A proper understanding of the processes of friction and wear can only be reached through a detailed study of the contact interface. Empirical laws, such as Archard's, are often used in numerical models. They give good results over a limited range of conditions when their coefficients are correctly set, but they cannot be predicted: any significant change of conditions requires a new set of experimental coefficients. In this paper, a new method, the use of Discrete Element Models (DEM), is proposed in order to tend to predictable models. As an example, a generic biphasic friction material is modelled, of the type used in aeronautical or automotive brake systems. Micro-scale models are built in order to study material damage and wear under tribological stress. The models show what could be achieved by these numerical methods in tribological studies, and how they can reproduce the behavior and mechanisms seen with real-life friction materials without any empirical law or parameter.