摘要
为了准确测算一定范围内的毛竹导热系数,同时改进现有的竹材导热系数研究方法,采用激光闪光法精确测量毛竹导热系数值,并以此为基础,建立毛竹导热系数随不同温度和密度变化的神经网络预测模型。由于原始BP算法收敛速度慢,使用Trainlm函数训练神经网络模型,确定了最佳隐层神经元个数,并对该模型的输出预测值进行线性分析及误差分析。实验结果如下:毛竹导热系数神经网络模型具有很高的预测精度,能准确预测一定条件范围内毛竹的导热系数,从而节省了以往常规试验所花费的大量时间和资源。本研究初步揭示了毛竹导热系数随温度、密度等因素的变化关系,为进一步研究毛竹热物理特性提供了理论依据。
In order to get the accurate thermal conductivity of Phyllostachys pubescens to a certain extent and to improve existing research methods of bamboo’s thermal conductivity,the author adopted laser flash method to accurately measure the thermal conductivity of Phyllostachys pubescens,and based on it,a prediction model of neural networks was formed,which made the Phyllostachys pubescens’thermal conductivity vary with different temperature and density.Because of the slow convergence rate of origin BP algorithm,the author made use of Trianlm function to train the prediction model of neural networks,thus obtained the ideal hidden layer neuron numbers,and analyzed the output prediction value of the model by linear analysis and tolerance analysis at the same time.The experimental results were as follows：the prediction model of neural networks was of higher accuracy which could predict Phyllostachys pubescens’thermal conductivity to a certain extent,which could save lots of time and resources compared with regular experimental methods.The article tried to reveal the relationship between Phyllostachys pubescens’thermal conductivity,temperature and density,providing theoretical evidences for the further study of Phyllostachys pubescens’thermal physical properties.
出处
《中国农学通报》
CSCD
北大核心
2011年第31期53-57,共5页
Chinese Agricultural Science Bulletin
基金
湖南省自然科学基金"毛竹竹材热力学参数的温度特性研究"(11JJ5020)
湖南省科技计划项目"毛竹热力学参数智能检测关键技术研究"(2011FJ3074)
关键词
毛竹
导热系数
神经网络
预测
Phyllostachys pubescens
thermal conductivity
neural network
prediction
