model based damage detection of concrete bridge deck using adaptive neuro-fuzzy inference system

concrete bridge deck damage detection by measurement and monitoring variables related to vibration signatures is one of the main tasks of any bridge health monitoring system (bhms). generally damage puts some detectable/discoverable signs in the parameters of bridge vibration behavior. however, differences between frequency and mode shape before and after damage are not remarkable as vibration signatures. therefore most of the introduced methods of damage detection cannot be used practically. among many methods it seems that models based on artificial intelligence which apply soft computing methods are more attractive for specific structures. in this paper an adaptive neuro-fuzzy inference system (anfis) is used to detect the damage location in a concrete bridge deck modeled by finite element method. some damage scenarios are simulated in different locations of the deck and accelerations as representatives of response at some specific points are calculated. excitement is done by applying an impact load at the center of the deck. in the proposed anfis damage detection model accelerations are inputs and location of the damage is output. trained model by simulated data can show the location of the damage very well with a few training data and scenarios which are not used in training stage. this system is capable to be included in real-time damage detection systems as well.