Dynamic Inference and Assessment of Structural Damage States Based on Naïve Bayesian Classifier

Abstract

The monitoring, inference and assessment of structural damage states have become a significant problem to be studied. In this paper, a framework based on Nai Bayesian Classifier (NBC) is proposed to infer and assess structural damage states. The framework includes data collection, feature extraction and damage states evaluation. NBC is employed to find and create the correlation between Damage Sensitive Feature (DSF) and Structural Damage States Transition (SDST). The correlation helps to shorten and fasten the mapping procedure from DSF to damage states evaluation. A numerical simulation experiment of a two-story frame structure based on seismic wave input is presented. Firstly, the framework collects data obtained from accelerometer sensors and extracts maximum structural response from it as DSF. Subsequently, Story Drift Ratio (SDR) can be calculated as damage indicator to determine the damage state level of this two-story frame structure. A NBC model is created to simplify the mapping procedure and improves the accuracy. Finally, the experiment results prove the feasibility and efficiency of this framework. Meanwhile, NBC model is proved to be useful and its accuracy meets the requirements of online real-time dynamic structure damage states evaluation