Skip to content

Latest commit

 

History

History
5 lines (4 loc) · 1.24 KB

README.md

File metadata and controls

5 lines (4 loc) · 1.24 KB

Bearing fault diagnosis using empirical modal decomposition and deep learning

Abstract

In the industry, to avoid monetary losses, it is important to know the state of the bearings of the machines since they represent 40 % of the total of the breakdowns. One way to know if the bearings have failures is through mathematical models that analyze the vibrations of the machine in operation. These models are composed of two stages, the first consists of extracting characteristics from the vibration and the second consists of using a classifier to identify the fault. This research proposes a fault diagnosis that extracts characteristics through empirical modal decomposition and classifies the failure thanks to deep learning. The model was tested with data from CWRU seeking to diagnose 10 types of bearing failures under varying operating conditions. The results obtained show that the model achieves an average accuracy of 97.23 %, where its strength is evident in the failures in early stages. Finally, the results show that the model is an excellent technique to carry out the predictive maintenance of bearings in industrial machineries.

Keywords: empirical mode decomposition, deep learning, autoencoder sparse, bearing fault diagnosis.