To investigate the fault states of automotive electronically controlled gear pumps under complex working conditions, it is necessary to achieve real-time monitoring of gear pump vibrations and extract the evolutionary characteristics of fault signals. Firstly, a dynamic mathematical model of gear pump was established, and the source and characteristics of gear pump vibration in normal operation and fault states were clarified through theoretical analysis, providing a theoretical basis for fault diagnosis. Then, based on the operating characteristics of the gear pump, an oil pump gear failure experiment platform was designed and built, which simulated the gear failure mode under various working conditions, conducted time-frequency analysis of the collected signals, and extracted the optimal indicators that reflected the fault information. parameter. Finally, conventional data processing methods and improved empirical modal decomposition of fully adaptive noise ensembles are used to compare and diagnose different fault modes under various operating conditions. The research results show that under the premise of ensuring the decomposition efficiency and fault identification accuracy, the proposed improved fully adaptive noise ensemble empirical modal decomposition method has increased by an average of 28.45% in decomposition time, and the fault of gear pumps at high speeds. The recognition accuracy rate has increased by 1.29%. The reliability and accuracy of this method are verified, and the theoretical basis and engineering reference for gear pump status monitoring and fault diagnosis technology are provided.