次优鲁棒函数观测器设计-计算数学专业论文.docx

Classified Index: TP13 U.D.C: 681.5 Dissertation for the Master Degree in Science Design of Suboptimal Robust Functional Observers Candidate： Jiang Yuebin Supervisor： Associate Prof. Zhang Biao Academic Degree Applied for： Master of Science Speciality： Applied Mathematics Affiliation： Department of Mathematics Date of Defence： June, 2009 Degree-Conferring-Institution： Harbin Institute of Technology 摘 要 本文提出了一种新的鲁棒函数观测器设计方法。首先通过求解 Sylvester 矩阵方程，得到系数矩阵 T，L 的参数形式。T，L 均为自由参数的线性组合 的形式，便于进行梯度计算。然后对于矩阵方程 K = NT + MC 进行转化，并 取定 N，求出矩阵 M 的参数形式。将求得的参数矩阵 T，L，M 带入优化指 标，对优化指标进行求导，得到指标梯度的显式表示。最后，利用共轭梯度 法进行指标优化，得到观测器设计中各个系数矩阵的最优数值解。 与已有方法相比，该方法的主要优点体现在：利用梯度方法求解优化问 题，收敛的速度很快；它的次优性在于将系数矩阵N事先取定，将减少系统 自由参数的个数，意味着观测器设计解的自由度减少，但通过这样处理，优 化问题中复杂的非线性约束转化为线性约束，使算法得到简化。 在文章的最后给出了上述鲁棒函数观测器设计方法的数值例子，并且在 第一个例子中分别给出了系统存在结构扰动和不存在结构扰动情况下的数值 模拟，通过计算结果可以看出文中提出的方法是有效的。 关键词 线性系统；鲁棒性；观测器设计；梯度 Abstract In this paper, a new approach for Design of Robust Functional Observers is presented. First, by solving Sylvester matrix equation, we obtain the parametric forms of T, L, which are in the form of linear combination of design parameters, and make the calculation of gradient much easy. Then the parametric solution of M can be obtained based on dealing with the equation K = NT + MC . Substituting T, L and M into objective function and differentiating the objective function, the explicit parametric representations of the objective function′s gradient is given. Finally, solving the optimization problem with Conjugate Gradient Method, the numerical solution of V, W and M are obtained. Compared with the other approaches, the main advantage of this approach is solving the optimization problem by applying gradient method, which leads to rapid convergence. Its disadvantage is that the matrix N must be determined before solving the equation K = NT + MC . This means the number of design parameters is reduced. However, by this way, the nonlinear constraint of the optimization problem is turned into linear one. Some n