Sparse signal recovery based on adaptive algorithms for debris detector

Inductive debris sensors are generally used in online debris detection and perform well in monitoring the wear condition of rotating facilities. The detection accuracy is restricted by the superposition and noise of the impedance signal. From the perspective of machine learning, superposition is underfitting and noise is overfitting, and both are caused by inappropriate model complexity. Therefore, a series of machine learning approaches is proposed in this paper to devise a sparse signal processing model that can adapt to model complexity. We propose that the algorithm applied to debris detection can recover the impulse signals from the impedance signals and help determine the material and the size of the debris. Numerical simulations are carried out to prove that the superposition resolution is 0.6 half-wave width and has the ability to resist noise with a signal-to-noise ratio of more than 10 dB. With the test results, we demonstrate how the proposed method can solve superposition and resist noise.