The case history of a road embankment constructed over soft soil stabilized with prefabricated vertical drains (PVDs) is presented. Modelling very complicated subsoil and structural elements could be assisted by numerical methods. Finite element analysis (FEA) was conducted in this study to verify the effectiveness of PVD modelling in the subsoil utilizing Plaxis 2D. The field settlement data were collected and compared to FEM predicted with different smear effect permeability ratio. Thereafter, predicted settlement data relatively accurate was used in the back-calculated to determine the ultimate settlement with the Asaoka and hyperbolic methods. The ultimate settlement predicted with the smear effects permeability ratio is 6 similar to measured. Asaoka method was found to be reliable in estimating settlement of ground foundation.

This paper presents an exhaustive review of the challenges faced in the construction of road embankments on soft ground and proposes a direction for future development. Frequently used techniques for soft ground improvement are discussed. The factors that contribute to the stability of the road embankment are reviewed by approach, results of past studies, and historical cases. The findings show that settlement, slope stability, and soil bearing capacity are all challenges to constructing the road embankment. Additionally, it is found that geometric data is a key factor in embankment design. Pre-loading with prefabricated vertical drain (PVDs) methods and lightweight fill were found to be widely used techniques in soft ground improvement. The information from this study can be used to develop design guidance systems, numerical modelling, and to give an overview and knowledge to other researchers who are or will conduct research in this field. Finally, future perspectives for research are related to predictions of factors that affect the stability of road embankment with an artificial intelligence approach. ABSTRAK: Kertas ini membentangkan ulasan kajian menyeluruh mengenai cabaran yang dihadapi dalam pembinaan benteng jalanraya di atas tanah lembut dan mencadangkan ke arah pembangunan kajian masa depan. Teknik-teknik penambahbaikan tanah lembut yang sering digunakan turut dibincangkan. Faktor- faktor yang menyumbang kepada kestabilan benteng jalanraya diulas dengan pendekatan kepada kajian lepas dan sejarah kes. Hasil kajian ini didapati bahawa enapan, kestabilan cerun dan keupayaan galas tanah merupakan cabaran dalam pembinaan benteng jalanraya. Selain itu, ia didapati bahawa data geometri merupakan faktor penting kepada rekabentuk benteng. Kaedah pra pembebanan dengan prefabrikasi saliran menegak (PVDs) dan isian ringan didapati teknik yang popular digunakan dalam pembaikkan tanah lembut masa kini. Maklumat dari kajian ini boleh digunakan untuk membangunkan sistem panduan reka bentuk, pemodelan berangka serta memberi gambaran dan ilmu kepada penyelidik lain yang sedang  atau akan menjalankan kajian dalam bidang ini. Akhir sekali, perspektif masa depan untuk penyelidikan berkaitan ramalan faktor-faktor yang mempengaruhi kestabilan embankment jalanraya dengan pendekatan kepintaran buatan.

<p><span lang="EN-US">This paper presents the slope stability for road embankment constructed on the soft ground treated with prefabricated vertical drains (PVDs). The slope stability was evaluated based on the factor of safety (FOS) through numerical analysis and modeled with an artificial neural network (ANN). The permeability ratio of the smear effect was verified based on a comparative analysis between field data and numerical simulation to develop the datasets used in ANN model training. A total of 75 datasets generated from numerical simulations were randomly selected into three groups for training, testing, and validation. The coefficient of determination (</span><em><span lang="EN-US">R²</span></em><span lang="EN-US">) and root mean square error (RMSE) were considered to evaluate the performance ANN model. It was found that the developed ANN model showed strong potential for predicting slope stability within the accepted range.</span></p>

The prediction of slope stability was performed using artificial neural networks (ANNs) in this work. The factor of safety determined by numerical analysis was used to develop ANN’s data sets. The inputs to the network are slope height, applied surcharge and slope angle. Correlation coefficients between numerical data and ANNs outputs showed the feasibility of ANNs for successfully modelling and predicting safety issues. The ANNs training phase is improved using a genetic algorithm (GA), and the results are compared to those obtained without GA trained ANNs. A sensitivity analysis is conducted to ascertain the relative contribution of different factors on slope stability. The slope angle and applied surcharge have a significant effect on slope stability.

No abstract