Deep Learning Model Based on TCN-LSTM for Projectile Initial Velocity Inversion of High-spin Projectiles

In the design process of high-spin projectile (HSP) shooting, it is necessary to obtain the accurate projectile initial velocity (PIV) with initial disturbances, for the subsequent external ballistics calculations and adjustment processes. Then, the PIV inversion model is proposed in this paper based on Temporal Convolutional Networks (TCN) and Long Short-Term Memory Networks (LSTM). And a database is also established by the 6-degree-of-freedom projectile dynamics model and external ballistics simulation with Runge-Kutta method. From the simulation results, the proposed TCN-LSTM model is compared with the BP model, LSTM model, and GRU model using the same dataset. It is noted that the TCN-LSTM model performs excellently PIV inversion, with a higher accuracy compared to the single-precision model.