Gear Fault Modelling by Using Acoustic Measurements and Artificial Neural Networks

Ulus Ş. , Suveren M.

5th. International Conference on Engineering and Natural Sciences (ICENS 2019), Praha, Czech Republic, 12 - 16 June 2019, pp.318-324

  • Publication Type: Conference Paper / Full Text
  • City: Praha
  • Country: Czech Republic
  • Page Numbers: pp.318-324


Rotating machine elements in mechanical systems such as gears and bearings have a major impact to maintain machine power transmission and machine working life in a healthy situation. Some mechanical failures including cracked tooth and pitting faults especially in gears may have a crucial effect on system failure and safety. However, fault condition monitoring (FCM) at rotating machine parts was studied mostly by using classical vibration monitoring in the literature. In this study, acoustic measurement results were obtained experimentally from a single stage gearbox. Experiments were performed by measuring different running speeds, loading conditions and fault characteristics under dry friction conditions of gears. According to results, an artificial neural network (ANN) design for predicting and modelling different possible faults of the system was created. Acoustic results were evaluated by using training data, test data and validation data. This work shows that ANN design represents a good accordance with experimental results and specify predictable information about fault varieties occurred at gearbox.