Prediction of Compaction Properties in Biopolymer-Stabilized Laterite Soils Using Artificial Neural Network

Shailendra Pandurang Banne; Pratima Vitthal Kalokhe; Ganesh Tarte; Mugdha Kshirsagar; Sudarshan Bobade

doi:10.5935/jetia.v12i61.3841

Authors

Shailendra Pandurang Banne Department of Civil Engineering, Pimpri Chinchwad College of Engineering, Nigdi, Pune 411044, India
Pratima Vitthal Kalokhe Department of Civil Engineering, Pimpri Chinchwad College of Engineering, Nigdi, Pune 411044, India
Ganesh Tarte Department of Applied Science and Humanities, Pimpri Chinchwad College of Engineering, Nigdi, Pune 411044, India
Mugdha Kshirsagar Department of Civil Engineering, Symbiosis Institute of Technology, Symboisis International Deemed Univerisity, Pune 412115, India
Sudarshan Bobade Department of Civil Engineering, Pimpri Chinchwad College of Engineering and Research, Ravet, Pune 412101, India

DOI:

https://doi.org/10.5935/jetia.v12i61.3841

Abstract

Biopolymers like xanthan gum, guar gum, cellulose, and chitosan, in varying proportions, can be used to stabilize weak laterite soil. By enhancing the soil's strength, resistance to erosion, and durability, these biopolymers improve its mechanical properties. This study presents artificial neural networks (ANN) to develop predictive models for estimating the Maximum Dry Density (MDD) and Optimum Moisture Content (OMC) of biopolymer-stabilized laterite soil, with model predictions validated against experimental results.
The incorporation of various biopolymers particularly xanthan gum, guar gum, cellulose, and chitosan significantly improves the compaction characteristics of laterite soil, with notable increases in Maximum Dry Density (up to 27.46%) and varied effects on Optimum Moisture Content, depending on the biopolymer type and concentration. A previously made available database was utilized to develop the prediction models. The correlation coefficient (R-value) was employed as the standard for acceptable performance in neural network development. The ANN-based prediction model for MDD (R = 0.941) demonstrates the strongest correlation and best model fit. In comparison, the OMC model shows a slightly lower yet still commendable level of accuracy (R = 0.925).