Parametric Tracking of Electrical Currents Components Using Gradient Descent Algorithm

TL;DR

This paper introduces a novel method for monitoring electrical motor currents by decomposing and tracking spectral components using a Vandermonde model and gradient descent, improving fault detection accuracy.

Contribution

It presents a new parametric tracking approach for electrical current components based on Vandermonde modeling and gradient descent, enhancing fault detection in motor current analysis.

Findings

Effective decomposition of current components achieved

Gradient descent successfully tracks spectral parameters

Improved fault detection potential demonstrated

Abstract

In the last few years, Motor Current Signature Analysis (MCSA) has proven to be an effective method for electrical machines condition monitoring. Indeed, many mechanical and electrical faults manifest as side-band spectral components generated around the fundamental frequency component of the motor current. These components are called interharmonics and they are a major focus of fault detection using MCSA. However, the main drawback of this approach is that the interference of other more prevalent components can obstruct the effect of interharmonics in the spectrum and may therefore impede fault detection accuracy. Thus, we propose in this paper an alternative approach that decomposes the different current components based on the Vandermonde model and implements the tracking of each distinct component in time and spectral domains. This is achieved by estimating their respective relevant parameters using the Gradient Descent algorithm. The results of this work prove to be promising and establish the parametric tracking of the electrical current components using the Gradient Descent algorithm as a reliable monitoring approach.