Compressive Sampling and Subspace Learning (CS-SL)

This chapter introduces a fault-diagnosis framework called compressive sampling and subspace learning (CS-SL). CS-SL based techniques combine CS and subspace learning techniques to learn optimally fewer features from a large amount of vibration data. With these learned features, a machine's health can be classified using a machine learning classifier. CS-SL receives a large amount of vibration data as input and produces fewer features as output, which can be used for fault diagnosis. Based on the CS-SL framework, the chapter introduces the following techniques: a recent fault-diagnosis framework called compressive sampling and feature ranking; fault-diagnosis framework called compressive sampling and linear subspace learning; and a fault-diagnosis framework called compressive sampling and nonlinear subspace learning. The compressive sampling and principal component analysis method receives a large amount of vibration data as input and produces fewer features as output, which can be used for fault classification of rotating machines.