1 Introduction
Concrete admixtures are used constantly in civil engineering projects. Most State Department of Transportation (DOT) specifications require these admixtures be tested and approved for quality and identification. These tests are important to ensure that the products have not been altered in any way to hamper their performance prior to application on the job site. One test method is to use infrared spectrophotometry scan (IR Scan) to verify the uniformity and equivalence of the job samples with the reference scan from manufacturer samples (ASTM C494/C494M-11).
In general concrete admixtures are used to enhance the concrete performance in the field. In this project, 23 of the most commonly used concrete admixtures in New Jersey Department of Transportation construction projects were selected. These include air-entrainers, accelerators, retarders, water reducers, and other combinations of these admixtures (NJDOT Material Specifications, 2011). Admixtures can accelerate/slow the setting time, improve workability, enhance frost and sulfate resistance, and help control strength development. About 80 % of concrete produced in North America contains one or more types of admixtures (Ramachandran
1995).
It is often required by the government agencies (DOT’s) to monitor the integrity of these admixtures so that they can guarantee the quality of materials that are used in their projects. ASTM (
2012) C494 requires testing of concrete admixture in accordance with Table
1 (ASTM 494) designated as Level 1 testing. It also requires Level 3 testing needed for uniformity and equivalency. Level 3 testing is established using the following requirements: (1) Infrared analysis, (2) Residue by oven drying, and (3) Specific gravity. The work done in this study focuses on the infrared analysis. ASTM C494/C494M-11 Sect.
6.1.1 requires that the absorption spectra of the initial sample and the test sample be essentially similar. This section does not provide specific criteria for acceptance or rejection of the test sample.
Table 1
Classification of approved concrete admixtures based on supplier and type.
MB-AE 90 | Master Builder (BASF) | Air | Air entraining |
MB-VR standard | Air | Air entraining |
Pozzolith 200-N | A | Water reducing |
Glenium 7500 | A and F | Water reducing and high range |
AIR MIX | Euclid | Air | Air entraining |
AEA92 | Air | Air entraining |
Eucon WR-91 | A and D | |
DARAVAIR | W.R. Grace | Air | Air entraining |
Daracem 55 | A | Water reducing |
WRDA with HYCOL | A | Water reducing |
Daracem 19 | F | High range |
Secton 6A | Great Eastern | Air | Air entraining |
Chemstrong A | A | Water reducing |
Chemstrong SP | F | High range |
Chemstrong R | D | Water reducing and retarding |
Sika Air | Sika | Air | Air Entraining |
Plastolcrete 161 | A | Water reducing |
Plastolcrete 161 FL | C and E | Water reducing and accelerating |
Catexol AE 260 | Axim | Air | Air entraining |
Catexol 1000 SP MN | A and F | Water reducing and high range |
Allegro 122 | A and F | Water reducing and high range |
Catexol 1000 R | B and D | Water reducing and retarding |
Catexol 3000 GP | A | Water reducing |
Infrared spectroscopy is used both to gather information about a compound’s structure and as an analytical tool of assessment for qualitative and quantitative analyses of the conformity of mixtures (Fernandez-Carrasco et al.
2012). These scans can be used to interpret both organic and inorganic compounds (Coates
2000). Infrared radiation is absorbed by molecules and is converted into energy of molecular vibrations. When the radiant energy matches the energy of a specific molecular vibration, absorbance occurs (Fernandez-Carrasco et al.
2012). This absorbance would then hold unique information of a specific sample (spectrum).
It is possible to obtain an IR spectrum from samples in many different forms, such as liquid, solid, and gas. However, many materials are opaque to IR radiation and must be dissolved or diluted in a transparent matrix in order to obtain spectra. Alternatively, it is possible to obtain reflectance or emission spectra directly from opaque samples (Sherman Hsu and Settler
1997).
IR spectroscopy is typically used in cases where the sample (or spectrum) is a ‘‘total unknown’’ and an identification is required, the sample (or spectrum) is an unknown and it needs to be characterized or classified, and the sample generally is known but the existence of a specific chemical class needs to be determined (Sherman Hsu and Settler
1997). IR spectroscopy can also be used when, and for the purpose of this project, the sample is a complete known and the interpretation is required to confirm the material composition and/or quality. This would include product quality control of chemical compounds such as concrete admixtures and structural steel paints.
