RUNTIME FAULT DIAGNOSIS OF BEVEL GEARS USING MLP NEURAL NETWORK ALGORITHM

Speaker:
Kucuk, Haluk; Marmara Universitesi Muhendislik Fakultesi; Turkey

Authors:
Demetgül, M.; Marmara University; Turkey
Keleşoğlu, C.; Marmara University; Turkey
Küçük, H.; Marmara University; Turkey

ID: ECNDT-0435-2018
Session: Structural Health Monitoring 2
Room: H1
Date: 2018-06-12
Time: 16:00 - 16:20

Haluk Küçük, Marmara University, Department of Electrical & Electronics Engineering, Istanbul, Turkey, halukkucuk@marmara.edu.tr
Cemal Keleşoğlu, Marmara University, Department of Mechatronics Engineering, Istanbul, Turkey, kelesoglucemal@gmail.com
Mustafa Demetgül, Marmara University, Department of Mechatronics Engineering, Istanbul, Turkey, mdemetgul179@gmail.com

Bevel gear mechanisms are key components in rotating machinery. Vibration and acoustic signatures of such systems has been widely studied for fault diagnostics purposes. Here a bevel gear test setup was developed in-house and open-field sound emission together with vibration data was monitored during specific runtimes with different shaft speeds, loading, oil-level and abrasive content in the mechanism. Vibration and sound signals were recorded followed by fast Fourier Transform and Power Spectrum Density computations to extract the features used in developing a Multi Layer Perceptron (MLP) based Neural Network for fault classification purposes. It has been shown that sound data together with vibration measurements can be confidently used to predict bevel gear faults for different mechanism runtimes under different operating conditions.