Research On Neural Network Identification And Active Vibration Control Approaches For Wind Tunnel Test Model

When conducting wind tunnel test of aircraft, disturbed by the unsteady aerodynamic interference,vibrations will generate on the experimental model. In order to perform the measurement of the aircraftaerodynamic forces during wind tunnel test, the test model is usually mounted on the pitching andyawing structure by back-support method. The combination structure-of testing object and supportsystem, similar to a cantilever beam, is usually flexible with low natural frequencies less than100Hz.During transonic testing, the fluctuation pressure caused by the airflow distributes in the range of lowfrequency bandwidth, and, furthermore, the distribution varies with different angle of attack, modelgeometry, and airflow Mach numbers, etc. In this situation, high magnitude vibration with lowfrequency is easily generated due to resonance of the flexible structure excited by the low frequencyfluctuation pressure. This high magnitude vibration would contaminate the test data and even damagethe instruments in a fatigue manner, and therefore should be suppressed. It is necessary to take effectivemeasures to suppress the vibration.Mathematical models of the test object and support system are established by model identificationmethod which is the ERA method (the eigensystem realization algorithm). Then, Linear QuadraticGaussian controller based on the low dimension model, an indirect adaptive neural controller based onthe off-line neural network NARMA system model and PID based on neural network controller aredesigned to suppress the vibrations. Finally, simulations based on Matlab platform and experiments arecarried out to verify the ability of the first two controllers to suppress the vibrations. For differentcontrol schemes, comparisons between them are given in terms of both simulation and experimentalresults.