Scholarlink

<p>In this Study, behaviour and response of self-centered post-tensioned concrete rocking wall was studied by machine learning techniques, developing multi-objective neural network (MONN) model. To generate required database for developing the proposed MONN, experimental results were employed, nine effective input parameters were selected for training the network model and sensitivity analysis of input parameters was conducted. Two learning algorithms were chosen for the neural network model and their performance were compared. Optimal number of neurons and the percentage of data used in the training and testing sets were determined for the MONN model. The model was verified by experimental results and showed that the proposed method was able to accurately predict lateral and moment strengths and neutral axis. The mean predicted-to-tested values for MONN were within range of 0.944 to 1.104 for lateral strength and 0.886 to 1.148 for moment strength of experimental results, while the existing procedures had relatively inaccurate estimations ranged from 0.694 to 1.408. Prediction of these parameters by MONN model was a new approach in field of designing these systems and a new effort for investigating their behaviour that can be used as an alternative to the complex modelling methods or expensive and time-consuming experimental tests.</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

Analytical study on behaviour of self-centered concrete walls using novel multi­objective artificial neural network

  • 期刊论文

  • 发表日期：

    2025年6月18日

查看全文
  • Amin Foyouzati,
  • Alireza Khaloo

摘要

In this Study, behaviour and response of self-centered post-tensioned concrete rocking wall was studied by machine learning techniques, developing multi-objective neural network (MONN) model. To generate required database for developing the proposed MONN, experimental results were employed, nine effective input parameters were selected for training the network model and sensitivity analysis of input parameters was conducted. Two learning algorithms were chosen for the neural network model and their performance were compared. Optimal number of neurons and the percentage of data used in the training and testing sets were determined for the MONN model. The model was verified by experimental results and showed that the proposed method was able to accurately predict lateral and moment strengths and neutral axis. The mean predicted-to-tested values for MONN were within range of 0.944 to 1.104 for lateral strength and 0.886 to 1.148 for moment strength of experimental results, while the existing procedures had relatively inaccurate estimations ranged from 0.694 to 1.408. Prediction of these parameters by MONN model was a new approach in field of designing these systems and a new effort for investigating their behaviour that can be used as an alternative to the complex modelling methods or expensive and time-consuming experimental tests.