This thesis discusses the use of 3D finite element method (FEM) in deep excavations, especially the simulation of sequential excavation method in 3D FEM model. It is an independent research project based on the Spaarndammertunnel project in Amsterdam. This thesis only focused on the excavation part of the tunnel project. Soil parameters used in models were derived from in situ soil investigation. Cone penetration tests and boreholes were carried out along the tunnel location. The soil data was compared with Amsterdam North-South metro line database as well. The excavation plan, retaining structure and building material were developed from the preliminary design of this project. The data for soil investigation and preliminary design were provided by CRUX Engineering BV. The research focus was to investigate the advantages and disadvantages of applying 3D FEM modelling for deep excavations. Since the construction of Spaarndammertunnel starts after the completion of this graduation project, no field data was available for verification. Instead, 2D models were made for comparison and verification purpose. The soil and building material parameters were first tested in 2D spring model and 2D FEM model. The deflections of sheet pile walls in both 2D models were at similar level. The 3D FEM model adopted the model set-up tested and adjusted in 2D FEM model. Parametric study was done in order to investigate how the domain and mesh set-up influence the results in 3D FEM model. Certain geometry (50m * 18m) of excavation section was determined so that the results from 3D FEM model and 2D FEM model were comparable. The excavation method applied in 2D and 3D FEM model was normal staged excavation. The excavation was designed to be dry excavation. After the comparison between 2D and 3D FEM model, the sequential excavation method was implemented into 3D FEM model. Different excavation rates, excavation directions and lateral support design were tested in order to optimize the sequential excavation model. The responses of sheet pile walls as well as surrounding soil were recorded and compared with results from normal 3D excavation model. The parametric study of 3D FEM model showed that the geometry of the excavation zone and the mesh set-up has great influence over the response of sheet pile wall. If the excavation section was longer than 50m, the whole domain can be considered as plain strain, so that the results in 2D and 3D FEM models were almost the same. If the section length was lowered from 50m to 30m, the bending of sheet pile wall could be reduced by 20%. This is because the sheet pile wall can benefit from the corner effect in 3D scenario. It is advised to apply 3D FEM model when dimensions and shape of the excavation zone allows the model to maximize the corner effect. The mesh set-up is also crucial to 3D FEM model. Coarse mesh leads to low accuracy results that certain stress points receiving much high stress than it should. Refine mesh requires a lot more time for calculation. For future research, it is still advised to perform field test for verification of 3D FEM models. It is also recommended to test how lateral support like struts and wailing beam influence the performance of sheet pile wall, especially the corners of excavation section. For the sequential excavation method in this project, the main benefit was to control the settlement of surrounding area. The effect of limiting the deflection and bending moments of sheet pile wall by using sequential excavation was very limited. It is recommended to carry out the sequential excavation when the surrounding area is very sensitive to settlement that control of settlement is of high priorities. But for Spaarndammertunnel, the road near the tunnel is outside the influenced zone even if normal staged excavation is used. The modelling of sequential excavation requires a lot of time and effort, the execution requires even more time and effort. So the disadvantages of applying sequential excavation method in Spaarndammertunnel outweigh the advantages.