7.7%.