Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Slope Stability Analysis of Vetiver Grass Stabilized Soil Using Genetic Programming and Multivariate Adaptive Regression Splines
Machine learning (ML) has been successful in predicting factors of safety of slopes; however, it has not yet been utilized for slope stabilization with vetiver grass, which offers a natural, cost-effective, and environmentally friendly solution for slope stability due to its deep and dense root systems. The study utilizes genetic programming (GP) and multivariate adaptive regression splines (MARS) to simulate the vetiver-stabilized slope. For this purpose, a dataset of 96 was collected from lab experiments. The performance of the ML models is enhanced by hyperparameter tuning and k-fold cross-validation. Statistical performance parameters conclude that both the ML models have robust performance; however, the MARS model (R2 = 0.983 and RMSE = 0.0228) outperforms the GP model (R2 = 0.969 and RMSE = 0.0310). The conclusion is reiterated in both internal and external validation. The developed ML models demonstrate significant potential as a novel alternative tool to assist engineers in predicting FS during the design phase of various engineering projects. Additionally, MARS and GP provide user-friendly equations to predict the output, which can be applied with ease in engineering applications.
Slope Stability Analysis of Vetiver Grass Stabilized Soil Using Genetic Programming and Multivariate Adaptive Regression Splines
Machine learning (ML) has been successful in predicting factors of safety of slopes; however, it has not yet been utilized for slope stabilization with vetiver grass, which offers a natural, cost-effective, and environmentally friendly solution for slope stability due to its deep and dense root systems. The study utilizes genetic programming (GP) and multivariate adaptive regression splines (MARS) to simulate the vetiver-stabilized slope. For this purpose, a dataset of 96 was collected from lab experiments. The performance of the ML models is enhanced by hyperparameter tuning and k-fold cross-validation. Statistical performance parameters conclude that both the ML models have robust performance; however, the MARS model (R2 = 0.983 and RMSE = 0.0228) outperforms the GP model (R2 = 0.969 and RMSE = 0.0310). The conclusion is reiterated in both internal and external validation. The developed ML models demonstrate significant potential as a novel alternative tool to assist engineers in predicting FS during the design phase of various engineering projects. Additionally, MARS and GP provide user-friendly equations to predict the output, which can be applied with ease in engineering applications.
Slope Stability Analysis of Vetiver Grass Stabilized Soil Using Genetic Programming and Multivariate Adaptive Regression Splines
Transp. Infrastruct. Geotech.
Kumar, Nitish (Autor:in) / Kumari, Sunita (Autor:in)
Transportation Infrastructure Geotechnology ; 11 ; 3558-3580
01.10.2024
23 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
| DOAJ | 2023
| Europäisches Patentamt | 2019