Predicting the initial setting time of self compacting concrete using artificial neural networks (ANNs) with the various of learning rate coefficient

This study focuses on development of Artificial Neural Networks (ANNs) in prediction of initial setting time of self compacting concrete (SCC). To predict the setting time of SCC six input parameters are identified. A total of 250 different data sets of SCC was collected from the ready-mix factory and concrete laboratory in Surabaya. Training data sets comprises 120 data entries, and the remaining data entries (130) are divided between the testing and validation sets. Different combinations of architecture, number of neurons in hidden layer, different coefficient for learning rate and momentum were considered and the results were validated using an independent validation data set. A detailed study was carried out, considering one hidden layer for the architecture of ANN. The performance of the 6-3-1 architecture was the best possible architecture. The error for the training set was 4.32 percent for the 120 training data points, at running time 41.765 seconds, 2.54 percent for the 80 testing data points, and 3.12 percent for the 50 verification data points. The results of the present investigation indicate that ANNs have strong potential as a feasible tool for predicting the setting time of concrete.