Loading...

Speed Estimation for Fault Diagnosis in Machinery Using Vibration Signal

Izanlo, Hassan | 2023

81 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55905 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Behzad, Mehdi; Arghand, Hesamaddin
  7. Abstract:
  8. Nowadays, condition monitoring and fault diagnosis are considered as one of the most important maintenance and diagnosis strategies in state-of-the-art industries. Serious damage such as machine and coupled equipment failure, machine failure in a neighborhood in case of severe damage, and production loss will be resulted in late fault detection. One of the critical information in condition monitoring is the machine's rotating speed. In many cases of condition monitoring, the rotating speed, and the machine's technical information (such as bearing characteristic frequency )is not available. In this study, a ruled-base model and an intelligent model are developed for machine running speed estimation and an intelligent model is develoed for fault detection of machinery without using technical information. Ruled-base model use the signal processing techniques and investigation of harmonics. The intelligent model is based on convolutional neural networks (CNNs). Scalled frequency domain and amplitudes are the entrance of the CNN-base model. This method indicates the probability of relating a candidate to running speed. The approach of intelligent fault detection is based on detection of impulse (impact) pattern in raw time signal. The impulses are related to rolling element bearing (REB) faults. The artificial time signal with wide range of running speed, natural frequency, damping and fault frequency are generated. The CNN model is trained by combination of artificial and laboratory signals and its performance is tested by laboratory and industrial signals.The accuracy of the ruled-base method is 93% and the test accuracy of CNN-based model of speed estimation is 91%. The test accuracy of the CNN-based model for fault detection of experimental data is 96% and for industrial data is 89%. both of speed estimation models are efficient in harsh noisy industrial environment and also usefull in the case of slowly varing speed machinery. The performance of intelligent fault detection model is considerable and the model is efficient for fault detection of REB without technical information
  9. Keywords:
  10. Rolling Bearing ; Convolutional Neural Network ; Speed Estimation ; Intelligent Fault Detection ; Impulse Pattern ; Harmonics Investigation

 Digital Object List

 Bookmark

No TOC