Edinburgh Napier University

  Early detection of faults in electric motors is crucial to prevent unplanned downtime and expensive repairs. Transient analysis through time-frequency transforms reveals important information on the motor condition. Computational time of these transforms becomes a problem when dealing with thousands of motors in just one industry. Researchers focus on obtaining the best quality results, usually using Gabor functions as t-f atoms. This paper shows how Frequency B-Splines can reach the same quality with nearly 40% less computational time. To achieve this goal, the slope criterion is applied to select the optimal atoms parameters. A formula that relates the slope of the Heisenberg Box of a FBS with its parameters is deduced. The proper time interval where the FBS must be defined is also determined. Results are shown not only with lab tests, but also through a field case.