Finite element model updating using multi-objective optimization with kriging surrogate model

Abstract

As the number of the existing infrastructure has been gradually increased, its maintenance is one of the most important issues worldwide. By accurately assessing the state of the structures, it is possible to predict the upcoming issues, and conduct safety assessment. One example of assessing the state of the structures, FE model can be adaptive in many fields such as safety assessment, risk prediction, and risk assessment. However, there is a discrepancy between FE model and existing structure due to deterioration of material, unexpected loading, and so on. Therefore, FE model updating has been widely performed to reflect the existing structures. Based on measurement data, updating parameters are tuned to fit the target output. Updated model can be used as response prediction and condition. FE model updating can be divided into two categories: (1) single-objective optimization and (2) multi-objective optimization. In case of single-objective optimization, weighting factors can be set by user’s knowledge and experience. In other words, optimal solution cannot be obtained properly. Multi-objective optimization does not have to consider weighing factors. From Pareto-optimal front, optimal solution can be found. It needs excessive calculation time to obtain the solution. Therefore, Kriging surrogate model can apply on FE model updating to reduce the calculation time and obtain accurate result. In this study, AMALGAM of the multi-objective optimization algorithms is used as FE model updating with Kriging surrogate model. It is validated with field test data which are obtained in a steel plate girder bridge and PSC I girder bridge.