The paper focuses on the development of a predictive numerical tool for the assessment of the filling performance of engine lubrication systems. Filling analyzes are typically carried out by means of multi-phase 3-D CFD models but, despite allowing detailed and reliable results, they require very demanding computational requirements. On this basis, a procedure for the lumped parameter modelling of the fluid domain is proposed, allowing the discretization of complex systems that cannot be straightforwardly attributable to elementary submodels. The presented criteria are then applied to the lubrication system of a heavy-duty engine, for which the filling of the circuit plays a fundamental role. Different temperature conditions are simulated, and the predictive capabilities of the numerical model are presented in terms of flow pattern and filling time of the circuit branches. The same simulations are also carried out by means of a 3-D CFD model, permitting a result comparison. The comparative analysis concerns both the overall distribution of the lubricant over time, and the local phenomena within the oil domain, in order to assess the approximation of the lumped parameter approach with respect to the more accurate three-dimensional models.