Detección de daño en rodamientos en máquina de inducción mediante MODWT

In this work we present the development of a method of detection of damage in bearings by means of MODWT and edge detection in images, which is capable of distinguishing 3 classes of damage different from a healthy bearing, which allows to reach an overall accuracy greater than 90 %, the tests are p...

Descripción completa

Detalles Bibliográficos
Autor: Victor Manuel Aviña-Corral
Tipo de recurso: tesis de maestría
Estado:Versión aceptada para publicación
Fecha de publicación:2019
País:México
Institución:Instituto Nacional de Astrofísica, Óptica y Electrónica
Repositorio:Repositorio Institucional del INAOE
Idioma:español
OAI Identifier:oai:inaoe.repositorioinstitucional.mx:1009/1788
Acceso en línea:http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/1788
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/Inspec/Tread
info:eu-repo/classification/Inspec/Induction machine
info:eu-repo/classification/Inspec/MODWT
info:eu-repo/classification/Inspec/Edge detection
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/22
info:eu-repo/classification/cti/2203
Descripción
Sumario:In this work we present the development of a method of detection of damage in bearings by means of MODWT and edge detection in images, which is capable of distinguishing 3 classes of damage different from a healthy bearing, which allows to reach an overall accuracy greater than 90 %, the tests are performed acquiring signals under 14 different operating conditions for each type of damage, the parameters that are changed is the origin of the electrical power supply and the frequency of operation of the machine as well as the load to which the motor is subjected. The signals are acquired through a DAS stage based on conditioning, conversion and an FPGA card that acquires and sends the signals to a PC; This algorithm applies the MODWT to the three current signals of the stator, from which, together with the original signal, the magnitude of current is acquired, which is used to generate a two-dimensional periodic array, the arrangement is smoothed by a Gaussian _ltering stage allowing it to observe patterns in the contour lines or edges, these patterns are accounted for by a mask that sweeps and analyzes the image in search of them, the results obtained have a lognormal distribution, which is characteristic according to the condition of the bearing. The overlap region is almost null between the damaged bearing and the healthy bearing for most cases. The studies of the signals are made by means of the signature of current and square current of the motor, in general both methods present similar results. The test signals were analyzed with other methods making a comparison between the results obtained with this method and other methodologies, in general their overall accuracy decreased when considering the 14 modes of operation of the motor, this method was also tested with a database analyzed with another method already published, the accuracy reached values higher than 98% with the same signals.