The Louisiana DOT outlines test methods for infrared spectrophotometric analysis ( Louisiana DOT
1994). The method is used for a variety of materials such as paint, epoxy resin systems, anti-strip additives, concrete admixtures, thermo-plastics, solvents and other materials that occur as a solid, low volatile liquid, or highly volatile liquid. DOTD TR 610M-94 outlines sample preparation procedures for solid samples and liquid samples. The interpretation of results is qualitative based on a favorable comparison of the infrared spectrum to that of the original sample. According to LADOT memo DOTD TR 610M-94, a sample is considered rejected if its IR spectrum exhibits significant nonconformity to the IR spectrum of the original sample, i.e. if there are different absorption valleys in the two spectra or if an absorption valley in one spectrum is significantly displaced from that in the other one.
The California Department of Transportation (Caltrans) published tests methods for concrete admixtures in 2007. In their CA Test 416, Caltrans outlines the testing procedure for IR scan of concrete admixtures. This procedure is somewhat different from ASTM C494/C494M-11. According to the Caltrans criteria (2007), test results are used for comparison purposes only and each spectrum is compared with samples run previously. Two materials are considered similar if all of the absorption peaks match the wavelength and relative magnitude (California Department of Transportation
2007).
The Illinois Department of Transportation (IDOT) published a list of approved concrete admixtures and specifications that outlines the submittal process for the approval of new concrete admixtures (Illinois DOT Bureau of Materials and Physical Research (
2011). Among these specifications are those for the submittal of an infrared spectrophotometer trace (IR) of current production material, no more than 5 years old. The IR scan should be labeled with the date the scan was performed, the product name, and the manufacturer’s name. However, the IDOT specifications do not provide information on quantitative methods for acceptance of IR scans of concrete admixtures.
The New Jersey Department of Transportation (NJDOT) uses a quantitative assessment of IR scans based on correlation to accept or reject job samples. They use a correlation value equal to 0.975 for all admixtures based on the manufacturer recommendation. Although this may seem like a fairly high and relatively safe correlation to abide by, every admixture possesses their own unique chemical and physical properties, and may not have the same acceptable correlation values. Furthermore, the basis for using this correlation coefficient for quantitative assessment of concrete admixtures quality control was not established.
This study is seeking to establish acceptance criteria based on a rigorous testing program and statistical analyses to establish acceptable correlations as basis for quantitative assessment of infrared scans. This will help verify whether the concrete admixtures received from the job sites are acceptable using a quantitative approach.
The objective of this investigation is: (1) establish correlations coefficients and acceptable tolerances for standard manufacturer samples of concrete admixtures, (2) verify the established acceptance criteria by testing job samples, and (3) provide interpretations of IR scans of concrete admixtures including the factors that may influence them. This paper will present the findings from this study, discuss limitations and applications of developed correlation coefficients, and make recommendations for future tests that can be preformed to better understand and identify what causes the nonconformity of the IR Scans for concrete admixtures.
3 Experimental Work
In order to accomplish the objectives of this study, a number of admixtures were selected for this experimental study. Twenty-three concrete admixtures, most commonly used by contractors on construction projects in New Jersey were selected for this study (NJDOT
2007). The admixtures, their suppliers and their function are shown in Table
1.
Once all of the concrete admixtures have been identified, IR Scans were performed on the manufacturer’s samples to establish baseline data, correlations, and acceptable tolerances. The experimental procedure for the IR scans followed the ASTM C494/C494M-11 specifications which will be discussed later. To create the correlations coefficients for the selected admixtures, it was decided at the onset of the research program that three batches provided by the suppliers at three different dates will be used. For each admixture in each batch, four scans were performed. With these scans, an extensive data library was created and had enough scans to establish acceptable correlation coefficients for all the concrete admixtures. After all of the samples have been scanned and correlation coefficients have been found, job samples were tested to verify the applicability of the tolerances found from the research. Five job samples from the concrete admixtures were selected to be compared to the created correlation library. Three scans from each job sample were prepared.
