中文摘要：

耦合局部最优法作为一种新型的优化技术，既具有高效的搜索速度又具有全局搜索能力。然而，对于大规模优化问题，该方法容易陷入局部最优；另外，梯度信息在该项技术中起着重要作用，而对于复杂问题往往不能得到精确的梯度信息，从而使得该算法的全局搜索能力下降。本文分别从初始种群的确定、变步长搜索、自调节种群三方面对原算法进行了改进，提出了自适应耦合局部最优法，使之具备解决多变量复杂优化问题的能力。通过两个测试函数验证了改进算法比原有算法更易于得到全局最优解并保持较高的计算效率。最后，采用一个试验算例验证了自适应耦合局部最优法的有效性。

英文摘要：

As a new optimization technique, coupled local minimizers （CLM） possesses a high search speed and global search capability. However, this technique is prone to plunge into local optimum for large scale optimization problem such as that rendered by structural model updating. Furthermore, the gradient information plays an important role for CLM, and when this information can not be accurately obtained, which is usually the case in dealing with real complex problems, the global search capacity of CLM may be limited. In this paper, amendments on the CLM are made in three aspects, including the selection of initial population, the variable-step searching technique and the adaptive population. These amendments give an adaptive capability of CLM, and enable the adaptive CLM to deal with large scale complex optimization problems. The adaptive CLM method is demonstrated to be more efficient and more prone to a global optimum than the CLM method by two test functions. Finallv the adaDtive CLM is emDloved to uDdate the finite element model of a beam-llke structure.