Blade vibration parameter identification under small-angle probe layouts: a multi-resolution ESPRIT-based frequency extension approach

Abstract Condition monitoring of rotating blades is essential for turbomachinery reliability. Blade tip timing (BTT) is a promising non-contact measurement technique, but its inherent under-sampling makes accurate resonance parameter extraction challenging. Although the ESPRIT algorithm provides super-resolution frequency identification, its performance may deteriorate under the restricted small-angle probe layouts commonly encountered in practical tests. To address this issue, a dual-baseline multi-resolution ESPRIT (MR-ESPRIT) method is proposed. By exploiting the spatio-temporal sampling characteristics of the probe layout, a dual-baseline observation model is constructed within a unified ESPRIT framework to obtain both unambiguous coarse estimates and aliased fine estimates. A prior-guided mechanism then uses the coarse estimate to select the physically consistent extension coefficient of the fine estimate, thereby recovering the resonance frequency. The method is also combined with a short-time quasi-stationary treatment to improve its applicability to non-stationary cases, including speed fluctuations and run-up sweeps. Numerical simulations and centrifugal compressor experiments indicate that MR-ESPRIT provides more stable identification than classical ESPRIT under the tested high-noise and small-angle conditions, suggesting its potential for BTT monitoring in restricted-space applications.