Long-Term Performance of Epoxy Adhesive Anchor Systems (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

59 The purpose of this chapter is to describe the procedures used to reduce the experimental data into usable results. The tests were labeled with a series of letters and numbers. The short-term load tests are identified as TS-A-ST-R, where: TS: Test Series (01–16, 21, 22); A: Signifies the adhesive type (A, B, or C); ST: Signifies short-term test; and R: Test repetition number (1–13). The sustained load tests are identified as TS-A-PP-R, where: TS: Test Series (01–16, 21, 22); A: Signifies the adhesive type (A, B, or C); PP: Signifies stress level percentage (85, 75, 65, etc.); and R: Test repetition number (1–15). The tests on the effects of early-age concrete (TS17) are identified as DDD-A-ST-R, where: DDD: Day of testing (D04, D07, D14, D21, D28); A: Signifies the adhesive type (A, B, or C); ST: Signifies short-term test; and R: Test repetition number (1–5). Short-Term Anchor Pullout Load Testing The short-term load tests were conducted as described in Chapter 2. The following provides the test results. Short-term Load Test Results The load-displacement graphs along with the peak load and displacement values for the short-term load tests con- ducted at the University of Florida and the University of Stuttgart are included in Appendix I. Rejection of Outliers The modified Thompson tau technique was used to test for outliers. In this method, the absolute value of the deviation (di) of a data point from the mean is compared against the standard deviation (sx) times Thompson’s tau value (t), which is tabu- lated by number of data points and can be found in most sta- tistics textbooks. The modified Thompson’s tau value is 1.572 for five data points and 1.798 for ten data points (Wheeler and Ganji, 2004). A data point is rejected if di > sxt. If a data point is rejected, the mean and standard deviation are recalculated from the remaining values. The following data points (Table 19) were determined to be outliers by the Thompson tau technique and chosen for rejection. These are assumed to have failed at lower bonds stresses due to incomplete curing issues with adhesive C. During installation of the above two series of five anchors (series 7 and 16) there were times at which the installer stopped the continuous injection and set the cartridge gun down for a few minutes. During this set-down period it appears that some unequal mixing of components occurred in the mixing nozzle. Adhesive C was significantly more difficult to dispense by hand during installation, compared to the other adhesives. The difficulty in dispensing indicated that at least one of the components was very viscous. If one component was signifi- cantly more viscous than the other, it is possible that during the set-down period there could have been an abundance of the other component (which flowed more easily) in the mix- ing nozzle. When the cartridge gun was picked back up and the installation resumed, the adhesive in the nozzle had an improper ratio of adhesive components. This resulted in the following repetition being poorly mixed and at low strength. Any subsequent repetitions seemed to be at full strength as the poorly mixed adhesive in the nozzle had been replaced. A qualification test for evaluating this effect is presented in the proposed AASHTO Standard Method of Test in Appendix O. Statistical Analysis The results of a statistical analysis for each test are pre- sented in Table 20 through Table 25 for the tests conducted C H A P T E R 3 Findings and Applications

60 Test Series Adhesive Repetition Value(kips) Mean (kips) i (kips) sx (kips) Result 7 Moisture (installation) C 4 8.5 21.1 12.6 11.1 REJECT 16 Test setup (unconfined) C 5 5.87 9.03 3.16 3.04 REJECT Table 19. Results of modified Thompson tau technique. Test Series Mean (kips) Std. Dev. (kips) COV Alpha- Reduct. Factor t-test1 p value Significantly Different? 1 Baseline 19.8 1.1 0.06 7 Moisture (installation) 16.2 0.9 0.06 0.82 0.00 YES 9 Hole cleaning (reduced) 18.4 0.8 0.04 0.93 0.01 YES 12 Concrete mix (DOT) 16.6 2.6 0.15 0.84 0.02 YES 13 Type of drilling (cored) 11.9 1.5 0.12 0.60 0.00 YES 14 Concrete Mix (FA) 18.5 1.2 0.07 0.93 0.04 YES 15 Concrete Mix (BFS) 17.4 0.7 0.04 0.88 0.00 YES 16 Test setup (unconfined) 10.4 0.3 0.03 0.53 0.00 YES 1Student’s t-test is one-sided at a confidence level of 90%. Table 20. Statistical analysis for short-term tests on adhesive A at University of Florida. Test Series Mean (kips) Std. Dev. (kips) COV Alpha- Reduct. Factor t-test1 p value Significantly Different? 1 Baseline 25.7 1.3 0.05 7 Moisture (installation) 24.1 1.4 0.06 0.94 0.03 YES 9 Hole Cleaning (reduced) 23.8 1.4 0.06 0.93 0.02 YES 12 Concrete Mix (DOT) 22.4 2.0 0.09 0.87 0.01 YES 13 Type of Drilling (cored) 18.7 1.7 0.09 0.73 0.00 YES 14 Concrete Mix (FA) 23.4 2.9 0.12 0.91 0.08 YES 15 Concrete Mix (BFS) 25.5 0.5 0.02 0.99 0.39 NO 16 Test setup (unconfined) 11.0 1.0 0.09 0.43 0.00 YES 1Student’s t-test is one-sided at a confidence level of 90%. Table 21. Statistical analysis for short-term tests on adhesive B at University of Florida.

61 Test Series Mean (kips) Sd. Dev. (kips) COV Alpha- Reduct. Factor t-test1 p value Significantly Different? 1 Baseline 26.3 1.7 0.06 72 Moisture (installation) 24.2 0.9 0.04 0.92 0.01 YES 9 Hole Cleaning (reduced) 21.3 1.4 0.06 0.81 0.00 YES 12 Concrete Mix (DOT) 25.1 1.0 0.04 0.95 0.05 YES 13 Type of Drilling (cored) 23.2 0.2 0.01 0.88 0.00 YES 14 Concrete Mix (FA) 26.5 0.6 0.02 1.01 0.37 NO 15 Concrete Mix (BFS) 24.8 0.8 0.03 0.94 0.02 YES 163 Test setup (unconfined) 9.8 0.9 0.09 0.37 0.00 YES 1Student’s t-test is one-sided at a confidence level of 90%. 2Repetition 4 of test series 7 is considered an outlier and is not included in statistical calculations. 3Repetition 5 of test series 16 is considered an outlier and is not included in statistical calculations. Table 22. Statistical analysis for short-term tests on adhesive C at University of Florida. Test Series1 Mean (kips) Std. Dev. (kips) COV Alpha- Reduct. Factor t-test2 p value Significantly Different? 2 Baseline 14.7 0.6 0.04 5 Installation direction (horizontal) 15.8 0.5 0.03 1.07 0.01 YES 6 Installation direction (overhead) 16.1 0.6 0.04 1.09 0.00 YES 1Test series 3, 4, and 8 were determined from other criteria as discussed later. 2Student’s t-test is one-sided at a confidence level of 90%. Table 23. Statistical analysis for short-term tests on adhesive A at University of Stuttgart. Test Series Mean (kips) Std. Dev. (kips) COV 2 Baseline 19.3 0.7 0.04 Table 24. Statistical analysis for short-term tests on adhesive B at University of Stuttgart. Test Series Mean (kips) Std. Dev. (kips) COV 2 Baseline 18.5 1.0 0.05 Note: Adhesive C was not used for sustained load investigation. Table 25. Statistical analysis for short-term tests on adhesive C at University of Stuttgart. on adhesives A through C at the University of Florida and the University of Stuttgart, respectively, to compare the baseline short-term test results to the short-term test results for each parameter. The statistical analysis includes the mean, stan- dard deviation, and coefficient of variation for each data set. An alpha-reduction factor is also calculated as the mean of a particular test series divided by the mean of its respective baseline test series. A one-sided student t-test with a confidence interval of 90% was conducted on each test series against its respective baseline test series to determine if the results of a particular test series were significantly different from its respective baseline test series. A one-sided t-test was chosen with the null hypothesis so that the mean of a test series was not less than the mean of

