Artificial Neural Network Equations for Predicting the Modified Proctor Compaction Parameters of Fine-Grained Soil

• Published in: Transportation infrastructure geotechnology Vol. 10; no. 3; pp. 424 - 447
• Main Authors: Verma, Gaurav, Kumar, Brind
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2023; Springer Nature B.V
• Subjects: Algorithms; Artificial neural networks; Building Materials; Compaction; Comparative analysis; Correlation coefficient; Correlation coefficients; Datasets; Engineering; Fine-grained soils; Foundations; Geoengineering; Geotechnical Engineering & Applied Earth Sciences; Highway construction; Hydraulics; Laboratory tests; Mathematical models; Neural networks; Parameter modification; Parameters; Performance measurement; Road construction; Soil; Soil compaction; Soils; Technical Paper; Optimum moisture content; Modified Proctor compaction; Fine-grained soil; Maximum dry density; Artificial intelligence; Artificial neural networks
• ISSN: 2196-7202; 2196-7210
• DOI: 10.1007/s40515-022-00228-4

In this study, a novel application of artificial neural network (ANN) was utilized to develop the predictive equations for the modified Proctor compaction parameters of fine-grained soil. A total of 532 in situ soil samples were collected from a highway construction work site and numerous geotechnical parameters were obtained from the laboratory testing. Besides the index properties test, modified Proctor compaction tests were conducted on the collected soil samples through BIS specifications. ANN algorithm code, written in Python V3.7.9 platform, was adopted for the analysis. Several performance measurement parameters such as MAE, RMSE, R , and R 2 were used to examine the performance of each of the models. The developed ANN equations present the correlation coefficient of 0.88 and 0.93 for MDD and OMC, respectively. Additionally, the selected model can predict the MDD and OMC of fine-grained soil within ±4% and ±12% variations, respectively. The results achieved for the validation dataset reveals that the proposed model is well efficient in predicting the unseen dataset. Eventually, it has also been perceived from the comparative analysis results of the present study model and previously existing models that the present study ANN model is more superior to those literature’s models.