Loading...

Fault Diagnosis of Rotating Machinery using Multi-Sensor Data

Hekmat Golmakani, Saeed | 2023

66 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56016 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Arghand, Hesamoddin; Behzad, Mehdi
  7. Abstract:
  8. Vibration signals have been used for fault diagnosis frequently. These signals are captured using sensors, which are mounted on different locations of the equipment. Using data captured from different locations of the equipment, will result in roubustness of the fault diagnosis model. But sometimes the predictions of sensors mounted on different locations, happen to conflict. These conflicts among sources of information, make it hard to decide about the overal health state of the equipment. In fact, these conflicts causes a kind of uncertainty called epistemic uncertainty. A powerful mathematical framework for managing epistemic uncertainties, is Dempster-Shafer theory of evidence. An improved form of this theory is utelised in this thesis inorder to come up with a rotating machinery fault diagnosis model based on information fusion. Through this document, the mentioned model and the related prerequisites are explained in the first place. After that a measure of closeness to label (CTL) is introduced to evaluate the models. The fault diagnosis model is then used on experimental data. The corresponding results state that using fusion techniques will result in an improvement to the model in comparison to single source models. For instance, the CTL for the prediction of four sensors is 19%, 65%, 71%, 77% respectively, while it is 98% for the fusion of them. In another section, a model based on the fusion of three machine learning models is proposed for the fault diagnosis of an industrial roller beaing. Also in this case, the performance of the fusion model shows an improvement in comparison to primary models separately
  9. Keywords:
  10. Condition Monitoring ; Diagnosis ; Data Aggregation ; Dempster-Shafer Theory ; Uncertainty

 Digital Object List

 Bookmark

No TOC