Frost Durability Indexes of Segmental Retaining Wall Units

(ProQuest Information and Learning: ... denotes formulae omitted.)

INTRODUCTION

The frost durability of segmental retaining wall (SRW) concretes (Fig. 1) is currently assessed by the methods of ASTM C 1262,1 which involves freezing-and-thawing cycling of specimens in either water or 3%

sodium chloride solution up to a prescribed number of cycles (typically 100). At this point, the condition of the specimens is evaluated, and their mass loss is determined and compared to specifications such as ASTM C 1372.2 The overall test may require 2 to 3 months, however, making it desirable to predict freezing-and-thawing performance on the basis of simpler and faster tests of related material properties.

In ordinary concretes, for example, instead of expensive and time-consuming freezing-and-thawing tests, the spacing factor as determined by the methods of ASTM C 4573 has been commonly used as an indicator of frost resistance with a transition between durable and non-durable concretes existing somewhere between a spacing factor of 200 to 300 ?m (0.008 to 0.012 in.).4 Although ASTM C 457 tests can be performed on SRW concretes, there is a questionable relationship between spacing factor and frost resistance for SRW concretes even though specific surface may be of value, as detailed in a later section. The lack of correlation between spacing factor and frost resistance is in part due to the different nature of voids comprising SRW concretes compared with voids in ordinary concretes.5 As an alternative, the National Concrete Masonry Association (NCMA) has suggested an index to predict frost resistance of SRW concretes based on the fundamental parameters of compressive strength, 24-hour water absorption, and unit weight of SRW units.6 These parameters, routinely obtained from ASTM C 140 testing,7 are combined in the following manner

... (1)

In NCMA testing, units possessing higher values of I were, in general, more likely to exhibit less than 1% mass loss after 100 freezing-and-thawing cycles when tested in accordance with ASTM C 1262. For instance, for freezing-and-thawing tests in water, approximately 10% of specimens with I < 58 MPa^sup 1/2^ (700 psi^sup 1/2^) met specifications. However, 43% of specimens with I of approximately 133 MPa^sup 1/2^ (1600 psi^sup 1/2^) and 71% of specimens with I > 208 MPa^sup 1/2^ (2500 psi^sup 1/2^) exhibited less than 1% mass loss after 100 cycles.6

Other attempts have been made to correlate the frost resistance of SRW units to individual material properties or to combinations thereof. At the most basic end of this spectrum cement content (by mass) has been shown to correlate with freezing-and-thawing durability of no-slump concrete. In a review of previous research work, Hance8 noted that minimum cement contents of 252 to 395 kg/m^sup 3^ (425 to 665 lb/yd^sup 3^) were generally associated with frost resistance of no-slump concrete immersed in water and 320 to 380 kg/m^sup 3^ (540 to 640 lb/yd^sup 3^) associated with frost resistance in the presence of deicing salts. At the more sophisticated end, neural networks have been used to build prediction tools for SRW frost durability in water and saline solution.9 Out of a number of parameters characterizing SRW properties and mixture design, water absorption, density and percentage of non-connected voids were found to be most reliable for estimating frost performance in water. In saline solution, compressive strength, percentage of connected voids, percentage of non-connected voids, and whether a freezing-and-thawing enhancing admixture was used were most reliable for estimating frost performance. Bremner and Ries10 have also attempted to correlate SRW frost resistance to material properties and found no correlation between ASTM C 1262 mass loss and concrete density and water absorption. A general decrease in mass loss with increasing cementitious material content and compressive strength was, however, observed by Bremner and Ries.

This paper describes independent analyses of common SRW material characteristics as frost durability indexes, in which the research team obtained all data by performing tests on 10 separate populations of SRW units received from five manufacturers. Properties included compressive strength, absorption, and unit weight (required by ASTM C 140) plus boiled absorption and volume of permeable voids (ASTM C 642).11 ASTM C 457 air-void system parameters were also evaluated. The utility of the NCMA Index as a frost durability index was also assessed for the data collected.

