Scholarlink

<p>Accurately predicting factor of safety (FoS) is a crucial step in the process of designing and stability assessment of embankments, particularly in those supported by <i>geogrid encased stone columns</i> (GESCs), by ensuring optimized design, preventing structural failure and reducing construction cost. In this regard, this study utilizes the potential of four machine learning models presented in probabilistic neural network (PNN), generalized regression neural network (GRNN), radial basis function neural network (RBF), and artificial neural network (ANN) in predicting the FoS of embankments supported by stone columns encased with geogrid. The adopted models were evaluated in the training, validation, and testing phases using various statistical matrices and graphical analysis. The findings showed that the GRNN model offers performance across all phases, particularly in the testing phase, with an R-value of 0.985, indicating a high linearity between the actual values and the predicted value induced, as well as lower prediction deviations. The ANN also obtains a reliable prediction by offering high prediction accuracy (<i>R</i> = 0.970) and lower prediction deviations, highlighting it as the second-best model and a strong alternative to the GRNN model. Conversely, the RBF model showed the lowest performance across all phases, showing higher prediction deviations and lower prediction accuracy. The analysis revealed that the key factors affecting safety are the diameter of the stone column, the embankment angle, and the stiffness of the geogrid.</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

Predicting the factor of safety in embankments improved by geogrid-encased stone columns using various neural network architectures

  • 期刊论文

  • 发表日期：

    2025年9月30日

查看全文
  • Omar H. Jasim,
  • Faidhalrahman Khaleel,
  • Deiaaldeen Khaleel,
  • Mohammed Y. Fattah,
  • Ali AbdulJabbar Alfahad,
  • Alaa H. AbdUlameer,
  • Haitham Afan

摘要

Accurately predicting factor of safety (FoS) is a crucial step in the process of designing and stability assessment of embankments, particularly in those supported by geogrid encased stone columns (GESCs), by ensuring optimized design, preventing structural failure and reducing construction cost. In this regard, this study utilizes the potential of four machine learning models presented in probabilistic neural network (PNN), generalized regression neural network (GRNN), radial basis function neural network (RBF), and artificial neural network (ANN) in predicting the FoS of embankments supported by stone columns encased with geogrid. The adopted models were evaluated in the training, validation, and testing phases using various statistical matrices and graphical analysis. The findings showed that the GRNN model offers performance across all phases, particularly in the testing phase, with an R-value of 0.985, indicating a high linearity between the actual values and the predicted value induced, as well as lower prediction deviations. The ANN also obtains a reliable prediction by offering high prediction accuracy (R = 0.970) and lower prediction deviations, highlighting it as the second-best model and a strong alternative to the GRNN model. Conversely, the RBF model showed the lowest performance across all phases, showing higher prediction deviations and lower prediction accuracy. The analysis revealed that the key factors affecting safety are the diameter of the stone column, the embankment angle, and the stiffness of the geogrid.