The goal of this thesis is to improve the predictive capability of Finite element (FE) by gathering data from experimental test and implement the characteristics into the material model that is used. FE is a commonly used method to predict the mechanical behavior of materials and components during applied forces. Therefore, it’s an important part of product development since it gives an opportunity to lower the costs as well as saving resources since it reduces the number of experimental tests. The method for this thesis was to first simulate tensile tests in Abaqus and then to analyze its results. Once all the simulations were done, we replicated the simulation with experimental tests. This was done with DIC (Digital Image Correlation) to help gather data. Since the goal of this thesis is to see how the predictive capability of the FEM-simulation can be improved the results are compared and discussed to see what from the FEM-simulation matches the DIC results and what does not. This will help understand what in the material model that needs to be changed to better match the testing. DIC is a non-contact method that is used to measure deformations and strain locally over an area which results in a more detailed view of the mechanical behavior of the material. The idea of using DIC during this thesis is to sample enough valuable data and apply it to the original material model of the FE-simulations to increase the predictive capability. After the results from the experimental tests were analyzed it was clear that there were both resemblances and differences in the results, for example the Young’s modulus in the FEM-calculations was higher than it was for the experimental tests, Yield strength was lower in the FEM-calculations compared to the experimental tests, maximum load at fracture was lower in the FEM-calculations compared to the experimental tests and elongation was lower in the FEM-calculations compared to the experimental tests. The FEM-calculations were based of the assumptions that the material was homogenous but that wasn’t the case for the experimental tests. Due to the strain varying over the tests the material model could be improved by adding a statistical variation, to all the elements to give them varying mechanical properties simulate how the strain vary more correctly over the specimen.