RESEARCH SIGNIFICANCE

This study evaluated various SRW concrete material characteristics as frost durability indexes. SRW concretes possess a distinct microstructure consisting of irregularly shaped and interconnected voids that differ from the typically spherically shaped and discrete air voids in ordinary concretes. The different SRW microstructure is largely due to the use of low-slump material that is compacted into steel molds at block manufacturing plants. Due to a different microstructure in SRW concretes, the applicability of typical frost prediction parameters such as spacing factor to SRW concretes thus becomes questionable. This paper considers a number of common SRW material characteristics, draws qualitative and quantitative assessments of the relationship between frost performance and these characteristics, and compares these results to existing durability guidelines and specifications of these types of concretes.

METHODOLOGY

Databases of SRW concrete freezing-and-thawing performance and characteristics

As part of a larger test program on SRW durability (Federal Highway Administration [FHWA] Project No. DTFH61-02-R-00078, Durability of Segmental Retaining Wall Blocks, Spring 2003 to Summer 2006), databases of ASTM C 1262 freezing-and-thawing mass loss and material characteristics from ASTM C 140, C 642, and C 457 tests were obtained for SRW units from five different manufacturers (identified as A to E). For Manufacturers A to D, two different types of SRW units were evaluated: 1) units that satisfied Department of Transportation specifications in the state of the project (DOT units); and 2) units that did not necessarily satisfy DOT specifications (nDOT units). DOT units tended to be denser with lower amounts of compaction voids compared with nDOT units. For Manufacturer E, two types of nDOT SRW units were evaluated: 1) wall; and 2) cap units. Figure 1 illustrates the difference in application between these units. Hence, a total of 10 different types of SRW units were evaluated.

Table 1 shows the scope of ASTM C 1262 freezing-and-thawing tests conducted on these SRW units. For DOT and nDOT units from Manufacturers A to D, tests were conducted in water and 3% NaCl solution inside two different types of temperature-controlled freezers: a large walk-in freezer and a Tenney freezer (a cabinet style freezer). For the units from Manufacturer E, the tests were carried out inside a commercial chest freezer. These various types of freezers were all capable of meeting temperaturetime requirements of ASTM C 1262. For Manufacturers A to D units, five replicate specimens were tested per set, whereas for Manufacturer E units, three replicate specimens were tested per set. The data available from these tests consisted of the percent mass loss after 100 cycles.

Table 2 shows a summary of the standard material property tests performed on the SRW units. The number of replicate specimens tested varied for each type of SRW unit and ranged from three to 12 specimens. With respect to ASTM C 457 compositional parameters, the total (air and compaction) void and paste contents refer to the volumetric fraction of these phases as determined from microscopical examinations. Air and compaction voids were distinguished during the microscopy tests using the following decision rule: a compaction void was defined as a void in which less than 3/4 of its boundary was a paste-void interface, while an air void was defined as a void in which more than 3/4 of its boundary was a paste-void interface irrespective of shape of void as shown in Fig. 2. With respect to the ASTM C 457 air-void parameters, the specific surface is defined as the boundary surface area of air voids divided by the volume of the voids, whereas the spacing factor is the average half-distance between two voids. It is noted that these air-void parameters were developed for ordinary concretes in which spherical bubbles are assumed. One of the aims of this study was to assess the applicability of these parameters (determined in the usual way for concretes) for SRW concretes.

In addition to the aforementioned characteristics, other parameters were also computed from the characteristics shown in Table 2. These are as follows

... (2)

... (3)

... (4)

and NCMA Index I, as in Eq. (1).

Several other relationships were also explored without much success. This paper reports only the results obtained with the parameters shown previously.

Synthesis of data

The overall experimental program involved shared testing responsibilities between two laboratories. While the majority of freezing-and-thawing tests were conducted in one laboratory, the majority of material property testing was conducted in another laboratory. Thus, attempts to relate freezing-and-thawing mass loss to material characteristics by simply plotting single (x, y) data points (where x is the average value of a given material characteristic and y is the average mass loss) for each SRW unit were deemed inadequate. This is because specimens for freezing-and-thawing testing and for material evaluation were extracted from separate SRW units (from the same population) in the two laboratories, and it has been demonstrated elsewhere that between-unit variability may be significant (for example, variations in 24-hour water absorption from unit to unit could be up to 20%).12 A different data analysis approach, which focused more on the range of data, was consequently employed as shown in Fig. 3. Herein, it is shown that for each type of SRW unit (for example, manufacturer A DOT), there was a range of freezing-and-thawing mass-loss values among the three or five specimens tested in the set. Similarly, for other SRW specimens from the same population, there was a range of values of a given material property among the set of three to 12 specimens. Hence, for this particular SRW unit, the mass loss versus property relationship could be represented by any point in the shaded box bound by the minimum and maximum mass-loss values in the ordinate and the minimum and maximum values of the material property in the abscissa. This set of data points bound by the shaded box was then reduced in one of two ways explained in the following.