62 at the University of Florida and at the University of Stuttgart. The results are presented in Appendix I and Table 26. Due to the 20% increase in baseline strengths for the specimens at the University of Stuttgart, the alpha factors and sustained load tests for TS 3, 4, 8, 10, and 11 were referenced to the later short-term tests. The short-term results and resulting alpha factors for these tests are presented in Table 27 and Table 28. The 20% increase in bond strengths at the University of Stuttgart is most likely due to the increase in concrete strength between the two testing dates. While the concrete strengths for the specimens at University of Florida stayed consistent within the baseline. A 90% confidence interval was chosen as it is the common practice in ACI 355.4 and indicates a 10% signifi- cance level (a = 0.10 in t-test table). Therefore, if the p value from the t-test was less than the significance level, then the null hypothesis was rejected and the parameter test data sets were significantly different than the baseline test data sets. As the short-term tests for TS 3, 4, 8, 10, and 11 were con- ducted later in the testing program, a few baseline short-term tests were conducted near the end of the project to investigate if the bond strength changed over time. Two repetitions of adhesive A and three repetitions of adhesive B were conducted Test Series Mean (kips) Std. Dev. (kips) COV Alpha- Reduct. Factor t-test1 p value Significantly Different? 2 Baseline 22.9 0.4 0.02 3 Service temperature [>120°F (49°C)] 23.1 0.4 0.02 1.01 0.29 NO 4 Service temperature [70°F (21°C)] 27.2 0.6 0.02 1.19 0.00 YES 8 Moisture (service) 24.4 0.7 0.03 1.07 0.00 YES 1Student’s t-test is one-sided at a confidence level of 90%. Table 28. Statistical analysis for late short-term tests on adhesive B at University of Stuttgart. Lab – Adhesive1 Date of Testing Mean (kips) Standard Deviation (kips) COV Ratio of Final/Initial UF – A (initial) 8/2010 19.8 1.1 0.06 0.93 UF – A (final) 4/2012 18.3 0.4 0.02 UF – B (initial) 8/2010 25.7 1.3 0.05 0.94 UF – B (final) 4/2012 24.1 3.4 0.14 US – A (initial) 8/2010 14.7 0.6 0.04 1.17 US – A (final) 4/2012 17.2 0.8 0.05 US – B (initial) 8/2010 19.3 0.7 0.04 1.19 US – B (final) 4/2012 22.9 0.4 0.02 1UF = University of Florida, US = University of Stuttgart. Table 26. Comparison of late baseline tests to initial baseline tests. Test Series Mean (kips) Std. Dev. (kips) COV Alpha- Reduct. Factor t-test1 p value Significantly Different? 2 Baseline 17.2 0.8 0.05 10 Installation temperature (mfr min/mfr min) 18.9 0.8 0.04 1.10 0.06 YES 11 Installation temperature [mfr min/110°F (43°C)] 14.8 0.6 0.04 0.86 0.05 YES 1Student’s t-test is one-sided at a confidence level of 90%. Table 27. Statistical analysis for late short-term tests on adhesive A at University of Stuttgart.

63 of Florida (UF) were very close, 98% and 99%, respectively. The ratio of the means for Adhesive C was 92%. Figure 42 compares the bond stress results from both labo- ratories. The means are plotted with an error bar indicating one standard deviation spread above and below the mean. This shows that the bond stress results between the two labo- ratories are statistically equivalent. The stress–displacement graphs along with the peak stress and displacement values for the short-term load tests con- ducted at the University of Florida and the University of Stuttgart are included in Appendix I. Selection of Adhesive for Sustained Load Investigation The determination of which adhesive to test for sustained load performance for test series 5 through 7, 9, and 12 through 16 was based on the lowest alpha-reduction factor. A summary of the alpha-reduction factors is presented in Table 30 and accompanying Figure 43. The adhesives chosen for sustained load evaluation are highlighted in Table 30. The NCHRP project panel recommended testing two adhe- sives for sensitivity to installation direction, therefore two adhesives were chosen for test series 5 and 6. Adhesive A was chosen as it exhibited the lowest alpha-reduction factor. Near the testing period, the concrete strengths at the University of Stuttgart increased approximately 50% over the course of the project. This is most likely due to the CEM I 32.5R cement used in Stuttgart which was a blended cement with pozzolans. Bond Stress Analysis As the tests at the University of Florida and the University of Stuttgart were conducted with different anchor diameters and embedment depths, the static bond stress was calculated for the baseline tests series 1 and 2 for comparison (Table 29). For adhesives A and B the means of the bond stresses deter- mined by the University of Stuttgart (US) and the University Lab – Adhesive Mean (psi) Standard Deviation (psi) COV Ratio Of Means US/UF UF – A 3,226 180 0.06 0.98 US – A 3,153 129 0.04 UF – B 4,182 218 0.05 0.99 US – B 4,125 156 0.04 UF – C 4,293 277 0.06 0.92 US – C 3,949 204 0.05 UF = University of Florida, US = University of Stuttgart. Table 29. Bond stress analysis. Figure 42. Bond stress analysis.

64 Test Series AdhesiveA Adhesive B Adhesive C UF 2 US2 3 Service temperature (>120°F (49°C))3 1.01 X 4 Service temperature (70°F (21°C))3 1.19 X 5 Installation direction (horizontal)4 1.07 X 6 Installation direction (overhead)4 1.09 X 7 Moisture (installation) 0.82 0.94 0.92 X 8 Moisture (service)3 1.07 X 9 Hole Cleaning (reduced) 0.93 0.93 0.81 X 10 Installation temperature (mfr min/mfr min)3 1.10 X 11 Installation temperature (mfr min/110°F (43°C))3 0.86 X 12 Concrete Mix (DOT) 0.84 0.87 0.95 X 13 Type of Drilling (cored) 0.60 0.73 0.88 X 14 Concrete Mix (FA) 0.93 0.91 1.01 X 15 Concrete Mix (BFS) 0.88 0.99 0.94 X 16 Test setup (unconfined) 0.53 0.43 0.37 X 1Adhesives chosen for investigation of sensitivity to sustained loading are highlighted. 4Adhesive A was chosen for test series 5 & 6 based on separate preliminary short-term tests. 3Test series 3, 4, 8, 10, & 11 used other criteria besides the lowest alpha-reduction factor to select the the product selected was determined. 2UF = University of Florida, US = University of Stuttgart. product for sustained load investigation as discussed below. Therefore only the alpha-reduction for Table 30. Summary of alpha-reduction factors.1 Figure 43. Summary of alpha-reduction factors per test series.

65 Discussion on Unconfined Results At the time of installation and testing for test series 16 (unconfined setup) the concrete compressive strength was 4,360 psi. The confined bond strengths as determined from the short-term baseline tests were as follows: • tconfined,adhesive A = 3,225 psi • tconfined,adhesive B = 4,180 psi • tconfined,adhesive C = 4,290 psi Due to the high confined bond strength of these adhesives it was anticipated that concrete breakout failure would occur for the standard 5⁄8” × 3.125” anchor used in the unconfined short-term tests. Taking a coefficient for mean concrete breakout strength of k = 35 from Fuchs et al. (1995), the predicted concrete break- out strength (Ncb) was: 35 4,360 3.125 12,800 12.8 1.5 1.5 N k f h N psi in N lbf N kips cb c ef cb cb cb ( ) = ′ = = = Assuming a 0.75 ratio (ACI 355.4-11 §10.4.5.1) of uncon- fined bond strength to confined bond strength to determine the unconfined bond strength (Na) from a series of confined tests for each adhesive was: 0.75 N d h N d h a unconfined ef a unconfined ef = τ pi = τ pi 0.75 3,225 0.625 3.125 14,800 14.8 , 2 , , N psi in in N lbf N kips a adhesive A a adhesive A a adhesive A ( ) ( )( )= pi = = 0.75 4,180 0.625 3.125 19,200 19.2 , 2 , , N psi in in N lbf N kips a adhesive B a adhesive B a adhesive B ( ) ( )( )= pi = = the end of the project, it was decided not to test a second adhe- sive for test series 5 and 6 as (1) the results from the sustained load results for adhesive A did not show a significant difference from the baseline and (2) it took longer to complete the test- ing program due to the longer than anticipated test durations. The adhesives chosen for sustained load investigation for test series 3 and 4 were based on the lowest glass transition tempera- ture (Tg). The glass transition temperatures for each adhesive based on DSC analysis conducted at the University of Florida are presented in Table 31. Adhesive B was chosen for investiga- tion for sustained load sensitivity for test series 3 and 4. The adhesive chosen for sustained load investigation for test series 8 was based on the adhesive that was most sensitive to alkalinity. The manufacturers provided the results from the alka- linity sensitivity slice tests found in ICC-ES AC308 Section 9.8. The results are summarized in Table 32. Adhesive B was chosen for investigation for sustained load sensitivity for test series 8. It was initially decided to choose the adhesive for sustained load investigation for test series 10 and 11 based on the lowest degree of cross-linking. The values for the degree of cross- linking for each adhesive based on DSC analysis conducted at the University of Florida are presented in Table 33. However, the adhesive with the lowest degree of cross-linking had a relatively high temperature for the lowest permissible instal- lation temperature. Table 34 summarizes the lowest permis- sible installation temperatures. Adhesive A was chosen for investigation for sustained load sensitivity for test series 10 and 11 as it had the second lowest degree of cross-linking and the lowest permissible installation temperature. Parameter Adhesive A (°C) Adhesive B (°C) Adhesive C (°C) Week cure 52 51 55 Notes: Values obtained from DSC tests performed at the University of Florida. Table 31. Glass transition temperatures. Parameter Adhesive A Adhesive B Adhesive C Alkalinity sensitivity 0.95 0.86 1.00 Notes: Values provided by the manufacturers. Table 32. Alkalinity sensitivity reduction factor. Parameter Adhesive A (%) Adhesive B (%) Adhesive C (%) Week cure 95.4 96.3 87.8 Notes: Values obtained from DSC tests performed at the University of Florida. Table 33. Degree of cross-linking. Parameter Adhesive A (°C) Adhesive B (°C) Adhesive C (°C) Installation Temperature 0 5 10 Table 34. Lowest manufacturer specified installation temperature.

