Uma abordagem neural no diagnóstico de falhas em rolamentos de motores de indução trifásicos

Abstract

The three phase induction motor occupies a prominent position in the production of electromotive force and this makes it widely used in industrial applications. Consequently, it is also subjected to the conditions of operation and maintenance of the machines as a whole, as well as faults which they are subject. Thus, this paper proposes an alternative method to traditional in fault detection in bearing of induction motors connected directly to the power grid. The objectives consist in using a neural approach able to classify the existence of bearing fault with a high percentage of correct. Analyzing for this, in the time domain, one half cycle of the voltages and currents of stator the motor in study. The proposal is validated through experimental tests on a computer and monitoring on-line embedded in a DSP. As a result, the work has the creation of a database of failure, with more than a thousand trials involving the main flaws found in three phase induction motors. These tests are performed considering the conditions of voltage supply unbalanced and with several situations of mechanical load on the machine shaft.