Strength Prediction of High-Strength Concrete by Fuzzy Logic and Artificial Neural Networks

High-strength concretes (HSC) were prepared with five different binder contents, each of which had several silica fume (SF) ratios (0–15%). The compressive strength was determined at 3, 7, and 28 days, resulting in a total of 60 sets of data. In a fuzzy logic (FL) algorithm, three input variables (SF content, binder content, and age) and the output variable (compressive strength) were fuzzified using triangular membership functions. A total of 24 fuzzy rules were inferred from 60% of the data. Moreover, the FL model was tested against an artificial neural networks (ANNs) model. The results show that FL can successfully be applied to predict the compressive strength of HSC. Three input variables were sufficient to obtain accurate results. The operators used in constructing the FL model were found to be appropriate for compressive strength prediction. The performance of FL was comparable to that of ANN. The extrapolation capability of FL and ANNs were found to be satisfactory.