66 15,600 15.6 N lbf N kips cb cb = = The short-term tests results for test series 16 verification tests at 110°F (43°C) are presented in Table 35 and Figure 44. The short-term tests results for test series 16 verification tests at 80°F (27°C) are presented in Table 36 and Figure 45. The anchors were installed in test slabs with a minimum edge distance of 4hef and spacing from the anchor to the test frame of 2hef which is greater than or equal to the 2hef requirement in ASTM E488-10. The mean of the unconfined tests results of the verifica- tion tests was 9.7 kips at 110°F (43°C) and 11.1 kips at 80°F (27°C), which are similar to the previous test results of 9.8 kips and well below the expected concrete breakout strength of 15.6 kips and the expected unconfined bond strength using a 0.75 ratio of unconfined to confined of 19.7 kips. The alpha-reduction factor for the verification tests was 0.41 at 110°F (43°C) and 0.40 at 80°F (27°C). This indicates that for unconfined tests, temperature does not have an effect and the alpha-setup factor is well below the assumed 0.75, and lies within the range of 0.35 to 0.55 for these products. Sustained Load Anchor Pullout Testing The sustained load (creep) tests were conducted as described in Chapter 2. The following provides the test results of the sustained load tests. Modification to Testing Program It was initially decided to test the baseline series at 85%, 75%, 65%, and 55%. However due to very early failures in the 85% and 75% stress levels, it was decided to test adhesives A, B, and C at 45% and adhesive A at 35%. Near the end of the project, several more tests were conducted at the higher stress levels ~65% to 85% to reexamine the early failures. Additionally, test series 3 to 16 were initially scheduled to be tested at three different stress levels. Based on the above 0.75 4,290 0.625 3.125 19,700 19.7 , 2 , , N psi in in N lbf N kips a adhesive C a adhesive C a adhesive C ( ) ( )( )= pi = = For the tests, high-strength steel was used to prevent yield- ing during testing (ASTM A354 grade BD) with a tensile strength fu = 150 ksi and a yield strength fya = 130 ksi. The steel yield strength of a 5⁄8” diameter (Ase = 0.226 in2) threaded rod (Nsa) is: 0.226 130 29.4 2 N A f N in ksi N kips sa se ya sa sa ( )( ) = = = As a result, the unconfined short-term tests were expected to exhibit concrete breakout at around 13 kips. The short-term tests results for test series 16 (unconfined setup) are presented in Table 20, Table 21, and Table 22. The anchors were installed in test slabs with a minimum edge dis- tance of 2.56hef and spacing from the anchor to the test frame of 2hef which is greater than or equal to the 2hef requirement in ASTM E488-10. As indicated in Table 20, Table 21, and Table 22, the failure loads were less than the expected 13 kips from concrete breakout for all products. The anchors had an apparent bond failure mode charac- terized by a shallow cone at the top and a bond failure along the lower portions of the anchor. As the alpha-setup ratios (0.53, 0.43, and 0.37) were much less than the accepted ratio of 0.75, a series of verification tests was conducted as described below. A series of short-term tests with adhesive C was conducted in higher strength (6,550 psi) concrete to verify the short- term results for test series 16. For the new concrete blocks, the predicted concrete breakout strength (Ncb) was: 35 6,550 3.125 1.5 1.5 N k f h N psi in cb c ef cb ( ) = ′ = Test Setup Test Repetition (kips) Mean STD COV Alpha-Setup 1 2 3 Confined 22.7 24.8 1.5 * 23.8 1.4 0.06 Unconfined 10.3 8.9 10.1 9.7 0.7 0.08 0.41 Notes: * Test repetition 3 for the unconfined tests was considered an outlier and was not used in the calculation of the mean. Prior to testing, the adhesive was still tacky after a week of curing. After testing, the anchor was removed from the hole and the adhesive was still tacky, indicating that it was not fully cured. Table 35. Test series 16 (unconfined setup) short-term verification tests results with adhesive C in higher strength concrete at 110°F (43°C).

67 Figure 44. Test series 16 (unconfined setup) short-term verification tests results with adhesive C in higher strength concrete at 110°F (43°C). Test Setup Test Repetition (kips) Mean STD COV Alpha-Setup 4 5 6 Confined 26.4 29.3 28.1 27.8 2.1 0.07 Unconfined 10.2 10.9 12.2 11.1 1.0 0.09 0.40 Table 36. Test series 16 (unconfined setup) short-term verification tests results with adhesive C in higher strength concrete at 80°F (27°C). Figure 45. Test series 16 (unconfined setup) short-term verification tests results with adhesive C in higher strength concrete at 80°F (27°C).

68 before failure. The photos and discussion can be found in Appendix H. Only a few typical examples will be discussed here. Several anchor tests that were terminated prior to failure were investigated and two different events occurred when splitting the core sample. The adhesive B samples (Figure 47) fractured through the concrete on one side of the anchor indicating that the adhesive bond between adhesive B and the steel and the concrete as well as the internal cohesive bonds were stronger than the tensile strength of the concrete. Adhe- sive C samples (Figure 48) separated between the steel and the adhesive indicating that the bond between the adhesive and the concrete was stronger than the bond between the adhesive and the steel. For short-term and long-term tests where failure occurred, two common failure modes were loss of adhesion with the concrete (Figure 49) and shearing failure along the threads (Figure 50). A common variant of the adhesive bond failure was seen in many tests in which, in some cases, it appears the adhesion discussion, the stress levels were reduced to 70%, 55%, and 40%. Due to the longer than anticipated failure durations, it was decided by the researchers with the approval of the NCHRP panel to test only some series at two stress levels. Sustained Load Displacement versus Time Test Results The displacement versus time results for the anchor pullout tests conducted at the University of Florida and the University of Stuttgart are presented in Appendix K. A sample is provided as Figure 46. It can be seen from the sample plot that the higher stress level tests have steeper slopes (creep rate) and fail more quickly than the lower stress level curves with shallower slopes. Core Sample Analysis Several anchors were cored and then split open for inves- tigation of the failure surface. Most were anchors that had failed but a few were anchors from tests that were terminated Figure 46. TS02B (US Baseline B) displacement versus time plot.

69 Figure 47. Typical terminated sample for adhesive B. Figure 48. Typical terminated sample for adhesive C. Figure 49. Typical adhesive bond failure. Figure 50. Typical shearing failure at threads.

70 Model Equation for Stress versus Time-to-Failure Relationship The SvTTF projection as listed in AASHTO TP 84-10 rec- ommends a logarithmic model. For comparison a logarith- mic model (s = m ln(t) + b) and a power model (s = AtB) were both evaluated and they resulted in essentially the same coefficient of determination (R2). It was decided to use the logarithmic model as recommended in AASHTO TP 84-10. Exclusion of Short-Term Tests in Stress versus Time-to-Failure Relationship The short-term tests were initially expected to be included on the SvTTF curve, but based on the distribution of stress along the borehole, analysis of the test results, and investiga- tion of failure modes, it was decided to not include the short- term test results in the SvTTF projection. Based on analytical work by McVay et al. (1996), Figure 4 to Figure 7 show that at low stress levels (<30% of MSL) the adhesive is still in the elastic range. At about 70% of MSL, the adhesive has undergone inelastic redistribution of stress along the entire length of the anchor. Under high stress level sus- tained load conditions, the coupling of creep strains caused by the sustained load and strains caused by inelastic redistri- bution of bond stress seem to hasten the failure. As an example, Figure 53 and Figure 54 show the results of TS01B (Baseline B) with the short-term tests excluded and with the concrete failed and a “plug” of the anchor with the adhesive still attached slipped in the hole. In other cases, portions of the adhesive also fractured within the bond line. The adhesive “plug” eventually stopped due to friction and reduction in load as the spring relaxed. As the anchor remained in the chamber, portions of the adhesive “plug” appeared to reattach to the concrete. When the sample was cored and split, either portions of the adhesive detached from the threads (Figure 51) or portions of the concrete fractured (Figure 52). The reattachment can be supported by the fact that many of the samples in Appendix H show large displacements (½”–1”) after failure with one side still attached to the core after splitting. It is not reasonable that the adhesive could dis- place this much and stay bonded to the concrete and steel. Rather the adhesive would have to debond and/or fracture, shift, and then reattach. Many of the samples remained in the 110°F (43°C) chamber for a few days prior to removal. Addi- tionally, many of the cores were not made until months after the tests concluded. This provided ample time at elevated temperatures for the adhesive to reattach. Anchor Pullout Testing Stress versus Time-To-Failure Test Results The stress versus time-to-failure (SvTTF) results for the anchor pullout tests conducted at the University of Florida and the University of Stuttgart are presented in Appendix L. Figure 51. “Plug and reattachment” failure with thread separation. Figure 52. “Plug and reattachment” failure with concrete fracture.

