Keywords: artificial neural networks, high performance concrete, permeability, porosity, strength.

This paper presents the application of artificial neural networks for the prediction of strength and durability such as permeability and porosity of high performance concrete. High-performance concrete (HPC) is relatively a new terminology used in the concrete construction industry. HPC is designed to give optimized performance characteristics for the given set of materials, usage and exposure conditions, consistent with requirements of cost, service life and durability. HPC should be determined in terms of both strength and durability performance under anticipated environmental conditions.

In order to produce HPC a very dense homogeneous concrete microstructure especially in the interface region between hydrated paste and aggregate is required [1,2]. This is generally achieved through the use of low water-binder ratio between 0.20 and 0.30 with the help of superplasticizers that can produce slumps ranging from 70 to 130 mm. Additional densification and homogeneity of the interfacial region are achieved through the incorporation of mineral admixtures which improve concrete microstructure. Therefore, in addition to the three basic ingredients in conventional concrete, i.e., Portland cement, fine and coarse aggregates, and water, the making of HPC needs to incorporate supplementary cementitious materials, such as fly ash (FA), silica fume (SF) and/or blast furnace slag, and chemical admixture, such as superplasticizer. HPC can be manufactured involving up to 10 different ingredients whilst having to consider durability properties in addition to strength.

Since the number of ingredients and the number of properties of HPC, which needs to be considered in its design, are more than those for ordinary concrete. Therefore, it is difficult to predict the properties of this type of concrete using statistical empirical relationship. An alternative approach is to use an artificial neural network (ANN). The ANN approach is good for modelling non-linear systems. A neural network model is a computer model whose architecture essentially mimics the learning capability of the human brain. An ANN is a layered network consisting of an input layer, an output layer and at least one layer of non-linear processing elements known as hidden layer. The input layer of the neural network receives signals from the external environment. The hidden layer receives signals from the input layer and transmits an output signal based on a transfer function to a subsequent layer.

In this investigation, ANN based on the radial basis function (RBF) have been used. A RBF neural network is a layered network consisting of an input layer, an output layer and at least one layer of non-linear processing elements known as hidden layer. Non-linear autoregressive model with exogenous inputs (NARX) was used [3]. The input layer of the neural network receives signals from the external environment. The hidden layer receives signals from the input layer and transmits an output signal based on a transfer function to the subsequent layer.

The network developed in this investigation has eight units in the input layer and two units in the output layer. The experimentally obtained data have been divided into two sets, one for the network learning called learning set, and the other for testing the network called testing set. Each set is composed of dozens of pairs of input vectors and output vectors (vectors in the input layer called input vectors, and in the output layer called output vectors). An input vector consists of 8 components which influence the output vectors (compressive strength, tensile strength, permeability and porosity). The input parameters used are cement content, FA, SF, water content, superplasticizer, fine aggregate, coarse aggregate and age of testing.

The predicted values obtained using artificial neural networks for compressive strength, tensile strength, permeability and porosity have been plotted against their respective experimentally obtained values. It has been demonstrated that there is a good correlation between experimental values and those predicted using neural networks. Therefore, it is possible to predict these properties of concrete using artificial neural networks.

1

P.C. Aitcin and A. Neville, "High-Performance Concrete Demystified", Concrete International, 15(1), 21-26, 1993.

2

O.E. Gjorv, "High Strength Concrete", Advances in Concrete Technology, (Malhotra, V. M. Ed.), CANMET, Energy, Mines and Resources, Canada, 19-82, 1994.

3

I.J. Leontaritis and S.A. Billings, "Input-output Parametric Models for Nonlinear Systems Part-1: Deterministic Nonlinear Systems", Int. J. Control, 41(2), 303-328, 1985. doi:10.1080/0020718508961129

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £82 +P&P)