Abstract: To address the key challenges in rotating machinery fault diagnosis, including model mismatch in physics-based approaches, the excessive dependence of data-driven methods on labeled samples, and limited physical interpretability, this study proposes a fault parameter identification framework that integrates model updating, time-frequency feature extraction, and an enhanced Kriging surrogate model. The proposed framework integrates simulation and experimental information through a three-stage procedure of “mechanism alignment-data enhancement-inverse identification.” First, the finite element model of the rotor system is updated using experimental responses. Sobol global sensitivity analysis and a genetic algorithm-differential evolution (GA-DE) optimization strategy are employed to identify the structural and support parameters that have significant effects on the dynamic responses, thereby establishing a high-fidelity dynamic model capable of accurately reproducing the natural frequencies and critical speeds of the actual rotor system. Second, fault response data under different fault parameters and operating conditions are generated using the updated model to compensate for the limited availability of experimental fault data. Time-domain statistical features, as well as the amplitudes and phases of the rotational frequency and its harmonics, are jointly extracted from the vibration signals. The time- and frequency-domain information is then integrated to construct an enhanced Kriging surrogate model, enabling a rapid and accurate mapping between fault parameters and dynamic responses. Finally, the Kriging surrogate model is used to replace repeated evaluations of the computationally expensive finite element model during fault identification, while GA-DE is employed to quantitatively identify multiple fault parameters, including fault location and severity. Numerical simulations and experimental results demonstrate that the proposed rotor-system fault identification framework can effectively improve the agreement between the numerical model and the actual dynamic characteristics of the rotor system. The proposed approach achieves accurate and robust fault parameter identification while significantly reducing computational cost and improving computational efficiency. It also enhances the utilization of limited experimental information by integrating physics-based simulations with measured vibration responses, providing a useful methodological reference for intelligent fault diagnosis and condition-based maintenance of complex rotating machinery.