Method for detecting a structural fault of an electric motor

A method for detecting a structural fault of an electric motor, comprising the steps:

• (i) acquiring, during an acquisition window of predetermined duration, a measurement signal of a physical parameter representative of the rotation of the electric motor, at a predefined sampling frequency,
• (ii) obtaining the high frequency component of the measurement signal, resulting in a corrected signal,
• (iii) applying a set of band-pass filters to the corrected signal, each band-pass filter having a central frequency and a bandwidth, the central frequencies of the band-pass filters being in ascending order and the bandwidth being separated from each other, each band-pass filter having an index corresponding to the rank of the central frequency in a sorted list of the central frequencies, resulting in a set of filtered corrected signals,
• (iv) determining a stator frequency and a rotor frequency of the electric motor,
• (v) computing a frequency signature vector comprising a plurality of components, each component of the frequency signature vector being a predefined linear combination of the stator frequency and of the rotor frequency, from the equation:
• FSVk=ak*Fs+bk*Fr,
• wherein FSVk is a component of the frequency signature vector, Fs is the stator frequency, Fr is the rotor frequency and the coefficients ak, bk are predefined integers,
• (vi) computing a temporal symptom vector from the computed frequency signature vector and from the set of filtered corrected signals, the temporal symptom vector comprising a plurality of components, the computation comprising the sub-steps:
• (vi-1) for each time sample of the acquisition window and each component of the frequency signature vector, selecting the band-pass filter of central frequency which is closest to the value of the component for the time sample,
• (vi-2) affecting the output value of the selected band-pass filter to the component number k of the temporal symptom vector for the time sample,
• (vi-3) iterating steps (vi-1) and (vi-2) for all time samples of the acquisition window, resulting in a temporal symptom vector,
• (vii) detecting a structural fault of the electric motor from the determined temporal symptom vector and from a classifier model.