APPLICATION OF ARTIFICIAL NEURAL NETWORK (ANN) FOR PREDICTION OF DRAG COEFFICIENT OF AXISYMMETRIC BOATTAIL MODELS

The utilization of longitudinal grooves has demonstrated effectiveness in reducing drag on axisymmetric boattail models. However, the intricate relationship between groove parameters and drag force defies simple physical equations, necessitating the discovery of an optimal parameter set. Conventional experimental and numerical simulation approaches prove impractical due to their resource-intensive nature. Instead, Artificial Neural Networks (ANNs) offer a promising alternative. In this study, a three-layer ANN is trained using 192 examples generated by Ansys Fluent with the Reynolds-Averaged Navier-Stokes (RANS) method and the k-ω SST model. Subsequently,48 examples are employed for network validation. Comparison between ANN-predicted values and CFDdetermined drag coefficients reveals an average difference of less than 0.76%, validating the network's reliability.The ANN successfully identifies optimal groove parameters across various boattail angles, and numericalsimulations conducted on models featuring these optimal grooves further validate the ANN's accuracy in predicting drag coefficients, with a negligible deviation of only 0.98%. Additionally, analysis of flow characteristics and aerodynamics aids in understanding factors contributing to drag reduction.