RESEARCH ON INTELLIGENT IDENTIFICATION ALGORITHMS FOR SHORT-TERM AERODYNAMICS OF HYPERSONIC WIND TUNNELS

Pulse combustion wind tunnel force measurement is an important step in the research and development process of hypersonic aircraft, and with the development of hypersonic aircraft technology, large-scale and heavy-load aircraft test models has become the trend of hypersonic pulse combustion wind force test. During the effective test time of several hundred milliseconds, large-scale force measurement system stiffness weakened and other issues will seriously lead to poor aerodynamic identification accuracy. The large-scale measurement model poses a challenge to the accurate aerodynamic identification of the short-term pulse combustion wind tunnel. To solve this problem, a new intelligent aerodynamic identification algorithm based on traditional signal processing combined with deep learning is presented in this paper. The algorithm framework is mainly divided into two stages for signal processing: (1) signal decomposition, (2) data training. In the signal decomposition stage, the original data is decomposed into different modal sub-signals through variational modal decomposition (VMD). In the training stage, the effective features in the remaining datasets containing characteristic sub-signals are extracted by deep learning model, and the real aerodynamic signals are obtained. In addition, in order to enhance the robustness and applicability of the algorithm, different optimization methods are used to optimize the hyperparameters in the algorithm at different stages of the algorithm framework to obtain the optimal parameter combination. This algorithm model has obtained relatively ideal results in terms of aerodynamic recognition accuracy and anti-interference. Finally, the algorithm is validated on a suspended force test bench, and the results show that the algorithm model can effectively identify and filter out the interference components that are difficult to eliminate by the traditional methods brought by the large-scale model. Finally, the algorithm is successfully applied to the large scale model force measurement system of pulse combustion wind tunnel. Accuracy of aerodynamic identification of large-scale model force measurement system is effectively improved.