Using Electrical Torque for Fault Diagnosis of Rotating Machines: AI and Model-Based Approaches-Slides

Abstract:Rotating machines are important devices in manufacturing processes. Particularly, gearboxes offer mechanical power, and their high performance must be preserved. Being that most of the failures in industrial processes appear in mechanical devices, the development of innovative solutions for fault diagnosis is continuously under research. Usually, vibration signals are one the common source of data to this purpose, however the placement of vibration sensors could be not easy to couple in the machines. Given that gearboxes are powered by induction motors, this talk shows the use of the electrical torque as useful signal to get proper data for fault diagnosis in gearboxes. This approach is since mechanical faults of the motor load are transmitted to its electrical variables through the shaft between the motor and the gearbox. In this study, the analysis of the electrical torque is tackled from two points of view: one by using machine learning and the other by the development of a state observer-based model. In the first, we demonstrate that common statistical indicators and alternative signal processing by using Poincaré plots provide useful features for fault classification. In the second, we propose a model-based solution by using state observers to generate proper residual signals which deviate enough their values from zero to warn about the arising of a fault process. Thus, this last approach also serves as a fault severity measure. Please note: This webinar will be held in Spanish.