The development of empirical models to predict concrete expansion produced
by sulfate attack is described. Data collected by the U.S Bureau of Reclama
tion over a 40+ year period of non-accelerated testing form the basis for t
he model. In the nonaccelerated test program concrete cylinders were contin
uouslly submerged at room temperature in 2.1% sodium sulfate solution which
corresponds to severe field exposure conditions. Expansion measurements we
re made periodically. Over 8000 expansion measurements were collected for 1
14 specimens cast from 51 different mixtures.
Analysis of the data showed the significance of water-cement ratio (w/c) an
d C(S)A content of the cement, with the data revealing distinct behavior fo
r mixtures containing cements with low C(S)A content (<8%) and high C(S)A c
ontent (>10%). As a result, two models are proposed to predict expansion by
sulfate attack as a function of w/c, duration of exposure, and C(S)A conte
nt. The random effects method was used to capture unobserved heterogeneity
in the data set, and the results from a simple regression and a regression
including the random effects method are compared.