abstract
- © 2021 Elsevier LtdDetection, diagnostic and isolation of crankshaft failures is of vital importance to ensure reliable operation during the course of their services. However, many times the fault may be submerged in a strong background noise making the feature extraction process difficult. To solve this problem, this paper shows a new method for extracting features from the signal immersed in background noise. This method is based on linking the dyadic Wavelet transform and the Welch-Bartlett classic periodogram, obtaining a modified Welch-Bartlett periodogram; which provides a dyadic spectrum of frequencies with multiple sensitivities. The validation process was based on two scenarios. The first one did not generate a fault, while the second one presented a fault with a signal to noise ratio of 0.3269 dB. The results show that the dyadic wavelet function to be used is the Daubechies 45 as it provides an attenuation of the rejection bands close to 150 dB. Through the new modified Welch-Bartlett periodogram, an engine vibration signature detected, 29 Hz, the component 120 Hz corresponding to the deformation of the magnetic fields of the engine and the component of 177.5 Hz inherent in squirrel-cage induction motors. A final result showed that the new modified Welch-Bartlett periodogram maximizes sensitivity in frequency component analysis. In addition, the results were superior to those achieved with the procedures reported in the literature reviewed.