C SUI (1), M Y RAFIQ (1), D EASTERBROOK (1), G BUGMANN (1) and G ZHOU (2)
(1) University of Plymouth, UK
(2)HarbinUniversityof Technology (HIT), China

The nature of model updating procedure is, in general, a reverse problem solving process. Due to the complexity of the structure and the number of variables involved, a number of models would exist that partially satisfy necessary requirements. Investigation into model updating process shows that due to compensatory effects of large number of variable involved, values of variables obtained during model updating process are different or even sometimes some of the variable values appear to be illogical. It is crucial to identify those variable values that not only improve the prediction of the model, but also do not violate basic established rules. In this paper, an extensive parametric study has been conducted in which an evolutionary computation (EC) technique is used to explore a population of feasible models for masonry wall subjected to lateral load. A non-linear finite elements analysis (FEA) program is called by the objective function of the EC to calculate failure load and load deflection values at various critical locations over the surface of the panel. These values are then compared with the experimental results and the error between the two values are minimised. Domain knowledge is used to select those models with parameter values that do not violate established design rules. The final parameter values are then used to predict failure load, failure pattern and load deflection values of any unseen panel with or without openings.

Key words
Uniqueness, model updating, optimisation, masonry panel, lateral load