Study On Thermal Conductivity Of Wood Based On Neural Network

The law of wood thermal conductivity is one of the most significant theoretical basis of wood application research.Wood heat transfer capacity is the key index to measure wood floor as heat transfer medium,and the thermal physical property constant is the main parameter to characterize the heat transfer law and capacity.At present,the transient plane heat source method is the most popular research method to obtain the thermal physical property constant of the target object,and the thermal constant analyzer developed based on this principle has also been applied in relevant experiments.However,due to the complexity of the algorithm used in this kind of instrument,especially in obtaining the conclusion with high accuracy,it takes a long time and takes a large amount of computing resources.So as to solve this trouble,this paper establishes a regression prediction model of wood thermophysical properties based on neural network around the calculation method of wood thermal constants.Based on the neural network model,the heat transfer law of wood is studied,which can provide a light calculation scheme for this kind of instrument,and provide theoretical basis for setting standards of wood floor heating industry.Therefore,this paper focuses on the following three aspects:First,the basic work of data regression modeling is carried out.This paper studies the computer system of the transient plane method,combs the functional relationship of each parameter,and then determines the type and quantity of the input and output variables of the model.Based on the wood specimen museum of the experimental team,130 kinds of common commercial wood specimens were selected and manufactured according to the size needs.Through the hot disk thermal constant analyzer in the team,130 kinds of experimental samples were tested repeatedly to obtain the original data.In the meantime,we should deal with the collected data to provide data basis for modeling,such as screening and filtering.Secondly,according to the data collected from the experiment and the parameters obtained from the analysis,two regression models for the calculation of wood thermal constants are established based on adaptive neural fuzzy inference system(ANFIS)and BP neural network(BP-NN).To establish the regression model of wood heat transfer law,we need to complete two aspects: one is to establish the volume specific heat calculation model of wood experimental samples,the other is to establish the thermal conductivity calculation model of wood experimental samples.In the first aspect,the adaptive neural fuzzy inference system(ANFIS)is used to model,while in the second aspect,the BP neural network is used to model.Two kinds of different neural networks are used to model considering the data quantity and data structure of the two data models,giving full play to the advantages of ANFIS in small sample regression and the advantages of BP neural network in dealing with multi input and large data quantity.We can get the conclusion that the two neural network algorithms can achieve a good balance between the calculation accuracy and the calculation resource consumption.At the same time,genetic algorithm is introduced as the parameter optimization algorithm of BP neural network,which improves the prediction accuracy and convergence speed of the model.Finally,particle swarm optimization algorithm is used as the optimization algorithm to classify the "families" of experimental wood samples based on probabilistic neural network.Because the wood samples of the first two groups of experiments involved in different "families" contain different wood species,this paper selects four "families" which contain the most wood species as the data set.According to the results of the first two experiments,the classification model was trained and tested.We can get a conclusion that the probability neural network optimized by particle swarm optimization can distinguish the species of "family" of experimental wood samples.It shows that the model established by thermophysical parameters is feasible to deal with wood classification.This paper concentrate on the regression and classification of neural networks.Through the modeling experiments in the field of wood thermal conductivity,it shows that the neural network is widely used in various fields.At the same time,the related methods and conclusions provide theoretical basis and data support for the research of heat transfer law,material selection and related equipment development of wood floor heating.