Adaptive fault diagnosis in rotating machines using indicators selection

Didier Rémond (1) , Ilyes Khelf , Lakhdar Laouar , Abdelaziz M. Bouchelaghem , Salah Saad

1 LaMCoS - Laboratoire de Mécanique des Contacts et des Structures [Villeurbanne]

Over the past two decades, condition monitoring and faults diagnosis in rotating machinery have been widely studied and reported. In the present paper an algorithm for fault diagnosis in industrial rotating machines facing new operating conditions emergence is developed on the basis of input indicators, extracted from vibrations spectrums. Indicators selection is used to improve diagnosis performances by the help of a hybrid approach using several selection criteria and different classifiers. To validate the performances of this algorithm, experimental tests were conducted on two industrial systems with various operating conditions. The results have proved the effectiveness of the developed algorithm compared to the "J48 decision tree" and also reveal the need to re-select the indicators for reliable monitoring of working conditions.

Mots clés: SUPPORT VECTOR MACHINE; DECISION TREE; NEURAL-NETWORKS; ROLLER BEARING; VIBRATION SIGNAL; CLASSIFIER; SYSTEM; ALGORITHM; FEATURES; SPEED

Domaines: Mécanique des structures [physics.class-ph]; Mécanique des structures [physics.class-ph]

Didier Rémond, Ilyes Khelf, Lakhdar Laouar, Abdelaziz M. Bouchelaghem, Salah Saad. Adaptive fault diagnosis in rotating machines using indicators selection. Mechanical Systems and Signal Processing, 2013, 40 (2), pp.452-468. ⟨10.1016/j.ymssp.2013.05.025⟩. ⟨hal-00905082⟩