基于BP神经网络的涤纶水刺非织造布孔径及其分布预测

Predicting pore size and its distribution of polyester spunlaced nonwoven based on BP neural network

以涤纶短纤维为原料,以水刺压力和水针作用距离的不同组合制备30种涤纶水刺非织造布,采用泡点法测取其孔径,探讨制备工艺参数对涤纶水刺非织造布的孔径及其分布的影响规律;以水刺压力和水针作用距离为输入,建立基于BP神经网络的模型对涤纶水刺非织造布的孔径及孔径变异系数进行预测。结果表明:随着水刺压力的增大,孔径及其变异系数均呈减小趋势;随着水针作用距离的增大,孔径及其变异系数均呈增大趋势;BP神经网络经过2000次迭代运算后,其性能函数值最小误差达0.010563,测试样本的网络输出值与网络目标值的相关系数达到0.99478;基于BP神经网络的预测,涤纶水刺非织造布的孔径预测值与实测值之间的绝对百分比误差的平均值仅为2.78%,孔径变异系数预测值与实测值之间的绝对百分比误差的平均值仅为3.14%;BP神经网络模型的预测准确度非常高,可以用于涤纶水刺非织造布孔径及其分布的预测。

Thirty kinds of polyester spunlaced nonwovens were prepared with polyester staple fiber as raw material under different combinations of spunlace pressure and water needle action distance.Their pore diameters were measured by bubble point method,and the influence of preparation process parameters on the pore size and its distribution of polyester spunlaced nonwovens was discussed.The model based on BP neural network was established to predict the pore size and its variation coefficient of polyester spunlace nonwovens with the input of spunlace pressure and the water needle action distance.The results showed that the pore size and its coefficient of variation decreased with the increase of the spunlace pressure and increased with the increase of the water needle action distance;after 2000 iterations of BP neural network,the model had the minimum error of performance function value of 0.010563 and the correlation coefficient between the network output value and target value of the test sample up to 0.99478;based on the prediction of BP neural network,the average of the absolute percentage error between the predicted pore size and the measured pore size of polyester spunlaced nonwovens was only 2.78%,while the average of the absolute percentage error between the predicted value and the measured value of the pore size variation coefficient was only 3.14%;and the BP neural network model was very high in prediction accuracy,which can be used to predict the pore size and its distribution of polyester spunlaced nonwovens.