Prediction of soil's compaction parameter using artificial neural network

Abstract EN

This research tackles the feasibility of using Artificial Neural Networks to capture nonlinear interactions between various soil parameters.

In this study an attempt was conducted to predict the compaction parameter (γdmax & O.M.C) using database comprising a total of 177 case records of laboratory measurements.

Eight parameters are considered to have the most significant impact on the magnitude of compaction parameters have been used as the model's inputs; liquid and plastic limits, plasticity index, specific gravity, soil type, gravel, sand, and fines content.

The model output is the maximum dry unit weight and optimum moisture content.

A Multi–layer perceptron trainings using the back–propagation algorithm, are used in this work.

A number of issues in relation to ANN's construction such as the effect of ANN's geometry and internal parameters on the performance of ANN's models are investigated.

A parametric study was conducted for the three models to investigate the effect of the input variables on the output of the model.

Based on statistical criterion, it was found that ANN's have the ability to predict the compaction parameter with a good degree of accuracy.