Neuroevolution for bearing diagnosis
Rita Sleiman  1  , Amani Raad  1@  , Souhayb Kass  3, 2  , Jérôme Antoni  4  
1 : Faculty of Engineering I, Lebanese University, Lebanon
3 : Doctoral School of Science and Technology, Lebanese University, Tripoli, Lebanon
2 : Laboratoire Vibrations Acoustique, Univ Lyon, INSA-Lyon, LVA EA677, F-69621 Villeurbanne, France
INSA Lyon
4 : Institut national des sciences appliquées de Lyon  (INSA Lyon)  -  Website
INSA Lyon
20 Avenue Albert Einstein, 69621 Villeurbanne cedex -  France

The monitoring of machinery and especially ubiquitous bearings in all means of transport has gained importance for decades in the industry because of the need to increase the reliability of machines and reduce the possible loss of production due to failures caused by the different faults. Many of the available techniques currently require a lot of expertise to apply them successfully. New techniques are required that allow relatively unqualified operators to make reliable decisions without knowing the mechanism of the system and analyzing the data. Reliability must be the most important criterion of the operation. Artificial intelligence is the revolutionary answer in all areas of industrial control. The main goal of this paper is to propose new solutions for bearing diagnosis based on deep neural networks (DNN). However, in general the optimization of the neural network architecture is done by trial and error, and the features reduction problem is solved by using the principal component analysis. In this paper, the application of the neuro-evolution is proposed for bearing diagnosis where the optimization of the neural network topology as well as the features reduction are done by an evolutionary genetic algorithm. An application of the general procedure is proposed for real signals; that shows the superiority of the combination between neural networks and genetic algorithms for bearing diagnosis.

References

- J.Ferner & M.Fischler, S.Zarubica & J.Stucki. (2017). Combining Neuro-Evolution of Augmenting Topologies with Convolutional Neural Networks.

- B.Samantha.(2004)Gear fault detection using artificial neural networks and support vector machines with genetic algorithms

Mechanical Systems and Signal Processing, Volume 18, Issue 3, p. 625-644.


Online user: 1