Modeling the Interaction Between Fly Ash and Lime Under Water Curing by an Artificial Neural Network Saikat Maitra, w Narayan Bandyopadhyay, and Ananta K. Das Government College of Engineering and Ceramic Technology, Kolkata 700010, India An artiﬁcial neural network was used to model the chemical interaction between ﬂy ash and lime with different ratios in water-cured compacts. As inputs for the model, different process parameters like pozzolanicity, surface area, unburnt carbon con- tent of the ash samples, water-curing periods, and the proportion of initial lime content in the ﬂy ash–lime mixes were used. Free lime remaining in the compacts after different curing periods was used as the output parameter. A generalized feedforward back- propagation three-layered neural network model was used with a tanhyperbolic transfer function at both the input and the output layer with 400 exemplars. For the training data, after 3500 it- erations, the error value was found to be the minimum for the prediction mode. When the model was tested for the test data, the difference between the actual value of the strength and the pre- dicted value of the strength was found to be within 715%. I. Introduction W ATER-cured ﬂy ash–lime compacts have been ﬁnding an increased use as an eco-friendly replacement of traditional clay-based ﬁred building materials in recent years. The manu- facturing process of ﬂy ash-based bricks does not generate greenhouse gas carbon dioxide unlike the traditional clay bricks, and the valuable topsoil that is generally used for making clay bricks can also be preserved by making ﬂy ash bricks. The major advantage of ﬂy ash as a raw material is that it can be used as a pozzolanic material as it reacts with lime in the presence of moisture to develop different lime-bearing hydrated phases like calcium silicate hydrates (C–S–H), calcium aluminate hydrates (C–A–H), calcium alumino-silicate hydrates (C–A–S–H), etc. These phases give rise to strength development in ﬂy ash–lime compacts. Fly ash-based bricks/blocks are suitable for use in masonry just like common burnt clay bricks at a lower cost with added advantages. The advantages include uniformity in shape, smoothness in ﬁnish, requirement of no plastering for building work, lighter weight, and less porosity than ordinary building bricks, etc. Many workers have carried out research works on ﬂy ash–lime compacts. A few of these works are mentioned below. Barbier 1 studied the possible uses of coal ﬂy ash in the brick industry with respect to the availability and characteristics of the ﬂy ash. Song et al. 2 studied the manufacture and properties of coal ﬂy ash–clay bodies. Tsunematsu et al. 3 studied the hydro- thermal reactivity of ﬂy ash with lime and gypsum with respect to the mineral composition. Kumar 4 conducted a perspective study of ﬂy ash–lime–gypsum bricks and hollow blocks for low-cost housing development. Ma and Brown 5 also studied the hydrothermal reaction of ﬂy ash with Ca(OH) 2 and CaSO 4  2H 2 O. Muntcan et al. 6 studied the autoclaved limestone materials with addition of ﬂy ash. The physico-mechanical prop- erties of the resulting siliceous limestone with 10%–30% ﬂy ash were found to be superior compared with the limestone obtained from lime and sand only. Wang et al. 7 studied the reaction mechanism of a ﬂy ash–lime–water system. The reaction mech- anism of the ﬂy ash–lime–water system was studied and a model of the reaction was established. Maitra et al. 8,10 and Basumajumdar et al. 9 carried out works on steam-cured ﬂy ash– lime compacts and developed a pozzolanic index to characterize ﬂy ash samples. They also determined the optimum temperature of heating for the ﬂy ash–lime compacts to develop the optimum level of strength. In the present investigation, the reaction between ﬂy ash and lime in compacted form was studied by measuring the free CaO remaining in the mixes after different periods of water curing. An artiﬁcial neural network (ANN) model was developed using input parameters like the pozzolanic nature of the ﬂy ash, un- burnt carbon content, surface area, lime content, and the reac- tion time between ﬂy ash and lime to predict the amount of unreacted lime that remained in the mixes after different periods of water curing. II. Experimental Procedure In the present investigation, 12 dry ﬂy ash samples (A–L) from different power plants of India were collected. Collection of the ash samples, their particle size distribution, measurement of bulk density, and speciﬁc gravity were carried out following the procedures described in IS: 1528–1974. The bulk density of the ash samples varied between 700 and 900 kg/m 3 . The speciﬁc gravity of the ash samples varied between 2.00 and 2.5. The surface area of the ash samples was determined by Blain’s air permeability method, and the surface area of all the ash samples was more than 300 m 2 /kg. Particle size distributions of the sam- ples were measured by screen analysis, and the result is given in Equivalent Spherical Diameter (Micron) Cumulative Mass Percent Finer 0 20 40 60 80 FlyAsh-K Fly Ash-C Fly Ash-G Fly Ash-I Fig. 1. Particle size distribution of different ﬂy ash samples. T. Uygunoglu—contributing editor w Author to whom correspondence should be addressed. e-mail: maitrasaikat@ rediffmail.com Manuscript No. 23073. Received April 11, 2007; approved June 21, 2007. J ournal J. Am. Ceram. Soc., 90 [11] 3712–3716 (2007) DOI: 10.1111/j.1551-2916.2007.01974.x r 2007 The American Ceramic Society 3712