Precision of in-place concrete strengths predicted using core strength correction factors obtained by weighted regression analysis

A two-step method for converting a concrete core compression test result to the in-place strength of the corresponding volume of concrete is presented. The strength of a non-standard core is first converted to the equivalent strength of a standard core, and then the standard core strength is converted to the equivalent in-place strength. Strength correction factors required for these conversions, obtained from weighted linear and nonlinear regression analyses presented elsewhere, are summarized. The accuracy of the predicted in-place strength is affected by the inherent error of the core strength measurement itself, and by the uncertainty of the various strength correction factors. It is shown that confidence intervals on the estimates of the strength correction factors obtained by regression analysis underestimate the true model error because the underlying models are imperfect. Instead, the accuracy of the strength correction factors is determined by a weighted regression analysis of ratios of observed-to-predicted values which accounts for the non-uniform variances of the dependent and independent variables. The coefficient of variation of the in-place strength predicted from a test of a 100 or 150 mm diameter core is between 4 and 5.5%. If the in-place strength is predicted from a test of a 50 mm diameter core, the coefficient of variation of the predicted in-place strength is approximately 12.5%. These error estimates do not account for possible variation of in-place strength throughout the volume of the element being cored. (C) 1997 Elsevier Science Ltd.