Cement Strength Prediction Based On Artificial Neural Network Model

The grading of cement is mainly based on its 28 day intensity value. Currently this value is mostly predicted by accelerating cement hydration on raising the specimen' s upkeep temperature, thus fast achieving a higher intensity to predict 28 day intensity on a given conversion. However, the prediction of cement intensity being a multivariable, non-linear and high time-lag problem, simplification of a highly non-linear relationship among the cement intensity and various measurements to a linear function in the end affects the accuracy of prediction. Therefore, this paper attempts to predict cement intensity on modeling of artificial NN.Here prediction models are built for BP NN and RBF NN separately, and simulated in test, with the result that BP model is inferior to RBF model in network practice speed and accuracy of prediction etc. Hence, the prediction is modeled after RBF NN.The prediction software is developed on modeling with Visual C++6. 0 for human interfacing, MATLAB program called by VC on conversion of M file scripting into COM module using in-built Combuilder, and SQL Server 2000 in background database development. This paper can be of useful guidance in the cement production.