Partager

Document IIF

Une nouvelle méthode d'inspection de la qualité des compresseurs basée sur les méthodes Deep SVDD et CWGAN-XGBoost

A novel quality inspection method of compressors based on Deep SVDD and CWGAN-XGBoost.

Auteurs : WANG J., JIN X., LYU Y., JIA Z.

Type d'article : Article de la RIF

Résumé

To ensure the quality of manufactured compressors, the quality inspection should be performed before they leave the factory. The quality inspection process consists of two steps: quality assessment and fault diagnosis. However, the traditional fixed threshold range setting method used for quality assessment cannot fully consider the correlation among the operating parameters, and the test data of faulty compressors are often inadequate for fault diagnosis. To address these two issues, a novel quality inspection method based on Deep Support Vector Data Description (SVDD) and Conditional Wasserstein Generative Adversarial Nets (CWGAN)-Extreme Gradient Boosting (XGBoost) is presented. The Deep SVDD model is setup and trained for compressor quality assessment, while the CWGAN model is used to generate fake faulty samples, and the generated faulty samples are then used to train the XGBoost model for fault diagnosis. The presented method is validated by quality record data of two types of compressors, and the results show that the Deep SVDD model can significantly improve the accuracy of compressor quality assessment. Comparing with the fixed threshold range setting methods, the quality assessment accuracy for these two types of compressors increases by more than 3.77 % and 3.12 % respectively. The validation results also show that CWGAN model can generate suitable faulty samples for training XGBoost model, and the trained XGBoost model can accurately diagnose the manufacturing faults. The fault diagnosis accuracy for these two types of compressors is 81.43 % and 81.9 % respectively.

Documents disponibles

Format PDF

Pages : 159-171

Disponible

  • Prix public

    20 €

  • Prix membre*

    Gratuit

* meilleur tarif applicable selon le type d'adhésion (voir le détail des avantages des adhésions individuelles et collectives)

Détails

  • Titre original : A novel quality inspection method of compressors based on Deep SVDD and CWGAN-XGBoost.
  • Identifiant de la fiche : 30032096
  • Langues : Anglais
  • Sujet : Technologie
  • Source : International Journal of Refrigeration - Revue Internationale du Froid - vol. 157
  • Date d'édition : 01/2024
  • DOI : http://dx.doi.org/10.1016/j.ijrefrig.2023.11.005

Liens


Voir d'autres articles du même numéro (17)
Voir la source

Indexation