Prediction Of Thermal Conductivity Of Wood Using Fractal And Neural Network Model

The thermal conductivity of wood was studied in this thesis using several theoretical and experimental methods. A fractal model and an artificial neural network (ANN) model were proposed to predict the thermal conductivity of wood, and the experimental data was used to validate them.First, the traditional transient plane source (TPS) method was improved. The transient heat conduction differential equation of one-dimensional slab with limited thickness was solved using Laplace transform. Moreover, the short time formula that describes the temperature of the flat wood as a function of time and thickness was obtained. The experimental equipment for measuring the thermal conductivity of wood at constant heat flux was also built. Subsequently, the thermal conductivity of the samples of larch and Korean pine with different moisture contents were measured using the improved transient measurement method, and the measured results were compared with the available literature data. In addition, the other thermal properties such as thermal diffusivity and specific heat can be measured simultaneously.Secondly, a fractal model was proposed to predict the transverse thermal conductivity of wood. The porous structure of wood samples on their cross sections was observed using the scanning electron microscope (SEM) images and analyzed via the fractal theory. The box-counting dimensions of four kinds of wood samples were calculated, which are equal to 1.38, 1.49, 1.38 and 1.44, respectively. Moreover, the heat conduction process of a single cell was studied and the tangential and radial thermal conductivity of wood were obtained by the thermal resistance network method. The relationship between the thermal conductivity and the porosity was achieved, and the formula that describes the thermal conductivity changing with fractal dimension was also gained. According to the proposed fractal model, the radial thermal conductivity of wood was calculated and compared with the experimental and the available literature data.Finally, to predict the variance of the thermal conductivity versus the physical properties of wood, a model based on artificial neural network was proposed. The...