Statistical Implications of Methods of Finding Characteristic Strengths

• Published in: Journal of structural engineering (New York, N.Y.) Vol. 122; no. 2; pp. 202 - 209
• Main Authors: Hunt, Richard D, Bryant, Anthony H
• Format: Journal Article
• Language: English
• Published: Reston, VA American Society of Civil Engineers 01.02.1996
• Subjects: Applied sciences; Buildings. Public works; Computation methods. Tables. Charts; Exact sciences and technology; Structural analysis. Stresses; TECHNICAL PAPERS; Statistical method; Properties of materials; Statistical distribution; Sampling; Strength of materials
• ISSN: 0733-9445; 1943-541X
• DOI: 10.1061/(ASCE)0733-9445(1996)122:2(202)

Various statistical methods of estimating the characteristic strengths of materials or structural components from the results of experimental tests can be assessed in the figures and tables of this paper. The methods are tested on normal, log-normal, and two-parameter Weibull populations as well as data sets from timber engineering activities. The results show how the required sample size depends on the coefficient of variation and the likely errors if an incorrect distribution is assumed. Two statistical methods that effectively fit distributions to the lower tails of experimental results, with the consequence that the estimations of characteristic strength are relatively insensitive to the form of the underlying distribution, are recommended because their estimations are almost as good as those obtained when the form of the underlying distribution is known. These methods also have the practical advantage that testing can be stopped when all members of a sample have been loaded to a level so that about 10% have failed, leaving 90% undamaged.