Interactive Visualisation — A Support for Model Updating

Accurate modelling of masonry wall panels has proved to be extremely difficult. One reason for this is the complex and highly composite nature of the material constituents that the masonry is composed of. In the previous research by the authors `corrector factors' were introduced to model variation in the properties of the masonry wall panels subjected to lateral loads. The concept of corrector factors was based on a numerical model updating approach, which was aimed at minimising the error between the measurement and analytical load deflection values over the entire surface of the panel. Due to the inverse problem nature of the model updating process, a number of scenarios could result in similar error. In other research by the authors, multi-objective evolutionary computing (EC) methods were used to identify best fit solution(s). One of the common approaches in EC is to combine objectives using a weighted sum of objectives. This approach needs careful selection of weights for various objectives which is an iterative approach and may add to uncertainties inherent to model updating techniques. This paper uses an interactive visualisation clustering genetic algorithm tool (the IVCGA) that identifies clusters of good solutions that best suit the problem. This tool allows users to interact with the system to explore the suitability of solutions within these clusters using their expert knowledge of the problem. This paper used individual runs of the IVCGA for each objective independently and identifies a `mutually inclusive region' for all objectives. The paper demonstrates that further exploration of solutions inside this mutually inclusive region can lead to identification of solutions that perform much better than the weighted sum approach and this could result in to finding the correct model for the problem.