Uncertain optimization for initial disturbance of projectile for moving tank based on stochastic programming and deep learning surrogate model

Abstract

To enhance the hitting accuracy of tank with moving firing, an uncertain optimization method based on stochastic programming is adopted to reduce the initial disturbance of projectile. Firstly, the firing dynamics model of moving tank is modeled to simulate the initial disturbance of projectile. Secondly, the controllable interior ballistic parameters such as projectile structure parameters and propellant parameters are treated as random design variables. The surrogate model for the firing dynamics model of moving tank is constructed by using the neural network based on deep learning. The uncertain optimization problem is transformed into a deterministic optimization problem by stochastic programming method. Then multi-objective genetic algorithm is adopted to settle optimization model, and reasonable design interval of random design variables is obtained. Finally, a six-degree-of-freedom rigid external ballistics model is used to establish a hitting accuracy evaluation model of moving tank based on interval uncertainty analysis. Through this model, the effectiveness of the optimization method is demonstrated.