<p>The rotor of the turbopump is a critical component of the rocket propulsion system, which is responsible for realizing the pressurization and delivery of rocket propellants, while the operating condition of the pump rotor has a direct impact on the performance and reliability of the engine. Unbalancing fault is the main problem that triggers and promotes other faults and affects the operating state of the pump rotor. This paper focuses on the vibration responses caused by unbalancing faults. The test data of vibration responses can be obtained by adjusting unbalance parameters of the rotor system, which is regarded as the core foundation for the BP neural network model training. By training the model with the data, the mapping relationship between vibration responses and unbalanced faults is approximately fitted. Based on this mapping relationship, the unknown unbalance parameters can be directly calculated by measuring the vibration responses of the rotor system in subsequent steps, so as to achieve accurate fault identification. In this way, we propose a novel type of unbalanced fault detection method that can reverse-identify the unbalanced parameters from the vibration response of the pump rotor. The simulation results show the accuracy of the proposed method.</p>