| Title |
Machine Learning Based Estimation of Structural Capacity of Unreinforced Masonry Buildings |
| Authors |
음영채(Eum, Yeong Chae) ; 신동현(Shin, Dong-Hyeon) |
| DOI |
https://doi.org/10.5659/JAIK.2025.41.7.223 |
| Keywords |
Unreinforced Masonry Building; Structural Capacity; Machine Learning Model; Damage States |
| Abstract |
Most unreinforced masonry(URM) buildings have been built without considering seismic design requirements. In order to enhance the seismic
performance of such URM buildings by adopting a retrofit method, it is required to precisely understand the capacity of URM buildings.
However, there is significant variability of predicted capacity of masonry walls due to various uncertainty sources including material
characteristics of components which are hard to be quantitatively defined. In this reason, the main purpose of this study is to develop a
machine learning based structural performance evaluation model for URM walls. Four machine learning models including ensemble, support
vector machine(SVM), kernel, deep neural network(DNN) were trained using geometrical dimensions and material properties of masonry
components as input features for predicting their capacities. The machine learning based performance prediction models were quantitatively
evaluated using mean squred error(MSE), root mean squred error(RMSE), mean absolute error(MAE), and R-squred(R²). Of considered
machine learning based models, the ensenble model achieved the best overall prediction accuracy across all evaluation parameters. The trained
model was validated using independent experimental data not included in the training process, resulting in an average error rate of less than
8% compared to the test results. Based on this machine learning based model, this study finally suggest the structural performance evaluation
procedure for URM buildings. |