ROTATIONAL FREQUENCY ESTIMATION FROM SENSED VIBRATIONS BASED ON A SUPERVISED LEARNING METHOD

Apparatus and associated methods relate to generating a frequency spectrum weighting function for use in estimating a rotational frequency of the rotating member. The estimation of the rotational frequency is based on vibrations sensed by an accelerometer remotely located from a rotating member. The frequency spectrum weighting function is generated by a supervised learning method. The method includes receiving a set of test vectors. The test vectors include a rotational frequency value of the rotating member and a vibrational frequency spectrum corresponding to vibrations propagated to the accelerometer. The vibrations include vibrations caused by the rotating member rotating at the rotational frequency. The method includes calculating a test weighting function, and then weighting the vibrational frequency spectra by the test weighting function. The method includes calculating a vector score indicative of whether the weighted vibrational frequency spectra promote the identification of the rotational frequency of the rotating member. The score relates the amplitude of the weighted vibrational frequency spectrum corresponding to the rotational frequency or speed of the rotating member to the maximum amplitude of the weighted vibrational frequency spectrum corresponding to frequencies not corresponding to the frequency of the rotating member.