A concrete mix proportion design algorithm based on artificial neural networks

Abstract

The concepts of five parameters of nominal water–cement ratio, equivalent water–cement ratio, average paste thickness, fly ash–binder ratio, grain volume fraction of fine aggregates and Modified Tourfar's Model were introduced. It was verified that the five parameters and the mix proportion of concrete can be transformed each other when Modified Tourfar's Model is applied. The behaviors (strength, slump, et al.) of concrete primarily determined by the mix proportion of concrete now depend on the five parameters. The prediction models of strength and slump of concrete were built based on artificial neural networks (ANNs). The calculation models of average paste thickness and equivalent water–cement ratio can be obtained by the reversal deduction of the two prediction models, respectively. A concrete mix proportion design algorithm based on a way from aggregates to paste, a least paste content, Modified Tourfar's Model and ANNs was proposed. The proposed concrete mix proportion design algorithm is expected to reduce the number of trial and error, save cost, laborers and time. The concrete designed by the proposed algorithm is expected to have lower cement and water contents, higher durability, better economical and ecological effects.