Prediction of mechanical and fresh properties of self-consolidating concrete (SCC) using multi-objective genetic algorithm (MOGA)

Compressive strength and concrete slump are the most important required parameters for design, depending on many factors such as concrete mix design, concrete material, experimental cases, tester skills, experimental errors etc. Since many of these factors are unknown, and no specific and relatively accurate formulation can be found for strength and slump, therefore, the concrete properties can be improved to an acceptable level using the neural networks and genetic algorithm. In this research, having results of experimental specimens including soil classification parameters, water to cement ratio, cement content, super-lubricant content, compressive strength, and slump flow, using the MATLAB software, the perceptron neural network training, general regression neural network, and radial base function neural network are considered, and then, with regard to coefficient of determination (R2) criteria and mean absolute error, the above networks are compared, and the proper neural network was identified, and finally, using the multi-layer perceptron neural network as the chosen network as well as multi-objective genetic algorithm fitting function, the 28-day compression strength and slump flow of self-compacting concrete are simultaneously optimized.