71 Figure 53. Baseline TS01B SvTTF plot with short-term tests excluded from the projection. Figure 54. Baseline TS01B SvTTF plot with short-term tests included in the projection (ST  short term, LT  long term).

72 This reduced expected failure stress level for short-duration loads appears to result from a dual requirement placed on the polymer. The magnitude of the load causes the polymer to undergo inelastic deformation as it redistributes the load down the anchor, and the sustained nature of the load causes the polymers to migrate within the adhesive. These two actions occurring simultaneously reduce the capacity. The lower stress level sustained load tests provide suffi- cient time for the polymer strands within the adhesive to slide past each other. This is supported by the much larger deformations seen in the sustained load tests than in the short-term tests as polymer strand migration leads to creep deformation and higher rupture displacements. For the UF baseline tests, the peak displacements in the sustained load tests were approximately 1⁄3 higher than the peak displace- ments in the short-term tests for adhesive A and double for adhesives B and C (Table 38 and Figure 55). If the peak displacements in the sustained load are compared to the limit- ing displacement (Dlim) at loss of adhesion as calculated in ACI included in the projection, respectively. By inspection it can be seen that the trend of the sustained load tests on Figure 53 does not intersect the data points of the short-term tests at 100% of MSL. Appendix L (pages L-2 to L-7) provides SvTTF figures showing all baseline data with trendlines including and not including the short-term tests. Using the constants from the regression analysis, the expected failure stress level for a 5-minute load duration for TS01B is 79% of MSL. Table 37 summarizes the expected fail- ure stress levels at a 5-minute load duration from the regres- sion analysis for the six baseline tests and from three baselines created by combining the results from US and UF. Test Series Expected Failure Stress Level at 5-Minute Load Duration (%MSL) TS01A 78 TS01B 79 TS01C 80 TS02A 71 TS02B 88 TS02C 76 A-combined 75 B-combined 82 C-combined 78 Table 37. Expected failure stress level at 5-minute load duration for baseline tests with short-term tests excluded in the SvTTF projection. Test Series ST Mean (in.) ST COV LT Mean (in.) LT COV Ratio LT/ST Baseline A 0.043 0.09 0.059 0.15 1.4 Baseline B 0.051 0.08 0.100 0.26 2.0 Baseline C 0.046 0.11 0.102 0.29 2.2 Table 38. Peak displacement data for short-term (ST) and sustained load (LT) tests for UF baselines. Figure 55. Ratio of sustained load test failure displacements to short-term test failure displacements for UF baselines.

73 Combined SvTTF Baseline Curves Figure 72 to Figure 74 present the individual and com- bined baseline curves from UF and US for the three adhesives, respectively. Since different anchor diameters and embed- ment depths were used at the two laboratories, the stresses have all been normalized by the average of the 15 short-term bond stresses (10 at UF and 5 at US). At the time of publication, TS10 had one stress level underway. Due to this limited data, TS10 has a SvTTF chart included in Appendix L, but there was not sufficient data to generate an experimental baseline. Rejection of Failures During Loading Several of the tests failed during the loading period prior to reaching the desired sustained load. It was decided that those tests that failed during loading were not reliable and were excluded from the time-to-failure projection. Tests Terminated Prior to Failure Several tests were terminated prior to failure per approval by this project’s NCHRP panel. These tests were identified as not likely to fail during the remainder of the testing program and their continued monitoring would not provide any more meaningful results than had already been obtained. The tests identified for early termination are listed below and are identi- fied in the SvTTF plots with a diamond and their test durations listed in the tables in Appendix L. Test Series ST Mean (in) ST COV LT Mean (in) LT COV Ratio LT/ST Baseline A 0.042 0.15 0.059 0.15 1.4 Baseline B 0.034 0.19 0.100 0.26 3.0 Baseline C 0.035 0.14 0.102 0.29 2.9 Table 39. Displacement data at loss of adhesion per ACI 355.4 for short-term (ST) tests and peak displacement data for sustained load (LT) tests for UF baselines. Figure 56. Failure displacement versus time to failure for all three UF baseline tests. 355.4, the ratio for adhesives B and C approaches 3 (Table 39 and Figure 55). Figure 56 and Figure 57 present the displacements versus time to failure and %MSL respectively for the short-term and sustained load tests for all three UF baseline tests. These fig- ures show that failure displacements are larger for lower stress levels and longer time to failures. Based on the above discussion, it was decided to exclude the short-term test results from the SvTTF relationships in anchor tests. Subsequently, the analysis for sustained load sensitivity to various parameters was based on projections derived only from sustained load test results. It should be noted that projections were also performed including short-term tests in the projec- tions for each test series and similar conclusions were drawn. Figure 58 to Figure 71 present the SvTTF results for test series 3 through 16 respectively with short-term results excluded from the projections. The same graphs as well as the data are presented in Appendix L.

74 Figure 57. Failure displacement versus %MSL for all three UF baseline tests. Figure 58. SvTTF TS03-B service temperature (120°F).

75 Figure 59. SvTTF TS04-B service temperature (70°F). Figure 60. SvTTF TS05-A installation direction (horizontal).

76 Figure 61. SvTTF TS06-A installation direction (vertical). Figure 62. SvTTF TS07-A moisture during installation.

77 Figure 63. SvTTF TS08-B moisture in service. Figure 64. SvTTF TS09-C reduced hole cleaning.

78 Figure 65. SvTTF TS10-A installation temperature (mfr minimum/mfr minimum). Figure 66. SvTTF TS11-A installation temperature (mfr minimum/110°F).

79 Figure 67. SvTTF TS12-A standard DOT mix. Figure 68. SvTTF TS13-B core drilling.

80 Figure 69. SvTTF TS14-B fly ash. Figure 70. SvTTF TS15-A blast furnace slag.

81 Figure 71. SvTTF TS16-C unconfined setup. Figure 72. Combined baseline SvTTF for adhesive A normalized by the average bond stress of the short-term tests from UF and US.

82 Figure 73. Combined baseline SvTTF for adhesive B normalized by the average bond stress of the short-term tests from UF and US. Figure 74. Combined baseline SvTTF for adhesive C normalized by the average bond stress of the short-term tests from UF and US.

83 Tests Still Running at Time of Publishing Those tests that were still running at the time this report was completed were included in the SvTTF plots with the current test duration and are identified with a circle and their test durations listed in the tables in Appendix L. Adhesive-Alone Testing Short-Term Results The short-term test results for the dogbone specimens are presented in Appendix I. For the baseline test series, both adhe- sives A and C exhibited brittle failures. Adhesive B was more ductile, failing at much higher strains and loads. For test series 22 using MPII cure time, adhesive A showed a slight increase in strength. Adhesive B had strengths about one half of what was seen in the week-cured specimens with significant scatter in the results (COV = 0.47) due to the fact that the testing tempera- ture was so close to the glass transition temperature. Adhesive C would break in the grips of the testing machine due to its high brittleness and was therefore unable to be tested. The alpha-reduction factors for the influence of cure time are presented in Table 40. Adhesives A and B were initially both chosen for sustained load investigation for test series 22, but due to the large ductility in the adhesive B manufacturer- cured samples the tests could not be conducted. DMTA Results For the initial characterization of the adhesives, measure- ments were conducted using a sinusoidally oscillating stress. The resulting strain was measured, and the component of the strain (e′) in-phase with the applied stress (s) was recorded and used to calculate the storage modulus (G′) by: G′ = σ ′ε The first set of experiments was a temperature ramp at a speed of 5 degrees per minute at 1 Hz on the DSR machine and the results of samples A, B, and C are shown in Figure 75, Figure 76, and Figure 77, respectively. Adhesive A showed a very broad and slow decrease of its shear storage modulus (G′), which is the typical behavior of a non-crosslinked vinyl ester with wide range of molecular weight distribution. Both adhesives B and C displayed a plateau in their storage modu- lus at a temperature above their Tg, which indicated a cross- linked system. A narrower half width of the tan delta peak in sample C suggested a more homogenous crosslink network. Crosslinks are chemical bonds formed in the adhesive during the curing reaction that cause the adhesive to harden. Cross- links restrict molecular motion, and thus crosslinks cause increased resistance to creep. Test Series AdhesiveA Adhesive B Adhesive C 22 Manufacturer Cure Time 1.05 0.54 --- Note: Short-term test results for adhesive C were not able to be obtained for the manufacturer cure time due to the brittleness of the material and its tendency to break when loaded into the Instron. Table 40. Summary of alpha-reduction factors. Figure 75. DMTA test results for adhesive A.

