Third-order spectral techniques for the diagnosis of motor bearing condition using artificial neural networks

摘要: This paper addresses the development of a novel condition monitoring procedure for rolling element bearings which involves combination signal processing, analysis and artificial intelligence methods. Seven approaches based on power spectrum, bispectral bicoherence vibration analyses are investigated as pre-processing techniques application in diagnosis number induction motor bearing conditions. The conditions considered normal with cage inner outer race faults. methods bispectrum, bicoherence, bispectrum diagonal slice, summed bicoherence. Selected features extracted from signatures so obtained these used inputs to an neural network trained identify Quadratic phase coupling (QPC), examined using magnitude biphase, is shown be absent system it therefore concluded that structure results inter-modulation effects. In order test proposed procedure, experimental data rig develop example diagnostic system. Results show can diagnosed high success rate, particularly when signatures.