Prediction of the chloride diffusivity of recycled aggregate concrete using artificial neural network

Abstract

The incorporation of recycled coarse aggregate (RCA) could reduce recycled aggregate concrete(RAC) durability. Chlorine penetration is an important indicator for concrete durability evaluation. In order to conduct a systematic study of chloride permeation in RAC, this paper demonstrates the feasibility of artificial neural network (ANN) in the study of chloride ion penetration in RAC. Totally 112 sets of test data collected from various literatures were used to train and validate the proposed ANN model, and 9 characteristic factors affecting RAC resistance to chloride penetration were selected as input variable. By analyzing model performance and simulation results of out-of-sample data sets, it can be concluded that the established model is feasible and has potential in predicting the chloride ion diffusivity of RAC. Using the proposed model, sensitivity analysis was carried out on the input parameters, and the influence index of a single variable was obtained, which is affected by the properties of RA. These findings suggest that, ANN is an effective tool in analyzing the chloride ion permeability properties of RAC from various sources.