Earthquake Damage Assessment Based on Fuzzy Logic and Neural Networks

Abstract

Potential damage assessment is fundamental for defining mitigation procedures and risk management strategies. Damage assessment involves the difficulties of defining, assessing, and modeling the variables involved, as well as handling uncertainty. Seismic damage estimation of structures does not only depend on the behavior of the structural system, but it involves other factors, which differ in nature. The paper presents a methodology for damage assessment of structures that combines systems theory, fuzzy logic, and neural networks. A feed-forward neural network supported on the systemic organization of information is used to assess the expected structural damage for a given earthquake. The methodology provides a very useful environment to consider the context of the building structure. The network has been trained using the damage observed in the recent earthquake that occurred in central Colombia. Several sets of structures were evaluated and the results compared to the damage observed. The model showed to be highly reliable and a good representation of experts' opinions. Computer software ERS-99 was developed and is currently being used for teaching and consulting purposes.