Machine health diagnosis based on approximate entropy

As a statistical measure for time domain signals, Approximate Entropy (ApEn) quantifies the regularity of a data sequence, thus can serve as an indicator for the severity of structural defects in a machine system, such as an electrical drive or a rolling bearing. This paper investigates the utility of ApEn for machine health diagnosis, using a realistic spindle-bearing test bed as the platform. Experimental results were consistent with the theoretical analysis, and confirmed that ApEn provides an effective measure for the severity of structural defect, with good computational efficiency and high robustness.