Predictions on compaction characteristics of fly ashes using regression analysis and artificial neural network analysis

For a preliminary assessment of the suitability of fly ashes required for a project, it is preferable to use prediction models for compaction characteristics of fly ashes on the basis of tests which are quick to perform, less time consuming, and cheap, such as the determination of sand size particle content of fly ashes. In this study, prediction models for compaction characteristics of fly ashes have been developed using regression analysis and artificial neural network analysis with the variables, sand fraction and fine fraction. Artificial neural networks have remarkable ability to derive meaning from complicated data. Hence neural network analysis has more potential as a forecasting tool than multiple regression analysis. The study indicates that the parameters, sand fraction and fine fraction, bear good correlation with maximum dry unit weight (cdmax) of fly ashes. However, even using neural network analysis, lesser correlation is observed between these parameters and optimum moisture content (OMC) of fly ashes. This could be due to the fact that in the case of fly ashes, the OMCs are not well defined, as compaction curves of fly ashes are relatively flat, when compared to soils so that the maximum dry unit weights remain the same for a wider range of moisture contents. Regression analyses on compaction characteristics of fly ashes suggest that generally speaking as ash becomes finer, its OMC reduces and cdmax increases, unlike natural soils.