Aplicação da técnica BoVW para identificar falhas em máquinas rotativas a partir de imagens termográficas de baixa resolução

Abstract

This master thesis aims to describe the implementation of a fault detection strategy in rotating machines using infrared thermography. It was based on images acquired with a low resolution camera and compared the results with those obtained with the use of a medium resolution camera to verify if it is possible to use lower resolution thermographic cameras without significant loss of performance, thus reducing costs and making this approach more accessible. The BoVW technique (bag of visual words) was applied for data treatment and fault categorization. The effectiveness of the presented method was validated experimentally from the diagnosis of the faults simulated based on the thermographic images of the two types of cameras, and, in addition, ratified by vibration analysis. The precision results were over 92% of the accuracy for both cameras, thus proving the viability of using low resolution thermographic cameras to identify faults in rotating machines.