With the increase in the complexity and reliability of mechanical systems， electromagnetic wear particle detection sensors have been widely studied and gradually applied in the field of wear-state monitoring for mechanical devices. Aiming at the problem that the velocity-randomness of metal wear particles in lubricating oil leads to poor consistency of detection results of electromagnetic wear particle detectors， the magnetic model of a spherical wear particle in an alternating magnetic field was established， and the distribution law of magnetic flux density and the method for calculating magnetic energy perturbation was studied. Meanwhile， the loss coefficient of magnetic energy for non-ferromagnetic wear particles was proposed and measured by experiments， and the characteristic signals of wear particles exported from the sensor are then compensated. The experimental results show that the method can eliminate the influence on the output signal amplitude by the speed of wear particles， which markedly improves the consistency of wear particle detection results.