Достижения в области прикладных научных исследований Открытый доступ

Абстрактный

Artificial neural network for the modulus of rupture of concrete

Onwuka. O. David and Awodiji. T.G. Chioma

In traditional method of concrete mix design, a normal concrete of required strength, can be achieved after several trials on mix proportion. This trial mix design approach has proven to be complex and difficult. This article demonstrates the applicability of artificial neural network (ANN) to the design of concrete of required modulus of rupture (MOR). The architecture of the neural network created is 4-20-1 (i.e. 4 input neurons, 20 neurons in the hidden layer and one output neuron). And, a feed forward back propagation learning algorithm was used for the training of the neural network. Sufficient set of mix proportions with their corresponding modulus of rupture, were generated experimentally and used for the training and testing of the neural network. The neural network toolbox of MATLAB software with TRAINGDM training function, LEARNGDM learning function and MSE regularization performance function, was utilized in creating the neural network. The results predicted by the network were in close agreement with corresponding experimental values. And a correlation coefficient of 0.9051 for all, shows that there is a linear relationship between the output and target during training.