Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 74
PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON THE APPLICATION OF ARTIFICIAL INTELLIGENCE TO CIVIL AND STRUCTURAL ENGINEERING
Edited by: B.H.V. Topping and B. Kumar
Paper 22

Prediction of Workability of Concrete incorporating Metakaolin and PFA using Neural Networks

J. Bai, S. Wild, A. Ware and B.B. Sabir

School of Technology, University of Glamorgan, Pontypridd, United Kingdom

Full Bibliographic Reference for this paper
J. Bai, S. Wild, A. Ware, B.B. Sabir, "Prediction of Workability of Concrete incorporating Metakaolin and PFA using Neural Networks", in B.H.V. Topping, B. Kumar, (Editors), "Proceedings of the Sixth International Conference on the Application of Artificial Intelligence to Civil and Structural Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 22, 2001. doi:10.4203/ccp.74.22
Keywords: neural network, modelling, prediction, concrete workability, metakaolin, pulverised-fuel ash.

Summary
Concrete has many advantages over other construction materials including low cost, adaptability and applicability under many conditions. The world production of cement has greatly increased since 1980. However, cement production generates a large amount of CO2 which damages the environment. Increasing use of by-products, such as Pulverised-fuel ash (PFA), to partially replace Portland cement (PC) in concrete not only reduces the amount of cement used, hence reducing the emission of CO2 and conserving existing resources, but significantly enhances the properties of concrete. Inclusion of PFA in concrete greatly improves workability. Research work done on Metakalin (MK) has shown that the partial replacement of PC with MK in concrete significantly enhances early strength[1]. However MK, unlike PFA, results in increasing water demand with increasing replacement level[2] due to its high chemical activity and high specific surface. The investigation into optimising the contributory effects of PFA and MK on concrete workability by introducing them in combination has been carried out by Bai et al [2]. The research results have shown that loss of workability due to the presence of MK can be compensated for by the incorporation of PFA. The degree of restoration of workability, provided by PFA, is influenced significantly by the cement replacement level. Comprehensive experimentally determined workability data on PC-PFA-MK blends[2] were enabled a comparison to be made of the effect of various replacement levels on the workability. It would be more useful if trends can be extrapolated to develop reliable predictive models. Such models are very important as they help designers to select PFA-MK combinations in performance terms. Thus, there is a need to explore existing data and develop models capable of learning from historical data and to predict concrete workability (slump, compacting factor and Vebe time) in a reliable but economic manner.

This paper presents neural networks for the prediction of workability of concrete incorporating MK and PFA. The neural network models are validated using independent data sets and give high prediction accuracy. The predictions produced reflect the effect of variations in pozzolanic replacement of portland cement by graduated replacement levels of up to 15% metakaolin and 40 show that the models developed are reliable and accurate and they can be used to predict the workability parameters of slump, compacting factor and Vebe time across a wide range of Portland cement-PFA-MK. These demonstrate that using neural networks to predict concrete workability is practical and beneficial.

The following conclusions may be drawn from the results of the work:

  • Development of neural network models for workability, measured by slump, compacting factor and Vebe time for concrete incorporating PFA and MK is described. The neural network models constructed provide an efficient, quantitative and rapid means of obtaining optimal solutions to workability for concrete mixtures using PC-PFA-MK blends as binder. Such models would prove extremely useful for deployment in computer based applications.
  • Based on the experimental data, comprehensive training and tests were performed. The neural network models for predicting workability were found with high accuracy and were validated on independent validation data sets. Values of workability (slump, compacting factor and vebe time) were calculated using the trained models and gave good agreement with observed experimental data. It indicates that the models are reliable, accurate and can be used in practice to predict the workability of PC-FA-MK blends.
  • On the basis of the models developed, effects of MK and PFA on workability were analysed and iso-slump maps were plotted for and using trained neural networks. They make it possible for the designer to produce mixtures with various blend compositions for a given range of slumps.

References
1
Bai J, Sabir BB, Wild S, Kinuthia J. "Strength development in concrete incorporating PFA and Metakaolin", Magazine of Concrete Research, 2000, 52, No.3, 153-162.
2
Bai J, Wild S, Sabir BB, Kinuthia J. "Workability of concrete incorporating pulverised fuel ash and metakaolin", Magazine of Concrete Research, 1999, 51, No.3, 207-216.

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