Complexity measure: A Nonlinear time series analysis technique for machine health monitoring

This paper presents a non-linear time series analysis technique for machine health monitoring, based on the complexity measure. As a statistical parameter, the complexity measure quantifies the randomness or regularity of a time series. After introducing the theoretical framework, numerical simulation of an analytic signal, which is mathematically formulated through a pair of Fourier transform and inverse Fourier transform operations on a measured vibration signal from a rolling bearing, is presented to quantitatively establish the relationship between the severity of signal degradation and the complexity values. The simulation results are then evaluated through experimental study of vibration signals measured during a bearing's run-to-failure test. It has shown that the complexity measure provides an effective technique for monitoring machine health status.