Header menu link for other important links
Online automatic degradation diagnosis of gas turbine bearings based on unsupervised machine learning
Kakati P., Dandotiya D., Savanur R.
Published in American Society of Mechanical Engineers (ASME)
Volume: 2
The bearing prognostic analysis plays a critical role in improving the reliability of any rotational engine. In data-driven methods, used in such bearing prognosis, real time bearing vibrational data can be acquired from a set of sensors. In literature, many supervised learning techniques have been proposed to analyze this data. However, the labelled training data required for such a supervised technique is not always available in real life. Therefore, in this work, an unsupervised learning technique based on adaptive resonance theory 2 (ART2) has been used for bearing vibrational signal analysis. The benefit of this method is that no separate training is required for the prognosis purpose. The gas turbine is studied in this work is a GE MS 3002 used in natural gas transportation in Algeria. This method is used to study the bearing vibrational signal emitted at the high-pressure turbine side. The use of online method helps us updating the model, as new observations are available. This method also offers better performance under noisy environment. Copyright © 2019 ASME
About the journal
Published in American Society of Mechanical Engineers (ASME)
Open Access
Impact factor