This paper proposes a real-time identification method for auto-regressive with exogenous inputs and state-dependent parameters (ARX-SDP). This model is always non-linear. We conveniently adapt Young's off-line approach to an real-time approach with reduced computational cost. Young's approach focuses on discovering state-parameter dependence. This implies to unveil the nonlinear structure of the system. The proposed real-time method focuses on monitoring the non-linear state-parameter dependence, detecting structural changes and updating the model structure in real time. Finally, applications to the real-time identification of state-dependent parameters are presented and structural changes are simulated in order to test our approach.