Bearing Fault Diagnosis Using Torque Observer in Induction Motor

This study introduces a sensorless fault diagnosis method for efficiently detecting bearing faults in induction motors. The proposed method eliminates the need for torque sensors, frequency sensors, thermal cameras, or real-time Fast Fourier Transform (FFT) tools. Induction motors are commonly utilized in a variety of industrial applications, including fans, pumps, and home appliances, due to their straightforward construction, affordability, and robust reliability. Traditional bearing fault diagnosis methods often rely on additional hardware such as vibration or thermal sensors. Additionally, approaches employing Artificial Intelligence (AI) and real-time FFT processing require advanced and expensive hardware capabilities. However, many V/f control systems are primarily intended for cost-effective and simple implementations, making resource-intensive approaches undesirable. Therefore, such methods present limitations for these use cases. To address these challenges, this paper presents a sensorless detection technique that estimates torque via a flux observer, removing the dependence on external sensors. The estimated torque is processed using an offline FFT to identify amplitude changes within bearing fault frequency bands. Here, the FFT-based frequency analysis is performed offline and is used to design a targeted band-pass filter (BPF). The torque signal, after passing through the BPF, undergoes a straightforward threshold-based logic to assess the existence of faults. Compared to AI- or data-driven approaches, the proposed method provides a lightweight, interpretable, and sensorless solution without the need for additional training or high-end processors. Despite its straightforward approach, the technique achieves effective detection of bearing faults across various components and speeds, making it ideal for embedded and economically constrained motor applications.