“YANGI AVLOD ILMIY TADQIQOTLARI” respublika ilmiy-amaliy konferensiyasi
PHYSICS-INFORMED SPECTRAL MEMORY IDENTIFICATION AND FLUTTER-REGIME DIAGNOSTICS FOR A PITCH–PLUNGE AEROELASTIC SYSTEM
Fergana State University
Hereditary effects in aeroelastic systems are governed by Volterra-type integro-differential equations whose memory kernel G(t) is not directly measurable, and its recovery from noisy displacement records is a severely ill-posed inverse problem. We present a physics-informed identification framework in which the matrix-valued kernel is parameterized as a learnable sum-of-exponentials (spectral) measure, G(t) ≈ ∑k Wk exp(−skt) and the governing equation is enforced through a once-integrated residual that requires only smoothed displacements and velocities, eliminating the noise-amplifying second derivative. Flutter-regime diagnostics are obtained from the exact augmented state-space eigenvalue problem of the identified model, combined with robust empirical growth-rate estimation. In a fair twenty-seed synthetic benchmark in which all competing models are fitted to identical noisy kernel observations, the proposed spectral fit attains a mean flutter-speed error of 0.020%, kernel relative error 0.51%, outperforming Prony-2 0.352%, Prony-3 0.068%, and the memoryless baseline 0.905%. Applied to the open Springer pitch-and-plunge (PAPA) wind-tunnel dataset, the identified model reproduces the post-flutter modal frequency at the held-out airspeed U = 9.50 m/s with a relative error of 1.563% and supports the censored stability statement Uf < 7.00 m/s. The framework is reproducible, fully seeded, and reports only experimentally supported claims.