Development of a predictive maintenance model which provides fault identification and diagnostics on electrical gearmotor systems

Abstract

Industrial electric motors are at the heart of almost every industry. They are a 47 billion USD market in 2020 and consume 70% of all industrial electricity. They are generally paired with a gearbox and load, which is referred to as an electric gearmotor system. Being so essential means that it is important to avoid failures, 98% of 300 researched companies by ITIC reported a cost of 100.000 USD every hour of downtime. To detect failures a technology called fault detection and diagnosis (FDD) is used, which is a method to foresee failures or faults in a system that deteriorates over time through evaluating the state of the system. It is an extensively academically research subject, however, has hardly been adopted in industrial settings where electric gearmotor systems are applied. Thus, a FDD model was developed to provide insight, knowledge, and a practical example into the necessities of an industrial FDD model. This was achieved through conducting an analysis on the state-of-the-art of FDD in industrial settings and conducting a literature review on FDD. Based on an analysis of the conclusions a hybrid diagnostics model was developed. For the fault detection a model-based solution was used, it compared the predicted torque to the measured torque of a motor to create a health indication value. If this value crosses a pre-determined threshold an alarm would go off. For fault identification a decision tree machine learning algorithm is used to identify: blockage, bearing, gear or random failure in a system. To verify and validate the hybrid diagnostics model it was applied to a client of SEW Eurodrive where data was available of a known system. The system had a fault detection accuracy as high as 94% and could classify failures with an accuracy of 93%.