Evaluation of Optimization Algorithms for Calibration of Numerical Models of Engineering Structures

Resumo

In civil engineering, the finite element method (FEM) is used to simulate the behavior of buildings, bridges, and other structures. However, there are several uncertainties that make it difficult to obtain a representative computational model suitable for evaluation and monitoring of existing structures. To overcome this, the calibration of the numerical model based on experimentally obtained modal data is employed. This step allows for the updating of parameters that are difficult to obtain directly, in order to approximate the dynamic response of the computational model to that of the real structure. For this purpose, various algorithms can be employed, presenting differences in approach, effectiveness, and computational cost. This study aims to compare the performance of different optimization algorithms applied in the calibration of numerical models. To achieve this, the model of a structure will be established beforehand, with initial uncertainties represented by masked parameters; a set of modal parameters will be extracted, and then a new calibrated model will be developed through distinct optimization processes. The calibrated models will be compared with the initial model to evaluate their performance in predicting structural behavior, considering aspects such as accuracy in reproducing initially unknown parameters and the computational cost of each solution. Thus, it will be possible to enhance the calibration process by identifying the most appropriate algorithm.