Isomap and intrinsic mode function feature energy-based multi-fault recognition method of rolling bearing

Aiming at the nonlinear and non-stationary characteristics of vibrations from faulty parts, an approach was proposed to solve the problem by means of Isomap and intrinsic mode function (IMF) energy. The approach firstly conducts empirical mode decomposition to the vibration signals and calculates the energy of each IMF component to form feature vectors. Then Isomap reduces the vector dimensionality, by which bearing conditions can be visually diagnosed. And BP neural network is used to classify fault types. The approach was applied to measured vibration signals from normal condition and outer race, inner race and ball defects of rolling bearing. The results demonstrate that the approach can recognize fault type with high accuracy and high clustering. © 2016, ICIC Express Letters ICIC International.