Abstract
The evolutionary periodogram (EP) is applied to signals collected from an on-line turbine flowmeter for condition monitoring. Firstly, the signal acquired for a normal (fault-free) flowmeter is synchronously averaged to reduce any possible transients or background noise, and then the data is treated to remove irrelevant components and enhance important details that may lead to detecting the presence of any flaws. Secondly, another set of data is collected for an operating faulty flowmeter and processed in a similar way. To simulate common faults, the tips of three rotor blades for the normal turbine flowmeter are cut short by a small amount; the normal rotor blade is then replaced with the faulty one. Time-frequency distribution is then calculated using the EP for both the normal and faulty cases. The answer obtained using the EP is compared with both the conventional short-time Fourier transform (Spectrogram) and the Choi-Williams methods in terms of time-frequency resolutions including suitability for this application.