Back propagation artificial Neural Network modeling of flexural and compressive strength of concrete reinforced with polypropylene fibers

Abstract

The production of fiber reinforced concrete involves a complex reaction system. This imposes an immense challenge in deriving appropriate material proportions of concrete to achieve desired mechanical properties. In order to facilitate selection of a design matrix for fiber reinforced concrete, a novel artificial Neural Network models for compressive and flexural strengths using back propagation feed-forward algorithm were proposed in this research. A wide range of varied concrete design mixes of cylindrical and beam samples was respectively tested for compressive and flexural tests. A polypropylene type of fiber reinforcement was used in the preparation of samples that were cured for 28 days in a water-saturated lime. Results showed that the compressive and flexural strength models provided predictions in good agreement with experimental results as described by high correlation values of 99.46% and 98.57% respectively. Mean squared errors of 0.0024 and 0.44 were obtained respectively in selecting the best fit model for compressive and flexural strengths. In the parametric analysis conducted, the proposed models were able to describe analytically the constitutive relationships of the material components and capture the dominant characteristics of concrete samples.