High-Performance Concrete Compressive Strength's Mean-Variance Models

摘要

The actual concrete compressive strength (MPa) for given mixture compositions for a specific age (days) is completely unknown. Based on an appropriate probabilistic model, it is only possible to identify the optimal level combinations of the mixture components to obtain the maximum concrete strength. In quality engineering, the most important problem is to predict the operating conditions that optimize concrete compressive strength (CCS) and simultaneously minimize the process variability. The considered CCS is a mixture of seven ingredients: cement, blast-furnace slag, fly ash, water, superplasticizer, coarse aggregate, and fine aggregate. It is found that the positive response variable CCS distribution is gamma and its variance is nonconstant. Thus, joint generalized linear models analysis is used to derive the joint mean and variance models. The present analysis has derived the following: (1) age, and all the marginal effects of the mixture components except superplasticizer, are significant either in the lognormal or the gamma mean model, (2) one third-order and six second-order interaction effects are significant in the final selected gamma mean model, (3) all the marginal effects of the mixture components are significant in both the variance models, and (4) six second-order interaction effects are significant in the final selected gamma variance model. A nonlinear stochastic third (second)-order mean (variance) model of CCS has been derived for seven ingredients along with the age. Effects of the ingredients along with the age on CCS have been derived from the presented derived mean and variance models.