Statistical analysis of symbol sequence distributions for machine condition monitoring

This paper introduces a novel method of investigating the frequency distributions of symbol sequences of discrete signals that have been generated from time series measurements of machine components. The approach is different from conventional spectral methods as the focus is on the time domain, and is therefore suited for monitoring systems that generate non-stationary measurements. A method of statistically analyzing symbolic time series methods is presented. Theoretical background of the method has been introduced and its efficacy is studied through experimental investigation of vibration signals recorded from a rolling bearing elements. Results indicate that the method is robust and can effectively characterize defects and varying operating conditions.