Statistical analysis towards the identification of accurate probability distribution models for the compressive strength of concrete

Structural design deeply involves material strengths, which are obviously affected by uncertainty. As far as the material “concrete” is concerned, building codes (Italian codes, Eurocodes, ACI regulations) typically state that it should be identified and classified by means of its conventional characteristic uniaxial compressive cubic strength at 28 days.Although the fundamental importance of the problem of the evaluation of the characteristic compressive strength of concrete in structural design, the statistical analysis geared to identify an accurate probabilistic model of the concrete strength has not been a central issue of the research works in the field for many years.This paper describes the results of an investigation performed to obtain the compressive strength statistic characteristics of a production of about half a million cubic meters of concrete. This amount of concrete was produced over a five-year period. The results obtained indicate that the statistical distribution that best captures the characteristics of the available experimental data is the Shifted Lognormal. On the other hand, the Italian code and, to some extent, the Eurocode substantially base the evaluation of concrete properties upon a Normal distribution. It is therefore advisable that design codes will encompass the possibility for the engineer to evaluate the concrete properties based upon these more refined statistical models.