84 Figure 76. DMTA test results for adhesive B. Figure 77. DMTA test results for adhesive C. A thermogravimetric analysis was conducted on the sam- ples that were heated under air from 68°F to 1,470°F (20°C to 799°C) at a rate of 18°F/min (10°C/min) while recording the weight change. These experiments were done to provide basic characterization of the adhesives. Adhesive polymers will start to decompose into gas at temperatures greater than their decom- position temperature [typically higher than 734°F (390°C)]. The decomposition was considered complete as evidenced by the fact that there was no weight change in the 1,290°F to 1,470°F (699°C to 799°C) range. The inorganic filler loading was calculated as the ratio of the final weight to the initial weight. The inorganic filler loading for the three samples was 60.0% for adhesive A, 38.9% for adhesive B, and 46.4% for adhesive C. Sustained Load Strain versus Time Results Figure 78, Figure 79, and Figure 80 present the baseline strain vs. time plots of adhesives A, B, and C obtained through the sustained load creep test on dogbone specimens. Because of the different stress levels, the strain plots are scattered.

85 Figure 78. Adhesive A baseline strain vs. time plot for dogbone specimens. Figure 79. Adhesive B baseline strain vs. time plot for dogbone specimens. Sustained Load Compliance versus Time Results Normalizing the strains according to their stress, the com- pliance vs. time plots for all three adhesives are shown in Figure 81, Figure 82, and Figure 83. As the majority of the compliance curves of adhesive C overlap each other, this gives a good indication that time–temperature superposition should work for adhesive C. For adhesives A and B, the com- pliance curves at different stress levels differ from each other, indicating a nonlinear creep rate dependence on stress level. Figure 84 presents the short-term creep response (log- log plot) of adhesive A obtained through DSR creep tests at different temperatures. Figure 85 presents the shifted creep

86 Figure 80. Adhesive C baseline strain vs. time plot for dogbone specimens. Figure 81. Adhesive A baseline compliance vs. time plot for dogbone specimens. response master curve at 110°F (43°C). Due to the nonlinear behavior observed during the long-term creep test of sample A and B, the short-term creep curves might also be accelerated by the applied stress or the resulted strain. If the shift fac- tors were directly obtained through shift of the creep curve, the shift factor will be coupled with the applied stress dur- ing the short-term creep experiment. Since the frequency sweep measurements were conducted at very low strain lev- els, where the nonlinear behavior’s effect is negligible in the linear viscoelastic region, the shift factor for time–temperature superposition was obtained by shifting the frequency sweeps at different temperatures. As a result, eight frequency sweeps from 0.1 Hz to 10 Hz were performed at the same temperature of the sustained load creep tests with a maximum oscillation

87 Figure 82. Adhesive B baseline compliance vs. time plot for dogbone specimens. Figure 83. Adhesive C baseline compliance vs. time plot for dogbone specimens. strain at 0.05% and shifted with 110°F (43°C) as the refer- ence temperature. Using the shift factor obtained this way, the short-term creep tests of adhesive A shown in Figure 84 were shifted accordingly. The resulting master curve is shown in Figure 85. Figure 86 presents the comparison of the long-term creep compliance curve of adhesive A to the shifted compliance curves from the short-term creep test. Please note that the compliance of the curves from the creep test is the shear compliance and when compared with tensile compliance obtained through the sustained load creep tests, the shear compliance was divided by 2(1 + n) where n is the Poisson’s ratio of the epoxy (taken as 0.4). From this comparison, it can be seen that the prediction from the DSR creep test captured

88 Figure 84. Compliance vs. time plot for DSR creep test of adhesive A at different temperatures. Figure 85. Shifted master compliance curve for adhesive A using 43°C as a reference temperature. the overall trend and shape of the adhesive creep. However, due to the dependence of compliance on the stress level, the prediction from the DSR creep test could not be used quan- titatively for adhesive A. Using the same treatment as described above for adhesive A, the short-term DSR creep tests of adhesive B shown in Fig- ure 87 were shifted accordingly. The resulting master curve is shown in Figure 88. The comparison between the compliance obtained from the sustained load creep test and the compliance predicted from the DSR creep tests is shown in Figure 89. As with adhe- sive A, the prediction from the DSR creep test could not be used to directly predict the creep behavior of the dogbone. An apparent trend in Figure 89 is, as the load was increased during the long-term test, the compliance increased at higher speeds over time, which should be due to the accelerating of the creep mechanism due to stress. To investigate if there was any simple stress time super- position relation, five DSR creep tests on adhesive B with shear stress ranging from 72.5; 2,180; 2,900; 3,630 to 5,080 psi (0.5, 15, 20, 25, 35 MPa) were tested on the DSR machine. Any further increase of the test stress level resulted in the failure of the test specimen. The raw and shifted curves are shown in Figure 90 and Figure 91. Only when the horizontal dis-

89 Figure 86. Comparison between predicted compliance from the DSR creep test and the sustained load creep tests on dogbone samples for adhesive A. Figure 87. Compliance vs. time plot for DSR creep test of adhesive B at different temperatures. tance between the overlapping regions of two curves was a constant can satisfactory superposition be possible. The poor matches of the shifted curve indicated that no simple time– stress superposition relationship existed for adhesive B. In effect, the horizontal distance between any two curves was not a constant but a function of compliance. The horizontal distance was calculated between any two pairs of compliance curves in Figure 91 resulting in a total of 10 pairs, and this value was plotted against the log10 of the cor- responding compliance regions and is shown in Figure 92. Note that if a simple stress–time superposition relation existed, there would be a few horizontal lines at different heights, which depended on the stress level difference. For each curve in Figure 92, three numbers were labeled, indicating the stress levels of the pair of compliance curves (labeled beside each curve) and the difference between the two stress levels (labeled on each curve). Note that here 72.5 psi (0.5 MPa) was treated as 1,450 psi (10 MPa) during the calculation of the

90 Figure 88. Shifted master compliance curve for adhesive B using 43°C as a reference temperature. Figure 89. Comparison between predicted compliance from the DSR creep test and the sustained load creep tests on dogbone samples for adhesive B. stress difference as the calculated stress difference agreed with the curves they overlapped. A possible reason for the equiva- lence of 72.5 psi (0.5 MPa) with 1,450 psi (10 MPa) could be that the nonlinear behavior of adhesive B was very low under low load, allowing 72.5 psi (0.5 MPa) to be treated as 1,450 psi (10 MPa) in this analysis. An interesting observation is that the curves with the same stress level difference seem to lie on top of each other. In addition, curves with higher stress level differences had higher shift factors. Similar treatment was done to the sustained load creep compliance curves of adhesive B, which is shown in Figure 93. Here the sustained load compliance axis was plotted linearly and the logs of the shift factor became a linear function of the compliance after a certain compliance value. A linear fit was performed for all the plots using the data points where compliance equaled 345 ksi-1 (5e-10 Pa-1) and was plotted in green on Figure 93. Depending on the stress level, it took about 2 to 10 hours for the compliance to reach

91 Figure 90. Compliance versus time for the DSR creep test of adhesive B at different stress levels. Figure 91. Shifted master compliance curve for adhesive B using 72.5 psi (0.5 MPa) as the reference stress.

92 Figure 92. Shifted factor as a function of compliance for each pair of compliance creep curves for adhesive B at different stresses for short-term DSR creep tests. Figure 93. Shifted factor for adhesive B as a function of compliance for each pair of compliance creep curves at different stresses for the sustained load creep drawn in blue semi-log plot. The linear fit of each curve is shown in green (shown in color in online version).

