eduzhai > Applied Sciences > Engineering >

Residual Generation Using Physically-Based Grey-Box Recurrent Neural Networks For Engine Fault Diagnosis

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Data-driven fault diagnosis is complicated by unknown fault classes andlimited training data from different fault realizations. In these situations,conventional multi-class classification approaches are not suitable for faultdiagnosis. One solution is the use of anomaly classifiers that are trainedusing only nominal data. Anomaly classifiers can be used to detect when a faultoccurs but give little information about its root cause. Hybrid fault diagnosismethods combining physically-based models and available training data haveshown promising results to improve fault classification performance andidentify unknown fault classes. Residual generation using grey-box recurrentneural networks can be used for anomaly classification where physical insightsabout the monitored system are incorporated into the design of the machinelearning algorithm. In this work, an automated residual design is developedusing a bipartite graph representation of the system model to design grey-boxrecurrent neural networks and evaluated using a real industrial case study.Data from an internal combustion engine test bench is used to illustrate thepotentials of combining machine learning and model-based fault diagnosistechniques.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×