5 Analysis and Results
To interpret the data obtained from the IR scans, correlation coefficients were determined for each admixture based on all scans from all batches. These correlation coefficients were then used to establish acceptance criteria and tolerances. If a specific sample achieves the established correlation threshold, this would indicate that the concrete admixture has not been altered during the manufacturing process or storage prior its use. The formula for determining the correlation coefficient of a typical admixture is based on the following statistical relationship given in Eq. (
1):
$$ r = {\text{correl}}\,(X,Y) = \frac{{\sum \left( {X - \overline{X} } \right)\left( {Y - \overline{Y} } \right)}}{{\sum \left( {X - \overline{X} } \right)^{2} \left( {Y - \overline{Y} } \right)^{2} }} $$
(1)
where, r = correlation factor, X = absorbance values of scan A of admixture/paint, \( \overline{x} \) = average of the absorbance values of scan A of admixture/paint, Y = average absorbance values of all scans from all three batches of admixture/paint \( {\bar{\text{Y}}} \) = average of the average absorbance of all scans from all three batches of admixture/paint.
The correlation coefficients obtained for the concrete admixtures from Eq. (
1) are given in Table
2. These correlations coefficients correlate the average absorbance values of all 12 scans of an admixture to the individual absorbance values of each scan of that admixture. These correlation coefficients are very close to 1.0 as expected.
Table 2
Correlation coefficients of all admixtures from all batches.
AEA-92 | 0.98310 | 0.97932 | 0.99554 | 0.99559 | 0.96255 | 0.97173 | 0.95924 | 0.95797 | 0.99096 | 0.99271 | 0.96369 | 0.97775 |
AIR MIX | 0.99060 | 0.98967 | 0.98742 | 0.99218 | 0.96462 | 0.95547 | 0.99289 | 0.91205 | 0.99177 | 0.99501 | 0.92930 | 0.90000 |
Eucon WR-91 | 0.98818 | 0.99280 | 0.99847 | 0.99590 | 0.99319 | 0.99777 | 0.99755 | 0.99228 | 0.99775 | 0.99756 | 0.99411 | 0.99298 |
MB-VR standard | 0.94894 | 0.94431 | 0.96918 | 0.89841 | 0.99290 | 0.99458 | 0.99240 | 0.97899 | 0.99167 | 0.99813 | 0.97963 | 0.98184 |
MB-AE 90 | 0.97359 | 0.97651 | 0.91430 | 0.94170 | 0.99428 | 0.99292 | 0.99287 | 0.98980 | 0.99673 | 0.99245 | 0.99742 | 0.99051 |
Pozzolit 200-N | 0.98316 | 0.97605 | 0.93357 | 0.91774 | 0.99338 | 0.99168 | 0.98666 | 0.99288 | 0.99008 | 0.99635 | 0.97845 | 0.99444 |
Glenium 7500 | 0.97332 | 0.98750 | 0.96369 | 0.90907 | 0.99406 | 0.99066 | 0.99377 | 0.99019 | 0.99059 | 0.98724 | 0.99575 | 0.99372 |
Daracem 55 | 0.99349 | 0.98156 | 0.98242 | 0.99589 | 0.90675 | 0.91322 | 0.98451 | 0.98679 | 0.98776 | 0.99122 | 0.98662 | 0.99212 |
WRDA with HYCOL | 0.99204 | 0.98877 | 0.99483 | 0.96995 | 0.95371 | 0.95501 | 0.97644 | 0.99592 | 0.97941 | 0.95937 | 0.99698 | 0.99222 |
DARAVAIR 1000 | 0.97846 | 0.98664 | 0.99425 | 0.99316 | 0.95600 | 0.98123 | 0.97285 | 0.86774 | 0.99523 | 0.99701 | 0.99652 | 0.99531 |
Daracem 19 | 0.99162 | 0.99268 | 0.99587 | 0.96570 | 0.98036 | 0.95795 | 0.99826 | 0.99739 | 0.99723 | 0.95828 | 0.99327 | 0.98780 |
Secton 6A | 0.97236 | 0.94886 | 0.87012 | 0.92054 | 0.98781 | 0.98065 | 0.98788 | 0.98103 | 0.98920 | 0.97570 | 0.91925 | 0.91164 |
Chemstrong A | 0.98165 | 0.97771 | 0.96748 | 0.92071 | 0.98240 | 0.98230 | 0.95930 | 0.93586 | 0.98134 | 0.97999 | 0.98366 | 0.98999 |