93 this value. Clearly the curve fits are very good for all curves and all the linear fit plots roughly converge at the origin of the coordinate system. The slope of each curve fit and the stress difference for each of the fits was plotted in Figure 94. A fairly good linear fit was obtained with R2 value equal to 0.909 and slope = 44.7. In addition, the intercept on the y axis of the linear fit was very close to origin. As a result, we postulate that the shift factor for adhesive B can be approximated as: log . 510 a C D Eq= × ∆σ × where a is the shift factor, Ds is the difference in stress, D is the compliance at which the shift factor is calculated, and C is a materials dependent constant which can be calculated from the slope of the solid line of Figure 94. Since the shift factor can be regarded as the viscosity ratio of two tests, from Eq. 5 it can be seen that the viscosity is proportional to et which indicates an Eyring type of viscosity stress relationship [Lee et al. (2009)]. If there is no D term in Eq. 5, a simple stress–time superposition is sufficient to describe the stress dependence of the creep behavior. Since stress–time superposition assumed an unchanged creep mechanism, it is believed the presence of a D term indicates that there is a dependence of the underlying creep mecha- nism on the current state of the polymer during the creep test. The reason is still currently unknown, but the apparent linear relationship of the log10 of the shift factor with D makes this a very interesting problem for further investigation. Figure 95 shows the DSR creep test for adhesive C and Figure 96 shows the resulting master curve. The shift was conducted by the built-in TTS processing function of the DSR instrument. Very good agreement between the overlapping regions of the shifted curves showed that the time-temperature superposition was valid for adhesive C. The DSR master creep curve was overlaid with the com- pliance curves of adhesive C and is shown in Figure 97. The compliance of the sustained load curve did show agreement with the prediction curve at the early stage of creep but the creep predicted from DSR creep test grew faster than the ten- sile creep, which may be due to additional cure of the long- term sample during the 1,000-hour long creep tests. To test this hypothesis, the DSR creep samples were allowed to cure at 122°F (50°C) for 2 days and the master curves were con- structed again as shown in Figure 98. The discrepancy in the master curves and the long-term creep compliance happened 100 hours later than the previous result, which confirmed the effect of the additional curing. Discussion and Suggestions Time–temperature superposition can be a powerful tool for accelerated polymer testing. However, for epoxy resins used in commercial adhesive products, due to the complex- ity of their formulas, one should not assume that time– temperature superposition always works. For adhesive C used in this study, time–temperature superposition appears Figure 94. Difference in stress vs. the slope of the fit for each of the plots from Figure 93. The solid line is a linear fit to the data. R2  0.909.

94 Figure 95. Compliance vs. time for DSR creep test of adhesive C at different temperatures. Figure 96. Shifted master compliance curve for adhesive C using 43°C as reference temperature. to be a reasonable method for long-term creep prediction from very short-term DSR creep tests. As for adhesives A and B, it was observed that linear viscoelastic behavior was not valid and the creep behavior depended on the applied stress. Although the time–stress supposition method was reported to be valid for a few polymers, we found the effect of stress to be more complicated. For adhesive B, the dependence of the shift factor on the stress was quite different between the sustained load creep tests and the short-term DSR creep test. Nevertheless, it was found that after the creep compliance passed a certain point, the relationship between the shift fac- tor, compliance, and stress became simple and apparently followed Eq. 5. As a result, we suggest the following steps to predict the sustained load creep rate from a set of relative short-term tests. First, construct a master compliance curve from a DSR creep test within the linear viscoelastic range for very low stress levels following the steps as shown in Figure 87 to Fig- ure 88. The duration of each DSR creep test can be as short as 30 minutes. Second, perform DSR creep tests under different stresses to see if there is any stress dependence on the compliance. If

95 Figure 97. Comparison between predicted compliance from the DSR creep test and sustained load creep tests on dogbone specimens for adhesive C. Figure 98. Comparison between predicted compliance from the DSR creep test and sustained load creep tests on dogbone specimens for adhesive C with higher curing temperature.

96 of magnitude. The strain measured in the anchor tests is not a direct measurement of adhesive strain but rather a measurement of the total system. The total displacement measured is composed of the strain in the adhesive as well as the strain in the anchor, and slippage of the adhesive/anchor “plug” within the hole. However, creep compliance curves for the anchor pull - out tests were generated and compared the dogbone speci- mens. Figure 100 to Figure 102 present the creep compliance comparisons for the three adhesives. Only the compliance comparison for adhesive C shows a similar trend between the two sets of tests, although separated by almost two such dependence does exist, attempt the stress–time supposi- tion first. If successful, measure the short-term creep tests at the desired stress level to obtain the shift factor and shift the master curve from step one accordingly. If the stress–temperature superposition does not work, it is still possible to predict the sustained load creep rate. In this case, the short-term creep tests must be conducted long enough so that Eq. 5 becomes valid, which can take about 2 to 10 hours based on testing adhesive B. With this data, deter- mine the C term in Eq. 5. Combined with the master compli- ance curve from step one, the sustained load creep rate at any stress level can be obtained. Note that this only predicts creep rate and not time to failure. Adhesive-Alone Testing to Anchor Pullout Testing Correlation Sustained Load Test Results This project investigated the existence of a correlation between the sustained load tests performed on the adhesive anchors in concrete and the dogbone samples. However, a direct comparison cannot be made as the strain in the dogbone specimens is a tensile strain and the strain in the anchor pullout tests is a shear strain calculated as the arc- tangent of the anchor displacement over the annular gap (Figure 99). The strains calculated in the anchor tests differ from the dogbone specimens by approximately two orders Figure 99. Shear strain in anchor tests. Figure 100. Creep compliance comparison between dogbone and anchor tests for adhesive A.

97 Figure 101. Creep compliance comparison between dogbone and anchor tests for adhesive B. Figure 102. Creep compliance comparison between dogbone and anchor tests for adhesive C.

98 tests and the dogbone tests. The SvTTF curves for the dogbones are presented in Appendix L (series 21 and 22). While the dog- bone tests did a very poor job predicting the SvTTF results for adhesives B and C, they did a better job for adhesive A. This is possibly due to the poor adhesion of adhesive A. Adhesive anchor systems with better adhesion can develop more fric- tion along the sides of the hole prior to failure as the adhesive/ anchor “plug” will have pieces of concrete attached to it. orders of magnitude. Adhesives A and B do not provide good comparisons. Stress versus Time-to-Failure Results for Anchor Pullout and Dogbone Tests Figure 103 to Figure 105 present a comparison of the SvTTF relationships determined from the combined anchor pullout Figure 103. SvTTF comparison between anchor pullout tests and dogbone tests for adhesive A (MPS  mean peak stress). Figure 104. SvTTF comparison between anchor pullout tests and dogbone tests for adhesive B.

99 does not have a more adverse effect at that point in time as compared to the short-term effect. Figure 106 presents the results of this analysis for the parameters investigated. As shown in Figure 106, some short-term tests (TS03, TS05, and TS08) indicated a slight increase in strength for the given parameter. As design standards should not increase the predicted short-term strength due to slight variations above the baseline for certain parameters, it would then seem Dogbone specimens do not have this additional frictional resistance. Influence on Sustained Load As discussed earlier, the influence of a given parameter on sustained load can be evaluated by evaluating the influence ratio. If this influence ratio is less than 1, then the parameter Figure 105. SvTTF comparison between anchor pullout tests and dogbone tests for adhesive C. Figure 106. Influence ratio for each test series.

100 Parameters with Adverse Sustained Load Influence TS03—120°F (49°C) service temperature. As discussed earlier, polymers exhibit high creep deformations at elevated temperatures. It is no surprise that the long-term tests con- ducted at temperatures above the baseline temperature showed increased creep displacements. TS03 indicates that the stress level predicted by the influence ratio is 122% than that to cause failure at an equivalent lifetime of 75 years. ACI 355.4 §8.5 provides tension testing at two temperature categories. Temperature Category A stipulates a long-term temperature of 110°F (43°C) and Temperature Category B has a long-term temperature greater than or equal to 110°F (43°C). The current testing temperature for Temperature Category A of 110°F (43°C) was based on temperature measurements provided in a CALTRANS study by Dusel and Mir (1991) of a bridge in Barstow, CA, in which 110°F (43°C) was noted to occur over a few hours during the day. In the CALTRANS study, there were no recordings greater than 115°F (46°C). If it can be shown that an anchor would be expected to be at or above 120°F (49°C) for significant portions of its service life, it is suggested that AASHTO require the adhesive anchor system to be tested and evaluated for Temperature Category B at a temperature equal to or greater than its high- est service temperature. TS22—cure time. While the influence ratio of TS22 for adhesive A was less than 1 (0.95), the effect of cure time seems to appropriate to evaluate the influence ratio of these param- eters against the baseline and not against an elevated baseline. This is identical to limiting the alpha-reduction factor to a maximum value of 1 (Figure 107). As it is more appropriate for design, the practice of limiting the alpha-reduction fac- tor to 1 in the analysis of sustained load influence ratio was adopted for the remaining analysis and discussion. For the parameters investigated, most showed a decreasing trend versus time and result in an influence ratio less than 1, indicating that these adhesive products are not affected more adversely by the given parameter under sustained load than under short-term load. Figure 108 presents the same information for a structure with 15 years at elevated temperature. ACI 355.4 assumes that a structure exceeds 110°F (43°C) for only 20% of its lifetime and, as a result, projects 110°F (43°C) test data to 10 years (20% of 50 years). For an AASHTO lifetime of 75 years, the influence ratio is therefore evaluated at 15 years (20% of 75 years). Of all the parameters tested, only two were identified as having an adverse effect on the sustained load performance of adhesive anchors: 120°F (49°C) service temperature and manufacturer’s cure time. The identification of these two parameters (TS03 and TS22) as having an adverse effect was based on not only the influence ratio but on an over- view of the respective test results. The highest three influence ratios were TS03 (1.22), TS08 (1.02), and TS22 (0.95). TS08 (in-service moisture) was not considered as explained below. Figure 107. Influence ratio for each test series with the alpha-reduction factor limited to a maximum value of 1.

