Diffusion coefficient is an important parameter used to predict the volatile organic compounds (VOC's) emission or sorption characteristics in porous building materials. Scale of the pores in building materials can be classified into macro and meso pores and the connection between them is very important to predict the diffusion coefficient. Many (theoretical and experimental) models have been proposed for the (macro-meso) pore connection. In contrast to these traditional models, a new numerical random generation macro-meso pores (RGMMP) method, based on geometrical and morphological information acquired from measurements or experimental calculations, is proposed here. This method not only reflects the stochastic macro-meso pores distribution characteristics of porous building materials but also can be used to calculate the mass (VOC) diffusion through building materials of very low porosity. The division of macro and meso pores' diameter and porosities is based on average pore diameter, mean free path and total porosity of the material. Lattice Boltzmann method along with the proposed structure generating tool RGMMP is validated with some theoretical values and then is applied for modeling and prediction of effective diffusion coefficient of wide range of porous building materials. The predicted effective diffusion coefficients are in good agreement with available experimental data. (C) 2015 Elsevier Ltd. All rights reserved.