Chemstrong SP | 0.98027 | 0.98107 | 0.98015 | 0.94295 | 0.98027 | 0.97784 | 0.98422 | 0.95306 | 0.99744 | 0.99547 | 0.95701 | 0.97587 |
Chemstrong R | 0.97634 | 0.98966 | 0.99723 | 0.99782 | 0.99555 | 0.99313 | 0.90430 | 0.90700 | 0.98760 | 0.99040 | 0.99454 | 0.99124 |
Sika Air | 0.94308 | 0.94187 | 0.96079 | 0.96890 | 0.96458 | 0.96516 | 0.97995 | 0.98260 | 0.96262 | 0.86390 | 0.95744 | 0.94597 |
Plastolcrete 161 | 0.91355 | 0.93069 | 0.99413 | 0.98323 | 0.94565 | 0.91742 | 0.97270 | 0.94722 | 0.97886 | 0.96491 | 0.96286 | 0.94246 |
Plastolcrete 161 FL | 0.93942 | 0.95225 | 0.99372 | 0.99431 | 0.97396 | 0.98917 | 0.98769 | 0.98434 | 0.98597 | 0.97983 | 0.98659 | 0.98692 |
Catexol AE 260 | 0.99371 | 0.98910 | 0.98844 | 0.99276 | 0.93069 | 0.96467 | 0.98305 | 0.98845 | 0.99683 | 0.99250 | 0.98409 | 0.99226 |
Catexol 1000 SP MN | 0.95934 | 0.98512 | 0.94906 | 0.93108 | 0.91099 | 0.88520 | 0.97735 | 0.95325 | 0.99484 | 0.98973 | 0.96238 | 0.97225 |
Allegro 122 | 0.97526 | 0.97162 | 0.98467 | 0.98160 | 0.90964 | 0.93359 | 0.98208 | 0.97186 | 0.98835 | 0.98298 | 0.96808 | 0.92662 |
Catexol 1000 R | 0.99690 | 0.99765 | 0.99776 | 0.99840 | 0.98424 | 0.98961 | 0.99539 | 0.98151 | 0.99392 | 0.99628 | 0.99683 | 0.99782 |
Catexol 3000 GP | 0.98987 | 0.97430 | 0.98852 | 0.93328 | 0.98634 | 0.97259 | 0.97329 | 0.94194 | 0.99351 | 0.99843 | 0.99494 | 0.99269 |
5.1 Correlation Coefficients for Concrete Admixtures
To determine the target correlation coefficient for each admixture, the Fisher’s r- to- Z transformation technique was used. Fisher realized that this transformation makes the variability of correlations which are close to ± 1.00 comparable to those of mid-range correlation values (Hotelling
1953). The Fisher r-to-Z transformation method is one of several procedures available to transform the correlation coefficients into additive quantities. In this method, a transformation parameter
Z is calculated using the following equation:
$$ Z = \frac{1}{2}{ \ln }\left( {\frac{1 + r}{1 - r}} \right) $$
(2)
The standard error in Z is given by Eq. (
3)
$$ {\text{SE}}_{\text{Z}} = \frac{1}{{\sqrt {n - 3} }} $$
(3)
The arithmetic mean of the Z values is obtained using Eq. (
4):
$$ \overline{Z} = \left( \frac{1}{n} \right)\sum\limits_{i = 1}^{n} {Z_{i} } $$
(4)
The Fisher weighted mean correlation coefficient of the 12 scans from the three batches is determined using Eq. (
5):
$$ r = { \tanh }\overline{Z} = \frac{{e^{{\overline{Z} }} - e^{{-\overline{Z} }} }}{{e^{{\overline{Z} }} + e^{{-\overline{Z} }} }} $$
(5)
The weighted mean correlation coefficient (
r) of all 12 scans from Eq. (
5) and the coefficient of determination
r
2
(or
R2) are shown in columns (1) and (2) in Table
3 respectively. Since these samples were delivered directly from the manufacturer and were stored in lab conditions until tested, it was expected that they will achieve high correlations. The goal is to develop these correlations experimentally for each product and use them to establish target correlations and acceptance criteria for job samples. The high correlation also verifies the accuracy, consistency, and care taken in performing the scan tests. Using weighted mean correlation is recommended especially for cases when the individual correlations are not high.
