Study On The Application Of Neural Network To The Prediction Of Wood Drying Quality

As an important part of the wood-processing process,wood drying can improve the physical properties of the wood,ensure the quality of the wood,reduce the transportation cost,and improve the corrosion resistance.The quality of dried wood products has a direct influence on the utilization rate of wood.Therefore,predict and study the quality of dried wood has a great significance on the development of Chinese wood industry and the construction of a resource-based country.In order to study the quality of dried wood in depth,according to the evaluation criteria of dried wood quality:wood moisture content and wood stress,neural network and support vector machine is adopted to predict wood moisture content and grade after wood drying.The paper has complete the following work:Firstly,the wood drying process and the mechanism of the drying process is researched and analyzed to determine the purpose,requirements and drying criteria of wood drying and study the influence of air temperature and humidity on wood drying.According to the mechanism of wood drying process,determine the water migration model and heat conduction model.Six sets of drying experiments with different conditions were carried out,the drying process was recorded and the drying data was collected,and the data was summarized and organized into a database for the next research.Secondly,the moisture content of the dried wood is studied.The back propagation neural network and the radial basis neural network is introduced,and the wood weight before and after drying(kg),initial moisture content and density(kg/m~3)of the wood are considered as inputs,the BP neural network and the RBF neural network were used to predict the moisture content of the wood respectively,then the probability density distribution curve was drawn.Through experimental comparison,it is found that the prediction of RBF neural network can get better results.Then,the wood drying quality grade is studied.When the moisture content is qualified,the wood drying grade is determined by the longitude change of the wood stress.The specific evaluation standard is determined by the Canadian National Wood Grade Rating Rules.Using support vector machines and probabilistic neural networks,the wood moisture content,density(kg/m~3)and bending(mm)before drying were used as inputs to establish prediction models to predict wood drying levels.Through the prediction results,it can be found that the probabilistic neural network has a greater advantage in the prediction of wood drying level.Finally,design the simulation platform of wood drying quality prediction through the MATLAB GUI visual interface,including simulation page layout and callback function program writing.Save the trained network model and design a simulation interface for wood drying quality prediction based on neural network.By directly inputting the data of the entire drying kiln,the probability density curve of the water content and the wood grade can be predicted.The neural network prediction wood drying model can be applied to the actual wood drying quality prediction,at the same time,it is convenient to operate and convenient to observe directly,so as to realize human-computer interaction.