The work presented in this dissertation involves an effort to develop a mix proportioning tool that can be used to determine the required type and amount of concrete components in a mixture based on the desired fresh and hardened concrete properties.

Concrete performance is affected by the quantity and quality of the paste, and aggregate systems. Therefore, this study analyzed the effect of binder systems with different types and content; the paste quality; and size, shape, texture and gradation of different aggregate systems on various fresh and hardened concrete properties.

In this experimental program, a total of 178 mixtures were prepared with 7 different gradation systems, 12 binder systems, 25 binder contents, 6 different water-to-binder ratio (w/b), and 3 different nominal air content.

Fresh properties of slump, air content, air-void system, setting time, unit weight, and temperature were tested. Hardened properties of compressive strength, rapid chloride penetration, surface resistivity, air permeability, and shrinkage were tested at various ages. However, to develop such a tool, this study overall focused on the assessment of workability, compressive strength, and durability as these three properties are commonly used as indicators of concrete performance. Durability was assessed by testing the rapid chloride penetration, and surface resistivity at 28-days.

As part of this study, an artificial neural network (ANN) approach has been used for concrete mix proportion design to analyze the complexity between concrete properties and concrete components.

Results have shown that development of a performance-based mix proportioning tool is possible for mixtures when aggregate gradation is not varied. Development of a mix proportioning tool with addition of the various aggregates systems generally was not as successful due to the increased variability of the mix design parameters such as size, shape, texture, and gradation of aggregates. The proposed mix proportioning tool is promising and achievable in terms of predicting the values of the tested properties based on the mix design variables. Although the proposed mix proportioning tool is not completely ready for prime time, the findings of this study can be implemented in real time when this approach is used with a larger data set.