Table 3
Weighted mean correlations parameters r and R2 of all admixtures.
AEA92 (A00158) | 0.96621 | 0.93357 | 6.6 |
AIR MIX (A00159) | 0.96206 | 0.92556 | 7.4 |
Eucon WR-91 (A00166) | 0.99154 | 0.98314 | 1.7 |
MB-VR standard (A00180) | 0.96821 | 0.93742 | 6.3 |
MB-AE 90 (A00181) | 0.97720 | 0.95492 | 4.5 |
Pozzolit 200 N (A00174) | 0.97267 | 0.94608 | 5.4 |
Glenium 7500 (A00189) | 0.97570 | 0.95200 | 4.8 |
Daracem 55 (A00229) | 0.96989 | 0.94069 | 5.9 |
WRDA with HYCOL (A00210) | 0.97247 | 0.94570 | 5.4 |
DARAVAIR 1000 (A00215) | 0.97655 | 0.95365 | 4.6 |
Daracem 19 (A00203) | 0.98202 | 0.96437 | 3.6 |
Secton 6A (A00226) | 0.93671 | 0.87743 | 12.3 |
Chemstrong A (A00222) | 0.95173 | 0.90579 | 9.4 |
Chemstrong SP (A00223) | 0.96360 | 0.92853 | 7.1 |
Chemstrong R (A00221) | 0.97880 | 0.95805 | 4.2 |
Sika Air (A00474) | 0.92189 | 0.84987 | 15.0 |
Plastolcrete 161 (A00144) | 0.92912 | 0.86327 | 13.7 |
Plastolcrete 161 FL (A00479) | 0.96853 | 0.93805 | 6.2 |
Catexol AE 260 (A00398) | 0.97648 | 0.95352 | 4.6 |
Catexol 1000 SP MN (A00400) | 0.93749 | 0.87889 | 12.1 |
Allegro 122 (A00397) | 0.94472 | 0.89249 | 10.8 |
Catexol 1000 R (A00402) | 0.99118 | 0.98244 | 1.8 |
Catexol 3000 GP (A00394) | 0.97297 | 0.94668 | 5.3 |
| | Average | 6.9 |
6 Acceptance Criteria
As mentioned in the introduction, state DOT’s are using different methods for the assessment of IR scan test results of concrete admixtures from job sites. Few State DOT’s have a quantitative assessment procedure in place for infrared analysis. The NJDOT is currently using a target correlation coefficient of 0.975 for acceptance criteria for all admixtures (Najm et al.
2011). This value was recommended by the manufacturer of the IR spectroscopy system; however, the basis of this target value was not established. The coefficient of determination (
r2) of the correlation coefficient provided by the manufacturer is (0.975)
2 = 0.9506. This means about 95 % of the total variation in absorbance can be explained by the linear relationship. Accepting a correlation coefficient of 0.975 thus means accepting that the other 5 % of the total variation remains unexplained or determined by other variables or by chance. These unexplained data can also be looked at as an “error” in
r2. Examining the data obtained in this study for
r and
r2 in Table
2 indicates that the coefficient of determination
r2 will vary from 0.98314 for admixture Eucon 91-R to 0.84987 for admixture Sika Air. The average error for all the admixtures in the last column in Table
2 is about 6.9 %. This average error is used to establish target value for r
2 as follows:
$$ r^{ 2} = { 1 - }\left( {\text{average error}} \right) \, = { 1\,-\,}0.0 6 9 { } = \, 0. 9 3 1 $$
Therefore, the corresponding target correlation r is given by the square root of
r2. In this case, the target correlation
r is equal to 0.965. Thus using average values from all admixtures and accepting an error of about 6.9 %, the target correlation value of all admixtures tested in this study will be 0.965. To be more specific, one can establish a target correlation for individual admixtures using the data in Table
3. For example, the target correlation for admixture Daracem55 will be 0.9698 with an error of about 6 % while that of admixture Glenium 7500 will be 0.9757 with an error of 4.8 %. The use of specific target correlation values for individual admixtures is more accurate. On the other hand, using an average correlation of 0.965 for all admixtures in this study is also acceptable given the value of the average error compared to the errors of the individual correlations.
