Coiled sheet metal often exhibits shape defects that result from the rolling process or the coiling operation. To meet the quality requirement, these defects need to be removed using leveling and straightening. The process efficiency strongly depends on several parameters like the machine design, the rollers penetration and the sheet metal. Consequently, the leveling process is very sensitive and it is difficult to find the appropriate setting using trial and error procedure. In this context, numerical simulation can be very helpful. The aim of our work is to predict the residual curvature of the sheet knowing its initial shape and the leveling process settings. The simulation is carried out in two steps to integrate the global and the local behavior of the strip along the leveling process. In the first step, a 2D finite element model is used to predict the sheet metal deformations under the rollers action. In this first step the strip curvatures along the leveling machine are predicted. The so obtained results are then used to simulate the alternating bending and the spring back of the strip with the help of a semi-analytical model using the MATLAB programming environment. To validate the proposed approach, leveling tests were carried out on a 2.5 mm thickness sheet of DX51 steel and the measured residual curvatures are compared with the predictions. These comparisons show that satisfactory predictions can be obtained with good computational efficiency.