Smart numerical technique for faults diagnosis based on a similarity algorithm experimentally investigated on AC machine stator windings

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Modern statistics are useful for setting up algorithms for digital protection and control systems. In this paper, a smart numerical technique based on a similarity algorithm is proposed for fault diagnosis in AC machine stator windings. The protection technique can detect online faults located on the three-phase stator windings, such as internal shunt and Turn-to-Turn Faults (TTTFs). The Pearson correlation-based algorithm amalgamates the functions of phase differential current and overcurrent relays into a single protection scheme. The method is investigated on a three-phase induction motor with 20 taps for each phase stator winding in order to facilitate managing extensive experimental tests to verify its effectiveness. The practical results exhibit the analysis and evaluation of the protection properties, revealing that the algorithm’s security and dependability rates are greater than 99%, and its reliability and accuracy percentages exceed 98%. Besides, the similarity approach provides a quick response, and a new configuration of relay operating characteristic curves is proposed. These curves can be used to consistently monitor TTTF and shunt faults, differentiate between internal and external shunt faults for the machine protection zone, specify the faulty phases, and classify different shunt faults located on the machine stator windings. Furthermore, the numerical method possesses the capability to ascertain the extent and intensity of disturbance and asymmetry in the three-phase machine currents, and to estimate the TTTF tripping time using a new design of correction-time curve. The comparative analysis demonstrates that this technique is significantly superior to other recently published techniques.

Article activity feed