摘要
针对木材干燥过程的非线性特性,以及环境因素对木材干燥过程的干扰,造成木材干燥建模困难的问题,通过对神经网络的非线性、并行结构,学习、推理和多变量处理能力的研究,以干燥窑的加热阀开度、喷湿阀开度、排潮阀开度3个控制信号作为输入量,以窑内温度、湿度2个量作为输出量,利用时延神经网络和动态递归神经网络分别建立了木材干燥过程中的温湿度控制模型和木材干燥基准模型。并通过干燥实验进行网络训练和测试。结果表明:时延神经网络建立的木材干燥温湿度模型和干燥基准模型比动态递归神经网络的误差小、网络输出接近于真实值,能够较好的逼近实际系统。
It is difficult to obtain an ideal model for wood drying because of the influences of nonlinearity and environmental variables during wood drying. To solve this problem, neural network was applied to the modeling of wood drying due to its nonlinearity and parallel structure, and the advantages of study, deducibility, and muhiattribute. A temperature-humidity model and a wood drying schedule model were established based on time-delay neural network and dynamical recurrent neural networks, with the open angle of heater, humidifier and drain tap as input variables and temperature and humidity in dry kiln as output variables. Result indicates that the temperature-humidity model and the wood drying schedule model could well simulate the real drying system because the output was approximate to the actual value.
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2009年第8期50-52,共3页
Journal of Northeast Forestry University
关键词
时延神经网络
动态递归神经网络
木材干燥
模型
Time-delay neural network
Dynamical recurrent neural networks
Wood drying
Model
