In this thesis, fires in road tunnels with longitudinal and transverse ventilation systems are investigated numerically and experimentally. The fire smoke is simulated as a buoyant plume obtained by injecting a mixture of air and helium into ambient air. With this simplified representation, the radiation and the heat losses at the walls are not taken into account, but the model can nevertheless provide relevant information on phenomenology and data fields that can be compared to real fires. The study aims to meet various objectives, in particular increasing the efficiency of the mechanical ventilation systems and improving the safety of tunnels users in the event of fires. In the first part, experiments are conducted to measure, using Particle Image Velocimetry (PIV) system, the velocity fields induced by turbulent buoyant plumes released within a longitudinally ventilated tunnel. The aim is to study the non-Boussinesq effects (i.e. effects related to large density differences between the buoyant plume and the ambient air) on the dynamics of the momentum-driven releases and buoyancy-driven releases. Then, the effect of solid barriers, placed at the tunnel ceiling, on the propagation of smoke in fire events within longitudinally ventilated tunnels is studied. Two types of barrier are considered: small barriers" designed to be fixed in place and "large barriers" designed to be mobile in real tunnels. Experiments are carried out with and without vehicular blockage, which are modelled by blocks of different sizes and placed upstream of the source. It is found that the presence of barriers and/or blocks reduces the critical velocity, which is defined as the minimum ventilation velocity required to ensure that all the smoke remains downstream of the source, in the same direction as the ventilation flow. The reduction rate of the latter depends on the blocking rate created by the obstacles (barriers, blocks or both) located just upstream of the source. Subsequently, the effect of blockages on pressure losses inside the tunnel is investigated. It is shown that the large barriers are more effective than small ones because they reduce the critical velocity and induce less pressure losses in congested tunnels. In the second part, using a transverse ventilation system, the conditions of confinement of the smoke flow between two exhaust vents located on either side of a buoyant source are investigated. The effect of the shape and the position (with respect to the tunnel axis) of the dampers, including the specific case of full-width dampers, on the performance of the transverse ventilation system is evaluated. The extent of the backflow length beyond the extraction dampers, the confinement velocity and the stability of the smoke stratification are studied. Results show that the greater the proportion of the tunnel width the vent covers and the closer to the centre of the tunnel the vent is placed, the more efficient the ventilation system at confining the smoke to the extraction zone and ensuring the stability of the smoke stratification. The effect of solid barriers placed at the tunnel ceiling is also evaluated with transverse ventilation and it's found that large barriers can improve the efficiency of vents that do not cover the full width of the tunnel, by reducing the confinement velocity and enhancing the stability of the smoke stratification. In the last part, numerical simulations of fires in a tunnel with longitudinal and transverse ventilation are carried out using Fire Dynamics Simulator (FDS) software and Large Eddy Simulation (LES) approach. Several physical simulations are numerically reproduced to complete the interpretation of the experimental results. A good agreement is usually reached between the experimental and the numerical results. [.]