基于代理模型的制导火箭炮发射诸元计算方法

针对制导火箭炮发射诸元的快速计算问题，提出了一种结合大样本数据和代理模型计算发射诸元的新方法。运用代理模型建立射角、无控弹道侧偏与炮位纬度、炮位高程、射向、射程、目标点高程及药温之间的函数关系，并根据射程和无控弹道侧偏的预测值对射向进行修正。仿真结果表明，高阶多项式响应面、相关函数为高斯函数的Kriging、高阶单项式径向基函数、核函数为高斯函数的最小二乘支持向量机、激活函数为正弦函数的超限学习机以及由上述单一代理模型构建的组合代理模型均具有较高的预测精度，各种单一代理模型对射角和无控弹道侧偏的预测时间均小于1 ms，证明了基于代理模型的射角和无控弹道侧偏预测方法切实可行，且通过对射向进行修正有效减小了由于地球自转引起的无控弹道侧偏。 

Method for calculating firing data of guided rocket launcher based on surrogate model

Aimed at the problem of rapid calculation of firing data of guided rocket launcher, a new method for calculating firing data based on large sample data and surrogate model is proposed. The surrogate models are used to establish the functional relations between the firing angle, uncontrolled lateral range and six influencing factors, including latitude of artillery location, elevation of artillery location, target azimuth, range between artillery location and target location, elevation of target location and propellant temperature, and the target azimuth is corrected according to the range and predicted value of the uncontrolled lateral range. The simulation results show that the high-order polynomial response surface, the Kriging with Gaussian correlation function, the radial basis function with high-order monomial, the least squares support vector machine with Gaussian kernel function, the extreme learning machine with sine activation function and an ensemble of the above individual surrogate models have higher prediction accuracy, and the execution time of each individual surrogate model for a prediction of firing angle or uncontrolled lateral range is less than 1 ms, which verifies the effectiveness and feasibility of the proposed method. Moreover, the uncontrolled lateral range due to the earth's rotation is effectively reduced after the target azimuth correction.