Predicting compressive strength of roller-compacted concrete pavement containing steel slag aggregate and fly ash

Abstract

ABSTRACT This study presents the analytical models to predict the compressive strength of roller-compacted concrete pavement (RCCP) containing steel slag aggregate and fly ash. Based on the experimental results, three models were established in this study including multiple regression analysis (MRA), artificial neural networks (ANN) and fuzzy logic (FL). In the RCCP mixtures, cement was partially substituted by fly ash at four levels: 10%, 20%, 30%, and 40%; natural coarse aggregate was replaced by steel slag aggregate at ratio of 50% and 100%. The compressive strength was determined at 3-, 7-, 28-, 56- and 91-day ages. 75 sets of testing data were collected to build the target values set. With same seven input variables, the MRA model is less reliable than the ANN model in terms of predicting the compressive strength of RCCP. Besides, the use of triangular membership functions with three input variables (fly ash content, steel slag aggregate content and age) in the FL algorithm is sufficient to obtain accurate results. The performance of the FL model is as good as the ANN model. Additionally, a total of 33 fuzzy rules found for building the FL model can be applied to predict the compressive strength of RCCP. Highlights MRA, ANN, and FL were used to construct the models for predicting the compressive strength of RCCP containing steel slag aggregate and fly ash. The ANN model and FL model created reliable results in predicting the strength of RCCP. The MRA model is less reliable than the ANN and FL models in terms of predicting of RCCP compressive strength. The best model is the FL model because of its friendly and efficiency.