Development of a diagnosis tool, based on deep learning algorithms and infrared images, applicable to condition monitoring of induction motors under transient regime

Infrared thermography can be a very useful technique for condition monitoring because the most common faults suffered by induction motors cause a temperature rise in the motor's frame. Moreover, this technique is non-intrusive, affordable and very sensitive due to the substantial technical progress in the design and development of new thermal cameras. However, data interpretation and decision making from the resulting infrared images is one of the major limitations of this technique, because it is directly dependent on the operator's experience. Several automated expert systems have been developed using machine learning and, to a lesser extent, with deep learning algorithms. The objective of this paper is to develop a diagnosis tool, based on infrared imaging and deep learning algorithms, applicable to induction motors working in transient conditions. The developed classifier, after training, presents high accuracy levels, classifying the images into one of the five considered scenarios and even at the early stages of the transient state. This methodology can be applied in a broad variety of scenarios with substantial cost saving and offering high-safety standards.