ANNs of concrete is the multi-layer perceptron

ANNs can be used to model any complex relationships between inputs and outputs in data (Prasad et al., 2009, Bal and Buyle-Bodin, 2013). Three important steps must be considered in constructing a successful artificial neural network: network architecture, training, and testing. The basic aspects of network architecture consist of the number of hidden layers between the input and output layers, the number of processing units in each hidden layer, the pattern of connectivity among the processing units, and the activation (transfer) function employed for each processing unit. These aspects change from one network to another, leading to a wide variety of network types. The training process can occur in a supervised or unsupervised manner. Supervised training means that the network is provided with sets of training data that include the expected output for each set of input and the network is told what to learn. There are no target outputs available in unsupervised training and the network must modify its weights and biases in response to the inputs only by categorizing the input patterns into a finite number of classes (Nehdi et al., 2001).There are several methods and techniques to train a network. The most applicable network in modelling the performance of concrete is the multi-layer perceptron (MLP) with back propagation for minimizing error. The network includes an input layer, one or more hidden layers and an output layer. The optimization is conducted by using the minimization of mean-squared error (MSE), MSE measures the average of the squares of errors. The aim of this chapter is to predict drying shrinkage of SCC, for this propose ANNs model with one hidden layer was constructed, and developed using software MATLAB (R2009a). Training, testing and validation of the model were performed using 147 sets of the database obtained from different published previous studies, then the predicted results were compared with the experimental results.