This work presents a data-driven modeling approach for discovering governing equations of the coast-down process in reactor coolant pumps. Based on the measured flow rate and rotational speed, the sparse identification of nonlinear dynamics (SINDy) is employed to identify the form of governing equations and associated coefficients from a library of candidate functions. The polynomial library allows the identified model to be seen as a generalization of the conventional momentum equations used in the coast-down analysis. The numerical results demonstrate that SINDy successfully recovers the dominant structure and contributions of the governing equations even in the presence of measurement noise. Experimental validation indicates that the identified model captures the low-flow-rate behavior, which is not adequately reflected in the conventional modeling approach. Moreover, the results demonstrate that accurate predictions of the experimental coast-down process are possible when applied to different initial conditions and inertia parameters.
Lee et al. (Fri,) studied this question.