AI-Based Bearing Defect Detection Using Variable Reluctance Sensor Signal

This paper presents a method for the detection of bearing wear and the prediction of bearing failure within a rotational assembly, utilizing commonly employed sensors. In the context of a fluid measurement device, a straightforward variable reluctance sensor is employed to register the passage of a turbine vane, ensuring volumetric measurement. The proposed methodology harnesses neural networks to classify various types of bearing damage by analyzing the raw sensor output signal. The results showcased herein underscore a remarkable level of accuracy, even when applied to a relatively constrained data set.