PREDICTION OF COMPRESSIVE STRENGTH CONCRETE BY ARTIFICIAL NEURAL NETWORKS, FUZZY LOGIC AND MULTIPLE REGRESSION

In the present paper, articial neural networks (ANN)and regression analysis for predicting compressivestrength of cubes of concrete containing silica fume(SF), y ash,Copper slag are developed at the ageof 7,28 days. For building these models, trainingand testing using the available experimental results for66 specimens produced with 6 dierent mixture proportionsare used. The data used in the multi-layer feedforward neural networks models,linear regressionmodel are designed in the format of seven input parameterscovering the age of specimen, cement, ne aggregate,coarse aggregate, y ash, silica fume,copperslag. According to these input parameters, in the multilayerfeed forward neural networks, models are used topredict the compressive strength,durability values ofconcrete. It was shown that neural networks have highpotential for predicting the compressive strength anddurability values of the concretes containing silica fume(SF), y ash,copper slag. Results show that thevalues obtained from the training,testing in ANN-I(LM Algorithm) model are very closer to the experimentalresults. The results show that ANN has strongpotential as a feasible tool for estimating the ingredientsof concrete to meet the design requirements. Also,multiple regression (MR) is a statistical technique thatallows us to predict someone's score on one variable onthe basis of their scores on several other variables. MR isemployed to learn more about the relationship betweenseveral independent or predictor variables,a dependentor criterion variable. Therefore, MR analysis wascarried out using a MATLAB 2013 package to correlatedetermined fc value to the seven concrete parameters.The data used while developing the ANN model (i.e.,66 data sets) were used in the development of the MRmodel. However, the obtained indices make it clear thatthe ANN model is more capable with a higher predictionperformance compared to the MR model.