A Fuzzy-Neuro Model for Normal Concrete Mix Design

Concrete mix design is a process of proportioning the ingredients in right proportions. Though it is based on sound technical principles and heuristics, the entire process is not in the realm of science and precise mathematical calculations. This is because of impreciseness, vagueness, approximations and tolerances involved. This paper presents the development of a novel technique for approximate proportioning of standard concrete mixes. Distinct fuzzy inference modules in five layers have been framed to capture the vagueness and approximations in various steps of design as suggested in IS: 10262-2003 and IS456-2000. A trained three layer back propagation neural network is integrated in the model to remember experimental data pertaining to w/c ratio v/s 28 days compressive strength relationship of three popular brands of cement. The results in terms of quantities of cement, fine aggregate, course aggregate and water obtained through the present method for various grades of standard concrete mixes are in good agreement with those obtained by the prevalent conventional method. Details of the system model are described and comparative graphs are presented.