Evolutionary design of generalized GMDH-type neural network for prediction of concrete compressive strength using UPV

Abstract

The main purpose of this paper is to predict the insitu compressive strength of concrete by means of non-destructive approach using ultrasonic pulse velocity (UPV) method. For this purpose generalized GMDH-type (group method of data handling) neural network was developed based on various data obtained experimentally. Evolutionary algorithms (EAs) are deployed for optimal design of GMDH-type neural networks. A set of experimental data for the training and testing the evolved GMDH-type neural network is employed in which ultrasonic pulse velocity (UPV), concrete age, water–cement ratio and fine/coarse aggregate ratio are considered as inputs and concrete compressive strength is regarded as the output variables. Sensitivity analysis has also been carried out on one of the obtaining models to study the influence of input parameters on model output. The results show that generalized GMDH-type neural network has a great ability as a feasible tool for prediction of the concrete compressive strength.