Prediction of compressive strength of concrete using multilayer perception network, generalized feedforward network, principal component analysis network, time lagged recurrent network, recurrent network

The present work aimed to study the artificial neural network (ANN) and its effectiveness for prediction of compressive strength (f_c). Genetic algorithm (GA) was used for optimization of five different types of ANN networks viz. multilayer Perception network (MLP), generalized feedforward network (GFF), principal component analysis network (PCA), time lagged recurrent networks (TLRN), recurrent networks (RN). A 272 data set of f_c was obtained from the various literatures and used for training, testing and validation. Mean square error (MSE), mean absolute error (MAE) and correlation coefficient (R) used as validation criteria. Water to cement (w/c) ratio, maximum size of aggregate, curing days and cement content etc. were used as input parameter for prediction of f_c. The result reveals that MLP has more precise compared with GFF, PCA, TLRN, RN, the observed values of R is 0.97, MSE is 42.30 and MAE is 5.57, which indicates the model is best fir for prediction of f_c.