Data mining-based damage identification of a slab-on-girder bridge using inverse analysis

Abstract Classical damage detection methods such as visual inspections have many limitations, i.e. time consuming procedure, costly process and ineffective for large and complex structural systems. To overcome these difficulties, a data mining-based damage identification approach is developed in this study. First four natural frequencies which obtained from the experimental modal analysis of a slab-on-girder bridge structure are used as an input database. The laboratory work is carried out through single-type and multiple-type damage scenarios. The applicability of machine learning, artificial intelligence and statistical data mining techniques are here examined using Support Vector Machine (SVM), Artificial Neural Network (ANN) and Classification and Regression Tree (CART) to predict the model behavior and damage severity. Then, a hybrid algorithm is proposed in the deployment step of Cross Industry Standard Process for Data Mining (CRISP-DM) model. According to the obtained results, the hybrid algorithm performs a better accuracy in compare to ANN technique itself.