Data representation by centroids-As a first approach, the shaded box in Fig. 3 could be reduced to a single data point representing the geometrical center of the box, that is, the mid-range value in both axes, hereafter referred to as centroid (Fig. 4). The use of the center of the box for this purpose was deemed reasonable for several reasons. First, various data sets consisted of only three to four data points, which was not sufficient to fully describe a statistical distribution. For these particular sets, the centroid was used as an approximate representation of the region bound by the maximum and minimum values in each axes. For larger data sets (five or more data points), the data points appeared to be normally distributed as confirmed using the statistical test of normality (refer to Fig. 5 for an example), and the centroid was thus taken as a representation of the shaded box. Repeating this process of representing shaded boxes by centroids for all other SRW units produced a data series as shown in Fig. 4, in which each data point represented one type of SRW unit. The disadvantage of this method, however, is that, by using single data points, the significant scatter in the data is ignored.

Data representation by boundary points-The other approach used to represent the shaded box of Fig. 3 was to focus on its boundary points. This method would take into account the range in test results and be more representative of the actual nature of the data. Hence, for a given SRW type, its mass-loss versus property relationship would be then represented by four boundary points (Fig. 6). Repeating this process for all other SRW unit types, a series of data points representing the corners of all boxes was produced as shown in Fig. 6.

A data series (mass loss versus material characteristic) constructed by either one of the methods shown in Fig. 4 or 6 provided a qualitative representation of the behavior of the variables considered. To obtain a quantitative interpretation of the dependence between variables, a relationship was fitted through the data points using curve-fitting software (also shown in Fig. 4 and 6). As seen in several of the following graphs, the curve-fits often served to demonstrate that no reliable relationship is indicated. The curve-fitting analysis yielded a list of equations ranked in order from highest to lowest correlation coefficient R^sup 2^. Typically, the equation with highest R^sup 2^ value was selected as being representative of the relationship between mass loss and a given material characteristic. These R^sup 2^ values were also used as a basis to compare the correlation strength between freezing-and-thawing mass loss and the various material characteristics considered. In addition to R^sup 2^ values, standard errors expressed in percent mass loss were also obtained for each curve fit. From the previous plots, it was also possible to determine approximate values of the material property above or below which the mass loss would exceed a particular limit (for example, 1% as in ASTM C 1372), as illustrated in Fig. 4 and 6.

RESULTS AND DISCUSSION

Freezing-and-thawing tests in water

Results are shown in Fig. 7 to 19 for ASTM C 1262 freezing-and-thawing mass loss in water versus the following material characteristics. Figure 7: compressive strength; Fig. 8: 24-hour absorption; Fig. 9: unit weight; Fig. 10: NCMA Index; Fig. 11: boiled absorption; Fig. 12: volume of permeable voids; Fig. 13: saturation coefficient; Fig. 14: total air and compaction voids content; Fig. 15: paste content; Fig. 16: paste-to-total voids ratio; Fig. 17: specific surface; Fig. 18: spacing factor; and Fig. 19: (specific surface)/total voids content. In each of these figures, the upper graph corresponds to data representation by boundary points while the lower graph corresponds to data representation by centroids. With respect to the number of data points in each graph, it is noted that for units from Manufacturers A to D, tests were carried out in two types of freezers, and as such, two sets of data per SRW unit type from these manufacturers were available. Hence, the total number of boxes in the upper graphs or centroid points in the lower graphs was equal to 18 (four manufacturers [A to D] ? 2 unit types [DOT and nDOT] ? 2 freezers [walk-in freezer and Tenney freezer] + Manufacturer E ? 2 unit types [wall and cap] ? 1 freezer [chest]).

