Bearing diagnostics under strong electromagnetic interference based on Integrated Spectral Coherence

•Two methods based on CSCoh are proposed for bearing fault detection under strong EMI.•Two criteria are proposed for the estimation of IES.•IESFOgram indicates the optimum band for the estimation of IES.•The methods are compared to SOTA techniques on real vibration.•The methods achieve good results for 2 faults and 3 different loads. Rolling element bearing fault diagnostics has been a topic of intensive research in recent decades, as they are critical components of rotating machinery and therefore their failure may result in sudden breakdown of machines and industrial installations. The early and accurate detection of incipient faults on bearings can reduce the production cost by allowing maintenance engineers to schedule a replacement at the most convenient time. Envelope Analysis is a widespread powerful method in bearing diagnostics, often used along with Fast Kurtogram. However, the presence of ElectroMagnetic Interference (EMI) and generally speaking of impulsive and non gaussian noise, increases the complexity of bearing fault diagnosis and may lead to rather poor diagnostic performance. EMI is often present in mechanisms and machines, where motors are controlled by Variable-Frequency Drives (VFD) and can present a vibration signature similar to that of bearing faults. Therefore, the main aim of this paper is the proposal of advanced signal processing techniques, which can detect bearing faults under the presence of strong ElectroMagnetic Interference or other impulsive noise (where state of the art methods fail). Two novel diagnostic methodologies are proposed based on the Cyclic Spectral Coherence (CSCoh). The integration of the CSCoh, over the full spectral frequency axis or over a specific spectral frequency band, results respectively in the Enhanced Envelope Spectrum or in the Improved Envelope Spectrum. The two novel diagnostic methodologies allow for the automatic selection and integration of the optimal bands on the CSCoh under heavy impulsive noise, such as EMI, resulting in a spectrum with enhanced characteristic bearing fault frequencies, without any human intervention required besides the knowledge of the characteristic fault frequency which is under investigation. The methods are applied on vibration data, captured on an epicyclic gearbox with seeded bearing faults, operating under the influence of strong EMI. The methods are tested and evaluated on different fault cases and achieve improved performance compared to state of the art diag