Title:
Statistical Properties of Techniques for Predicting Concrete Strength and Examples of Their Use
Author(s):
John K. Moore and Michael A. Taylor
Publication:
Symposium Paper
Volume:
56
Issue:
Appears on pages(s):
259-284
Keywords:
accelerated tests; calibrating; compressive strength; com-puter
programs; quality control; regression analysis; statistical
analysis.
DOI:
10.14359/6507
Date:
10/1/1978
Abstract:
This paper contrasts the statistical properties of two statistical techniques for predicting the ultimate strength of concrete, using data from two preliminary calibration experiments. The techniques are: (1) prediction from the mean early strength by the usual regression line; (2) prediction from a conservative estimate of early strength by a conservative relationship of early strength to ultimate strength. It is shown that the latter method guarantees (for almost all calibration experiments) a, high percentage of conservative predictions of ultimate strength -- as concrete users desire -- and that the former method, which is commonly used, does not. The use of the techniques is illustrated by predictions of 28-day cylinder strength from the strength of cylinders autogenously cured for 48 hours.