Time-Domain Frequency Estimation with Application to Fault Diagnosis of the Unmanned Aerial Vehicles' Blade Damage

In this article, a novel time-domain frequency estimation approach is proposed based on the well-known subspace identification method. Differing from the fast Fourier transformation method, the nonlinear identification for frequency estimation is reformulated into the eigenvalue identification problem, which avoids the spectrum peak search. The recursive frequency estimation approach is proposed based on the updating/downdating of the Cholesky decomposition, where the extended observability matrix can be identified from a small matrix with lower computation cost. In addition, a gap metric-oriented performance indicator is proposed as a test statistic for fault detection as well as the evaluation of fault severity. The effectiveness of the proposed methods is verified for frequency estimation and fault diagnosis performance through numerical simulations and the experimental measurements from a laboratory unmanned aerial vehicle rig with partial blade damage.