• DocumentCode
    3230346
  • Title

    An ANS based helicopter transmission diagnostic system

  • Author

    Xu, Xiaoshu ; Vanderveldt, Hans ; Allen, Robert

  • Author_Institution
    American Joining Inst., Knoxville, TN, USA
  • Volume
    4
  • fYear
    1997
  • fDate
    9-12 Jun 1997
  • Firstpage
    2434
  • Abstract
    From the development effort for the CH-46 Helicopter Aft Transmission Fault Detector project the following can be concluded. The optimized-entropy training algorithm, successfully configured and optimized an ANS for processing vibration signals, albeit test signals from a test stand transmission with implanted faults. The ANS successfully analyzed the test data with, no false alarms, no failures to detect, and no misclassifications of original test data. Additionally, no false alarms, and no failures to detect and a very small percentage of misclassification rates (0.16%) on an expanded test data set. It is worth noting that diagnostic systems in use today are reported to have false alarms rates as high as 40 percent. The total detection time needed for classification to occur per instant is less than 0.002 seconds. When combining each instant with others into a time buffer the final classification always takes less than 1 second. The development of an ANS for use as a helicopter transmission fault detector diagnostic system is both feasible and has a high probability of success
  • Keywords
    fault diagnosis; helicopters; learning (artificial intelligence); neural nets; pattern classification; time series; CH-46 Helicopter Aft Transmission Fault Detector project; classification; detection time; helicopter transmission fault detector diagnostic system; optimized-entropy training algorithm; vibration signals; Accelerometers; Acoustic noise; Acoustic sensors; Artificial neural networks; Costs; Fault detection; Fault diagnosis; Helicopters; Sensor systems; Testing;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Neural Networks,1997., International Conference on
  • Conference_Location
    Houston, TX
  • Print_ISBN
    0-7803-4122-8
  • Type

    conf

  • DOI
    10.1109/ICNN.1997.614461
  • Filename
    614461