Scholarlink

<p>The unreinforced masonry (URM) walls serve as the primary load resisting elements in URM structures, supporting gravity loads from the floors and lateral loads during earthquakes. The in-plane lateral strength and drift capacity of URM walls are key parameters necessary in the seismic vulnerability assessment of existing URM buildings. Analytical and empirical expressions have been developed to estimate the ultimate in-plane strength and drift capacity of URM walls. However, based on test results and observations of damaged buildings in earthquakes, URM walls typically exhibit a hybrid failure mode. Therefore, there is a need for a more comprehensive predictive model that can accurately estimate the in-plane strength and drift capacity of URM walls regardless of the dominant failure mode. By developing advanced data analysis and artificial intelligence (AI) algorithms and applying them in structural engineering, this study investigates the effectiveness of these algorithms in predicting the in-plane behavior of URM walls. In this regard, a dataset of in-plane tests on 191 URM walls is compiled. Moreover, six well-known machine learning (ML) algorithms including Decision Tree, Random Forest, Gradient Boosting, Extreme Gradient Boosting, Light Gradient Boosting Machine, and Bagging Regressor are employed to predict the in-plane strength and drift capacity of the walls. The input variables considered in this study are the geometrical characteristics of the walls, boundary conditions, and the compressive strength of the masonry. According to the analysis results, despite the high level of uncertainties associated with masonry material, the machine learning algorithms could estimate the in-plane strength and drift capacity of the walls with an acceptable level of accuracy. Among the examined algorithms, Gradient Boosting showed the best performance in predicting both the ultimate strength and drift capacity of the walls. In addition, a finite element (FE) model for URM walls was developed and a parametric study was conducted. Comparing the in-plane strength of the walls obtained from the FE study and the proposed models demonstrates the acceptable accuracy of the developed machine learning-based predictive models.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predictive models for the in-plane strength and drift capacity of unreinforced masonry walls with various types of failure modes using machine learning algorithms

  • 期刊论文

  • 发表日期：

    2026年2月10日

查看全文
  • Mohammad Amir Najafgholipour,
  • Mahsa Zohrei

摘要

The unreinforced masonry (URM) walls serve as the primary load resisting elements in URM structures, supporting gravity loads from the floors and lateral loads during earthquakes. The in-plane lateral strength and drift capacity of URM walls are key parameters necessary in the seismic vulnerability assessment of existing URM buildings. Analytical and empirical expressions have been developed to estimate the ultimate in-plane strength and drift capacity of URM walls. However, based on test results and observations of damaged buildings in earthquakes, URM walls typically exhibit a hybrid failure mode. Therefore, there is a need for a more comprehensive predictive model that can accurately estimate the in-plane strength and drift capacity of URM walls regardless of the dominant failure mode. By developing advanced data analysis and artificial intelligence (AI) algorithms and applying them in structural engineering, this study investigates the effectiveness of these algorithms in predicting the in-plane behavior of URM walls. In this regard, a dataset of in-plane tests on 191 URM walls is compiled. Moreover, six well-known machine learning (ML) algorithms including Decision Tree, Random Forest, Gradient Boosting, Extreme Gradient Boosting, Light Gradient Boosting Machine, and Bagging Regressor are employed to predict the in-plane strength and drift capacity of the walls. The input variables considered in this study are the geometrical characteristics of the walls, boundary conditions, and the compressive strength of the masonry. According to the analysis results, despite the high level of uncertainties associated with masonry material, the machine learning algorithms could estimate the in-plane strength and drift capacity of the walls with an acceptable level of accuracy. Among the examined algorithms, Gradient Boosting showed the best performance in predicting both the ultimate strength and drift capacity of the walls. In addition, a finite element (FE) model for URM walls was developed and a parametric study was conducted. Comparing the in-plane strength of the walls obtained from the FE study and the proposed models demonstrates the acceptable accuracy of the developed machine learning-based predictive models.