Certain investigations on fault detection in induction motor using infrared thermography with machine learning and time frequency signal analysis

Abstract

As the majority of the industry uses induction motors, this technology newlinehas become simple, reliable, and plays an essential role in manufacturing, newlinetransportation, and other applications. For monitoring motor health, newlinemanufacturing operations utilized various tools to visualize, analyze, and newlineacquire sample data collected through vibration monitoring and signal newlineprocessing techniques. The induction motor may be faulty based on the two newlineconditions: one is a motor fault and the one is an electrical fault. In this newlineresearch work, improved image processing and signal processing methods are newlinepresented to discover induction motor faults. Two works are presented in this newlineresearch work for the induction motor faults detection. newlineIn the first work, infrared images are used to discover faults in an newlineinduction motor. Nowadays, Infrared imaging can be used for energy and newlinecondition monitoring as the thermographic patterns can vary according to the newlinemachine condition or fault. However, a limited number of machine faults are newlinepresently examined employing the thermal imaging process. Therefore, in newlinethis work, a new automatic motor fault identification method is presented newlineusing infrared thermography (IRT) with combined image processing and newlinemachine learning methods, focusing on energy efficiency. Initially, the color newlineand texture features are extracted from the motor s infrared image through the newlineGabor filter and GNS (global neighborhood structure) map. The proposed newlinemethod integrates the faster R-CNN (Region-based Convolutional Neural newlineNetwork) and Speeded Up Robust Features (SURF) algorithm to enhance newlinefault detection and classification accuracy, in which SURF is used as a feature newlinedescriptor to faster R-CNN for object detection and fault classification newlinedepending on the extracted features. newline