Prediction of pressure distribution and aerodynamic coefficients for a variable-sweep wing

Abstract

To satisfy the performance requirements across multiple speed ranges, a variable-sweep wing (sweep angle range from 25° to 40°) is derived from the BQM-34 “Firebee” drone model. However, predicting aerodynamic characteristics across various flight conditions and sweep angles is a challenging task. Traditional methods like CFD and wind tunnel testing are both time consuming and expensive. In order to efficiently predict the pressure distributions and aerodynamic coefficients, a novel network that combines a Radial Basis Function Network (RBFN) and a Convolutional Auto-Encoder (CAE) is proposed. Two distinct loss function methods, the standard Pressure-Targeted Method (PTM) and the newly developed Comprehensive Evaluation Method (CEM), are employed to optimize the network's predictive performance. These methods are evaluated on datasets with both trained and untrained sweep angles. The results show that while both PTM and CEM accurately predict pressure distributions, the enhanced CEM provides more uniform and reliable predictions. Moreover, the CEM method significantly outperforms PTM in predicting aerodynamic coefficients, reducing errors by over 50%. The proposed RBFN-CAE network with the CEM loss function offers an effective way to predict the aerodynamic characteristics of a variable-sweep wing, improving predictive models in aerodynamic applications.