基于BP神经网络的棉秆重组方材性能预测研究  

Performance prediction of reconsolidated square materials of cotton stalk based on BP neural network

作　　者：盖玲[1] 宋孝周[1] 

机构地区：[1]西北农林科技大学机械与电子工程学院,杨凌712100

出　　处：《木材加工机械》2014年第2期43-46,14,共5页Wood Processing Machinery

基　　金：陕西省农业攻关项目(2012K02-16);西北农林科技大学基本科研业务费专项资金资助项目(QN2009044)

摘　　要：为了实现对不同热压工艺参数条件下棉秆重组方材力学性能的设计和预测,减少预实验的次数,本文采用正交试验方法进行了棉秆重组方材制备工艺试验,以正交试验数据为样本,利用BP神经网络构建棉秆重组方材主要工艺参数密度、施胶量、热压温度、热压时间与材料力学性能抗弯弹性模量、抗弯强度和顺纹抗压强度之间的关系模型,并利用训练好的模型对材料力学性能进行预测。结果表明:利用网络模型预测出的方材力学性能与实测结果基本相符;在施胶量10%、热压温度180℃、热压时间40min保持不变,密度值为0.6g/cm3、0.65g/cm3、0.7g/cm3条件下压制棉秆重组方材并测试其抗弯弹性模量、抗弯强度和顺纹抗压强度值,网络的预测值与实际测量值的平均误差分别为2.4%、1.94%和7.4%;利用BP神经网络模型预测不同工艺条件下棉秆重组方材的力学性能,能够产生较好的预测结果,可以弥补试验的不足。The purpose of this paper is to achieve the design and prediction of mechanical performance of cotton stalk reconsolidated square materials under the condition of the different hot pressing process parameters and reducing the frequency of pre-experiment. The preparation technology of reconsolidated square materials of cotton Stalk was explored by orthogonal test. The data of orthogonal test were used to build the relationship between mechanical performance of cotton stalk reconsolidated square materials（modulus of elasticity, bending strength and compression strength） and main technology parameters（density,glue consumption,hot-pressing temperature and hot-pressing time） through 3-layer Back-Propagation neural network（BP neural）. The optimized 3-layer BP neural models was used to predict the mechanical performance of reconsolidated square materials of cotton stalk at the given technology conditions. The result showed that the BP neural could properly predict the mechanical performance of cotton stalk reconsolidated square materials. The predicted values were obtained favorably accorded with experimental results under the given technology conditions（glue consumption10%,hot-pressing temperature 180℃,hot-pressing time 40 min, density 0.6 g/cm3,0.65 g/cm3,0.7 g/ cm3）. Under the above conditions, the average error of modulus of elasticity,bending strength and compression strength parallel to grain are 2.4%,1.94% and 7.4%,respectively. The prediction results of BP neural model can make up for the inadequacy of the experiment during the study of cotton stalk reconsolidated square materials.

关 键 词：棉秆重组方材 正交试验 BP神经网络 性能预测 

分 类 号：TS653[轻工技术与工程]