Damage Detection Based on Artificial Immune System

摘要

This study has underlined that a negative selection algorithm, the greedy algorithm, is able to detect efficiently the occurrence of damages in a structure directly based on raw acceleration data. According to the chosen combinations of parameters, it is possible to reach more than 95% of detection. Moreover, by using the differences of relative accelerations between successive stories, the localization of damages is also possible due to the link between the highest detection rate and the location of the damage. Furthermore, in order to obtain a larger difference between detection rates for the localization step, it is necessary to work on more sensitive signals. Another representation of data can also be planned for the future, using for example real-valued vectors or even a hybrid representation.