针对某导弹平台调平过程难以建立解析模型的问题,根据平台调平过程历史数据,采用支持向量回归机分别对平台的俯仰角θ₁、偏航角θ₂、滚动角θ₃、加速度计当量w₁、w₂、w₃、调平输出当量C₁₄七个参数的时间序列进行模型辨识。利用参数的前5组数据作为输入,后1组数据作为输出,利用最小二乘法对支持向量回归机进行训练。仿真结果证实,支持向量回归机模型辨识预测误差2.56%,满足导弹平台调平过程建模要求。

To overcome the difficulty of analytic modeling of missile platform, support vector regression (SVR) was adopted to identify the model of platform parameter, including pitch angle θ₁, yaw angle θ₂, roll angle θ₃, accelerometer equivalent w₁,w₂,w₃, shifting output equivalent C₁₄. The inputs of SVR were four groups' data, and the output was one group of data. The SVRs were trained by least square algorithm. The simulation demonstrates that the model identification testing precision is 2.56% which satisfies requirement of missile platform shifting process modeling.