C SUI (1), M Y RAFIQ (1), D EASTERBROOK (1), G BUGMANN (1) and G ZHOU (2)
(1) University of Plymouth, UK
(2) Harbin University of Technology (HIT), China
It is not easy to construct an accurate FEA model for laterally loaded masonry panels. Model updating has been an effective way to improve an existing FEA model, however, model updating is a reverse problem solving process, which means there maybe many solution, some solutions are the local optima which mislead the model to be a wrong physical meaning, or lost the physical meaning. Therefore, the uniqueness study to the solutions is very necessary. In this paper, the uniqueness of the solution of updating a FEA model of laterally loaded masonry panel by genetic algorithm (GA) was studied, so as to clarify whether the obtained solutions reflect the real condition of the structure and could be used to make effective revision to the model. The study is carried out in the following ways: Firstly, compare solutions under all possible permutations of assumed boundary conditions in which the permutation of two kinds of constraint function was used. Secondly, compare solutions under all possible permutations of assumed boundary conditions in which the stiffness close to the supported boundaries are changed. Thirdly, compare the solutions under the condition of changing the mesh sizes. It has been found that the solutions indicate the same physical meaning and reveal the same problem of the existing FEA model; they also make similar revision to the model although their values were different. Consequently, solutions could be regarded as unique and the model updating results by the proposed model updating methodology is credible for improving the existing FEA model.
Uniqueness, Model updating, optimisation, masonry panel, FEA, genetic algorithm