Time-Frequency Analysis for Transfer Function Estimation and Application to F18-SRA,” submitted to the Automatica

A transfer function estimation procedure that relies on the time-frequency analysis of input and output signals is described. This method was developed in an attempt to better identify the aeroelastic behavior of NASA Dryden’s F-18 systems research aircraft and to predict its utter boundaries using in- ight experimental data. Numerical experiments on eld data show that exploiting the time-frequency characteristics of the excitation inputs can bring enhanced accuracy and con dence when identifying multi-input/multi-output transfer functions. In particular, the proposed approach complements many well-established black-box identi cation procedures by providing an independent way to obtain transfer function estimates. A computational tool implementing this approach is now being evaluated for practical use at NASA Dryden Flight Research Center. Nomenclature b.w / = passband lter h i = transfer function from input i to the outputQh l;i = estimated transfer function from left input to the output for i data set Qhr;i = estimated transfer function from right input to the output for i data set Ji j = cost function = time-frequency representation of a signal = denoised time-frequency representation of a signal Ui = discrete Fourier transform (DFT) of a denoised time-frequency representation of an input signal u i = input signals Ou. j! / = DFT of a signal u.t / = signal in time domain Y = DFT of a denoised time-frequency representation of an output signal y = output signal 1T = sampling time I