E. N. Bachelder and D. H. Klyde, Wavelet-Based Analysis of Roll Ratchet Using a Flight Test Database, presented at AIAA Atmospheric Flight Mechanics Conference, Austin, TX, August 11-14, 2003. (AIAA 2003-5692, STI-P-607)Degraded lateral axis handling qualities have at times resulted in sustained high frequency, small amplitude pilot-vehicle system oscillations or roll ratcheting. The degraded handling qualities can result from a combination of high roll damping, high levels of roll control sensitivity, and poor feel system dynamics. In some cases the roll ratchet phenomenon has been considered more of a nuisance that primarily affects ride quality, while in the more severe cases it significantly impacts aircraft handling qualities. Past ground based simulation research has indicated a peaking of the human pilot neuromuscular mode around the ratchet frequency. This link between the dynamics of the human pilot and the roll ratchet phenomenon was further explored using the flight research database from a lateral handling qualities study conducted with the USAF variable stability NT-33A. In this analysis Fast Fourier Transform techniques (FFT) were used to compute pilot and pilot-vehicle system describing functions. Strong evidence was generated to support the theory that the pilot neuromuscular mode couples with the feel system dynamics to produce an effective limb-manipulator mode. However, the smoothing of the frequency domain data over a given time window that results from the FFT analysis may have obscured the pilot-vehicle dynamic coupling that was changing over time. In this paper a wavelet-based system identification technique is used to reevaluate a subset of the same flight test database and to identify key pilot parameters that vary in time. These new results appear to indicate that: a lightly-damped mode in the open-loop system is introduced by the human pilot at frequencies higher than the task bandwidth; a direct correspondence exists between the peak amplitude of the lightly-damped mode and the occurrence of roll ratchet; and the wavelet-based identification technique is a powerful tool for assessing pilot-vehicle system performance over time intervals that were previously not feasible with FFT-based analysis.

E. N. Bachelder, D. H. Klyde, P. M. Thompson, Ph.D., and C. Harris, System Identification Methods for Improving Flutter Flight Test Techniques, presented at Atmospheric Flight Mechanics Conference, Providence, RI, June 10, 2004. (AIAA 2004-5065, STI-P-638)Classic flutter flight testing involves the evaluation of a given configuration at a stabilized test point before clearance is given to expand the envelope further. At each stabilized point flight test data are compared with computer simulation models to assess the accuracy of predicted flutter boundaries. Because of the time constraints associated with these procedures, the Air Force has been seeking methods to improve current flight test methods. This paper describes a technique that provides a rapid, on-line tool for the identification of aeroservoelastic (ASE) systems. The technique involves the use of discrete wavelet transforms to compute the impulse response (Markov parameters) of the estimated system. This is then used in the Eigensystem Realization Algorithm (ERA) method to compute the discretized state-space matrices. The technique used herein includes metrics that are used to assess the validity of the identified system. Although the method does require that the identification begin from stabilized initial conditions, it has been shown to be relatively insensitive to input forcing function. A model of a modern naval fighter aircraft was used to evaluate the capabilities of the identification method. The identification techniques were evaluated with and without an active oscillation controller in place

D. H. Klyde, P. M. Thompson, Ph.D., E. N. Bachelder, Ph.D., and T. J. Rosenthal, Evaluation of Wavelet-Based Techniques for Detecting Loss of Control, presented at Atmospheric Flight Mechanics Conference, Providence, RI, August 16-19, 2004. (AIAA 2004-4702, STI-P-633)Although great strides have been made in the design process toward understanding and eliminating pilot-vehicle system loss of control in the form of pilot-induced oscillations (PIO), significant and sometimes catastrophic events continue to occur with both military and commercial aircraft. This has led to the development of several real-time methods for on-board detection of impending loss of control. In this paper the piloted simulation evaluation of a wavelet-based system identification technique is described. The evaluation was conducted using a facility that featured a modern fly-by-wire military transport aircraft model and flight control computers in-the-loop. Signal transfer over the data bus and system redundancy was therefore representative of the actual aircraft. The pilot conducted a number of closed-loop tasks in both cruise and approach flight conditions. Flight control system changes were introduced as the pilot performed the given task in attempt to induce loss of control. Changes in the pilot-vehicle system were monitored with a time-varying computation of airplane bandwidth frequency and phase delay. Detection times, normalized by the PIO period, were computed when a key airplane bandwidth criteria parameter changed by 60%. The normalized detection time in the pitch axis ranged from 1.3 to 1.9 PIO cycles and from 0.9 to 2.1 PIO cycles in the roll axis. With refinements and appropriately defined pilot alerting schemes, the wavelet-based techniques can provide an effective on-board mechanism to alleviate loss of control.

P. M. Thompson, Ph.D., D. H. Klyde, E. N. Bachelder, Ph.D., T. J. Rosenthal, and M. J. Brenner, Development of Wavelet-Based Techniques for Detecting Loss of Control, presented at Atmospheric Flight Mechanics Conference, Providence, RI, August 16-19, 2004. (AIAA 2004-5064, STI-P-634)Wavelet transform methods were developed to predict loss of control for a broad range of refractory automatic and manual control system problems, including threats to flight safety. The systems being controlled are characterized by dynamics that change with time, so that loss of control or potential loss of control occurs only under unusual or rare circumstances, and the dangers can escape detection by conventional design criteria and methodologies. Recent technical developments are presented that were used for the real time implementation of wavelet transform methods for frequency response estimation. These developments include a description of the real time algorithm, the variance of the power spectral estimate, and methods used to reduce the variance of the power spectral estimate. An example is presented where the sudden onset of extra delay in a pitch stability and control augmentation system (SCAS) is detected. Wavelet transform methods are shown to have a faster detection time than Fourier transform methods