基于人工神经网络的非球形破片阻力系数预测模型

The trajectory of non-spherical fragments is closely related to the drag coefficient from supersonic to subsonic velocities. The non-spherical fragments tumble or rotate during ballistic flight, and the drag coefficient changes with the attitude of fragment. In order to obtain the average fragment drag coefficient under tumbling state from the drag coefficients corresponding to the finite fragment attitudes, a regular icosahedron-based averaging method is proposed. The average fragment drag coefficient under tumbling state is obtained by averaging the drag coefficients corresponding to 32 specific fragment attitudes. The error between the average drag coefficients of the cubic and cylindrical fragments obtained by the proposed method and those obtained by the ballistic gun test is within 10%. On this basis, the effect of fragment morphology, that is, sphericity, on the fragment drag coefficient in the full Mach number range is studied. The drag coefficients of a large number of non-spherical fragments are calculated by using the averaging method, with a sphericity of 0.35-1.00. A drag coefficient prediction model based on Mach number and fragment shape is established by artificial neural network. The test results show that the prediction model has high accuracy.It is found that the sphericity is the most important shape factor affecting the fragment drag coefficient, and its influence is most obvious at subsonic velocity.The dependence of fragment drag coefficient on the sphericity is significantly reduced at supersonic velocity.