Enhanced Diagnostics Empowered by Improved Mechanical Vibration Component Extraction in Nonstationary Regimes

When analyzing vibration and sound signals from rotating machinery, accurately tracking individual orders is crucial for diagnostic and prognostic objectives. These orders correspond to sinusoidal components, also known as deterministic signals, whose amplitude and phase are modulated in response to the angular speed of the machine. The extraction of these components leads to a more comprehensive approach to differential diagnostics. When the machine operates under varying conditions, consistently tracking the orders becomes challenging, particularly in nonstationary regimes with very fast variations. Typically, this issue is addressed using common techniques such as Vold-Kalman filter (VKF), where the bandwidth of the selective filter is adjusted to handle the speed variations. However, in the presence of high-speed fluctuations, manual adjustment of these weights becomes difficult to balance the compromise between achieving accurate tracking by effectively filtering around the speed variations, and maintaining a low estimation bias by reducing noisy errors. To overcome this constraint, the proposed methodology is driven by the need to integrate speed fluctuations into an optimal solution using VKF. This adaptation involves the consideration of angular acceleration profiles within the innovation process. In this context, the bandwidths are automatically adjusted to their optimal values according to the machine’s regime. Optimality is achieved by crafting a model dependent on the order signal-to-noise ratio (SNR) and the auto-regression coefficient. This optimization allows for a practical adjustment tailored to the distinctive characteristics of each order. A comprehensive analysis of the resulting model transfer function reveals crucial insights into the impact of the given order SNR and the speed fluctuations. Subsequently, the methodology undergoes performance assessment through simulations and synthetic cases, showcasing its viability and effectiveness across various regimes. Notably, its practical application is highlighted in envelope-based bearing diagnosis, during operations characterized by variable-speed conditions, thus underlining its promise in real-world applications.