101 duration) then the parameter was said to have an adverse effect on the sustained load performance. Most of the fol- lowing test series have the same or more favorable in-service conditions compared to the baseline but vary by installation condition. It appears that once the adhesive has cured, any reduction in strength due to the installation condition can be completely defined by the alpha-reduction factor from short- term testing. As long as the in-service conditions are not worse than the baseline, there should not be any further reduction in strength over the service life. TS04—70°F (21°C) service temperature. Polymers will exhibit higher creep displacements at higher temperatures, especially as the temperature approaches the glass transition temperature. As discussed above, as long as the in-service conditions remain the same as the baseline, there should not be any further reduction in strength over the service life. In the case of TS04, the in-service temperature is lower than the baseline. A lower in-service temperature is a condition that is more favorable for sustained load performance. With an influ- ence ratio of 0.73, this parameter is considered not adverse to sustained load performance. TS05—installation direction (horizontal). Quality prod- ucts, for example, those that have passed ACI 355.4-11 crite- ria, that are to be used for horizontal installations must have passed the sensitivity to installation direction test (ACI 355.4-11 §7.18). In this test series the short-term load strength of a hori- zontally installed anchor must be at least 90% of the strength of an anchor installed in the downward direction. be product specific as adhesive B resulted in an alpha-reduction factor of 0.54 and was unable to be adequately tested for sus- tained load at manufacturer’s cure time to very high deforma- tions. For sustained load applications it is important that the adhesive is sufficiently cured. A practical solution would be to require a cure time for sustained load applications beyond the minimum required by the manufacturer. Research by Cook and Konz (2001) tested 20 anchor sys- tems at 24 hours and at 7 days. Almost one-half of the systems obtained 90% of the 7-day strength at 24 hours and the aver- age of all 20 obtained 88% of the 7-day strength at 24 hours. ACI 355.4-11 §8.7 has a required test method for cure time at standard temperature in which anchors tested at the man- ufacturer’s minimum cure time must achieve 90% of the strength of anchors tested at the minimum cure time plus 24 hours. It seems reasonable to require that anchors used in sus- tained load applications be required to cure for an additional 24 hours beyond the manufacturer’s minimum cure time before loading. Parameters without Adverse Sustained Load Influence The testing criterion for evaluating influence on sustained load was based on the alpha-reduction factor determined from short-term testing. If the reduction in strength at any point in time was greater than at 2 minutes (short-term test Figure 108. Influence ratio for each test series at 15 years exposure to elevated temperature (75-year design life).

102 temperature and tested at 110°F (43°C) as was the case for the baseline and evidenced by the alpha-reduction factor of 1.10. TS11—installation temperature [mfr minimum/110°F (43°C)]. It appears from the tests conducted at low tempera- ture that as long as the adhesive is installed at a temperature at or above the minimum permitted by the manufacturer that there are no adverse effects under sustained load compared to the baseline. This was noticed in TS10, which tested at the manufacturer’s minimum permitted temperature and in TS11, which tested at 110°F (43°C). It is definite that the adhesive underwent additional cure over the 24 hours as the specimens were conditioned from the installation temperature to the ele- vated testing temperature. However, the 0.86 alpha-reduction factor indicates that it was not as cured as the baseline that was installed at room temperature. However, the low influence ratio of 0.71 indicates that this parameter is not adverse to sustained load performance. TS12—DOT concrete mix. As discussed earlier, it appears that as long as the in-service conditions are the same as the baseline, the alpha-reduction factor obtained from short- term testing for the influence of concrete mix is sufficient to conservatively evaluate the sustained load performance. With an influence ratio of 0.53, this parameter is considered not adverse to sustained load performance. TS13—core drilling. While core drilling created a reduc- tion in short-term bond strength (a = 0.73) as time progressed, the reduction in strength over time was no worse than in the short term. It is believed that the short-term reduction is due to reduced friction along the smoother core drilled hole after loss of adhesion. For the lower stresses experienced in the sustained load tests, the anchor is not as dependent on friction along the sides of the hole. With an influence ratio of 0.63, this parameter is considered not adverse to sustained load performance. TS14—fly ash. It is believed that the addition of fly ash to the concrete mix does not adversely affect the sustained load performance for the same reasons discussed for the DOT mix (TS12). With an influence ratio of 0.69, this parameter is con- sidered not adverse to sustained load performance. TS15—blast furnace slag. It is believed that the addition of blast furnace slag to the concrete mix does not adversely affect the sustained load performance for the same reasons discussed for the DOT mix (TS12). With an influence ratio of 0.60, this parameter is considered not adverse to sustained load performance. TS16—unconfined setup. It was shown earlier that the alpha-setup factor of 0.75 is not appropriate for some adhe- sives. The three adhesives in this study had alpha-setup fac- tors in the range of 0.35 to 0.55. As all the points in the TS16 SvTTF lie above the aST-baseline, and due to the low influence ratio of 0.56, it appears that sustained load in unconfined setup is not an adverse condition as long as it is assumed that the correct alpha-setup factor for the product is used (i.e., 0.37 not 0.75). If a product passes this test then installation direction can be considered to not affect the short-term strength. Once the adhesive has cured, if the only difference between an anchor installed horizontally to one installed in the downward direc- tion is orientation (i.e., same concrete, moisture condition, temperature, etc.) then the application of sustained load should reasonably have the same effect for both conditions. Due to the discussion above, with an influence ratio of 0.93, this param- eter is considered not adverse to sustained load performance. TS06—installation direction (vertical). The sensitivity to installation direction test (ACI 355.4-11 §7.18) discussed above also tests for anchors installed vertically. It is believed that verti- cal installation does not adversely affect the sustained load per- formance for the same reasons discussed above for horizontal installation (TS05). Due to the discussion above, with an influ- ence ratio of 0.86, this parameter is considered not adverse to sustained load performance. TS07—moisture at installation. While moisture at instal- lation created a reduction in short-term bond strength (a = 0.82), the sustained load performance was no worse than the short-term reduction. This can be explained by the fact that the concrete began to dry after installation and eventu- ally dried out in the 110°F (43°C) chamber. The subsequent in-service conditions were the same as the baseline. With an influence ratio of 0.61, this parameter is considered not adverse to sustained load performance. TS08—moisture in-service. While the influence ratio is greater than 1 (1.02), the experimental line and the baseline appear to be the same line within scatter of that data (SvTTF curve in Appendix H, page H-13). It is therefore the research- ers’ opinion that this parameter is considered not adverse to sustained load performance. TS09—reduced hole cleaning. While reduced hole clean- ing created a reduction in short-term bond strength (a = 0.81), once the adhesive had cured, the reduction in adhesion due to the presence of dust on the sides of the borehole could be accounted for in the alpha-reduction factor. As time progressed, the amount of adhesion did not change and the reduction in strength over time was no worse than in the short term. Due to an influence ratio of 0.84, this parameter is considered not adverse to sustained load performance. TS10—installation temperature (mfr minimum/mfr min- imum). Currently data is only available for one sustained load stress level (70%MSL) for TS10 and it is not possible to develop an experimental trendline and subsequent influence ratio. However, based on the current results from the 70% stress level tests, it appears that this test series will not have an adverse effect on the sustained load performance. As discussed earlier and illustrated in Figure 14, adhesives respond to temperature slightly differently, but all show a decrease in bond strength as the temperature increases. As seen in Figure 14, it is possible for an anchor installed and tested at a very low temperature to have a higher bond strength than an anchor installed at room

103 it was still tacky with a dark gray glossy color indicating an improper ratio of the hardener and resin. Per NTSB (2007b), excess hardener is evidenced by a pliable consistency and a decrease in bond strength. The anchors of this adhesive that failed at higher bond stresses were also removed and exhib- ited hard fully cured adhesive with a flat whitish-gray color (Figure 111). All the holes were cleaned identically per the MPII at the same time. The same adhesive tube was used for all five rep- etitions for a given test day. The anchors for day 28 used a different tube than those for day 21. These three samples were considered anomalies and were not included in the determi- nation of the mean. These three test samples were not completely cured due to the previously discussed problem with setting down the cartridge gun during installation. Temperature and Humidity The four Sensiron temperature and humidity sensors that were cast in the control slab were destroyed in the casting pro- cess. Therefore, the two 9” long PVC pipes with PVDF filters on the embedded ends in each slab were used for temperature Early-Age Concrete Evaluation The short-term test load versus displacement and stress versus displacement results for the early-age investigation are presented in Appendix M. The results are summarized in Figure 109, which normalizes the results by the 28-day bond strength. It appears that on the basis of bond strength alone, adhe- sive A (vinyl ester) does not show any significant increase after 14 days, and adhesives B and C (epoxies) do not show any significant increase after 7 days. Discussion of Anomalies Several of the anchors for adhesive A failed not with a strength type failure but rather a stiffness type failure. Failure was defined as the point when the load–displacement curve dropped below a stiffness of 28.6 kip/in (5 kN/mm) as discussed earlier and illustrated in Figure 38. Test samples D21-C-ST-4 (Figure 110), D28-C-ST-3, and D28-C-ST-5 all pulled out at very low bond stresses. The anchors were removed from their holes for investigation. It was noticed that the adhesive had not completely cured as Figure 109. Normalized bond stress (by 28-day value) versus concrete age.

