Deep Neural Network Models for the Prediction of the Aggregate Base Course Compaction Parameters

Laboratory tests for the estimation of compaction parameters, namely maximum dry density (MDD) and optimum moisture content (OMC) are time-consuming costly. Thus, this paper employs artificial neural network technique prediction OMC MDD aggregate base course from relatively easier index properties tests. The grain size distribution, plastic limit, liquid limits used as inputs development ANNs. In study, multiple ANNs (240 ANNs) tested to choose ANN that produces best predictions. This focuses on studying impact three different activation functions: number hidden layers, neurons per layer predictions, heatmaps generated compare performance every with settings. Results show hyperparameters change depending predicted parameter. Additionally, hyperbolic tangent is most efficient function it outperforms other two functions. simplest architectures results in deteriorates increase layers or layers.