【作者】 王建宏； 王道波； 王志胜；

【Author】 Jianhong Wang,Daobo Wang,Zhisheng Wang College of Automation Engineer,Nanjing university of Aeronautics and Astronautics,Nanjing,210016

【机构】 南京航空航天大学自动化学院；

【摘要】 对于同时存在输入和输出观测噪声的飞行器颤振试飞试验的随机模型,本文借助于频域极大似然估计原理推导了该随机模型的极大似然代价函数的简化形式。在此基础上为了降低收敛于局部最小的可能性,弱化算法对初始值的依赖性,利用全局优化理论推导了全局优化的迭代卷积平滑辨识方法。该辨识方法通过引入随机抑制项来调整迭代算法,使算法每次都收敛到全局最小值。最后利用试飞试验数据辨识飞行器的系统参数,验证了该方法的有效性。更多还原

【Abstract】 For stochastic models with input and output measurement noises in aircraft flutter experiment,the maximum likelihood cost function’s simple form is firstly proposed by means of frequency domain maximum likelihood estimation principle.Then a global optimization iterative convolution smoothing identification method is derived to significantly reduce the possibility of convergence to a local minimum and weakly dependent of the starting values’ choice by using the global optimization theory.The identification method modifies the iterative method with a stochastic perturbation term and guarantees the algorithm converge to a global minimum.The simulation with real flight test data shows the efficiency of the algorithm.更多还原