APPLICATION OF NEURAL NETWORK FOR PREDICTION OF COMPRESSIVE STRENGTH OF SILICA FUME CONCRETE

Use of silica fume as mineral admixture has become more common these days for production of silica fume concrete. The extent of replacement has been different depending upon the exposure and prevailing conditions of the project and most of the silica fume concrete are tailor made. In this work, an exhaustive study was done experimentally optimizing the replacement extent of cement with silica fume. From more than 150 results from the study, this paper aims in predicting compressive strength of silica fume concrete using Artificial Neural network (ANN). The constituent materials added for production of concrete are taken as inputs. There are five different parameters: weight of cement, silica fume, fine aggregate, coarse aggregate and water in kgs that are considered in this prediction analysis. ANN is an effective tool in predicting the output if the training is done with proven sets of data. In this work, a portion of experimental data was used in the training phase of ANN. The compressive strength was fed as the target. The ANN was trained with the experimental data till the Mean Square Error (MSE) was consistent. After the training, few unknown sets were given as inputs to the ANN and the simulations were carried out. The compressive strength was predicted and the values were close to the experimental results. Hence we conclude that ANN can be used to predict the compressive strength for various values of input instead of conducting experiments