Experimental Validation of Thermoelastic Neural Network-Based Inverse Model on a Slender Hypersonic Vehicle

Achievement of regular hypersonic flight is proving itself to be a considerable engineering challenge. One of the feats that has yet to be properly addressed relates to effective estimation of in-flight loads and their corresponding aerodynamic state so that guidance, navigation, and control can be attained. We investigate here a novel vehicle-as-a-sensor concept, where the vehicle’s deformed state is internally measured to infer the loading acting upon it. A neural network-based inverse model is introduced to connect strain and temperature measurements on the internal structure of the vehicle to its outer surface load state. It is validated experimentally with a representative testbed of a slender hypersonic vehicle. The benchtop model is instrumented with fiber optic strain sensors, force is applied with two linear actuators, the model is thermally excited with a resistance heat tape, and temperature measurements are obtained through an infrared camera. The predicted forces of the neural network-based inverse model are compared with experimental values. While the results look promising, it was observed that the method requires further refinement regarding the modeling and measurement of the thermal effects.