104 The Sensiron sensors in the control slabs reported a con- sistent 100% relative humidity (RH) reading for the entire month. For testing, separate slabs were used for a given day and then discarded. The RH readings from the sensors in the test slabs were all greater than 96%. It seems reasonable that since all the concrete slabs were cast at the same time and kept together prior to testing there would be a consistency in RH readings with each other. However, when the sensors were switched between slabs, the RH reading would be less than the previous slab and would show a sharp increase and it would take several days for the readings to stabilize. Initially, the slabs would be changed out on Monday morning, the anchors installed on Thursday, and tested on Friday. Except for day 14, this did not provide sufficient time for the read- ings to stabilize prior to testing. In response to this, the slabs for day 28 were changed out on the Friday before testing pro- viding a full week of readings and the RH readings began to stabilize (within the daily fluctuation of the ambient RH) on the testing day. The RH data is presented in Table 42. It does not seem reasonable that the RH at day 21 should be higher than at day 14 as the RH in concrete should decrease and humidity monitoring. During testing, two sensors were placed in the two pipes of the control slab and left for the dura- tion of the month-long testing period. The remaining two sen- sors were placed in the test slab of the anchors being tested. The temperature readings from the control slab and the individual testing slabs were within a 2°F (1°C) agreement with each during the testing program. The internal concrete temperature for both slabs followed the daily temperature fluctuation of the laboratory and were within 4°F (2°C) (and less than) the ambient temperature of the laboratory. The only exception was that on day 4, the concrete internal temperature was 7°F (4°C) below the ambient temperature of the labora- tory. The temperature readings are presented in Table 41. Figure 110. D21-C-ST-4 showing failure surface of incompletely cured specimen. Figure 111. D21-C-ST-5 showing failure surface of fully cured specimen. Test Day Control Slab Test Slab Ambient 4 78°F (26°C) 77°F (25°C) 84°F (29°C) 7 74°F (23°C) 72°F (22°C) 75°F (24°C) 14 72°F (22°C) 70°F (21°C) 74°F (23°C) 21 73°F (23°C) 73°F (23°C) 75°F (24°C) 28 77°F (25°C) 76°F (24°C) 79°F (26°C) Table 41. Temperature readings for the early-age concrete evaluation.

105 Initially, the surface absorption of the top formed surface and the sides of the hole showed similar rates. The top formed surface drastically increased in surface absorption over the first 2 weeks and then leveled off (within the scatter of the data). For this concrete specimen, as the concrete dehy- drated, the top surface increased in absorption but reached equilibrium with the environment after 2 weeks. The surface absorption of the sides of the hole remained fairly consistent over the first 2 weeks as the moisture several inches from the surface was not as easily lost to the environment. Eventually after 2 weeks, the surface absorption began to increase as the process of dehydration slowly dried out deeper and deeper portions of the concrete specimen. Initial surface absorption is indirectly a measure of inter- nal moisture. If the internal moisture is high, the surface absorption will be lower. Without accurate internal humid- ity data, the initial surface absorption data is the only indi- cation we have on the relative measure of internal moisture. Based on these tests it appears that there is a threshold of internal moisture above which the bond stress is not affected. Hardness The rebound and indention hammers used to determine hardness generated similar trends of increasing hardness over time. Both hammers had conversion charts to predict the 6” cube compressive strength for which the indention hammer had good agreement for the first 14 days and then under- estimated the strength. The rebound hammer consistently overestimated the concrete strength. The rebound hammer produced values that were 20% to 45% higher than the inden- tion hammer (Table 44). Figure 113 presents the data for the hardness tests as well as the compression and split tensile tests. All show similar trends of increasing value over time. Summary This chapter presented the findings from the experimental program. The major findings were: • The ratio of unconfined tests to confined tests (asetup) of 0.75 assumed in ACI 355.4 is not conservative for high- strength adhesives. Test results in this project obtained alpha factors for unconfined setup in the rage of 0.35 to 0.55. These results were confirmed in a series of verifica- tion tests. • Short-term test results should not be included in a stress versus time-to-failure relationship with results from sustained load tests. The short-term tests (which failed over time as hydration progresses and moisture is lost to evaporation at the surface. The datasheet for the sensors indicate that they are accurate to ± 4% RH in the range of 90 to 100% RH. Based on control slab RH readings of 100% and the limita- tions of the sensors (tolerance and time to stabilize) the only definitive conclusion that can be drawn from the RH data is that the RH was in the range of 96% to 100%. Initial Surface Absorption The initial surface absorption test (ISAT) samples were read at 10 minutes, 30 minutes, and 60 minutes after applying the water. For adhesive anchors, the 10-minute reading is the most relevant reading as the 30-minute and 60-minute read- ings measure the surface absorption of essentially saturated concrete, which is not a common condition for most adhesive anchor installations. Table 43 presents the 10-minute sample data from the ISAT program as well as the relative humidity recorded during testing. The data is based on three repeti- tions (one repetition for day three of the formed surface). In order to better evaluate trends, ISAT testing was conducted up to 35 days after casting. Figure 112 presents the ISAT data over the 35 day testing period for the top formed surface and the sides of the hole. Test Day Test Slab Relative Humidity (%) Ambient Relative Humidity (%) Comment 4 98.6 40 RH not stabilized 7 96.2 38 RH not stabilized 14 99.4 49 21 99.8 52 RH not stabilized 28 99.3 53 Table 42. Relative humidity readings for the early-age concrete evaluation. Age (days) Sides of Hole (ml/m2·s) Formed Surface (ml/m2·s) Ambient Relative Humidity (%) 3 0.036 0.030* -- 6 0.031 0.047 39 13 0.028 0.097 54 20 0.043 0.094 65 27 0.059 0.080 59 35 0.074 0.092 46 Note: *All ISAT data is based on an average of three repetitions except for the day-3 sample for the formed surface. Table 43. ISAT 10-minute sample data and relative humidity for sides of hole and formed surface.

106 nature of the load causes the polymers to migrate within the adhesive. These two actions occurring simultaneously reduce the capacity. • For the parameters tested in this project, only elevated ser- vice temperature [>120°F (>49°C)] and manufacturer’s cure time were shown to have an influence on the sus- tained load performance. • No consistent correlation between adhesive-alone (dog- bone or DSR creep) and anchor creep tests was discov- ered. Dogbone tensile specimens are poor predictors of long-term and short-term performance, and are not recommended for qualification testing for adhesives for anchors. • It was shown for the three adhesives tested that the bond strength did not increase significantly after 14 days (adhe- sive A) and 7 days (adhesives B and C). It is believed that the high level of internal moisture existent in early-age con- crete was the leading contributor to lower bond strengths in the earlier-age concrete tests. at 100% MSL and 2 minutes) plotted well above the SvTTF relationship generated from sustained load tests alone. The reduced expected failure stress level for short- duration loads appears to result from a dual requirement placed on the polymer. The magnitude of the load causes the polymer to undergo plastic deformation as it redis- tributes the load down the anchor, and the sustained Figure 112. ISAT 10-minute sample data and relative humidity for sides of hole and formed surface. Age (days) Rebound Hammer (psi) Indention Hammer (psi) Ratio Rebound/Indention 4 3,480 2,900 1.20 7 4,640 3,400 1.36 14 4,930 3,770 1.31 21 5,000 3,770 1.33 28 5,220 3,630 1.44 Table 44. Rebound and indention hammer results.

107 Figure 113. Hardness, concrete compression strength, and split tensile strength versus concrete age.