From Fig. 7 to 19, the scatter in test results was evident by the size of the constructed boxes. This scatter was particularly evident for freezing-and-thawing mass loss as reflected by the height of the boxes. Despite this scatter, trends in the data were generally discernable and as expected. For example, mass loss increased with increasing total voids content and absorption (24-hour and boiled), and with decreasing paste content and unit weight. Lower mass loss also corresponded to higher values of the NCMA Index, as reported in Reference 6. The decreasing mass loss with increasing specific surface was comparable to the trend expected for conventional concretes. As for spacing factor, however, the observed trend was actually opposite to that expected for ordinary concretes, where lower spacing factors are characteristic of systems with smaller and more closely spaced air voids and hence better frost durability. It is likely that this parameter cannot be applied to SRW concretes in the same manner as applied to ordinary concretes perhaps due to differences in the air-void structure. Mass loss also decreased with increasing value in the computed parameter (specific surface)/(total voids volume). This parameter could be envisioned as the amount of surface area afforded per unit volume of voids.

To supplement observations of data trends in these plots, curves were fit through the data points to provide a quantitative assessment of these trends. The inset boxes in Fig. 7 to 19 show the expressions for the best-fit curves, along with corresponding R^sup 2^ and standard error values for each material property considered. (It is noted that these equations are only valid within the range of values of SRW unit characteristics included by the test data sets.) It is apparent that these equations were generally of power or exponential form suggesting that mass loss was fairly sensitive to the characteristics considered. This sensitivity can also be visually detected from the shape of the curves where the mass loss exhibited steep changes relative to most of the material characteristics.

Values of R^sup 2^ in all cases were generally low (up to 0.33 for analysis with boundary points and up to 0.69 for analysis with centroids), due in part to scatter in test data. These R^sup 2^ values were nevertheless used to rank the various material characteristics for their correlation strength to mass loss. Ranking of material characteristics was carried out using two different sets of R^sup 2^ data:

* R^sup 2^ values obtained from the relationships between material characteristics and all available mass-loss values (up to 100%).

* R^sup 2^ values obtained from the relationships between material characteristics and mass-loss values only up to 5%, as illustrated in Fig. 20.

The reason for performing an analysis with mass loss of only up to 5% is because in practice, for SRW mixture qualification, freezing-and-thawing tests may be discontinued as soon as test specimens exhibit a substantial amount of mass loss (that is, several times greater than the maximum allowed by project specifications). The analysis using maximum 5% mass loss data thus reflects better-performing units that have shown little damage after 100 cycles, and focuses attention closer to the range of typically specified values. Ranking of material characteristics was thus possible under four separate analyses:

Mass loss up to 100% using:

* boundary points

* centroids

Mass loss up to 5% using:

* boundary points

* centroids

Table 3 shows a summary of various material characteristics ranked in order from highest to lowest R^sup 2^ value for these four different analyses (ranking of these characteristics based on standard error yielded similar results). Herein, it is seen that the paste-to-total-voids ratio (P/V) consistently ranked among the top three material characteristics in each of the four analyses performed. One of its components, paste content P, ranked among the top three material characteristics in two out of four analyses, while total voids content TV demonstrated highest correlation of all material characteristics considered in one of the analyses. These results demonstrate the significance of material compositional parameters on the durability of the SRW unit. The NCMA Index ranked among the top three material characteristics in two analyses and among top five in another analysis, which was likely due to the correlation strength of the 24-hour absorption (24A) and, to some extent, the unit weight OD. ASTM C 642 parameters (BA and VPV) ranked in the middle of the range of all material characteristics considered, while ASTM C 457 air-void parameters (a and L) ranked in the lower 1/3 of all material characteristics considered. The observation that mass loss correlated better with paste content than with strength was in agreement with the results from Ghafoori and Mathis13 and Hance.8 The saturation coefficient, which is commonly used as frost criterion for clay or shale bricks14 ranked lowest out of all material characteristics considered for the SRW units evaluated in this study. Hence, it is evident that typical frost performance criteria for other types of porous materials (spacing factor for ordinary concretes and saturation coefficient for bricks) may not be equally applicable to SRW concretes.

