Machine learning-based model for recognizing the failure modes of FRP-strengthened RC beams in flexure

Abstract

Plate end (PE) debonding and intermediate crack (IC) debonding are the main failure modes of fiber reinforced polymer (FRP) strengthened reinforced concrete (RC) beams in flexure. Different failure modes exhibit different failure characteristics. Therefore, accurately identifying the failure mode is of great significance in selecting methods to prevent the debonding of the strengthened beams. This study first established a primary indicator system through literature research. Then one hundred and eighty-eight FRP-strengthened RC beams containing PE and IC debonding were collected from the published literature of forty-eight researchers. After that, correlation analysis and grey correlation analysis were used to study the data of two failure modes. Finally, the indicator system for the prediction of failure modes was established. After the indicator system was established, six machine learning algorithms, including K-nearest neighbor algorithm (KNN), decision tree (DT), random forest (RF), back propagation neural network (BPNN), logistic regression (LR), and support vector machine (SVM), were used to build the prediction model of failure modes. The evaluation of the models shows that the coefficient of variation of the accuracy of the decision tree is merely 5.4%, which has the best robustness; the average accuracy of the random forest for the two failure modes reaches 93% and 98%, which has the highest precision.