Lateef/my-repo

The suitability of laterite for engineering constructions depends largely on the extensive examination of its compaction and strength characteristics but this is lacking these days due to their cumbersome, costly and time-consuming laboratory tests unlike index properties with cheap and simple tests. To overcome this limitation, artificial neural network (ANN), Gaussian process regression (GPR), and multilinear regression (MLR)models have been developed in this study to predict laterite compaction and strength characteristics from the index properties. To achieve this, laboratory tests were conducted on three hundred samples taken from thirty different laterite deposits within southwest, Nigeria and the experimental dataset obtained was randomly divided into modeling dataset comprising two hundred and forty data points which were used to develop the models and trialing dataset with sixty data points which were used to compare and validate the developed models. Specific gravity (SG), linear shrinkage (LS), liquid limit (LL), plasticity index (PI), and fines content (FC) were selected as the models’ predictors while maximum dry density (MDD), optimum moisture content (OMC) and unconfined compressive strength (UCS) are the targeted output. The developed models’ performances were appraised using various prediction performance metrics and based on this, ANN and GEP models provided satisfactory performance with ANN models taking the lead while MLR models presented poor performance. Hence, the developed ANN and GPR models can be used for practical purpose. The sensitivity analysis was conducted, and the analysis revealed that SG has the greatest influence on MDD and UCS while LS has the greatest influence on OMC.