论文部分内容阅读
针对导弹伺服机构液压源液面下降的问题,根据导弹伺服液面实测数据,采用支持向量机、神经网络集成和最小二乘多项式拟合3种数据驱动方法对伺服液面高度进行模型辨识。通过对实测数据仿真分析,发现不同输入维数对预测精度有所影响:当输入维数为4时,支持向量机预测误差最低;不同输入维数下,最小二乘多项式预测误差最稳定,且综合误差最小。
According to the measured data of the missile’s servo surface, three kinds of data driven methods based on the support vector machine, neural network integration and least square polynomial fitting are used to model the height of the servo liquid surface. Through the simulation analysis of the measured data, it is found that the different input dimensions have an impact on the prediction accuracy: the prediction error of the SVM is the lowest when the input dimension is 4; and the least squares polynomial prediction error is the most stable under different input dimensions The smallest integrated error.