Machine Learning-Based Failure Mode Recognition of Circular Reinforced Concrete Bridge Columns: Comparative Study

Abstract

The prediction of failure mode of columns is critical in deciding the operational and recovery strategies of a bridge after a seismic event. This paper contributes to the critical need of failure mode prediction for circular reinforced concrete bridge columns by exploring the capabilities of machine learning methods. Three types of failure mode such as flexure, flexure-shear, and shear are considered in this study, and 311 specimens are compiled from experimental studies on the circular columns. The efficiency of various machine learning models such as quadratic discriminant analysis, K-nearest neighbors, decision trees, random forests, naive Bayes, and artificial neural network is evaluated using a randomly assigned test set from the collected data. It is noted that artificial neural network has superior performance amongst all the machine-learning methods, and the comparison of this classification with the existing methods underscores the advantage of the artificial neural network in failure mode recognition. Classification based on artificial neural network is 91% accurate in identifying the failure mode of the collected experimental data.