Abstract
Creep of concrete is an important design consideration. National design codes therefore provide
empirically based models for the estimation of creep deformation. Such models estimate a creep
coefficient (φ) and an elastic modulus (E) of the concrete, both of which are used to predict the creep
strain at any age.
This paper assesses the accuracy of the creep coefficients (φ) predicted by fourteen “design code-type"
models, with a view to ascertain whether the estimated φ or E is responsible for the inaccuracy of some
of the models. The models considered are those contained in SANS 10100 (2000)/BS 8110 (1985),
SANS 10100 (2000) Modified, ACI 209 (1992), AS 3600 (2001 & 2009), CEB-FIP (1970, 1978 &
1990), the Eurocode EC (2004), Gardener and Lockman (2000 & 2004), Gardener and Zhao (1993) and
the RILEM B3 (1995) methods.
Laboratory creep tests were conducted on concrete prisms covering a range of mixes. The measured φ
values were statistically compared to those predicted by the models considered.
The results indicated that, for the range of concretes tested, the CEB-FIP (1990) method yielded the
most accurate predictions of creep coefficient, giving the lowest overall coefficient of variation (all) of
27,7 %. The least accurate method was the CEB-FIP (1978) which yielded an overall coefficient of
variation (all) of 112,5 %. Furthermore, the accuracy of the predicted φ values correlated highly
significantly (P = 0,001 %) with the accuracy of the predicted creep magnitudes.
The results of this investigation led to recommending the SANS 10100 (2000)/ BS 8110 (1985) model
for predicting creep coefficients for South African conditions.