Quantifying air-void distribution and associated uncertainty in fresh and hardened concrete

Abstract

Freeze-thaw damage in Portland cement concrete pavements (PCCP) has been a significant concern for regions experiencing (numerous) rapid freeze-thaw cycles along with frequent precipitation. In the early 1950s Powers recommended entraining a well-distributed air-void system (comprised of equally spaced, small air voids) through the cementitious paste to minimize frost damage in PCCP. There are several methods to measure total air content in fresh concrete, such as ASTM C231 (volumetric air content using a pressure meter) and ASTM C173 (volumetric air content using a roll-o-meter). Still, air content alone is insufficient to assess an air void distribution. Hardened air-void analysis (ASTM C457) is used to accurately measure distribution in terms of total air content, paste-to-air ratio, spacing factor, and specific surface area. But this test can only be performed in hardened concrete about 28 days after it has been placed on-site, thus limiting its use for making a real-time assessment. According to ACI 201.2R, a minimum 4% total (volumetric) air content, a paste-to-air ratio between 4 to 10, along with a specific surface area in the range of 25 to 45 mm⁻¹ (600 to 1100 in.⁻¹), and a spacing factor in the range 0.1 to 0.2 mm (0.004 to 0.008 in.), is required to ensure adequate freeze-thaw resistance of PCCP. Accordingly, limitations have been placed by various transportation agencies to ensure the required air­ void distribution in PCCP. Recently, the KT-86/AASHTO T 395 (Sequential Pressure Method, referred to as the Super Air Meter - SAM) and the pressure meter were proposed as field-ready devices to improve quality control/assurance (QC/QA) practices. However, several potential shortcomings are associated with using these devices to evaluate air-void distribution in fresh concrete in real-time on the field. To enrich the predictive capability of these traditional methods, complementary models have been proposed in this study to predict spacing factors using mixture design and fresh concrete test measurements as predictors. A set of 271 observations were used to develop the statistical models for predicting spacing factor (PSF) measured by a commercial air-void scanning system (RapidAir, based on the ASTM C457-16 method). Significant variable(s) for model development were selected using the LASSO, Binary logistic, and Bayesian probit regression analysis models. The predictive models were evaluated on their ability to distinguish between a well-entrained air-void system (spacing factor < 0.008 in. or 0.2 mm) from a coarsely-entrained air-void system (spacing factor > 0.008 in. or 0.2 mm). It was observed that the predictive models were better and more efficient in assessing the air-void distribution than total air content or SAM number alone. Given that the models were calibrated and validated for a data set dominated by a spacing factor less than 0.008 in. (0.2 mm), its validation was limited for coarsely entrained air-void systems. The given study focuses on examining current & alternate methods and the error associated with them to evaluate air-void distribution in fresh & hardened concrete. Distribution parameters measured by a RapidAir system are assumed to be an actual/accurate representation of the entrained air-void system. But these parameters are regulated by the RapidAir system settings (namely threshold and gain), selected by the operator during evaluation. Thus, to accurately identify and quantify this influence, a set of known air-void distributions were compared to their similar properties measured by the RapidAir system at different system settings. The known air-void distributions were generated as unique synthetic air-void structures using a Python script and printed as luster photographs using a 3200-dpi printer. A RapidAir system at different system settings scanned these photographs. The 'true' distribution parameters for these photographs were calculated from the exact size & location of air voids, recorded by the Python script. As expected, a significant error was observed between the two measurements. Later, these errors were used to minimize the influence of system settings on the distribution parameters measured for a set of concrete samples. As a result, a 'fair ground' to compare air-void structures of the same/different samples evaluated by the same/different systems and settings was provided. Previously, researchers have recommended using alternate parameters and spacing factor definitions to evaluate air-void distribution in hardened concrete. One such parameter is the 'dispersion parameter' (DP), formerly used to quantify the dispersion of fibers in composites. The current study considers this parameter to quantify the dispersion of air voids through its cementitious paste. DP is a scalar quantity calculated as the ratio of maximum work required to disperse air voids from their current locations to their location in 'best' and 'worst' dispersion scenarios. DP helps quantify the distribution of air voids across the sample and identify the location and presence of clusters, thereby offering a better characteristic distance parameter than the 'spacing factor' to quantify the quality of air-void distribution. Overall, the given study serves as a guide to enrich QC/QA practices by assessing the accuracy & uncertainty of current & proposed methods to evaluate air-void distribution in fresh & hardened concrete.