Effects of Significant Variables on Compressive Strength of Soil-Fly Ash Geopolymer: Variable Analytical Approach Based on Neural Networks and Genetic Programming

The identification of significant input variables to the output provides very useful information for mix design for soil-fly ash geopolymer in order to obtain the optimum compressive strength. The importance of input variables to the output of soil-fly ash geopolymer is quantified by Garson’s algorithm and connection weights approach in an artificial neural networks (ANN) model, whereas model analysis and fitness method are used in a genetic programming (GP) model. The former approaches in the ANN model use the connection weights among the input, hidden, and output layers to evaluate the importance of the input variables. The latter methods in the GP model assess the frequency of variables used in the model and the value of fitness for the evaluation. The assessment results identify the percentages of fly ash, water, and soil as important input variables to the output. The percentage of hydroxide and the ratios of silicate to hydroxide and alkali activator to ash are ranked as less important input variables. The positive or negative relationships between these input variables and the output demonstrate a very significant influence on the strength development of soil-fly ash geopolymer, showing a positive or negative effect on the compressive strength.