7 Job Samples
The established (target) correlation coefficients for all admixtures evaluated in this study were tabulated in column (1) in Table
3. Also tabulated are the coefficients of determination
r2. The established correlations will be used to quantitatively assess job samples from road and bridge construction sites. Five job samples of admixtures were tested against the established target correlations to observe the applicability and the reliability of these correlations in providing quantitative quality assurance and quality control of concrete admixtures. Three IR scan tests were performed for each job sample. The five job samples were designated as follows:
These admixtures were actual job samples supplied by the NJDOT from several of their construction projects. Three IR scans from each job sample were prepared and compared to the target correlation coefficient of each admixture. Comparison of the individual correlations of the three job samples (total 15 scans) to the target correlation are shown in Table
4. One way to compare the results is, if any one of the job sample individual correlation coefficients is equal to or higher than the established correlation coefficient, then the job sample will be approved (pass); otherwise it will be rejected (fail). Table
4 shows that when comparing the correlation values of the individual scans to the target correlation, 10 out of 15 scans passed (4 out of the 5 job samples). Comparison of the individual correlations of the three job samples (15 scans) to a proposed average target correlation value of 0.965 for all admixtures is shown in Table
5. The comparison in Table
5 shows 9 out of 15 scans pass (4 out of the 5 job samples). Finally, comparing the average correlations of the three scans of each of the 5 job samples to the target correlation (0.965) in Table
6 shows that 4 out of 5 samples pass.
Table 4
Quantitative assessment of job samples using individual correlations.
ADMX 1 | 0.96206 | 0.92503 | FAIL | 0.92505 | FAIL | 0.93099 | FAIL |
ADMX 2 | 0.99154 | 0.99336 | PASS | 0.92132 | FAIL | 0.93668 | FAIL |
ADMX 3 | 0.97720 | 0.99673 | PASS | 0.99370 | PASS | 0.99639 | PASS |
ADMX 4 | 0.97267 | 0.98759 | PASS | 0.99612 | PASS | 0.99498 | PASS |
ADMX 5 | 0.93671 | 0.98497 | PASS | 0.95242 | PASS | 0.96969 | PASS |
Table 5
Quantitative assessment of individual job samples using target correlation of 0.965.
ADMX 1 | 0.965 | 0.92503 | FAIL | 0.92505 | FAIL | 0.93099 | FAIL |
ADMX 2 | 0.965 | 0.99336 | PASS | 0.2132 | FAIL | 0.93668 | FAIL |
ADMX 3 | 0.965 | 0.99673 | PASS | 0.99370 | PASS | 0.99639 | PASS |
ADMX4 | 0.965 | 0.98759 | PASS | 0.99612 | PASS | 0.99498 | PASS |
ADMX 5 | 0.965 | 0.98497 | PASS | 0.95242 | FAIL | 0.96969 | PASS |
Table 6
Quantitative assessment of job samples average correlations compared to target correlations.
ADMX 1 | 0.965 | 0.92708 | FAIL |
ADMX 2 | 0.965 | 0.96761 | PASS |
ADMX 3 | 0.965 | 0.99580 | PASS |
ADMX 4 | 0.965 | 0.99377 | PASS |
ADMX 5 | 0.965 | 0.97210 | PASS |
Quantitative assessments using the average correlations of the job samples and a target correlation of 0.965 seems to be acceptable acceptance criteria for most admixtures. The average error this case will be 6.9 % based on 276 performed IR scans for a total of 23 admixtures. As shown in Table
3 earlier, error levels vary for different admixtures and for certain admixtures lower correlation values may be used based on observations from job sample tests. More testing of job samples is needed to verify the consistency of the test results and to have more confidence in using individual target correlations instead of using an overall target correlation coefficient equal to 0.965. Also further testing from additional manufacturer samples and batches is needed for further investigation and justification of the target correlations and for continuous improvement of the target correlations.