It is noted that, despite the helpfulness of curve fits in suggesting a quantitative interpretation of these data series, the actual behavior of mass loss relative to material characteristics is probably best discerned from the nature of the data itself. It is evident from Fig. 7 to 19 that the constructed boxes resembled some form of step function, whereby mass loss ranged from being negligible to almost 100% in the region to one side of a threshold value of the material property, but was low (less than approximately 1.5%) on the other side of this threshold value. These threshold values are summarized in Table 4 for the particular SRW units evaluated. Table 4 also compares ASTM C 140 threshold values to the requirements of ASTM C 1372. With respect to ASTM C 457 air-void parameters, the threshold value for specific surface (15 mm^sup 2^/mm^sup 3^ [375 in.^sup 2^/in.^sup 3^]) was found to be similar to those of concretes of satisfactory frost resistance (16 mm^sup 2^/mm^sup 3^ [400 in.^sup 2^/in.^sup 3^] in Reference 15). The threshold value for spacing factor (120 ?m [0.005 in.]), however, differed from the maximum 200 to 300 ?m (0.008 to 0.012 in.) values of frost durable concretes. In fact, these threshold values for spacing factor obtained in this study would not even be comparable to values for ordinary concretes because the trend in mass loss relative to spacing factor appeared to be different for SRW concretes than for normal concretes. For the SRW concretes evaluated in this study, as demonstrated in Fig. 18, mass loss decreased with increasing spacing factor, which is opposite to the trend observed in ordinary concretes in which frost damage increases with increasing spacing factor of air voids.

Freezing-and-thawing tests in saline

Analysis using mass loss data from freezing-and-thawing tests in 3% NaCl solution was also carried out in a manner similar to the one described in this paper (that is, data representation by centroids and by boundary points). Only the relationship between saline solution mass loss and paste-to-total-voids ratio is presented herein. Figure 21 shows the four boundary points and centroids plots where greater data scatter is observed compared to tests in water, as evidenced by the longer boxes in the vertical direction. This implied it was possible for some specimens to significantly out-perform other specimens obtained from the same SRW unit (same composition and manufacturer). Although mass loss generally decreased with increasing paste-to-total-voids ratio, which agreed with the trend observed in water, there were instances of units with relatively high paste-to-total-voids ratio that also showed significant mass loss. This observation occurred similarly for other material characteristics and indicates that units considered of high quality when in tested water may still display a substantial mass loss when tested in saline.

CONCLUSIONS

Various common SRW concrete material characteristics were assessed for their potential as frost durability indexes by evaluating their correlation to ASTM C 1262 freezing-and-thawing mass loss. Two methods of data analysis were presented, of which data representation by boundary points appeared more suitable due to its ability to reflect scatter in test data and identify threshold levels in material characteristics. Of all material properties considered, the paste-to-total-voids ratio exhibited the strongest correlation to freezing-and-thawing mass-loss in water. Its components, paste content and total voids content, also exhibited strong correlations demonstrating the importance of unit composition on frost durability. The NCMA index also ranked among top parameters in two analyses, which was likely attributed to the correlation strength of the 24-hour water absorption. The spacing factor and saturation coefficient displayed low correlation to mass loss implying that these parameters may not be applicable to SRW concretes in the same way that they apply to ordinary concretes and clay bricks, respectively. The threshold compressive strength value was determined to be well above the minimum ASTM C 140 specification value, while the threshold 24-hour water absorption was well below the maximum specification value. This implied that SRW units that barely meet standard specifications for these material characteristics may still not be frost resistant. Finally, it was demonstrated that trends and observations drawn from water tests may not be applicable to tests in saline solution due to the higher data scatter in saline tests. Units considered of high quality when tested in water may display substantial mass loss when tested in saline.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the financial support provided by the Federal Highway Administration (FHWA) in making this research possible (FHWA Project No. DTFH61-02-R-00078, Project Manager: M. Adams).

NOTATION

24A = 24-hour absorption

BA = boiled absorption

fm = compressive strength

L = spacing factor

NCMA = NCMA index

OD = oven-dry density

P = paste content

P/V = paste to total voids ratio

SC = saturation coefficient

TV = total air and compaction void content

VPV = volume of permeable voids

a = specific surface

a/TV = specific surface /total voidsx