基于改进增广径向基的固体姿轨控发动机推力快速预示

Fast Thrust Prediction Method for Solid Divert and Attitude Control System Based on Improved Augmented Radial Basis Functions

关键词:  固体姿轨控发动机  增广径向基函数  各向异性  代理模型  交叉验证  推力预示

Key words:  Solid divert and attitude control system  Augmented radial basis functions  Anisotropy  Surrogate model  Cross validation  Thrust prediction

摘要:

为实现沉积影响下的喉栓式固体姿轨控发动机推力快速预示，提出一种融合各向异性和快速交叉验证的增广径向基近似建模方法。基于正交条件构建混沌多项式增广的径向基混合模型，并提出基于样本局部密度的各向异性方法，进一步提高混合模型精度。采用递归演化的拉丁超立方试验设计生成样本点，根据通用交叉验证误差求解过程中高阶矩阵快速求逆方法，降低模型训练的计算复杂度。与其他常用代理模型方法相比，本文提出的方法具有更好的精度和稳定性。将本文方法应用于固体姿轨控发动机推力快速预示，预示结果与仿真结果偏差控制在2.5%以内，计算耗时由小时级降低至秒级。

Abstract:

In order to realize the rapid prediction of the thrust of solid divert and attitude control system （SDACS） with pintle under the influence of deposition， this paper proposes an augmented radial basis function metamodeling method based on anisotropic and fast cross validation. Firstly， a polynomial chaos expansions augmented radial basis functions metamodel is constructed based on orthogonal conditions， and an anisotropic method based on local density of samples is proposed to further improve the accuracy of the metamodel. Secondly， the recursive evolution Latin hypercube design is used to generate sample points. According to the fast inversion method of high-order matrix in the process of solving the universal cross-validation error， the computational complexity of model training is reduced. Compared with other common metamodels， the proposed method has better accuracy and stability. The proposed method is applied to the thrust fast prediction of solid divert and attitude control system. The deviation between the predicted results and the simulation results is controlled within 2.5%， and the calculation time is reduced from hour level to second level.