ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM IN STRUCTURAL DAMAGE ASSESSMENT

摘要

The non-linear dynamic analysis of constructions is a timernconsuming process. This paper proposes a method basedrnon an Adaptive Neuro-Fuzzy Inference System (ANFIS)rnfor automatic seismic structural damage classification. Arnset of 200 artificial accelerograms is utilized. The severityrnof these seismic excitations is described by 20 widelyrnused seismic parameters. The global structural damagernindex Maximum Inter-Storey Drift Ratio (MISDR) is usedrnas metric to quantify and classify the structural damagerncaused to a certain structure, into 4 categories (negligible,rnweak, moderate, severe). Results indicate that thernproposed model is suitable for adaptive predictive controlrnof the behavior of the construction for seismic signals andrnits accuracy rates up to 89.5%. The proposed method isrncompared to previous reported fuzzy techniques. Furtherrnstudy attempts to decrease the number of parameters usedrnand maintain the same level of classification performance.