IDENTIFICATION AND CONTROL OF ROTORCRAFT HUB LOADS USING NEURAL NETWORKS

The objective of the study was to develop a robust neural network based controller to minimize vibratory hub loads. A metric consisting of five vibratory hub load components (with their sine and cosine components) obtained from a wind tunnel test of a four-bladed rotor with individual blade control was used to characterize the hub loads. The neural network control procedure was bound by the following ground rules: the controller must. converge quickly in six iterations or less and gradient based optimization techniques must not be used. A simple and straightforward iterative procedure for neural control was applied. Two neural networks were used in the procedure requiring a plant model (using a radial-basis function neural network) and an "inverted neural network for control" model (using a back-propagation neural network). A simple half-interval calculation which halves the metric was used in order to speed up convergence. The neural network control procedure successfully achieved the objective within the given constraints. Finally, a limited-scope comparison of the results from the present neural control procedure with those from a one-step deterministic controller showed that the two control methods were roughly comparable, with neural control being slightly more robust.