谱约束的矩阵逼近法是结构设计中模型修正的一种重要方法。给出了广义中心对称谱约束解的存在条件和一般解的表达式,对于一个预估矩阵,提供了它的逼近解的表达式,讨论了该逼近解的误差界,并且提供了一个数值方法和数值例子,数值结果表明方法可行且是有效的。
The best approximation of a matrix with assigned spectra is an important method of correcting a model in the structural design. The solvability conditions for the assigned spectra of a generalized centrosymmetric matrix are established and an explicit expression of the solutions is derived. For an estimated matrix, the associated best approximation solution is deduced. Moreover, a theoretic error bound of the best solution is proposed. A numerical method is provided and an example is also carried out. Numerical examples show that the method is reliable and effective.