Non-Vibrational Fault Analysis of Turbojet Engine Bearings by using Deep Neural Networks

Abstract

This paper depicts the implementation of  deep neural networks in predicting common faults of the  turbojet engine bearings by training the model with images  and processing them by designing proper Deep Neural  Network model apart from conventional vibration analysis  methods, for faster detection of bearing health and  reusability. The turbojet engines have higher main-shaft  speeds operating at elevated temperature conditions,  reducing the bearing estimated life and thus the need of  schedule maintenance. This system can identify some of  the bearing damages like cracks, dents, fatigue, fretting  and smearing conditions prevailing due to thermal effects,  high axial and radial loads over the main-shaft, propeller  shank and auxiliary systems bearings. It finally assists the  aircraft maintenance engineers and technicians to reach to  the conclusions of bearing conditions by taking pictures of  bearings from any device and fetching them to the system  for better results of bearing conditions.