STRENGTH DATA AND CONCRETE QUALITY
A large number of specifications for concrete quality which are based on statistical data, limit the number of poor results in strength tests. The usual specification prescribes no more than V erroneous results out of n number of tests. The concept behind this specification is that the proportion of defective concrete in a structure must not exceed v/n. An attempt is made to show that if the number of tests n is small (as is often the case) there is a high probability that the proportion of defective concrete accepted according to the above rule, is greater than v/n. For example, if it is specified that no more than one test result out of ten can be erroneous and if 10, 20 or 30 tests are carried out, the probability of accepting a lowquality concrete (e.g. Concrete which is more than 10% defective) is 0.505, 0.375 and 0.302 respectively. On the other hand, the probability of eliminating a highquality concrete, if more than one test result out of ten is found erroneous, is very low. For the cases cited above, the probability is 0.110, 0.106, 0.104 respectively. It is proposed that, in the case where a proportion of lowstrength concrete equivalent, at the most to v/n is acceptable in a structure, the standard prescribed be modified so as to specify that no more than V1 out of n test results be erroneous. Therefore the probability of an error of acceptance is reduced and becomes approximately equal to the probability of an error of refusal. With reference to the cases cited above, if the objective is to obtain no more than 10% defective concrete, the specification for ten tests must be that no result can be erroneous, for 20 tests no more than one result, for 30 tests no more than two results. Therefore the probability of an error of acceptance is 0.313, 0.236 and 0.200 respectively, while the corresponding probability for an error of refusal is 0.375, 0.274 and 0.226 respectively.

Corporate Authors:
Rilem
12 Rue Brancion
75737 Paris Cedex, France 
Authors:
 Koufopoulos, T
 Publication Date: 1972910
Media Info
 Features: Figures; References; Tables;
 Pagination: p. 302306

Serial:
 Materiaux & Constructions/Materials & Structures
 Volume: 5
 Issue Number: 29
 Publisher: International Union of Testing and Research Laboratories for Mat & St
Subject/Index Terms
 TRT Terms: Compliance; Concrete; Errors; Performance based specifications; Probability; Quality control; Specifications; Statistical analysis; Statistical sampling; Statistics; Strength of materials; Testing; Tests
 Uncontrolled Terms: Quality; Sample size
 Old TRIS Terms: Probable error; Quality control testing
 ITRD Terms: 178: Compliance (specif); 4755: Concrete; 6440: Error; 6534: Probability; 9063: Quality; 177: Specifications; 6555: Statistics; 5544: Strength (mater); 6255: Test
 Subject Areas: Data and Information Technology; Highways; Materials; Planning and Forecasting; I32: Concrete; I72: Traffic and Transport Planning;
Filing Info
 Accession Number: 00099603
 Record Type: Publication
 Source Agency: Laboratoire Central des Ponts et Chausees (LCPC)
 Files: ITRD, TRIS
 Created Date: Nov 18 1975 12:00AM