Assessment of fine-grained soil compaction parameters using advanced soft computing techniques
The compaction parameters are the most important parameters for any civil engineering project. In this work, the sand content, fine content, liquid limit, plastic limit, and plasticity index were used as input parameters by the soft computing models to predict the compaction parameters of soil. Thes...
Gespeichert in:
Veröffentlicht in: | Arabian journal of geosciences 2023-03, Vol.16 (3), Article 208 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The compaction parameters are the most important parameters for any civil engineering project. In this work, the sand content, fine content, liquid limit, plastic limit, and plasticity index were used as input parameters by the soft computing models to predict the compaction parameters of soil. These soft computing models are gene expression programming (GEP), least square support vector machine (LSSVM), long short-term memory (LSTM), and artificial neural network (ANNs). For this purpose, three databases, i.e., training, testing, and validation, are prepared from the database available in the literature. Also, twelve soil samples are collected from and around Kota, Rajasthan, and tested in a laboratory for cross-validation of the best architecture models. The performance of models is measured by thirteen performance indicators, including three new indicators, i.e., a20-index, index of agreement (IOA), and index of scattering (IOS). The test performance comparison reveals that the polynomial LSSVM model MD15 (a20-index = 100.00%, IOA = 0.9371, IOS = 0.0519) and linear LSSVM model MD110 (a20-index = 100.00%, IOA = 0.9179, IOS = 0.0122) have the highest performance in predicting optimum moisture content (OMC) and maximum dry density (MDD), respectively. Also, models MD15 and MD110 have higher performance in the validation phase. Models MD15 and MD110 have predicted OMC and MDD of twelve soil samples with residuals of ± 1.776% and ± 0.044 g/cc, respectively. This study demonstrates ANN achieves high overfitting than the LSTM model in predicting the compaction parameters of soil. The LSSVM model represents overfitting in predicting OMC and underfitting in predicting MDD of soil. Finally, the present research introduces high-performance soft computing models for predicting the compaction parameters of fine-grained soil. |
---|---|
ISSN: | 1866-7511 1866-7538 |
DOI: | 10.1007/s12517-023-11268-6 |