This work is concerned with the model re-identification of processes controlled by MPC systems. The MPC system considered here has a two-layer structure, where in the upper layer a steady-state optimization algorithm determines a set of optimal targets for the process inputs and passes this set to the MPC controller that determines the best way to drive the process to such targets. This is the case of several commercial MPC packages applied in industry. In this paper, it is proposed two internal excitation approaches aiming to obtain closed-loop data sufficiently rich for process identifiability. Here, the term internal is used to indicate that the excitation test signal is applied within the MPC control structure. In the first excitation approach, the test signal is introduced as a weighting factor in the objective function of the target calculation layer. In the second approach, the test signal is injected as a dither signal in the objective function of the dynamic controller layer. These two approaches are compared to the usual method where the excitation signal is added to the controller output. The application of the methodologies is illustrated through numerical simulations carried out on a depropanizer column of the oil industry. The results show the effectiveness of the proposed approaches and their good potential to be applied in practice.