In the sausage industry, ripening is considered to be the most important phase of the production process. Numerical models were successfully used to study the mass transport in many food during baking, freezing, ripening and drying process. The aim of this research was to develop two finite element models of water diffusion inside a salami, taking account of the vapour exchange phenomena at the surface. One model was based on the real fat and meat distribution, as acquired by image analysis, while a second one, more simple, considering the salami material as homogeneous and imposing an equivalent value of diffusion coefficient, based on compositional equations (parallel, series, Maxwell and Krisher) or literature data. The vapour exchange phenomenon at the surface was described using a surface mass transfer coefficient (h(m).) calculated on the basis of the well-known Chilton-Colburn analogy. The two models were compared and validated with experimental mean moisture concentration. The agreement between mean simulated and experimental values was reported in terms of determination coefficients (R-2). For all models, the R2 value is higher than 0.95, supporting that the experimental and all calculated data are in good agreement. The less good data were obtained with the simplest model, based on equivalent diffusion coefficient coming from literature (R-2 = 0.955, RMSE = 1665 mol/m(3)). Very small difference was observed between results related to structural models. The finite element model based on the real fat and meat distribution underestimated experimental results, giving results not far from structural models, but giving the best result at the end of ripening time. The deviation from experimental results probably is due to geometrical reconstruction uncertainties and particularly on binarisation triggering on a grey scale. Considering the real fat distribution, a more complex model, but not more precise results than those calculated considering an equivalent homogeneous material, were observed. (C) 2015 Elsevier Ltd. All rights reserved.