Damage identification using experimental modal analysis and adaptive neuro-fuzzy interface system (ANFIS)

Abstract

The adaptive neuro-fuzzy inference system (ANFIS) is a process for mapping from a given input to a single output using the fuzzy logic and neuro-adaptive learning algorithms. Using a given input-output data set, ANFIS constructs a Fuzzy Inference System (FIS) whose fuzzy membership function parameters are adjusted using combination of back propagation algorithm with a least square type of method. The feasibility of ANFIS as strong tool for predicting the severity of damage in a model steel girder bridge is examined in this research. Reduction in the structural stiffness produces changes in the dynamics properties, such as the natural frequencies and mode shapes. In this study, natural frequencies of a structure are applied as effective input parameters to train the ANFIS and the required data are obtained from experimental modal analysis. The performance of ANFIS model was assessed using Mean Square Error (MSE) and coefficient of determination (R 2). The ANFIS model could predict the severity of damage with MSE of 0.0049 and correlation coefficient (R 2) of 0.9976 for traing data sets. The results show the ability of an adaptive neuro-fuzzy inference system to predict the damage severity of the structure with high accuracy. © The Society for Experimental Mechanics, Inc. 2012.