New method based on the optimal generalised sequential test for bearing localised defects detection and diagnosis

In modern industrial environments, continuous operation of rotating machinery is vital for production efficiency, equipment reliability and safety. Bearings, as key transmission components, are highly susceptible to localised defects that, if undetected in early stages, can cause severe failures and costly downtime. This paper presents an early fault detection and diagnosis framework based on the optimal generalised sequential test (OGST), a truncated form of the classical Wald test, adapted for bearing vibration analysis under stationary and non-stationary conditions. OGST accelerates decision-making by reducing the required sample size without sacrificing detection accuracy. The method transforms raw vibration signals into a binary ‘hypothetical sequence’, with zeros representing healthy states and ones indicating faults. This sequence is analysed using the fast Fourier transform (FFT) to detect characteristic defect frequencies, enabling precise identification of faulty bearing elements (inner race, outer race, rolling elements or cage). Designed for incipient fault detection, the method assumes known variances for both healthy and defective states and remains effective under high noise, including cases where signals are entirely submerged in noise. The proposed methodology is validated using both simulated signals and real vibration data acquired from a bearing fault test bench under variable operating conditions, representative of realistic operational scenarios. Results show that OGST outperforms conventional techniques such as envelope analysis and FFT-based demodulation in terms of detection speed, reliability and noise immunity, making it a powerful tool for predictive maintenance in critical rotating machinery systems. Moreover, the proposed method can be integrated into artificial intelligence frameworks for defect classification and learning, further enhancing the speed and accuracy of real-time decision-making.