结构动力模型修正方法的比较研究及评估

COMPARISON AND EVALUATION OF ANALYTICAL APPROACHES TO STRUCTURAL DYNAMIC MODEL CORRECTION*

在实际工程中,由结构动力模型得到的计算值与通过试验获得的测量值间往往存在偏差,为了能够精确预测结构的动力响应,依据测量信息修正存在的动力模型是非常必要的.对现有几种有效的用于结构动力模型修正的理论方法（包括基于敏感性分析的矩阵型法、基于神经网络算法的参数型法和基于遗传优化算法的方法）做了详细的综述;介绍了这些方法的步骤和研究进展;并分析了这些动力模型修正方法在工程运用中存在的一些实际问题,如不完整的模态测量值、模型修正的鲁棒性、模型修正的计算效率和收敛性等.最后,通过对一实际的五层钢框架的动力模型修正,比较了这几种方法的优缺点,提出了今后需要解决的问题.

Considerable discrepancies exist between predictions from a structural dynamic model and experimental results of a laboratory model or actual structure when the two are compared. Updating the existing dynamic model based on modal test data is very important in order to precisely predict actual behaviors of the structure via the structural dynamic model. This paper reviews the procedures for refining a structural dynamic model from modal test data by different approaches which are used effectively, including the modal sensitivity method, neural networks method and Genetic algorithm, together with the recent research advances in this field. Some problems of dynamic model correction encountered in the actual applications such as incomplete modal test data and robustness of correction, as well as the factors affecting the computational efficiency and solution convergence, are discussed. Merits and defects of these proposed methods are also discussed and some current problems needed to be solved in the future are pointed out by comparison of numerical results of a real 5-story-steel-frame model updating from limited modal test data.