摘要
在较大输入层样本数、较多输入层节点数的条件下,尝试使用单隐层BP神经网络模型与双隐层BP神经网络模型分别对精纺毛纱的条干不匀率与断裂强力进行预测,分析比较单、双隐层模型的预测性能。结果表明:隐含层节点数为9的双隐层BP神经网络模型预测性能最佳,相关系数值为0.920 5;对精纺纱的断裂强力进行预测时,隐含层节点数为8的双隐层BP神经网络模型预测性能最好,相关系数值为0.917 1。因此,在输入层样本数较大、输入层节点数较多的条件下,双隐层BP神经网络模型更适合对精纺毛纱的性能进行预测。
One-hidden layer and two-hidden layer BP neural network models are attempted to predict both unevenness value （CV） and breaking strength （BS） of worsted yarns under the condition of largescale input samples and high input dimensions. Additionally, prediction performances of one-hidden layer and two-hidden layer BP neural network models are analyzed. The experimental results show that two-hid-den layer BP neural network with 9 hidden layer nodes is demonstrated to be the best one in the prediction of unevenness value （CV）, the relative coefficient value is 0. 920 5. And two-hidden layer BP neural net- work with 8 hidden layer nodes is proved better than others in forecasting breaking strength （BS）, the relative coefficient value is 0. 917 1. Therefore, two-hidden layer BP neural network model was more suitable to predict the performances of worsted yarns on the case of large-scale input samples and high input dimensions.
出处
《浙江理工大学学报（自然科学版）》
2011年第3期347-350,共4页
Journal of Zhejiang Sci-Tech University(Natural Sciences)
基金
浙江省重大科技专项(2008C01069-3)
关键词
BP神经网络
精纺毛纱
单隐层
双隐层
BP neural network
worsted yarns
one-hidden layer
two-hidden layer
