Keywords: seismic parameters, damage indices, damage classification, fuzzy logic, artificial neural networks, neuro-fuzzy, adaptive neuro-fuzzy inference systems, support vector machines.

It is well known that post-seismic damage observations on buildings are interrelated with the seismic intensity parameters. In addition, the seismic response of buildings is directly dependent on the ground excitation. Consequently, the seismic response of buildings is directly dependent on the accelerograms used and their intensity parameters. From the several response parameters the focus is on the overall damage indices because they summarise the post-earthquake status of buildings on a single value, which can be easily handled. The global structural damage index of Park/Ang and the maximum inter-storey drift ratio are considered. Intervals for the values of the damage indices are defined to classify the damage degree as low, medium, large and total. This chapter first provides a systematic overview of seismic intensity parameters used in earthquake engineering. Twenty peak, spectral and energy parameters are presented. Alternative intelligent procedures for the damage classification on buildings using the intensity parameters of the accelerograms are presented and discussed. These are based on fuzzy logic, artificial neural network, neuro-fuzzy, adaptive neuro-fuzzy inference systems and support vector machines techniques. The techniques presented have been applied to the damage classification of two reinforced concrete frame structures, designed according to the rules of the Eurocodes 2 and 8. The nonlinear dynamic calculations have been considered as reference solutions, employing several natural and artificial accelerograms. The efficiency of the presented intelligent techniques was expressed by the correct classification rates of a set of accelerograms that were not used during the training process. Thus, the fuzzy logic procedures provided a correct classification rate of 84% for Gaussian and 83.5% for triangular membership functions. Further, using artificial neural networks, the correct classification rates are between 75% and 97.5%. In addition, the neuro-fuzzy model provides correct a classification rates between 71% and 93.5%. On the other hand, the adaptive neuro-fuzzy inference systems show a correct classification rate of 89.5%. Finally, using the support vector machines as classification technique, the correct classification rate is between 87.5% and 93.5%. The results depend on the used overall structural damage indices (maximum inter-storey drift ratio or the global damage index of Park/Ang) and the number of training samples. All these results lead to the conclusion that the presented intelligent techniques are appropriate procedures for the correct classification of seismic excitation based on the seismic intensity parameters of their accelerograms.

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £85 +P&P)