The computational fluid dynamics (CFD) has been applied to simulate heat transfer in complex and porous materials, and the governing partial differential equation (PDE) has been used to solve this kinds of models by means of numerical approach. The solution of the governing differential equations has been obtained using the Crank-Nicolson method. The main problem in these methods is lack of accurate estimation of thermal diffusivity in bulk complex fibrous materials. To overcome this problem, a novel modelling method of heat transfer in transient condition based on training of a propagation neural network has been applied to analyze heat transfer. The neural net model has been evaluated by experimental data of polyester samples of bulky fibrous layers. The results present that propagation neural net can estimate thermal diffusivity of samples with good accuracy.