Training feedforward neural networks for fault diagnosis of ball bearing

Vibration-based condition monitoring plays important roles for early fault detection and diagnosis of expensive rotating machinery. This paper presents the application of a novel metaheuristic approach named chaos-enhanced stochastic fractal search (CFS) to train feedforward neural networks (FNNs) f...

Full description

Bibliographic Details
Main Authors: Tuan Abdul Rahman, Tuan Ahmad Zahidi, As'arry, Azizan, Abdul Jalil, Nawal Aswan, Raja Ahmad, Raja Mohd Kamil
Format: Conference or Workshop Item
Language:English
Published: Centre for Advanced Research on Energy, Faculty of Mechanical Engineering, Universiti Teknikal Malaysia Melaka 2018
Online Access:http://psasir.upm.edu.my/id/eprint/65091/
http://psasir.upm.edu.my/id/eprint/65091/1/68-73-1.pdf
Description
Summary:Vibration-based condition monitoring plays important roles for early fault detection and diagnosis of expensive rotating machinery. This paper presents the application of a novel metaheuristic approach named chaos-enhanced stochastic fractal search (CFS) to train feedforward neural networks (FNNs) for monitoring a ball bearings system. The vibration response data are analyzed using statistical methods to characterize several defects of ball bearings and generate vibration signature features. Then, a novel CFS-based FNNs approach is applied to classify these ball bearings conditions. The results show that the proposed approach produces comparable classification accuracy as parameters of the FNNs were optimized systematically using CFS algorithm.