Thermal images and temperature matrices for the state assessment of rolling bearings

Abstract

In this paper, the potential that IRT may have for classifying the severity of failures of rolling bearings is investigated. Considering the promising results obtained within the Infrared Breast Thermography domain, thermal-based analysis (TBA) using temperature matrices, intensity-based analysis (IBA) using thermal images, and different approaches for combining the two methods are compared in the classification of the severity of outer-race defects of rolling bearings. An accuracy and F1-score of more than 90% were obtained through the concatenation of the temperature matrix to the thermal image through pseudo colors and their processing with the VGG deep learning algorithm.