导航

ACTA AERONAUTICAET ASTRONAUTICA SINICA ›› 2023, Vol. 44 ›› Issue (2): 426800-426800.doi: 10.7527/S1000-6893.2021.26800

Previous Articles     Next Articles

Intelligent migration diagnosis of mechanical faults driven by hybrid fault mechanism and domain adaptation

Gongye YU1,2, Weidong CAI3, Minghui HU1,2, Wencai LIU4, Bo MA1,2()   

  1. 1.Key Laboratory of Engine Health Monitoring-Control and Networking (Ministry of Education),Beijing University of Chemical Technology,Beijing 100029,China
    2.Beijing Key Laboratory of Health Monitoring and Self-recovery for High-end Mechanical Equipment,Beijing University of Chemical Technology,Beijing 100029,China
    3.Institute of Engineering Thermophysics,Chinese Academy of Sciences,Beijing 100190,China
    4.CNPC Research Institute of Safety & Environment Technology,Beijing 102249,China
  • Received:2021-12-09 Revised:2021-12-15 Accepted:2021-12-23 Online:2023-01-25 Published:2022-01-11
  • Contact: Bo MA E-mail:mabo@mail.buct.edu.cn
  • Supported by:
    National Natural Science Foundation of China(11802153)

Abstract:

The health and stability of aircraft engines play an important role in ensuring the safe operation of aircraft, and the establishment of intelligent diagnostic models with high accuracy for different engines is the key to the stable operation of aircraft. The existing fault diagnosis methods can achieve good results when fault data are available, but the actual application cannot always realize the construction of the diagnosis model because it only contains normal data. To address this problem, a hybrid fault mechanism and domain adaptive driven intelligent migration diagnosis method for mechanical faults is proposed. This method firstly establishes a virtual sample generation model for rotating mechanical faults based on the fault mechanism and source domain data, then the normal data of the target domain is used to realize the adaption of the generation model to the target domain, and finally, the target domain fault diagnosis model is obtained by virtual sample training. The proposed method is validated using standard data sets and laboratory bearing data, and the results show that the proposed method achieves an average accuracy of 88.61% when diagnosing different types of bearings, which is 41.22% higher compared with the comparison method.

Key words: fault diagnosis, mechanism model, personalized model, transfer learning, domain adaptation

CLC Number: