Field Performance of Corrugated Pipe Manufactured with Recycled Polyethylene Content (2018)

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.

9 2.1 Research Objectives The primary objective of this research project was to eval- uate the field performance of corrugated HDPE pipes man- ufactured with recycled materials. The project was based in part on the research conducted in NCHRP Project 4-32 and reported in NCHRP Report 696 (1). Supporting objec- tives of the research included the development of a service life prediction model for corrugated HDPE pipes manufac- tured with recycled materials and proposed manufacturing criteria to ensure these pipes meet the service life require- ments for culverts and storm drains in state and federal highway applications. A performance-based draft AASHTO specification for corrugated HDPE pipes manufactured with recycled materials was desired as an outcome of this research project. 2.2 Research Approach and Findings In NCHRP Report 696, it was reported that the primary failure modes that govern the service life of corrugated HDPE pipes manufactured with recycled materials are Stage II (slow crack growth) and Stage III (oxidation or chemical) failures (1), as shown in Figure 2-1. Stage III failures—typically related to prolonged UV exposure—can be prevented from occur- ring prior to the end of the intended service life of the pipe by compounding stabilizers such as antioxidants and carbon black with the base resin used to manufacture the pipes. These stabilizers may be pre-compounded with the base resin or added in the form of a masterbatch. In the case of manufac- turing pipes with recycled materials, the stabilizers will need to be added in the form of a masterbatch or compounded in the repelleting process since the recycled materials will have undergone a prior heat history that will have depleted a por- tion of the original antioxidants. Regardless, the prevention of Stage III failures is the same for pipes manufactured with or without recycled materials; therefore, guidelines previously established and proven for pipes manufactured with virgin materials are prudent for pipes manufactured with recycled materials as well. The level of antioxidants and carbon black in the compound can be measured via a thermal stability test [ASTM D3895 (25) or D3350 (26)] and the muffle furnace test [ASTM D4218 (27)], respectively. Because the prevention of Stage III failures is well understood, it was not a primary focus of this research project. Suggested material require- ments based on current practices to ensure the prevention of Stage III failures are included in Chapter 3. The primary focus of this research project was Stage II brittle failures from the SCG mechanism. While the preven- tion of Stage II failures is well understood for pipes manu- factured with virgin HDPE materials, the incorporation of recycled materials adds a new level of complexity because they increase the likelihood of contaminants that may serve as stress risers to initiate cracking (see Section 1.2). Because of this, a new test method, the BFF test (see Section 1.4), was proposed and reported in NCHRP Report 696 (1). The BFF test involved the testing of unnotched specimens under a constant load in an elevated temperature deionized water bath. Because the specimens did not contain an artificial notch, cracking was forced to initiate at the next highest stress riser, typically a contaminant, void or other imperfection in the test specimen. Based on the consistent and encouraging results of the BFF test presented in NCHRP Report 696 and its ability to gener- ate service life prediction models when conducted at multiple temperature/stress conditions, it became the basis for this research project. Given this background information, the research approach consisted of the following components: 1. Develop and formalize a new standardized test method based on the BFF test to evaluate the performance and predict the service life of corrugated HDPE pipe materials with recycled content relative to both the crack initiation and propagation phases of the Stage II SCG mechanism; C H A P T E R 2 Research Objectives, Approach, and Findings

10 2. Obtain and characterize several large-diameter corrugated HDPE test pipes manufactured with various blends of recycled and virgin materials typically used in the industry in accordance with this new test method; 3. Develop a model to predict the service life of these pipes in accordance with the new test method; 4. Validate this service life prediction model on full-scale pipe specimens both in the laboratory and in simulated field conditions; 5. Assess the performance of pipes manufactured with recycled materials relative to fatigue from cyclical live loads by install- ing pipes underneath an active commuter railroad; and 6. Based on these results, develop proposed guidelines for the AASHTO design methodology and material specifications for corrugated HDPE pipes manufactured with recycled materials. Each of these components are detailed in the following sections. 2.2.1 The UCLS Test Based on the encouraging and consistent results of the BFF test developed in NCHRP Project 4-32 and reported in NCHRP Report 696, it was decided to refine and formalize the test into a standard test method. Since the test method involves evaluating unnotched test specimens under a con- stant load (i.e., constant engineering stress), the new stan- dard test method was referred to as the unnotched constant ligament stress (UCLS) test. While there are some similarities to the NCLS test standardized in ASTM F2136 (4) and com- monly used to evaluate the SCR of corrugated HDPE pipe materials, there are some significant differences. First, the UCLS test is conducted on unnotched specimens rather than notched specimens, allowing evaluation of both the crack ini- tiation and crack propagation phases of the stress-crack mech- anism. Second, the UCLS test is conducted in deionized water at elevated temperatures, whereas the NCLS test is conducted in a solution of water and Igepal® or some other surfactant. The advantage of conducting the test in deionized water with no surfactant is that data from tests conducted at multiple tem- peratures and stresses can be bi-directionally shifted to predict the service life at various stresses and temperatures. This is not as straightforward for tests conducted in a solution containing surfactants, as such test media have been shown to increase the slope of the brittle failure curve, thereby making service life predictions at field conditions inaccurate (28). A draft standard test method for the UCLS test was bal- loted through ASTM International so that additional labora- tories could conduct the test and generate more data. The test method was published in 2016 as ASTM F3181 (5), and it is proposed that AASHTO references this ASTM standard rather than recreating a new standard. Due to copyright issues, the ASTM standard cannot be shown in this report. However, a draft of the balloted standard is shown in Appendix A. 2.2.2 Large-Diameter Test Pipes for research For the purposes of this research project, it was important to conduct tests on full-scale corrugated HDPE pipes manu- factured with blends of commonly available PCR materials. It was decided to evaluate 750 mm (30 in.) diameter pipes, as these are typical for culverts and storm drains in highway applications. Nine test pipes from three different manufac- turers were selected for the study, each pipe manufactured with a different blend of PCR and virgin materials typically used in the agricultural and land drainage industries. Pipes 1 through 5 were supplied by Manufacturer A. Of these, Pipes 1 and 5 were standard AASHTO M 294 pipes contain- ing 98% virgin materials with 2% carbon black added for UV protection (0% PCR materials); Pipes 2 and 4 were stan- dard ASTM F2648 (29) pipes typically used for land drainage applications and contained 49% PCR materials, 49% virgin materials, and 2% carbon black; and Pipe 3 was a custom- manufactured pipe containing 98% PCR materials and 2% carbon black. Pipes 1 and 2 were used for the commuter rail- road fatigue evaluation study, and Pipes 3, 4, and 5 were used for the other laboratory and field studies. Each pipe was manu- factured from a different material or blend of materials. Pipes 6 and 7 were supplied by Manufacturer B. Both of these pipes contained 98% PCR materials and 2% carbon black, but were manufactured on different dates from different lots of materi- als. Pipes 8 and 9 were supplied by Manufacturer C. Pipe 8 was a perforated pipe and contained 54% PCR materials, 44% virgin materials, and 2% carbon black. Pipe 9 contained 59% PCR materials, 39% virgin materials, and 2% carbon black. The nine test pipes were characterized in TRI/Environ- mental’s laboratory in Austin, Texas. A summary of the Figure 2-1. Illustration of failure modes for high-density polyethylene.

11 measured properties of all the test pipes is shown in Table 2-1 and sample test reports for the pipes are shown in Appen- dix B. Testing was done in accordance with the requirements of AASHTO M 294 (2), but note that M 294 does not cur- rently allow recycled materials. The pipes that contained recy- cled materials were not manufactured to or expected to meet the M 294 requirements. All materials testing was conducted on materials taken directly from the pipe wall and compres- sion molded into a plaque. Additionally, the NCLS test was also conducted on specimens taken directly from the pipe liner. The pipes passed the majority of the requirements of the current AASHTO M 294 standard (2). The main exceptions to this were the NCLS and density requirements. Seven of the nine pipes had NCLS values lower than required by AASHTO M 294. With regard to the density, the colorant present in mixed-color PCR materials slightly increased the density of the compound. AASHTO M 294 includes a correction to density for carbon black, but not for other colorants. Since SCG was identified as the primary performance- limiting property of the pipes with regard to service life, spe- cial attention was focused on the tests that assess stress-crack initiation and propagation, namely, the UCLS and NCLS tests. These are discussed in more detail in the following subsections. 2.2.2.1 NCLS Testing and Results NCLS testing was conducted in accordance with ASTM F2136 (4) on specimens prepared from compression-molded plaques of materials taken from the pipe wall as well as directly from the pipe liner. AASHTO M 294 requires the NCLS of compression-molded plaques taken from the pipe wall (plaque NCLS) to be a minimum of 24 hours, while the minimum requirement for specimens taken directly from the pipe liner (liner NCLS) is 18 hours (2). The reason the liner NCLS requirements are less than the plaque require- ments is to account for a reduction in performance due to the manufacturing process, as discussed by Hsuan in Property TestMethod M 294 Req’t. Pipe 1 Pipe 2 Pipe 3 Pipe 4 Pipe 5 Pipe 6 Pipe 7 Pipe 8 Pipe 9 PCR Content (%) Per Mfcr. 0 0 49 1 981 491 0 981 981 541 591 Density2 (g/cc) ASTM D1505 0.948– 0.955 0.963 1 0.9661 0.9401 0.9711 0.949 0.9581 0.9441 0.951 0.948 Melt Index (g/10 min) ASTM D1238 < 0.40 0.12 0.30 0.52 1 0.20 0.19 0.631 0.481 0.07 0.016 Flexural Mod. (ksi) ASTM D790 110 152.8 146.3 150.7 149.2 146.1 142.5 143.5 151.5 164.5 Tensile Strength (psi) ASTM D638 3000 4050 4062 3939 3743 3722 3832 4008 3982 3884 Carbon Black (%) ASTM D1603 2–4 2.15 2.57 1.90 1 2.90 2.29 2.05 2.63 2.70 2.73 Oxidation Induction Time (min) ASTM D3895 N/A N/A 3 N/A3 18.2 17.5 35.5 9.00 6.81 20.2 23.7 Pipe Stiffness (lb/in./in.) ASTM D2412 29 35.0 34.3 34.9 33.1 30.5 35.5 36.9 38.3 42.9 Pipe Flattening4 ASTMD2412 Varies Pass Pass Pass Pass Pass Pass Pass Pass Pass Brittleness ASTMD2444 No cracks Pass Pass Pass Pass Pass Pass Fail 1 Pass Pass Liner NCLS (h) ASTM F2136 18 87.9 18.4 3.9 1 16.91 32.5 8.31 5.81 18.1 14.11 Plaque NCLS (h) ASTM F2136 24 106.0 13.7 1 4.91 32.3 59.8 9.01 6.81 16.01 12.71 UCLS5 (h) ASTM F3181 N/A 1271 99.2 15.7 97.5 1432 19.4 19.3 141.0 70.2 1 Property does not meet current AASHTO M 294 requirements (cell classification 435400C based on ASTM D3350). Note that the pipes containing recycled materials were not manufactured to the requirements of M 294; these requirements are listed for reference. 2 Density is corrected for carbon black content using the equation in ASTM D3350. Note there is no correction for colorants other than carbon black. 3 Oxidation induction time testing was not requested or conducted on Pipes 1 and 2. 4 Pipe flattening requirements are a function of wall profile shape and pipe diameter. 5 UCLS results are the arithmetic average of five specimens at a test condition of 80°C and 650 psi stress. Table 2-1. Properties of test pipes.

12 NCHRP Report 631 (30). The results of the plaque and liner NCLS testing are tabulated in Tables 2-2 and 2-3, respectively, and shown graphically in Figure 2-2. The two virgin pipes (Pipe 1 and Pipe 5) met the AASHTO M 294 requirements, but all pipes manufactured with recycled materials failed either one or both liner and plaque NCLS requirements. 2.2.2.2 UCLS Testing and Results To evaluate the crack initiation phase of the SCG process, UCLS testing was conducted at three temperature/stress con- ditions on specimens prepared from the nine test pipes. The data are summarized in Table 2-4 (see Appendix C for a log- based table of the data). As expected, the pipes manufactured with only virgin materials (Pipes 1 and 5) had very long fail- ure times for each of the test conditions. This is because the UCLS test is evaluating both the crack initiation and propaga- tion phases of the SCG process, and the crack initiation phase is dependent on the presence of a contaminant or void in the polymer, both of which are minimized with virgin materials. 2.2.2.3 Relationship Between NCLS and UCLS Figure 2-3 shows the relationship between NCLS and UCLS for NCLS test data from pipe plaques and liners. There are three groups of data evident on this chart. The three sets of failures on the lower left portion of the chart correspond to Pipes 3, 6 and 7, all manufactured with 98% PCR materi- als; the four sets of failures in the middle of the chart cor- respond to Pipes 2, 4, 8 and 9, all manufactured with a blend of virgin and recycled materials; and the two sets of failures on the upper right of the chart correspond to Pipes 1 and 5, manufactured with only virgin materials. While the general trend of the data indicates a relationship between NCLS and UCLS, this is unlikely because the tests look at different aspects of the SCG mechanism. The NCLS test forces a notch to grow in a specific location, and only the material around the crack tip will influence the crack growth rate. On the other hand, the UCLS test allows cracks to grow wherever there is a critically sized defect present in the specimen. The two are related in the sense that materials that perform better in the NCLS test will perform better in the UCLS test, provided the number of defects is the same. Intuitively, it is feasible that a pipe could have a relatively low NCLS failure time but a high UCLS failure time if there were no contaminants present in the specimen. One way to test this would be to evaluate some pristine (clean) virgin materials with inherently low NCLS hours. Additional testing would be necessary to determine the relationship between NCLS and UCLS, if any. Pipe PCR % Specimen Failure Time (h) Avg. COV 1 2 3 4 5 1 0 102.8 104.6 100.76 117.72 104.65 106.1 0.063 2 49 13.71 13.71 13.7 13.7 13.7 13.7 0.000 3 98 5.3 4.5 5.3 5.3 3.9 4.9 0.131 4 49 34.8 26.7 33.4 34.8 31.8 32.3 0.104 5 0 57.5 63.1 58.2 57.9 62.5 59.8 0.046 6 98 5.8 7.2 7.4 17.4 7.2 9.0 0.053 7 98 6.4 7.4 7.4 6.7 6.0 6.8 0.090 8 54 16.0 15.8 16.2 16.2 16.0 16.0 0.012 9 59 13.4 12.8 13.4 12.1 11.8 12.7 0.057 Table 2-2. NCLS test results on specimens taken from compression-molded plaques of materials from the pipe wall. Pipe PCR % Specimen Failure Time (h) Avg. COV 1 2 3 4 5 1 0 87.0 74.98 83.71 104.1 89.8 87.9 0.121 2 49 18.3 20.4 18.5 18.4 16.3 18.4 0.079 3 98 4.0 4.0 3.6 3.9 3.8 3.9 0.043 4 49 17.5 16.9 16.9 17.1 16.3 16.9 0.026 5 0 32.7 32.1 31.1 33.9 32.9 32.5 0.032 6 98 9.2 7.2 8.0 8.7 8.6 8.3 0.090 7 98 5.8 6.7 5.4 5.9 5.3 5.8 0.096 8 54 18.2 17.8 19.1 18.6 16.8 18.1 0.048 9 59 14.4 14.4 14.4 13.7 13.8 14.1 0.025 Table 2-3. NCLS test results on specimens taken directly from the pipe liner.

13 106.1 13.7 4.9 32.3 59.8 9.0 6.8 16.0 12.7 87.9 18.4 3.9 16.9 32.5 8.3 5.8 18.1 14.1 0.0 20.0 40.0 60.0 80.0 100.0 120.0 1 2 3 4 5 6 7 8 9 N CL S (h ) Test Pipe Pipe Plaque Pipe Liner AASHTO M 294 Pipe Plaque NCLS Requirement = 24 Hours AASHTO M 294 Pipe Liner NCLS Requirement = 18 Hours Figure 2-2. Comparison of NCLS liner and plaque data for test pipes. Pipe Condition Specimen Failure Time (h) Avg. COV 1 2 3 4 5 1 80°C / 650 psi 1414.3 1523.1 1596.9 733.1 1088.8 1271.2 0.282 80°C / 450 psi N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 70°C / 650 psi N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 2 80°C / 650 psi 107.2 41.5 128.3 125.8 93.3 99.2 0.356 80°C / 450 psi N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 70°C / 650 psi N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 N/A1 3 80°C / 650 psi 18.2 15.7 11.6 20.7 12.4 15.7 0.244 80°C / 450 psi 143.6 171.8 79.0 129.0 160.1 136.7 0.264 70°C / 650 psi 68.7 41.2 113.6 157.9 119.0 100.1 0.456 4 80°C / 650 psi 130.0 75.7 125.9 98.6 57.1 97.5 0.323 80°C / 450 psi 773.9 250.6 372.4 702.2 705.7 561.0 0.416 70°C / 650 psi 932.9 661.2 539.1 378.4 351.0 572.5 0.415 5 80°C / 650 psi 1688.5 1972.4 1270.4 >2130 795.6 1431.7 0.358 80°C / 450 psi >2130 >2130 >2130 >2130 >2130 >2130 N/A 70°C / 650 psi >2130 >2130 >2130 >2130 >2130 >2130 N/A 6 80°C / 650 psi 14.0 22.0 15.0 21.5 24.5 19.4 0.239 80°C / 450 psi 132.8 108.6 54.3 87.8 228.4 122.4 0.538 70°C / 650 psi 113.5 125.8 135.0 28.1 92.2 98.9 0.432 7 80°C / 650 psi 22.2 16.4 21.5 18.6 17.6 19.3 0.130 80°C / 450 psi 86.3 105.2 110.6 34.8 63.0 80.0 0.393 70°C / 650 psi 65.2 31.0 44.3 94.7 51.9 57.4 0.422 8 80°C / 650 psi 134.9 91.0 67.5 174.6 237.0 141.0 0.480 80°C / 450 psi 1169.6 691.0 876.5 378.2 340.9 691.2 0.503 70°C / 650 psi 441.4 299.7 511.6 440.1 197.7 378.1 0.335 9 80°C / 650 psi 65.6 90.5 93.5 53.2 48.1 70.2 0.298 80°C / 450 psi 682.0 449.5 302.4 740.3 1081.0 651.0 0.458 70°C / 650 psi 648.5 285.5 384.3 358.1 560.3 447.3 0.338 AVERAGE 0.366 1 Testing was not conducted at the 80°C / 450 psi and 70°C / 650 psi conditions for Pipes 1 and 2, as these pipes were used for the live load fatigue study underneath railroads and were not intended to be utilized for the service life prediction model developed in the research project. However, these pipes were from the same manufacturer as Pipes 3, 4 and 5; Pipe 1 was similar in composition to Pipe 5 (both M 294 pipes) and Pipe 2 was similar in composition to Pipe 4 (both ASTM F2648 pipes). Table 2-4. UCLS failure data for test pipes.

14 2.2.3 Development of a Service Life Prediction Model Based on the UCLS Test As discussed earlier, one of the advantages of the UCLS test is that the elevated temperature test data can be bi-directionally shifted to predict the service life relative to Stage II brittle fail- ures at other temperatures. Two common ways of doing this for polyethylene materials are the Popelar shift method, or PSM (31), and the rate process method, or RPM (32). These two methods are detailed in the following subsections. 2.2.3.1 Popelar Shift Method for Shifting UCLS Data The PSM involves both vertical (stress) and horizontal (time) shift factors based on the difference between the test temperature and the desired service temperature. To shift stress from a higher test temperature (T2) to a lower service temperature (T1), the shift factor in Equation 2.1 is used, while to shift failure time from a higher temperature to a lower temperature, the shift factor shown in Equation 2.2 is used. = ( )−Stress Shift Factor (2.1)0.0116 T2 T1e = ( )−Time Shift Factor (2.2)0.109 T2 T1e Testing materials at three different elevated stress/ temperature conditions allows the shifting of three data points to develop a master curve at desired service temperature conditions (e.g., 23°C). When plotted on a log-log scale, the shifted data fall on a line, which allows the projected failure time at any applied stress to be determined. The Popelar stress shift factor for shifting data from 80°C to 23°C is shown in Equation 2.3 and the time shift factor is shown in Equation 2.4. The stress and time shift factors for shifting data from 70°C to 23°C are shown in Equations 2.5 and 2.6, respectively. Shift Factors for 80°C to 23°C: = = ( )−Stress Shift Factor 1.937 (2.3)0.0116 80 23e = = ( )−Time Shift Factor 499.2 (2.4)0.109 80 23e Shift Factors for 70°C to 23°C: = = ( )−Stress Shift Factor 1.725 (2.5)0.0116 70 23e = = ( )−Time Shift Factor 167.8 (2.6)0.109 70 23e Applying these shift factors to the average failure time for Pipe 3 at the 80°C and 4.48 MPa (650 psi) test condition, for example, the shifted failure time is 499.2  15.3 = 7637 hours at a stress of 1.937  4.48 = 8.68 MPa (1259 psi). In other words, a failure time of 15.3 hours at 80°C and 4.48 MPa (650 psi) stress is equivalent to a failure time of 7637 hours at 23°C and 8.68 MPa (1259 psi) stress. Similar calculations were done for the average failure times for each pipe at each condition, as shown in Table 2-5. Mastercurves based on the PSM for each pipe at a service temperature of 23°C are shown in Figures 2-4 through 2-9. The upper and lower confidence limits for each curve were calculated by applying a statistical adjustment to each shifted 10.0 100.0 1000.0 10000.0 1.0 10.0 100.0 1000.0 U CL S (h ) NCLS (h) Plaque NCLS Liner NCLS Figure 2-3. Relationship between NCLS and UCLS for NCLS test specimens taken from both compression-molded plaques and directly from the pipe liner. UCLS data at 80çC, 4.48 MPa (650 psi) stress condition.

15 1See Appendix C for log-based average failure times. Pipe Condition Log Avg. Failure Time1 (h) Avg. Failure Time (h) 23°C Shifted Stress (psi) 23°C Shifted Time (h) 23°C Shifted Stress - Log (psi) 23°C Shifted Time - Log (h) 1 80°C / 650 psi 3.104 1271.2 1259.0 634,583 3.100 5.802 80°C / 450 psi N/A N/A 871.7 N/A 2.940 N/A 70°C / 650 psi N/A N/A 1121.0 N/A 3.049 N/A 2 80°C / 650 psi 1.965 92.3 1259.0 46,077 3.100 4.663 80°C / 450 psi N/A N/A 871.7 N/A 2.940 N/A 70°C / 650 psi N/A N/A 1121.0 N/A 3.049 N/A 3 80°C / 650 psi 1.186 15.3 1259.0 7,637 3.100 3.884 80°C / 450 psi 2.121 132.1 871.7 65,952 2.940 4.819 70°C / 650 psi 1.956 90.4 1121.0 15,175 3.049 4.181 4 80°C / 650 psi 1.969 93.0 1259.0 46,447 3.100 4.667 80°C / 450 psi 2.711 513.7 871.7 256,462 2.940 5.409 70°C / 650 psi 2.729 535.8 1121.0 89,931 3.049 4.954 5 80°C / 650 psi 3.132 1354.5 1259.0 676,172 3.100 5.830 80°C / 450 psi > 3.220 N/A 871.7 N/A 2.940 N/A 70°C / 650 psi > 3.220 N/A 1121.0 N/A 3.049 N/A 6 80°C / 650 psi 1.277 18.9 1259.0 9,452 3.100 3.976 80°C / 450 psi 2.039 109.4 871.7 54,636 2.940 4.737 70°C / 650 psi 1.940 87.0 1121.0 14,608 3.049 4.165 7 80°C / 650 psi 1.282 19.1 1259.0 9,550 3.100 3.980 80°C / 450 psi 1.869 73.9 871.7 36,882 2.940 4.567 70°C / 650 psi 1.729 53.5 1121.0 8,987 3.049 3.954 8 80°C / 650 psi 2.107 127.9 1259.0 63,871 3.100 4.805 80°C / 450 psi 2.792 619.6 871.7 309,311 2.940 5.490 70°C / 650 psi 2.554 358.1 1121.0 60,101 3.049 4.779 9 80°C / 650 psi 1.830 67.7 1259.0 33,788 3.100 4.529 80°C / 450 psi 2.774 594.4 871.7 296,714 2.940 5.472 70°C / 650 psi 2.631 427.5 1121.0 71,749 3.049 4.856 Table 2-5. Average UCLS failure data shifted to 23çC. 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 Lo g St re ss (p si ) Log Time (h) UCL LCL Mastercurve at 23 deg. C y = -0.1709x + 3.7638 R2 = 1.0000 Figure 2-4. UCLS failure data for Pipe 3 shifted to 23çC using Pope- lar shifting; LCL and UCL curves based on a COV of 0.456, the maxi- mum determined from the three test conditions for Pipe 3.

16 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 Lo g St re ss (p si) Log Time (h) y = -0.2175x + 4.1196 R2 = 0.9936 UCL LCL Mastercurve at 23 deg. C Figure 2-5. UCLS failure data for Pipe 4 shifted to 23çC using Popelar shifting; LCL and UCL curves based on a COV of 0.416, the maximum determined from the three test conditions for Pipe 4. 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 Lo g St re ss (p si ) Log Time (h) y = -0.2052x + 3.9110 R2 = 0.9947 UCL LCL Mastercurve at 23 deg. C Figure 2-6. UCLS failure data for Pipe 6 shifted to 23çC using Popelar shifting; LCL and UCL curves based on a COV of 0.538, the maximum determined from the three test conditions for Pipe 6.

17 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 Lo g St re ss (p si ) Log Time (h) y = -0.2211x + 3.9513 R2 = 0.8813 UCL LCL Mastercurve at 23 deg. C Figure 2-7. UCLS failure data for Pipe 7 shifted to 23çC using Popelar shifting; LCL and UCL curves based on a COV of 0.422, the maximum determined from the three test conditions for Pipe 7. 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 Lo g St re ss (p si ) Log Time (h) y = -0.1904x + 3.9867 R2 = 0.8847 UCL LCL Mastercurve at 23 deg. C Figure 2-8. UCLS failure data for Pipe 8 shifted to 23çC using Popelar shifting; LCL and UCL curves based on a COV of 0.503, the maximum determined from the three test conditions for Pipe 8.

18 data point using the Student’s t-distribution. This is illus- trated in Equations 2.7 and 2.8. Since five test specimens were evaluated at each of three test conditions, each condition has an associated coefficient of variation (COV) of the data. The maximum COV of the three conditions was used for the anal- ysis. A Student’s t table is shown in Appendix D. = +  ( )−UCL COV (2.7)95% 1X t X nn  = −  ( )−LCL COV (2.8)95% 1X t X nn  where UCL95% = Upper 95% confidence limit LCL95% = Lower 95% confidence limit – X = Log-based average of five test specimens t(n–1) = Student’s t value at (n − 1) degrees of freedom = 2.132 COV = Coefficient of variation of five test specimens = (St. Dev.)/–X n = Number of test specimens at each condition = 5 Images of failures of specimens from Pipes 3 and 4 are shown in Figures 2-10 and 2-11, respectively. These are 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 Lo g St re ss (p si ) Log Time (h) y = -0.1703x + 3.8734 R2 = 0.9988 UCL LCL Mastercurve at 23 deg. C Figure 2-9. UCLS failure data for Pipe 9 shifted to 23çC using Popelar shifting; LCL and UCL curves based on a COV of 0.458, the maximum determined from the three test conditions for Pipe 9. Figure 2-10. UCLS failures for Pipe 3 (manufactured with 98% PCR materials), at 20ë magnification.

19 Figure 2-11. UCLS failures for Pipe 4 (manufactured with 49% PCR materials), at 20ë magnification. typical of all the observed UCLS failures. Note the presence of a void or contaminant as the origination point of the crack in each of the images, as indicated in the circled areas. Beach marks indicating the direction of crack growth are evident around the initiation site. The brittle crack prop- agates through the pipe wall until the specimen exhibits ductile yielding on the remaining ligament area. Materials with more or larger contaminants fail quicker than those with fewer or smaller contaminants. As shown in Table 2-4, the average UCLS failure time for Pipe 3 at 80°C and 4.48 MPa (650 psi) stress was 15.7 hours, while the average failure time for Pipe 4 was 97.5 hours. Note that the size of the contaminants in Pipe 3 are larger than those in Pipe 4. It also appears as if more crack initiation sites were pres- ent in Pipe 3, particularly in the specimen on the right (Figure 2-10). 2.2.3.2 Rate Process Method for Shifting UCLS Test Data Another method for shifting failure times at elevated tem- peratures to lower temperatures is to apply the RPM, detailed in ASTM D2837 (32). Section 5.2 of ASTM D2837 outlines a method used for validation testing for the hydrostatic design basis of pressure pipe materials. For determination of the brit- tle failure performance, which is of concern in the case of cor- rugated HDPE drainage pipes, Procedure I in ASTM D2837 requires testing at three temperature/stress conditions and using the data to solve the three-coefficient equation shown in Equation 2.9. = + +log (2.9)t A B T CLogS T where t = Failure time, h T = Absolute temperature, K S = Stress, psi A, B, C = Constants Once the three coefficients are solved, the failure time at any stress and temperature condition can be predicted. To illustrate the determination of the three coefficients in Equation 2.9, con- sider the failure data for Pipe 3. Equations 2.10 through 2.12 show the equations for the average failure times from Table 2-5 at each of the three conditions of testing, using imperial units. Condition I (80°C, 650 psi stress): ( ) = = + +log 1.186 353 650 353 (2.10)t A B CLog Condition II (80°C, 450 psi stress): ( ) = = + +log 2.121 353 450 353 (2.11)t A B CLog Condition III (70°C, 650 psi stress): ( ) = = + +log 1.956 343 650 343 (2.12)t A B CLog Subtracting Equation 2.11 from 2.10 results in the deter- mination of C = -2067. Equation 2.11 can be solved for B as follows:   353 2.121 2067 650 343 6233 353A (2.13) B A Log( ) ( ) = − − −    = −

20 Substituting into Equation 2.12 allows the calculation of A as follows: -  A 1.956 6233 343 353 343 2067 650 343 0.02915 0.73499 A A Log A 25.2 ( ) ( ) = − + − − => − = => = Equation 2.13 can be used to determine B = 6233 – 353 (−25.2) = 15,143. So, the RPM equation for Pipe 3 is shown in Equation 2.14: ( ) = − + + − log 25.2 15,143 2067 Log (2.14)t T S T Based on Equation 2.14, the time to brittle failure for Pipe 3 at any given stress and temperature can be predicted. An alternative method for the coefficients in the three- series RPM equations is to use matrices. Using matrix alge- bra, Equations 2.10 through 2.12 for Pipe 3 can be written as shown in Equation 2.15: ( ) ( ) ( )       ×       =       1 1353 650 353 1 1353 450 353 1 1343 650 343 1.186 2.121 1.956 (2.15) Log Log Log A B C Performing the math, this simplifies to Equation 2.16:       ×       =       1 0.00283 0.00797 1 0.00283 0.00752 1 0.00292 0.00820 1.186 2.121 1.956 (2.16) A B C The coefficients A, B, and C can then be solved by multi- plying both sides of Equation 2.16 by the inverse of the first matrix, as shown in Equation 2.17.       =       ×       −1.186 2.121 1.956 1 0.00283 0.00797 1 0.00283 0.00752 1 0.00292 0.00820 (2.17) 1A B C Using the MMULT function in Microsoft Excel, the RPM coefficients for each of the test pipes were calculated as shown in Table 2-6. Pipes 1, 2, and 5 were not analyzed since there were insufficient UCLS failure data for these pipes at the three test conditions. 2.2.3.3 Comparison of RPM and PSM Figures 2-12 through 2-17 show RPM curves for the UCLS failure data for Pipes 3, 4, 6, 7, 8 and 9 shifted to 23°C. Also shown for comparison purposes are the curves generated by the PSM previously shown in Figures 2-4 through 2-9. As evi- dent from the charts, the PSM is generally more conservative than the RPM. The exceptions to this are Pipes 7 and 8. Both of these pipes had more scatter in the UCLS failure data than the other pipes, and the shifted linear regression curves also had a significantly poorer fit than the other pipes (note the coeffi- cients of determination for the linear regression analysis of the PSM-shifted curves for Pipes 7 and 8 are around 0.88, while the other curves are all 1.00). The reason for this is not entirely known, but it appears that the average failure times from the lower temperature test condition (70°C / 650 psi stress) were lower than anticipated for these two pipes due to one outlying data point in each five-specimen data set, which resulted in a poorer fit of the data when performing the regression analysis. Figures 2-18 and 2-19 show the shifted curves for all six pipes on the same graph, using both the PSM (Figure 2-18) and the RPM (Figure 2-19). The slopes of the curves generated by the PSM are much more consistent than the slopes of the RPM curves. A summary of the slopes, y-intercepts and projected mean failure times at 3.4 MPa (500 psi) stress for each of the two methods is shown in Table 2-7. Summary charts comparing the PSM and RPM predicted failure times and brittle curve slopes for the data presented in Table 2-7 are shown in Figures 2-20 and 2-21, respectively. The pipes in the charts are grouped by PCR content. The predicted time to cracking via the RPM is considerably longer than via the PSM for all the pipes except Pipes 7 and 8, both of which exhibited some non-linearity in the PSM plots shown in Figures 2-7 and 2-8. There is also a clear difference in projected failure times for the pipes manufactured with 98% PCR materials as opposed to those manufactured with lower PCR content. The difference is more clearly defined when analyzing the data via the PSM vs. the RPM (See Figures 2-18 through 2-20). Fig- ure 2-18 shows two distinct groups of curves, the lower set of curves corresponding to the three pipes manufactured with 98% PCR content and the upper set corresponding to those Pipe RPM Equation Coefficient Slope of RPM Curve y-Intercept of RPM Curve A B C 3 −25.2 15,143 −2067.1 −0.1432 3.711 4 −24.1 13,820 −1640.3 −0.1805 4.074 6 −21.4 12,758 −1684.2 −0.1757 3.806 7 −14.0 9,061 −1297.1 −0.2282 3.779 8 −13.2 9,671 −1514.3 −0.1955 3.802 9 −25.6 15,559 −2085.7 −0.1419 3.823 Table 2-6. RPM coefficients for UCLS failure data shifted to 23çC.

21 y = -0.1432x + 3.7114 y = -0.1709x + 3.7638 2.6 2.7 2.8 2.9 3.0 3.1 3.2 Lo g St re ss (p si ) Log Time (h) RPM PSM 3.5 4.0 5.04.5 5.5 6.0 7.06.5 Figure 2-12. UCLS data for Pipe 3 shifted to 23çC using the rate process method and the Popelar shift method. y = -0.1805x + 4.0745 y = -0.2175x + 4.1196 2.6 2.7 2.8 2.9 3.0 3.1 3.2 Lo g St re ss (p si ) Log Time (h) RPM PSM 3.5 4.0 5.04.5 5.5 6.0 7.06.5 Figure 2-13. UCLS data for Pipe 4 shifted to 23çC using the rate process method and the Popelar shift method.

22 y = -0.1757x + 3.8064 y = -0.2052x + 3.911 2.6 2.7 2.8 2.9 3.0 3.1 3.2 Lo g St re ss (p si ) Log Time (h) RPM PSM 3.5 4.0 5.04.5 5.5 6.0 7.06.5 Figure 2-14. UCLS data for Pipe 6 shifted to 23çC using the rate process method and the Popelar shift method. y = -0.2282x + 3.7791 y = -0.2211x + 3.9513 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 Lo g St re ss (p si ) Log Time (h) RPM PSM 3.5 4.0 5.04.5 5.5 6.0 6.5 Figure 2-15. UCLS data for Pipe 7 shifted to 23çC using the rate process method and the Popelar shift method.

23 y = -0.1955x + 3.8018 y = -0.1904x + 3.9867 2.6 2.7 2.8 2.9 3.0 3.1 3.2 Lo g St re ss (p si ) Log Time (h) RPM PSM 3.5 4.0 5.04.5 5.5 6.0 7.06.5 Figure 2-16. UCLS data for Pipe 8 shifted to 23çC using the rate process method and the Popelar shift method. 2.6 2.7 2.8 2.9 3.0 3.1 3.2 Lo g St re ss (p si ) Log Time (h) RPM PSM 3.5 4.0 5.04.5 5.5 6.0 7.06.5 y = -0.1419x + 3.823 y = -0.1703x + 3.8734 Figure 2-17. UCLS data for Pipe 9 shifted to 23çC using the rate process method and the Popelar shift method.

24 Shallowest slope y = -0.1709x + 3.7638 Steepest slope y = -0.2175x + 4.1196 2.4 2.6 2.8 3.0 3.2 3.4 3.6 Lo g St re ss (p si) Log Time (h) Pipe 3 - 98% PCR Pipe 4 - 49% PCR Pipe 6 - 98% PCR Pipe 7 - 98% PCR Pipe 8 - 54% PCR Pipe 9 - 59% PCR 100 years 500 psi 2.0 3.0 4.0 5.0 6.0 7.0 Figure 2-18. UCLS failure data for all pipes shifted to 23çC using the Popelar shift method. 2.4 2.6 2.8 3.0 3.2 3.4 3.6 Lo g St re ss (p si ) Log Time (h) Pipe 3 - 98% PCR Pipe 4 - 49% PCR Pipe 6 - 98% PCR Pipe 7 - 98% PCR Pipe 8 - 54% PCR Pipe 9 - 59% PCR 2.0 3.0 4.0 5.0 6.0 7.0 Shallowest slope y = -0.1432x + 3.7114 Steepest slope y = -0.2282x + 3.7791 100 years 500 psi Figure 2-19. UCLS failure data for all pipes shifted to 23çC using the rate process method.

25 manufactured with around 50% PCR content. The difference is not as discernible when analyzing the data with the RPM (Figure 2-19). Figure 2-21 shows that the brittle slopes of all the pipe curves are steeper with the PSM than the RPM, again with the exceptions of Pipes 7 and 8. 2.2.4 Validation of Service Life Prediction Models 2.2.4.1 Laboratory Validation on Full-Scale Pipes Installed thermoplastic pipes typically experience a con- stant deflection (i.e., strain) rather than a constant load (i.e., stress). This is because the pipe is less stiff than the surround- ing soils and, as the pipe deflects, the load is shed to the sup- porting soils around the pipe (33). As the soils consolidate, the pipe deflection stabilizes. The result is a condition of rela- tively constant deflection. It is well known that viscoelastic materials such as HDPE will creep when held at a constant load or stress. This is the condition tested in the NCLS and UCLS tests—the specimens creep until rupture (either duc- tile or brittle, depending on the stress level). When held in a condition of constant strain, the stresses in the material relax over time due to rearrangements of adjacent segments in the molecular chains (18). Since installed pipes are typically in a condition of con- stant strain rather than constant stress, it was important to validate the service life prediction model in this condition and to correlate the results from the NCLS and UCLS testing to a condition of constant strain. To do this, full-scale pipe specimens were evaluated in fixtures that held the pipes at a constant deflection. The tests were conducted in collabora- tion with Ohio University. The stress-relaxation properties of the pipes were monitored over time using strain-gage- based load cells built into the loading rods on the test fix- tures. The bending strains in the pipe wall were monitored throughout the test, and an equivalent average stress was calculated to correlate the results to the UCLS data and to validate the service life prediction model. To accelerate failure times, the bending strains (and therefore stresses) were elevated by deflecting the pipes until the vertical inside diameter was reduced by 20%, four times the maxi- mum allowed in-field deflections for commuter railroad applications. As illustrated in Figures 2-4 through 2-9 and 2-12 through 2-17, increased stress in the pipe wall results Pipe Slope y-Intercept Avg. Predicted Time to Cracking @ 500 psi stress, 23°C (years) PSM RPM PSM RPM PSM RPM 3 −0.1709 −0.1432 3.764 3.711 195 1343 4 −0.2175 −0.1805 4.120 4.074 389 4783 6 −0.2052 −0.1757 3.911 3.806 92 228 7 −0.2211 −0.2282 3.951 3.779 53 6 8 −0.1904 −0.1955 3.987 3.802 663 50 9 −0.1703 −0.1419 3.873 3.823 899 9502 AVG −0.1959 −0.1775 3.934 3.833 382 2652 COV 0.11 0.18 0.030 0.033 0.89 1.44 Table 2-7. Summary of slopes, y-intercepts, and mean predicted failure times at a field condition of 23çC and 3.4 MPa (500 psi) stress for both RPM and PSM shifting. 1 10 100 1000 10000 Fa ilu re T im e (y rs .) Pipe PSM RPM ~50% PCR Pipes 98% PCR Pipes 3 6 7 4 8 9 Figure 2-20. Comparison of projected brittle failure times for various pipes at service conditions of 23çC and 500 psi stress for the Popelar shift method and rate process method.

26 ~50% PCR Pipes 98% PCR Pipes Pipe PSM RPM 3 6 7 4 8 9 -0.24 -0.22 -0.2 -0.18 -0.16 -0.14 -0.12 -0.1 Sl op e of B ri tt le C ur ve Figure 2-21. Comparison of projected brittle failure slopes for vari- ous pipes at service conditions of 23çC and 500 psi stress for the Popelar shift method and rate process method. Figure 2-22. Pipes [750 mm (30 in.) diameter] in constant deflection test frames at Ohio University. in decreased failure times relative to the SCG mechanism. Also, it should be noted that since the pipes were not buried in soil, there were no hoop compressive loads to offset the bending strains induced in the pipe wall. The temperature in the laboratory was 23°C (72°F). Figure 2-22 shows some images of the pipes in the load- ing fixtures; Figures 2-23 and 2-24 show some images of the strain gages used on the pipes and loading rods; and Fig- ure 2-25 shows the location of the strain gages around the pipe wall (odd-numbered gages were placed parallel to the corrugations, or circumferentially; even-numbered gages were placed perpendicular to the corrugations, or longitu- dinally). Figure 2-26 illustrates the before and after deflec- tion measurements. Pipes 3 (98% PCR), 4 (49% PCR),

27 Figure 2-23. Strain gages placed on loading rods to measure load reduction due to stress relaxation. Figure 2-24. Strain gages placed on the outer and inner walls of the test pipes.

28 Figure 2-25. Numerical designation for 12 gages placed around pipes. Odd- numbered gages were placed circumferentially; even-numbered gages were placed longitudinally. Figure 2-26. Approximate dimensions of pipe during constant deflection test: (A) initial outside diameter = 900 mm (35 in.); (B) initial inside diameter = 750 mm (30 in.); (C) deflected inside diameter = 600 mm (24 in.). and 5 (0% PCR) were tested to validate the service life pre- diction model. Figures 2-27 through 2-29 show the measured wall strains from the undeflected position to 20% vertical deflection for Pipes 3, 4, and 5. Positive strain indicates tension, and nega- tive indicates compression. As discussed above, the service life prediction models were based on test specimens held at a constant stress (i.e., the UCLS test), while the full-scale parallel plate test specimens were held at a constant strain. To use the parallel plate test as a validation test for the UCLS service life prediction model, it was important to determine how the constant strain in the pipe wall relates to stress. For linear elastic materials, the stress and strain can be related simply by the elastic modulus (i.e., E = δ/ε). Since HDPE behaves as a viscoelastic material at small strains and a nonlinear viscoelastic or viscoplastic material at large strains (34), the relationship between strain and stress is time dependent and not as straightforward. Chua (35) developed a power law model that determined the time-dependent uniaxial secant relaxation modulus valid for HDPE materials at 21°C (70°F) according to Equation 2.18, and Masada (36) determined a similar relationship for HDPE materials at 23°C (73°F), as shown in Equation 2.19. ( ) = + −7630 99,507 (2.18)Chua 0.097786E t t ( ) = −106,753.8 (2.19)Masada 0.08257E t t where E = Relaxation modulus, psi t = Time, min Graphs for both the Chua relaxation modulus (Equa- tion 2.18) and the Masada modulus (Equation 2.19) are shown in Figure 2-30. The two independently developed equations are nearly identical.

29 -4.00E+04 -3.00E+04 -2.00E+04 -1.00E+04 0.00E+00 1.00E+04 2.00E+04 3.00E+04 4.00E+04 St ra in (m ic ro st ra in ) Deflection (in.) SG1 (C) SG2 (L) SG3 (C) SG4 (L) SG5 (C) SG6 (L) SG7 (C) SG8 (L) SG9 (C) SG10 (L) SG11 (C) SG12 (L) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Figure 2-27. Pipe 3 (98% PCR) strain measurements for deflections from 0% (no deflection) to 20% (6 in. vertical deflection) in parallel plate test fixture. -4.00E+04 -3.00E+04 -2.00E+04 -1.00E+04 0.00E+00 1.00E+04 2.00E+04 3.00E+04 4.00E+04 St ra in (m ic ro st ra in ) Deflection (in.) SG1 (C) SG2 (L) SG3 (C) SG4 (L) SG5 (C) SG6 (L) SG7 (C) SG8 (L) SG9 (C) SG10 (L) SG11 (C) SG12 (L) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Figure 2-28. Pipe 4 (49% PCR) strain measurements for deflections from 0% (no deflection) to 20% (6 in. vertical deflection) in parallel plate test fixture.

30 -4.00E+04 -3.00E+04 -2.00E+04 -1.00E+04 0.00E+00 1.00E+04 2.00E+04 3.00E+04 4.00E+04 St ra in (m ic ro st ra in ) Deflection (in.) SG1 (C) SG2 (L) SG3 (C) SG4 (L) SG5 (C) SG6 (L) SG7 (C) SG8 (L) SG9 (C) SG10 (L) SG11 (C) SG12 (L) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Figure 2-29. Pipe 5 (0% PCR) strain measurements for deflections from 0% (no deflection) to 20% (6 in. vertical deflection) in parallel plate test fixture. Figure 2-30. Comparison of time-dependent modulus of elasticity for HDPE materials according to models developed by Chua (35) and Masada (36).

31 To see if either Equation 2.18 or 2.19 is valid for assess- ing the time-dependent modulus of elasticity of the blends of PCR and virgin HDPE materials used in manufacturing the pipes in this research project, the loads to hold the pipes at a constant deflection were monitored over time via strain- gage-based sensors in the loading rods. Figure 2-31 shows the responses of loading Pipes 3, 4, and 5 until the inside diam- eter was reduced from 0% to 20%, and Figures 2-32 and 2-33 show the decay in the measured load over the first 100 days and the first 250 days of loading, respectively. The deflections in the pipes were induced by reducing the distance between the parallel loading plates by mechanically turning down the fasteners on the threaded rods that held the plates together (see Figure 2-22). Since this process was done manually, the time to load the three pipes to 20% deflec- tion was not exactly the same and varied from around 8 to 11 minutes. Because of this, Pipe 4 reached a slightly higher load at 20% deflection than Pipes 3 or 5, as it was loaded at a faster rate than the other pipes (see Figure 2-31). All three pipes showed a relatively similar decay in load over the first 0 1000 2000 3000 4000 5000 6000 Lo ad (l b) Time (min) Pipe 3 - 98% PCR Pipe 4 - 49% PCR Pipe 5 - 0% PCR 0 5 10 15 20 25 30 Figure 2-31. Measured load response of Pipes 3, 4, and 5 from 0% to 20% vertical deflection. Pipe 3 - 98% PCR Pipe 4 - 49% PCR Pipe 5 - 0% PCR 0 20 40 60 80 100 0 1000 2000 3000 4000 5000 6000 Lo ad (l b) Time (d) Figure 2-32. Measured load response of Pipes 3, 4, and 5 held at 20% vertical deflection for 100 days.

32 Pipe 3 - 98% PCR Pipe 4 - 49% PCR Pipe 5 - 0% PCR 0 50 150100 200 250 300 350 Time (d) 0 1000 2000 3000 4000 5000 6000 Lo ad (l b) Figure 2-33. Measured load response of Pipes 3, 4, and 5 held at 20% vertical deflection for 250 days. 100 days of loading, with Pipe 3 (98% PCR) tracking slightly below Pipes 4 (49% PCR) and 5 (0% PCR). Additionally, Pipes 3 and 4 both showed more “noise” in the load reduction curves than Pipe 5. This could potentially be due to micro- cracking in the material because of the greater presence of contaminants in pipes manufactured with PCR materials, but could also be related to the different hardware and instru- mentation used on the three loading fixtures. As shown in Figure 2-33, the measured load on Pipe 3 started to deviate from the other two pipes after about 100 days. This coincided with the first cracks that were observed in Pipe 3. A significant longitudinal crack was noted in the springline of Pipe 3 after around 131 days of loading. This corresponded with another change in slope of the load-decay curve for Pipe 3. Figures 2-34 through 2-36 show the cracking that developed in Pipe 3 throughout the test period. The first crack was noted at the crown of the pipe after 101 days, with addi- tional cracks and crack propagation occurring throughout the duration of the test. No cracking was observed in either Pipes 4 or 5 after a year of loading. To see how well the load response curves shown in Fig- ure 2-33 related to the time-dependent modulus curves pro- posed by Chua and Masada for HDPE, the data were normal- ized and plotted along with the normalized Masada modulus curve (recall that the Chua and Masada models were nearly identical, so just the Masada curve was evaluated here). The resulting plot for the first 100 days of loading is shown in Figure 2-37. Both the load reduction and modulus curves show a decay of approximately 62% over 100 days of loading. Since the normalized load reduction curves for the full-scale pipes follow the Masada modulus model nearly identically, it is reasonable to conclude that the time-dependent stress in the pipe wall can be related to the strain by applying the time- dependent modulus curve to the measured strain readings. Note that this is just an approximation, as the load reduction observed in the parallel plate test is dependent not only on the modulus decay but also on slight changes in geometry as well as some redistribution of stresses over time. Addition- ally, the relationship between stress and strain is known to be nonlinear at high strains (34), though this approximation assumes linearity. However, the approximation seems reason- able given the good correlation between the load reduction and modulus curves shown in Figure 2-37, assuming the pri- mary contributor to the load reduction is stress relaxation in the material. Using this approximation, the average tensile stresses in the outer wall of the springline and inner wall of the crown of the pipes throughout the first 100 days of loading are shown in Figure 2-38, based on applying the Masada time-dependent modulus to the average peak measured strains at these loca- tions. As shown in Figures 2-27 through 2-29, the highest tensile strains for all pipes at 20% deflection occurred in the circumferential direction on the outer wall at the springline (see curves for strain gages SG1 and SG3). The strains at this location were comparable for all three test pipes, ranging from 2.9% to 3.1% at 20% deflection. The next highest tensile strain occurred in the circumferential direction on the inner wall at the crown and invert of the pipes, directly underneath the loading plates, and ranged from around 2.3% to 2.5% at 20% deflection for all three pipes. Note that these locations of peak measured wall strains also corresponded to the loca- tions of the observed cracks in Pipe 3.

33 A B Figure 2-34. First cracks observed in the crown of Pipe 3 underneath the loading plates: (A) crack observed after 101 days of loading; (B) crack propagation and an additional crack forming after 131 days. A B Black markings indicate crack propagation Figure 2-35. Crack in springline of Pipe 3 observed after (A) 131 days and (B) 173 days. Crack propagation noted by black markings.

34 Figure 2-36. Crack propagation observed in the Pipe 3 after 286 days of loading. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 N or m al iz ed M od ul us N or m al iz ed L oa d Time (d) Pipe 3 - 98% PCR Pipe 4 - 49% PCR Pipe 5 - 0% PCR Masada Modulus 0 20 40 60 80 100 Figure 2-37. Normalized reduction in load on Pipes 3, 4, and 5 for first 100 days of loading, plotted along with the normalized modulus of elasticity as determined by Masada in Equation 2.19.

35 The curves shown in Figure 2-38 were integrated and divided by the loading duration to determine an equivalent average stress in the pipe wall at the given locations, as shown in Equations 2.20 and 2.21. ∫ σ = −1756.8 100 (2.20) 0.083 0 100 x OWS ∫ σ = −1405.4 100 (2.21) 0.083 0 100 x IWC where sOWS = Equivalent average stress at the outer wall spring- line over 100 days loading, psi sIWC = Equivalent average stress at the inner wall crown over 100 days loading, psi Solving Equations 2.20 and 2.21 yields an equivalent aver- age stress on the outer wall of the pipe at the springline of y = 1756.8x-0.083 y = 1405.4x-0.083 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 St re ss (p si ) Time (d) Outer Wall Springline Inner Wall Crown 0 20 40 60 80 100 Figure 2-38. Approximation of stress relaxation over 100 days on pipe held at 20% vertical deflection based on application of Masada’s elastic modulus to measured wall strains. 9.03 MPa (1309 psi) and 7.22 MPa (1047 psi) on the inner wall of the pipe at the crown. The equivalent average modulus over the 100-day loading period is 301 MPa (43,660 psi). Again, these equivalent average stresses are based on the application of Masada’s time-dependent elastic modulus to the measured strains in the pipe wall. It is likely that some local strains around the pipe wall profile exceeded the measured values due to slight changes and variances in geometry. Based on the charts and equations shown in Figures 2-12 and 2-13 for Pipes 3 and 4, respectively, the failure time for these pipes at these calculated wall stresses were predicted using both the PSM and RPM. The results are shown in Table 2-8. A COV of 0.456 was used to determine the upper and lower bounds on the predicted failure times for Pipe 3, using Equations 2.7 and 2.8. This was based on the highest COV of the three UCLS test conditions. Similarly, a COV of 0.416 was used to determine the upper and lower bounds on the predicted failure times for Pipe 4, based on the highest COV of its three UCLS test conditions. In addition to the 9.03 Service Stress MPa (psi) Pipe 3 – 98% PCR Pipe 4 – 49% PCR PSM RPM PSM RPM 3.45 (500) 110–280 years 759–1926 years 234–543 years 2884–6681 years 7.22 (1047) 1.5–3.7 years 4.4–11.0 years 7.8–18.2 years 48–111 years 9.03 (1309) 143–364 days 332–843 days 2.8–6.5 years 13.9–32.1 years 10.30 (1500) 65–165 days 129–327 days 1.5–3.5 years 6.5–15.2 years 11.30 (1637) 39–99 days 70–178 days 1.0–2.3 years 4.0–9.3 years Table 2-8. Projected failure times for Pipes 3 and 4 based on equivalent average wall stresses determined from constant deflection test at 23çC.

36 and 7.22 MPa (1309 and 1047 psi) conditions, conditions of 10.30 MPa (1500 psi) and 11.30 MPa (1637 psi) are shown to account for the possibility of localized elevated strains around the pipe wall profile and a condition of 3.4 MPa (500 psi) is shown based on the maximum factored stress anticipated in the pipe wall for field installations. An example calculation for Pipe 3 using the PSM follows. From Figure 2-12, the equation of the extrapolated brittle failure curve for Pipe 3 at 23°C is shown in Equation 2.22. σ = − +0.1709 3.7638 (2.22)SVC t where sSVC = Log-based service stress, psi t = Log-based projected average failure time, h For a service stress of 1309 psi, for example, Equation 2.22 can be solved for the average log-based failure time, t, as shown in Equations 2.23 through 2.25. ( )σ = − = = − +1309 3.1169 0.1709 3.7638 (2.23)SVC Log t ( ) ⇒ = − − = 3.1169 3.7638 0.1709 3.785 (2.24)t ⇒ = = =Average Failure Time 10 6096 h 254 days (2.25)3.785 From Equations 2.7 and 2.8, the upper and lower bounds of the projected failure times can be calculated as shown in Equations 2.26 and 2.27. ( ) = +   = + = ( )−UCL COV 254 2.132 0.456 254 5 364 days (2.26) 95% 1X t X n n   ( ) = −   = − = ( )−LCL COV 254 2.132 0.456 254 5 143days (2.27) 95% 1X t X n n   where UCL95% = Upper 95% confidence limit, days LCL95% = Lower 95% confidence limit, days X – = Average calculated failure time, days t(n–1) = Student’s t value at (n − 1) degrees of freedom = 2.132 COV = Max. coefficient of variation of five test specimens = 0.456 n = Number of test specimens at each condition = 5 Similar calculations were performed for Pipes 3 and 4 at multiple stress conditions for both the PSM and RPM. The results are tabulated in Table 2-8. Based on the data shown in Table 2-8, one can conclude that the PSM provided a reasonable prediction of the fail- ure times observed in the constant deflection test. The first crack observed in the springline of the pipe occurred at 131 days, while the PSM model predicted cracking at this location between 143 and 364 days, based on the wall stresses calculated from measured strains. The PSM overpredicted the time to cracking in the crown of the pipe—the predicted failure time was between 1.5 and 3.7 years, while cracking was first observed after 101 days. Keep in mind that the stress calculations around the pipe are approximations based on measured strains in a specific location, and it is likely that some local wall strains and stresses exceeded those measured in the test. Also, it is likely that the stresses in the inner wall at the crown of the pipe increased over time due to redistribu- tion of stresses in the pipe and local buckling at the point of contact of the pipe with the loading plates. To evaluate the distribution of local strains and stresses in the pipe wall, a finite element analysis (FEA) was conducted by Ohio University (see Appendix E). Figure 2-39 shows an image of the strain distributions in the pipe at 20% vertical deflection, and Figure 2-40 shows how the strains vary along the length of the pipe. The peak local strain in the crown of the pipe cal- culated from the FEA was 3.75%. Using the same equivalent average 100-day modulus as discussed earlier (301 MPa, or 43,660 psi), the peak local stress in the inside crown of the pipe was 11.3 MPa (1637 psi). Based on this stress, crack- ing was predicted to occur in the crown of Pipe 3 between 39 and 99 days using the PSM and between 70 and 178 days using the RPM (Table 2-8). Both the PSM and RPM yielded relatively good predictions of the actual failure time of 101 days. The peak local strain at the outer springline of the pipe calculated from the FEA was 3.10%, corresponding to a stress of 11.3 MPa (1353 psi). Based on this stress, cracking was predicted to occur in the springline of Pipe 3 between 119 and 301 days using the PSM and between 265 and 673 days using the RPM. The first crack in the springline was observed at 131 days, indicating that the PSM model accurately pre- dicted the failure time, while the RPM overpredicted the failure time. Cracking was not predicted to occur in Pipe 4 with either the PSM or RPM within the range of testing (less than 1 year), even at the very high local stresses obtained from the FEA [e.g., 11.3 MPa (1647 psi)]. No cracking was observed in this pipe. The model was not evaluated for Pipe 5 since there were not sufficient UCLS failure data to establish a predic- tion. However, cracking was not anticipated for Pipe 5, and no cracking was observed.

37 Figure 2-39. Peak strains in 750 mm (30 in.) diameter pipe at 20% deflection via FEA evaluation. Figure 2-40. Local strains along pipe axis at crown (left) and springline (right) for 750 mm (30 in.) diameter pipe at 20% deflection. In general, the PSM was more conservative than the RPM for both pipes, particularly at low stresses. This is because of the greater slope of the brittle failure curve as predicted by PSM versus RPM. Additionally, the PSM prediction model showed better agreement with the experimental and FEA results than the RPM model. In particular, the PSM perfectly predicted the failure times at both the crown and the spring- line of Pipe 3 based on the local stress analysis obtained from the FEA results. Based on these test results, the service life prediction model based on application of the PSM to UCLS failure data is validated. Table 2-9 summarizes the parallel plate test results. 2.2.4.2 Simulated Field Test Validation of Model In addition to the parallel plate test validation of the UCLS-based service life prediction model, full-scale pipes were installed in a simulated field test to evaluate the model. To generate failures within a reasonable time frame (less than a year), the pipes were installed in an extreme condition

38 Pipe PCR Content Analysis Method Predicted Time to Crown Cracking Predicted Time to Springline Cracking Load Duration Results 3 98% FEA PSM1 33–99 days 119–301 days 365 days Cracking commenced at 101 days on the inner crown and 131 days on the outer springline RPM2 70–178 days 265–673 days SG PSM3 1.5–3.7 years 143–364 days RPM4 4.4–11.0 years 332–843 days 4 49% FEA PSM1 1.0–2.3 years 2.4–5.6 years 365 days No cracks after 1 year of testing RPM2 4.0–9.3 years 12–27 years SG PSM3 7.8–18.2 years 2.8–6.5 years RPM4 48–111 years 14–32 years 5 0% N/A N/A 365 days No cracks after 1 year of testing 1 Analysis via the Popelar shift method to predict cracking based on peak local strains obtained from finite element model 2 Analysis via the rate process method to predict cracking based on peak local strains obtained from finite element model 3 Analysis via the Popelar shift method to predict cracking based on physical strain gage measurements 4 Analysis via the rate process method to predict cracking based on physical strain gage measurements Table 2-9. Summary of parallel plate test results. designed to target tensile strains in the pipe wall in excess of 3.0% (resulting in average stresses of around 1200 psi or more). Creating such a condition was not a trivial task, as most field installations result in compressive strains that dominate and offset the tensile strains in the pipe wall (23). To generate high tensile strains, the pipes were installed on a firmly compacted bedding material with moderately com- pacted [85–90% standard Proctor density (SPD)] ASTM Class III soils around the pipe and no compaction in the haunch area. The simulated fill height was 9.1 m (30 ft). An analysis of this installation in accordance with the AASHTO design method (37) for the 750 mm (30 in.) diameter test pipes yields predicted pipe deflections of 12% at the time of loading and increasing to 15% as soils consolidate. The peak calculated circumferential tensile bending strain in the pipes for this condition is 3.5%, occurring on the inner wall of the invert and the outer wall of the springline of the pipe. See Appendix F for detailed calculations. This installation condi- tion also results in high longitudinal tensile stresses at the junction of the liner and corrugation, as shown by McGrath et al. (30) and illustrated in Figure 2-41. The simulated field test was conducted by installing pipes in precast reinforced concrete chambers and loading them via dead weights hanging from a double lever arm assembly with a 12:1 mechanical advantage. The precast concrete chambers were 3 m (10 ft) wide by 3 m (10 ft) long and 2.4 m (8 ft) high. This allowed enough for one diameter of pipe on each side of the installed pipe, which has been shown to be important for minimizing edge effects (30, 38). To further minimize frictional effects at the interface of the soil and sidewall of the chamber during loading, the sidewalls were lined with two sheets of HDPE material with grease between them, as Note: Image for illustration purposes only; stress magnitudes do not apply to the pipes in this study. Figure 2-41. Mechanism of liner bending resulting in longitudinal tensile stresses at the junction of the corrugation and liner, as illustrated by McGrath et al. and reported in NCHRP Report 631 (30).

39 suggested by Brachman et al. (38). To simulate 9.1 m (30 ft) of cover, weights were added to the loading arms to generate a surcharge of 175 kPa (25 psi). Steel plates [20 mm (0.75 in.) thick] were used to distribute the applied load, and a 150 mm (6 in.) layer of geogrid-reinforced compacted crushed rock was placed beneath the loading plates to prevent shear failure Sidewalls were treated with two sheets of HDPE with a grease film between to minimize frictional effects at the edges. Bedding was firmly compacted sand, and the backfill was an ASTM Class III material compacted to 85% SPD. Simulated cover height was 9.1 m (30 ft). Figure 2-42. Illustration of precast reinforced concrete loading chambers for simulated field test. of the ASTM Class III backfill materials. Figure 2-42 shows an illustration of the designed system, including the location of pressure cells, and Figures 2-43 and 2-44 show images of the test fixtures and loading mechanism. Pipes 3 through 9 (Table 2-1) were evaluated in four sepa- rate precast reinforced concrete chambers. Pipe 3 (98% PCR,

40 Figure 2-43. Precast reinforced concrete chamber assembly and preparation for pipe installation. Figure 2-44. Images of assembled load chambers.

41 Manufacturer A) was installed in Chamber 1. Pipes 4 and 5 (49% and 0% PCR, respectively; Manufacturer A) were coupled together with their inline bell and spigot joint and installed in Chamber 2. Pipes 6 and 7 (98% PCR, Manufac- turer B) were coupled together with an external split coupler and installed in Chamber 3. Pipes 8 and 9 (54% and 59% PCR, respectively; Manufacturer C) were coupled together via an inline bell and spigot joint and installed in Chamber 4. Table 2-10 shows a summary of the pipes in each chamber. Pipe deflections were measured throughout the duration of the test and are shown in Figures 2-45 through 2-51. Soil pressures were recorded in Chambers 1 through 3 to ensure general consistency in the installations, and strain gages were installed on Pipes 8 and 9. See Appendix G for soil pressure and strain gage measurements. As shown in Figures 2-45 through 2-51, the deflections were relatively similar for each pipe tested. Furthermore, they agreed with the target deflections calculated from the AASHTO design method very well. The vertical deflection of the pipes was −10% to −12% after initial application of the load, increasing to −13% to −15% after a year of loading. The horizontal deflection stabilized at around 10%. Pipes 4 and 5 (Chamber 2), manufactured with 49% and 0% PCR materi- als, respectively, showed the least deflection over time. Pipes 8 and 9 (Chamber 4), manufactured with 54% and 59% PCR materials, respectively, showed significant local buckling at the springline and had final vertical deflections of around 20% (negative) after 223 days. However, it should be noted that these pipes were installed and loaded for 83 days, then exca- vated and re-installed when the strain gages and data acquisi- tion system were found to be generating erroneous readings. Unfortunately, a permanent vertical deflection of −5% and horizontal deflection of 4% had set in the pipes after the first installation (note that Figures 2-50 and 2-51 show starting Chamber Pipe Mfr. PCR Content Pipe Length Joint Duration of Load 1 3 A 98% 3.0 m (10 ft) None 400 days 2 4 A 49% 1.5 m (5 ft) Inline bell and spigot 442 days5 A 0% 1.5 m (5 ft) 3 6 B 98% 1.5 m (5 ft) External split coupler 433 days 7 B 98% 1.5 m (5 ft) 4 8 C 54% 1.5 m (5 ft) Inline bell and spigot 306 days9 C 59% 1.5 m (5 ft) Table 2-10. Summary of test pipes in precast reinforced concrete chambers. -20.0% -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 15.0% D efl ec ti on Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection 0 100 200 300 400 500 Figure 2-45. Deflection measurements in Pipe 3 (98% PCR, Chamber 1) throughout 400 days of loading.

42 -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 15.0% D efl ec ti on Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection 0 100 200 300 400 500 Figure 2-46. Deflection measurements on Pipe 4 (49% PCR, Chamber 2) throughout 442 days of loading. Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection 0 100 200 300 400 500 -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% D efl ec ti on Figure 2-47. Deflection measurements on Pipe 5 (0% PCR, Chamber 2) throughout 442 days of loading.

43 -20.0% -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 15.0% D efl ec ti on Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection 0 100 200 300 400 500 Figure 2-48. Deflection measurements on Pipe 6 (98% PCR, Chamber 3) throughout 433 days of loading. -20.0% -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 15.0% D efl ec ti on Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection 0 100 200 300 400 500 Figure 2-49. Deflection measurements on Pipe 7 (98% PCR, Chamber 3) throughout 433 days of loading.

44 -20.0% -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 20.0% 15.0% D efl ec ti on Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection Note: Pipe was loaded, then excavated and re-installed due to instrumentation errors, resulting in vertical and horizontal pre-deflections of 5%. 80 130 180 230 280 330 Figure 2-50. Deflection measurements on Pipe 8 (54% PCR, Chamber 4) throughout 306 days of loading. -25.0% -20.0% -15.0% -10.0% -5.0% 0.0% 5.0% 10.0% 20.0% 15.0% D efl ec ti on Time (d) Average Vertical Deflection Average Horizontal Deflection Average Diagonal Deflection Note: Pipe was loaded, then excavated and re-installed due to instrumentation errors, resulting in vertical and horizontal pre-deflections of 5%. 80 130 180 230 280 330 Figure 2-51. Deflection measurements on Pipe 9 (59% PCR, Chamber 4) throughout 306 days of loading.

45 vertical and horizontal deflections of −5% and 4%, respec- tively). After re-installing the pipes, the vertical deflections increased from −5% to −20%, and the horizontal deflections increased from 4% to 15%, so the net change was similar to the other pipes. According to the AASHTO design methodology (37), the peak calculated circumferential tensile bending strain in the pipes for this installation condition is 3.5% (see Appendix F for calculations). Using the same 100-day equivalent average modulus of 301 MPa (43,660 psi) determined from the paral- lel plate test analysis, the peak 100-day average tensile stress in the pipe wall can be calculated to 10.5 MPa (1528 psi), as shown in Equation 2.28. ( )σ = ε = =i i301 0.035 10.5MPa 1528 psi (2.28)E where s = Peak equivalent tensile stress, MPa (psi) E – = Equivalent average modulus for 100-day loading period, MPa (psi) ε = Peak circumferential tensile bending strain, mm/mm (in./in.) Using this equivalent average stress, the failure times can be predicted for each pipe based on the curves shown in Figures 2-12 through 2-17, as illustrated previously for the parallel plate test. A summary of the predicted failure times using both the PSM and RPM is shown in Table 2-11. All pipes that were predicted to develop cracks within 1 year of loading did, validating the prediction model established by the UCLS test. As with the parallel plate test, the PSM more accu- rately predicted the failure times than the RPM. Pipe 3 started to exhibit cracking after 105 days of loading. The prediction model based on the PSM predicted cracking to occur on Pipe 3 between 58 and 148 days, while the RPM predicted cracking to occur between 113 and 288 days. Both circumferential and lon- gitudinal cracks were observed at the invert and springline of the pipe. Upon excavation of the pipe, it was evident that some Stage I ductile yielding occurred at the outer springline, and Stage II brittle cracks propagated from the ductile failure. This makes sense because the short-term stress at the outer spring- line, assuming a short-term modulus of 758 MPa (110 ksi), was 26.5 MPa (3850 psi), near the measured tensile yield strength of the material, which was 27.2 MPa (3939 psi). All circumferential cracking occurred at the junction of the corrugation and liner at the invert and crown of the pipe. This was due to the high lon- gitudinal strains created at this junction because the pipe was installed on firmly compacted bedding with minimal haunch support, as previously shown by McGrath et al. (30). Images of the cracks observed in Pipe 3 are shown in Figures 2-52 and 2-53. Pipes 4 and 5 were not predicted to exhibit cracking within the duration of the loading period, and no cracks were observed (Figure 2-54). Cracks were first noticed in Pipes 6 and 7 after 185 days of loading. The cracks were observed in the outer corrugations at the invert as the result of high tensile stresses created from local buckling. This was confirmed upon excavation of the pipes. Longitudinal cracking was observed at the springline of Pipe 6 after 400 days of loading, resulting from high tensile stresses in the outer wall and local buck- ling of the liner. Cracking was predicted to occur in Pipe 6 between 71 and 220 days using the PSM and between 70 and 219 days using the RPM. For Pipe 7, cracking was predicted to occur between 73 and 172 days using the PSM and between Pipe Mfr. PCR Content Predicted Time to Cracking PSM RPM 3 A 98% 58–148 days 113–288 days 4 A 49% 1.4–3.2 years 5.9–13.7 years 5 A 0% N/A N/A 6 B 98% 71–220 days 70–219 days 7 B 98% 73–172 days 10–24 days 8 C 54% 203–578 days2 18–52 days2 9 C 59% 139–357 days2 351–897 days2 1 This installation results in a peak calculated 100-day tensile stress of 10.5 MPa (1528 psi). Service life calculations are based on a temperature of 23°C. 2 Calculations for Pipes 8 and 9 conservatively assume a wall stress 10% greater than the other pipes [11.6 MPa (1700 psi)] to account for the increase in bending deflection due to the removal and re-installation of the pipes that resulted in increased vertical and horizontal deflections. Table 2-11. Predicted failure times for 750 mm (30 in.) diameter test pipes in simulated field test of 9.1 m (30 ft) of fill, ASTM Class III soils compacted to 88% SPD, with no compaction on the haunches and installed on firm bedding.1

46 Figure 2-53. Image of ductile failure on outer springline of Pipe 3, leading to brittle cracking and localized buckling of the liner. Note: Longitudinal cracks are circled in the image; circumferential cracks are boxed. Image on right is the first crack observed, a circumferential crack at the invert. Figure 2-52. Circumferential and longitudinal cracks developed in Pipe 3, starting at 105 days.

47 Figure 2-54. Pipes 4 and 5 showed no cracking after 422 days of loading. 18 and 52 days using the RPM. Again, the service life predic- tion model developed from the UCLS test and using the PSM correlated well with the field test results, with cracks occurring within the predicted time frame. Images of cracks observed in Pipes 6 and 7 are shown in Figures 2-55 and 2-56. Recall that Pipes 8 and 9 were deflected to 20% vertical rather than 15% due to the excavation and re-installation of the pipes. This resulted in severe local buckling at the spring- line. Pipe 9 developed small longitudinal cracks in the liner at the springline as a result. The cracks first became evident after 300 days of loading. Pipe 8 did not exhibit any cracking. Due to the complex stress field resulting from the local buckling at the springline, it is difficult to determine the actual local applied stress. Assuming a tensile stress of 11.6 MPa (1700 psi), cracking was predicted to occur on Pipe 9 between 139 and 357 days using the PSM and between 351 and 897 days using the RPM. Pipe 8 was not expected to crack using the PSM, but was when using RPM. As with the other pipes, the PSM provided a better prediction of failure times than the RPM. Figure 2-57 shows an image of local buckling and cracking failures in Pipes 8 and 9. Based on this testing, the UCLS-based service life predic- tion model using the PSM is again validated on full-scale pipe specimens under semi-constant strain conditions. This finding paves the way for the UCLS test to be used as a true performance-based test. A summary of the results from the simulated field test along with the predicted values is shown in Table 2-12. 2.2.5 Fatigue Failure Assessment from Cyclical Live Loads 2.2.5.1 Field Study of Pipes in Shallow Cover Underneath a Commuter Railroad To evaluate the pipes’ durability with regards to cyclical live loads, two 750 mm (30 in.) diameter pipes (Pipes 1 and 2 from Table 2-1) were installed with shallow cover underneath Figure 2-55. Cracks in Pipes 6 and 7 due to local buckling at invert, first occurring after 185 days of loading.

48 Figure 2-56. Cracking in Pipe 6 at springline (left) and invert (right) resulting from local buckling; invert cracking started at 185 days, and springline cracking at 400 days. Figure 2-57. Local buckling occurring at springline of Pipes 8 and 9 deflected to 20%. Pipe 9 (59% PCR) developed cracking after 300 days of loading. Pipe 8 (54% PCR) did not exhibit cracking after 306 days of loading. an active commuter railroad in the Doylestown subdivision north of Philadelphia. The railroad was owned and operated by the Southeastern Pennsylvania Transit Authority (SEPTA), and it collaborated with the research team in selecting a loca- tion for the pipes and installing them in accordance to its stan- dard practices. The project included the first large-diameter corrugated HDPE pipe manufactured with PCR content installed underneath any active railroad in the world. Data obtained from the field test were used to assess the static and dynamic loads that are transferred to the pipe in an aggres- sive loading condition, and the resulting strains in the pipe wall were used to develop an accelerated laboratory test to evaluate the long-term performance of the various types of materials relative to fatigue. The two pipes—one manufac- tured with 0% PCR content and the other with 49% PCR content—were installed end to end in the same trench and coupled at the centerline of the track with a bell-and-spigot- style watertight joint. This installation procedure ensured each pipe experienced the same loads from the railcars so that a direct comparison could be made. Figure 2-58 shows a schematic of the installed pipes, and Figure 2-59 shows a Google Maps® image of the test site. Prior to installation, the pipes were instrumented with strain gages and string potentiometers to measure the wall strains and deflections, respectively. The strain gages and string potentiometers were placed such that they would be located directly underneath one of the tracks upon installation of the pipes (Figure 2-58), approximately 725 mm (28.5 in.) from the spigot end on the ASTM F2648 pipe and 1030 mm (40.5 in.) from the bell end of the AASHTO M 294 pipe. Type J thermo- couples were also installed in each pipe to monitor tempera- ture. Vishay Precision Group® CEA-00-375-UW-350 strain gages were used to measure the strains, and Celesco® SPD-4-3 string potentiometers with a 120 mm (4.75 in.) stroke were used to measure the deflections. Table 2-13 shows a summary of the strain gages and string potentiometers installed in the test pipes, and Figure 2-60 illustrates their locations relative to the respective pipes. Two Dataq Instruments® DI-718Bx-S stand-alone data loggers were used to record the strain and deflection data from each pipe. Dataforth® DI-8B38-05 isolated strain gage amplifiers were used to amplify and condition the strain gage input signal, and DO-8B36 poten- tiometer input modules were used to amplify and condition each string potentiometer input signal.

49 Pipe Manufacturer PCR Content Predicted Time to Cracking Observed Time to CrackingPSM RPM 3 A 98% 58–148 days 113–288 days 105 days 4 A 49% 1.4–3.2 years 5.9–13.7 years No cracks after 422 days 5 A 0% N/A N/A No cracks after 422 days 6 B 98% 71–220 days 70–219 days 185 days 7 B 98% 73–172 days 10–24 days 185 days 8 C 54% 203–578 days1 18–52 days1 No cracks after 306 days 9 C 59% 139–357 days1 351–897 days1 300 days 1 Calculations for Pipes 8 and 9 conservatively assume a wall stress 10% greater than the other pipes [11.6 MPa (1700 psi)] to account for the increase in bending deflection due to the removal and re-installation of the pipes that resulted in increased vertical and horizontal deflections. Table 2-12. Summary of test results for simulated field test based on 750 mm (30 in.) diameter pipes installed on firm bedding to a simulated cover height of 9.1 m (30 ft), backfilled with ASTM Class III soil compacted to 88% SPD, with no compaction in the haunches. Figure 2-58. Illustration of pipe installation underneath active commuter railroad. The strain gage manufacturer’s recommendation of flame- burnishing the polyethylene surface prior to installing the strain gage was used. The string potentiometers were limited to a stroke of 120 mm (4.75 in.), so braided wires were used to complete the span of the pipe diameter. Figure 2-61 shows the installed string potentiometers and strain gages. The pipes were installed in accordance with SEPTA stan- dard installation practices. The existing track was removed and the trench excavated for the installation of the new pipes, as shown in Figure 2-62. Sloped trench walls were used in accordance with SEPTA practices. The trench width at the bottom of the trench was 2.4 m (8 ft), which allowed 0.75 m (2.5 ft) on each side of the installed pipe to ensure adequate compaction of the backfill materials. SEPTA’s design requirements specified a minimum of 1.7 m (5.5 ft) of cover from the top of the pipe to the bottom of the railroad tie for main line installations. However, SEPTA engineers allowed this minimum to be reduced to 0.6 m (2 ft) for the purposes of this test project. This was advantageous for the research as it allowed the observation and measure- ment of pipe behavior in an extreme installation. The trench was excavated to a depth of 1.8 m (6 ft) below the railroad tie to allow the placement of 300 mm (12 in.) granular bedding material at the bottom of the trench. This excavation depth resulted in 0.6 m (2 ft) of cover from the top of the pipe to the bottom of the railroad tie after the pipe was placed on the bedding material [the pipe outside diameter was 900 mm (35 in.)]. Figure 2-63 shows the measured trench depth after

50 Figure 2-59. Google Maps® image of test location (pipes shown are not to scale). excavation. The trench was graded with a slight downward pitch from the center to each end to ensure adequate drainage. After the trench was excavated, approximately 300 mm (12 in.) of bedding material was installed in the bottom of the trench. This material was classified as modified 2-A granular material (a mix of course stone, no greater than 2 in., and fines) according to Pennsylvania DOT (PennDOT) spec- ifications and was the same material that was placed in the pipe envelope. The pipes were then lowered into the trench and placed on top of the bedding material. The instrumented section of Pipe 1 (virgin) was joined to the instrumented sec- tion of Pipe 2 (49% PCR) via the inline bell and spigot con- nector prior to lowering the pipes into the trench. Following the placement of the pipes in the trench, care was taken to ensure the joint was centered between the rails and that the instrumentation in each pipe lined up directly underneath each rail. The pipe was rotated such that the strain gages and string potentiometers were positioned properly. See Fig- ure 2-64 for images of the transportation, installation, and alignment of the pipes. The PennDOT modified 2-A granular backfill material was installed in 200 to 300 mm (8 to 12 in.) lifts up to the spring- line of the pipe, then 300 to 600 mm (12 to 24 in.) lifts until the backfill extended to 300 mm (12 in.) above the top of the pipe. A vibratory rammer was used to compact the backfill materials in the haunches of the pipe, and each lift was com- pacted with a Wacker Neuson® vibratory plate soil compactor until the backfill reached up to the springline of the pipe. A Wacker Neuson vibratory sheepsfoot trench roller was used to compact each lift of backfill material above the springline. The track ballast material (standard SEPTA ballast, 2 to 3 in. crushed granite) was then placed on top of the PennDOT modified 2-A granular backfill material. The construction and pipe backfill processes are shown in Figures 2-65 through 2-67. Standard SEPTA procedures were followed to re-install the track after final placement of ballast material. This involved placing the track and railroad ties that had been removed and re-joining them with the existing track (Figure 2-68). The ballast material was then vibrated into place with a ballast tamper and the track leveled. The track was re-leveled after a

51 Pipe Label Location Description Measurement Model Type / Number Pipe 1 – Virgin SG1-V 0° (Crown), Outer wall Strain CEA-00-375-UW-350 SG2-V 0° (Crown), Inner wall Strain CEA-00-375-UW-350 SG3-V 315°, Inner wall Strain CEA-00-375-UW-350 SG4-V 270°, Outer wall Strain CEA-00-375-UW-350 SG5-V 270°, Inner wall Strain CEA-00-375-UW-350 SG6-V 180° (Valley), Inner wall Strain CEA-00-375-UW-350 SG7-V 90°, Inner wall Strain CEA-00-375-UW-350 SG8-V 45°, Inner wall Strain CEA-00-375-UW-350 SP1-V Vertical: 0°–180° Deflection SPD-4-3 120 mm SP2-V Horizontal: 90°–270° Deflection SPD-4-3 120 mm SP3-V Diagonal: 45°–225° Deflection SPD-4-3 120 mm SP4-V Diagonal: 135°–315° Deflection SPD-4-3 120 mm T1-V NW End of Pipe Temperature Type J Thermocouple Pipe 2 – 49% PCR SG1-R 0° (Crown), Outer wall Strain CEA-00-375-UW-350 SG2-R 0° (Crown), Inner wall Strain CEA-00-375-UW-350 SG3-R 45°, Inner wall Strain CEA-00-375-UW-350 SG4-R 90°, Outer wall Strain CEA-00-375-UW-350 SG5-R 90°, Inner wall Strain CEA-00-375-UW-350 SG6-R 180° (Valley), Inner wall Strain CEA-00-375-UW-350 SG7-R 270°, Inner wall Strain CEA-00-375-UW-350 SG8-R 315°, Inner wall Strain CEA-00-375-UW-350 SP1-R Vertical: 0°–180° Deflection SPD-4-3 120 mm SP2-R Horizontal: 90°–270° Deflection SPD-4-3 120 mm SP3-R Diagonal: 45°–225° Deflection SPD-4-3 120 mm SP4-R Diagonal: 135°–315° Deflection SPD-4-3 120 mm T1-R NW End of Pipe Temperature Type J Thermocouple Table 2-13. Instrumentation description for field test pipes. Figure 2-60. Instrumentation layout for test pipes (Pipe 2 labels designated with “-R”).

Figure 2-61. String potentiometers and strain gages installed on test pipes. Figure 2-62. Removal of track and excavation of trench for pipe installation. Figure 2-63. Measurement of trench depth after excavation.

53 Figure 2-64. Transportation and installation of pipes and alignment of instrumentation. Figure 2-65. Placement of PennDOT modified 2-A backfill material in trench and ensuring haunch support.

54 Figure 2-66. Compaction of the backfill material in lifts. Figure 2-67. Placement of ballast material over pipe envelope. Figure 2-68. Re-installation of track after pipe installation. week of live loading over the pipes, as some resettling of the track is typical on this type of construction project. A Harsco® Mark IV ballast tamper weighing 29.5 tons (65,000 lb) was used to consolidate the ballast. This equipment utilizes posi- tive displacement vibrators to penetrate and compact the bal- last. The vibrators extended to a depth of around 300 mm (12 in.) below the railroad tie, which was approximately 300 mm (12 in.) above the crown of the pipe. Data was col- lected during this process to evaluate the effects of the vibra- tion process on the wall strains and deflections in the pipe. Figure 2-69 shows images of the ballast tamper. The installation of the pipes was completed on October 12, 2013, with some final track grinding and leveling occurring on October 13. After installation, the pipes were thoroughly inspected and baseline inside diameter measurements were made to evaluate the post-construction deflection of the pipes. The results from the deflection measurements are

55 Vibrating ballast penetrators Figure 2-69. Ballast tamper passing over installed pipes. shown in Table 2-14. Some slight peaking [i.e., increase in vertical inside diameter (ID) and decrease in horizontal ID due to compaction of backfill materials alongside of the pipes] was noted in the installation. This is not uncommon and indicates a quality installation. Strain and deflection data from live loads were recorded periodically from October 2013 through June 2015. The first set of data was recorded on October 14, 2013, during the first live load that passed over the pipes. Strain gage data for this event are shown in Figures 2-70 and 2-71 for Pipe 1 and Pipe 2, respectively, and deflection data are shown in Figures 2-72 and 2-73. As evident from the figures, the measured wall strains on Pipe 2 were slightly larger than Pipe 1. The static strains (due to soil loads) ranged from approximately −600 to −2750 microstrain for Pipe 1 and −1500 to −3400 micro- strain for Pipe 2; the maximum peak-to-peak dynamic strains ranged from −50 to −360 microstrain for Pipe 1 and −100 to −400 microstrain for Pipe 2. As with the wall strains, the mea- sured deflections on Pipe 2 were slightly greater than Pipe 1, though all deflections were very small. The deflections after installation were less than 5 mm (0.20 in.), and the peak dynamic deflections measured during the first live load were less than 0.5 mm (0.02 in.). Comparisons of the static and dynamic wall strains and deflections for each pipe are shown in Figures 2-74 through 2-77. The strain gage at the 180° location of Pipe 2 (SG6-R) read significantly higher than its counterpart gage in Pipe 1 (SG6-V). Subsequent data collection periods suggested that the gage in Pipe 2 was not reading correctly, likely due to a poor bond with the pipe wall. Figures 2-78 and 2-79 show the ratios between the peak measured static and dynamic wall strains and deflections for Pipes 1 and 2. The static strain ratios were calculated by taking the average measured static strains (no live loads) at each gage in Pipe 2 and dividing by the average measured strains in their counterpart gages in Pipe 1. The dynamic strain ratios were calculated by taking the peak-peak mea- sured dynamic strains during the live load event at each gage in Pipe 2 and dividing by the peak-peak measured dynamic strains in their counterpart gages in Pipe 1. Similar calcula- tions were performed on the string potentiometer measure- ments from each pipe to determine the static and dynamic deflection ratios. Dynamic wall strains were 1.2 to 2.0 times greater for Pipe 2 than Pipe 1, and Pipe 2 deflections were 0.8 to 1.3 times those in Pipe 1. All of the Pipe 2 strains were greater than or equal to the Pipe 1 strains, for the dynamic loading condition as well as the static. There could be several reasons for this. First, it may be due to the difference in materials. Pipe 1 had a slightly greater pipe stiffness than Pipe 2 [241 kPa (35.0 pii) vs. 236 kPa (34.3 pii)], as well as a higher material flexural modulus [1054 MPa (152.8 ksi) vs. 1009 MPa (146.3 ksi)]. It could also be due to minor differences in backfill materials, compaction effort, or other installation nuances. Regard- less, strains and deflections on both pipes were very small, and much less than allowed by industry standards, which limit tensile strain to 5.0%, compressive strain to 4.1%, and the in-field deflection to 5% (37, Section 12). Because of the quality, well-compacted backfill material, the measured strains and deflections in both pipes were very small. Net compressive strains were measured around the entire circumference of the pipe, both during static and dynamic loading conditions. No tensile strains were observed. This is consistent with other buried-pipe studies in railroad Measurement Location Pipe 1 (virgin) Pipe 2 (49% PCR) Initial Vertical ID 765 mm (30.1 in.) 765 mm (30.1 in.) Initial Horizontal ID 765 mm (30.1 in.) 765 mm (30.1 in.) Installed Vertical ID 768 mm (30.3 in.) 768 mm (30.3 in.) Installed Horizontal ID 759 mm (29.9 in.) 759 mm (29.9 in.) Table 2-14. Pipe measurements before and after installation.

56 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 St ra in (m ic ro st ra in ) Time (s) SG 6V - 180 deg. Liner SG 5V - 270 deg. Liner SG 7V - 90 deg. Liner SG 3V - 315 deg. Liner SG 2V - 0 deg. Liner SG 4V - 270 deg. Crown SG 8V - 45 deg. Liner SG 1V - 0 deg. Crown 37.00 39.00 41.00 43.00 45.00 47.00 Figure 2-70. Pipe 1 (virgin) strain gage data during four-car train event on October 14, 2013. Time (s) 40.00 42.00 44.00 46.00 48.00 50.00 -4000 -3500 -3000 -2500 -2000 -1500 -1000 St ra in (m ic ro st ra in ) SG 7R - 270 deg. Liner SG 5R - 90 deg. Liner SG 4R - 90 deg. Crown SG 1R - 0 deg. Crown SG 6R - 180 deg. Liner SG 3R - 45 deg. Liner SG 2R - 0 deg. Liner SG 8R - 315 deg. Liner Figure 2-71. Pipe 2 (49% PCR) strain gage data during four-car train event on October 14, 2013.

57 -0.200 -0.150 -0.100 -0.050 0.000 0.050 D efl ec ti on (i n. ) Time (s) SP1V - Vertical SP3V - Diagonal 45-225 SP4V - Diagonal 135-315 SP2V - Horizontal 37.00 39.00 41.00 43.00 45.00 47.00 Figure 2-72. Pipe 1 (virgin) deflections during four-car train event on October 14, 2013. -0.200 -0.250 -0.150 -0.100 -0.050 0.000 0.050 D efl ec ti on (i n. ) Time (s) 40.00 42.00 44.00 46.00 48.00 50.00 SP1R - Vertical SP3R - Diagonal 45-225 SP4R - Diagonal 135-315 SP2R - Horizontal Figure 2-73. Pipe 2 (49% PCR) deflections during four-car train event on October 14, 2013.

58 -450 -400 -350 -300 -250 -200 -150 -100 -50 0 0 DEG. CROWN 0 DEG. LINER 45 DEG. LINER 90 DEG. LINER 180 DEG. LINER 270 DEG. LINER 315 DEG. LINER St ra in (m ic ro st ra in ) Pipe 1 - Virgin Pipe 2 - Recycled -311 -128 -237 -190 -53 -243 -172 -365 -185 -322 -273 -97 -404 -339 Figure 2-75. Comparison of peak-peak dynamic wall strains on Pipe 1 (virgin) and Pipe 2 (49% PCR) during 4-car train event on October 14, 2013. -4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 0 DEG. CROWN 0 DEG. LINER 45 DEG. LINER 90 DEG. LINER 180 DEG. LINER 270 DEG. LINER 315 DEG. LINER St ra in (m ic ro st ra in ) Pipe 1 - Virgin Pipe 2 - Recycled -2752 -2269 -2659 -1374 -606 -962 -2005 -2715 -3227 -2938 -2121 -2842 -1516 -3410 Figure 2-74. Comparison of average static wall strains due to earth load only (no live load) on Pipe 1 (virgin) and Pipe 2 (49% PCR) on October 14, 2013.

59 -0.250 -0.200 -0.150 -0.100 -0.050 0.050 0.000 VERTICAL HORIZONTAL DIAGONAL - 45-225 DIAGONAL - 135-315 D efl ec ti on (i n. ) Pipe 1 - Virgin Pipe 2 - Recycled 0.020 -0.177 -0.036 -0.143 0.033 -0.193 -0.097 -0.099 Figure 2-76. Comparison of average static deflections due to earth load (no live load) on Pipe 1 (virgin) and Pipe 2 (49% PCR) on October 14, 2013. -0.020 -0.016 -0.012 -0.008 -0.004 0.000 VERTICAL HORIZONTAL DIAGONAL - 45-225 DIAGONAL - 135-315 D efl ec ti on (i n. ) Pipe 1 - Virgin Pipe 2 - Recycled -0.014 -0.010 -0.011 -0.013 -0.018 -0.008 -0.013 -0.014 Figure 2-77. Comparison of peak-peak dynamic deflections on Pipe 1 (virgin) and Pipe 2 (49% PCR) during 4-car train event on October 14, 2013.

60 0.00 0.50 1.00 1.50 2.00 0 DEG. CROWN 0 DEG. LINER 45 DEG. LINER 90 DEG. LINER 180 DEG. LINER 270 DEG. LINER 315 DEG. LINER St ra in (m ic ro st ra in ) P2/P1 Dynamic P2/P1 Static 1.18 1.44 1.36 1.44 1.85 1.66 1.96 0.99 1.42 1.11 1.54 1.58 1.70 4.69 Figure 2-78. Ratio of Pipe 2 (49% PCR) to Pipe 1 (virgin) strain measurements based on data shown in Figures 2.74 and 2.75. VERTICAL HORIZONTAL DIAGONAL - 45-225 DIAGONAL - 135-315 D efl ec ti on (i n. ) 0.000 0.500 1.000 1.500 2.000 2.500 3.000 P2/P1 Dynamic P2/P1 Static 1.302 0.790 1.139 1.067 1.62 1.09 2.69 0.69 Figure 2-79. Ratio of Pipe 2 (49% PCR) to Pipe 1 (virgin) deflection measurements based on data shown in Figures 2.76 and 2.77.

61 applications. For example, in Plastics Pipe Institute (PPI) study of the performance of 1200 mm (48 in.) diameter corrugated virgin HDPE pipes under heavy freight rail loads with 1.2 m (4 ft) of cover, all of the measured wall strains were compressive, with the exception of one gage in the invert that read slightly in tension (39). While the pipes in the PPI study were buried twice as deep as the pipes in the SEPTA field study, the live loads were significantly higher (140-ton freight railcars in the PPI study vs. 68-ton commuter railcars in the SEPTA study). The measured static and dynamic strains and deflec- tions in the PPI study were comparable to those measured in the SEPTA study, though of slightly greater magnitude. For comparison purposes, Figures 2-80 and 2-81 show the peak static and dynamic strains that were measured in the PPI study. The track was re-leveled after a week of live loading, which is typical practice for SEPTA operations. The vertical deflection on both Pipes 1 and 2 shifted downward approximately 1.5 mm (0.06 in.) from the measurements taken on October 14, and there was also a slight shift downward on one of the diagonal measurements on both pipes. This was most likely caused by the re-leveling of the track and the compaction of the bal- last material above the pipe. The average static strain readings on both pipes were comparable to those taken on October 14, and the dynamic wall strains were slightly less than those Figure 2-81. Peak-peak dynamic strains during live load freight train event in PPI’s study, reprinted with permission from the author (39). Figure 2-80. Peak measured strains during live load freight train event in PPI’s study, reprinted with permission from the author (39).

62 -3707 -2556 -3270 -1890 -1062 -1403 -2481 -3330 -3407 -2607 -3172 -2279 -3973 -4500 -4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 0 DEG. CROWN 0 DEG. LINER 45 DEG. LINER 90 DEG. LINER 180 DEG. LINER 270 DEG. LINER 315 DEG. LINER St ra in (m ic ro st ra in ) Pipe 1 - Virgin Pipe 2 - Recycled Figure 2-82. Comparison of average static wall strains on Pipe 1 (virgin) and Pipe 2 (49% PCR) on October 24, 2013. measured on October 14. The strain gage located on the 45-degree position of Pipe 2, SG3-R, came unbonded and was no longer reading properly. Charts for the data are shown in Appendix H. Figures 2-82 and 2-83 show the static and dynamic strain comparisons between Pipes 1 and 2, respectively. The static strains in Pipe 2 are again slightly higher than those in Pipe 1, ranging from one to three times those measured in Pipe 1. Likewise, the dynamic strains in Pipe 2 were one to two times those measured in Pipe 1. Figure 2-84 shows the ratio of the Pipe 2 strains to the Pipe 1 strains. Figures 2-85 and 2-86 show the static and dynamic deflection comparisons between Pipes 1 and 2, and Figure 2-87 shows the ratio of the Pipe 2 deflection to the Pipe 1 deflection. Pipe 2 static deflections ranged from 0.67 to 3.14 times those measured in Pipe 1. Pipe 2 dynamic deflec- tions were 0.51 to 1.27 times those of Pipe 1. The temperature inside the pipes was 7.7°C (45.8°F), significantly cooler than the temperature inside the pipes on October 14, which was 20.8°C (69.5°F). Due to the high coefficient of thermal expansion for HDPE, this temperature change accounts for at least some of the downward shift in deflections and compressive strains. Strains and deflections were recorded on December 19, 2013, June 26, 2014, and June 10, 2015. The charts for these data are in Appendix H. Figures 2-88 and 2-89 show the his- torical static and dynamic strains and deflections on Pipe 1, respectively, and Figures 2-90 and 2-91 show the historical static and dynamic strains and deflections on Pipe 2. The static and dynamic strains and deflections for both pipes stayed relatively consistent throughout the data col- lection period. Some slight shifting in the measurements occurred during the first few months of service, likely due both to backfill consolidation as well as temperature changes. From June 2014 through June 2015, the readings were very stable. Three of the five strain gages on Pipe 2 came unbonded during the data collection period. SG3-R came unbonded sometime in the first 2 weeks of operation, SG5-R between October 24 and December 19, 2013, and SG6-R sometime prior to June 26, 2014. All of the strain gages on Pipe 1 remained func- tional throughout the data collection period. The cause of the unbonding is not known but could be that pipes manufac- tured with recycled materials do not respond as well to the flame-burnishing procedure for gage installation. An average of 36 trains passed over the pipe each day throughout the test period, and a typical train consist had three to six cars. The SEPTA railcars are 26 m (85 ft) long and weigh 75 tons unloaded. Each railcar has four axles (two trucks, each with two axles) with a capacity of 109 passen- gers. See Figure 2-92 for an illustration of the SEPTA railcars. Typical speeds of the trains at the location of the pipes were from 65 to 80 km/h (40 to 60 mph). The sampling frequency of the data acquisition system was set to 50 Hz to ensure all dynamic loading events were captured. Examination of Fig- ures 2-70 and 2-71 shows that strain responses from each axle were captured by the data acquisition system. Note that each large peak corresponds to a truck traveling over the pipe, and within each large peak are two small peaks, corresponding to each axle on the truck. The dynamic loads generated by

63 -300 -250 -200 -150 -100 -50 0 0 DEG. CROWN 0 DEG. LINER 45 DEG. LINER 90 DEG. LINER 180 DEG. LINER 270 DEG. LINER 315 DEG. LINER St ra in (m ic ro st ra in ) Pipe 1 - Virgin Pipe 2 - Recycled -223 -100 -158 -152 -40 -166 -145 -221 -110 -148 -84 -259 -193 Figure 2-83. Comparison of peak-peak dynamic strains on Pipe 1 (virgin) and Pipe 2 (49% PCR) during three-car train event on October 24, 2013. 0.00 0.50 1.00 1.50 2.00 2.50 0 DEG. CROWN 0 DEG. LINER 45 DEG. LINER 90 DEG. LINER 180 DEG. LINER 270 DEG. LINER 315 DEG. LINER St ra in (m ic ro st ra in ) 0.99 1.11 0.97 2.10 1.57 1.33 0.90 1.33 1.38 1.62 1.60 2.99 P2/P1 Dynamic P2/P1 Static Figure 2-84. Ratio of Pipe 2 (49% PCR) to Pipe 1 (virgin) strains on October 24, 2013.

64 -0.042 -0.177 -0.035 -0.189 -0.028 -0.193 -0.110 -0.143 -0.250 -0.200 -0.150 -0.100 -0.050 0.000 VERTICAL HORIZONTAL DIAGONAL - 45-225 DIAGONAL - 135-315 D efl ec ti on (i n. ) Pipe 1 - Virgin Pipe 2 - Recycled Figure 2-85. Comparison of static deflections for Pipe 1 (virgin) and Pipe 2 (49% PCR) on October 24, 2013. VERTICAL HORIZONTAL DIAGONAL - 45-225 DIAGONAL - 135-315 D efl ec ti on (i n. ) Pipe 1 - Virgin Pipe 2 - Recycled -0.022 -0.013 -0.020 -0.010 -0.017 -0.006 -0.025 -0.009 -0.028 -0.024 -0.020 -0.016 -0.012 -0.008 -0.004 0.000 Figure 2-86. Comparison of peak-peak dynamic deflections for Pipe 1 (virgin) and Pipe 2 (49% PCR) during three-car train event on October 24, 2013.

65 VERTICAL HORIZONTAL DIAGONAL - 45-225 DIAGONAL - 135-315 D efl ec ti on (i n. ) 0.762 0.508 1.271 0.961 0.66 1.09 3.14 0.76 0.000 0.500 1.000 1.500 2.000 2.500 3.000 3.500 P2/P1 Dynamic P2/P1 Static Figure 2-87. Ratio of Pipe 2 (49% PCR) to Pipe 1 (virgin) deflections on October 24, 2013. -8000 -6000 -4000 -2000 0 2000 7/1/13 7/1/14 7/1/15 SG1-V - 0 Deg. Crown SG2-V - 0 Deg. Liner SG3-V - 315 Deg. Liner SG4-V - 270 Deg. Crown SG5-V - 270 Deg. Liner SG6-V - 180 Deg. Liner SG7-V - 90 Deg. Liner SG8-V - 45 Deg. Liner -800 -600 -400 -200 0 200 7/1/13 7/1/14 7/1/15 St ra in (m ic ro st ra in ) St ra in (m ic ro st ra in ) SG1-V - 0 Deg. Crown SG2-V - 0 Deg. Liner SG3-V - 315 Deg. Liner SG4-V - 270 Deg. Crown SG5-V - 270 Deg. Liner SG6-V - 180 Deg. Liner SG7-V - 90 Deg. Liner SG8-V - 45 Deg. Liner (A) (B) Figure 2-88. Historical static (A) and dynamic (B) strain measurements on Pipe 1 (virgin).

66 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 7/1/13 7/1/14 7/1/15 D efl ec ti on (i n. ) SP1-V - Vertical SP2-V - Horizontal SP3-V - Diagonal 45-225 SP4-V - Diagonal 135-315 -0.08 -0.06 -0.04 -0.02 0 0.02 7/1/13 7/1/14 7/1/15 D efl ec ti on (i n. ) SP1-V - Vertical SP2-V - Horizontal SP3-V - Diagonal 45-225 SP4-V - Diagonal 135-315 (B)(A) Figure 2-89. Historical static (A) and dynamic (B) deflection measurements on Pipe 1 (virgin). -8000 -6000 -4000 -2000 0 2000 7/1/13 7/1/14 7/1/15 St ra in (m ic ro st ra in ) SG1-R - 0 Deg. Crown SG2-R - 0 Deg. Liner SG4-R - 90 deg. Crown SG6-R - 180 Deg. Liner SG7-R - 270 Deg. Liner SG8-R - 315 Deg. Liner -800 -600 -400 -200 0 200 7/1/13 7/1/14 7/1/15 St ra in (m ic ro st ra in ) SG1-R - 0 Deg. Crown SG2-R - 0 Deg. Liner SG4-R - 90 deg. Crown SG6-R - 180 Deg. Liner SG7-R - 270 Deg. Liner SG8-R - 315 Deg. Liner (A) (B) Figure 2-90. Historical static (A) and dynamic (B) strain measurements on Pipe 2 (49% PCR). -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 7/1/13 7/1/14 7/1/15 D efl ec ti on (i n. ) SP1-R - Vertical SP2-R - Horizontal SP3-R - Diagonal 45-225 SP4-R - Diagonal 135-315 -0.08 -0.06 -0.04 -0.02 0.00 0.02 7/1/13 7/1/14 7/1/15 D efl ec ti on (i n. ) SP1-R - Vertical SP2-R - Horizontal SP3-R - Diagonal 45-225 SP4-R - Diagonal 135-315 (B)(A) Figure 2-91. Historical static (A) and dynamic (B) deflection measurements on Pipe 2 (49% PCR).

67 Figure 2-92. Typical SEPTA passenger railcar. the commuter railcars in this shallow cover installation are greater than those generated by typical highway loads. After 3 years of service, no changes in the performance of either Pipe 1 (virgin materials) or Pipe 2 (49% PCR materials) were observed. No cracks or other performance limits were detected in either pipe, and the pipes remained functional and in excellent condition throughout the duration of the study. 2.2.5.2 Laboratory Study to Evaluate Long-Term Durability from Fatigue Loads To evaluate the performance of corrugated HDPE pipe materials manufactured with and without PCR materials relative to fatigue loading from cyclical live loads, specimens were sampled directly from the wall of the pipe and cycled in a servo-hydraulic universal testing machine at Villanova Uni- versity’s Structural Engineering and Research Laboratory. The specimens were taken from the junction of the outer corruga- tion and the inner liner, as this location had previously been identified as an area prone to cracking due to the presence of a stress riser (40). These are the same type of specimens used for FDOT’s 100-year service life protocol for corrugated HDPE Liner Valley Longitudinal Circumferential Corrugation removed Figure 2-93. Junction-style specimen used in FDOT 100-year service life protocol (23) and fatigue testing for this research project. pipe (23). An illustration of the type and location of the speci- men relative to the pipe is shown in Figure 2-93. The test specimens were cut directly from the end sections of the pipes using an ASTM Type IV die in accordance with ASTM D638 (41). An example of this procedure is illustrated in Figure 2-94. After the specimens were stamped from the pipes, they were instrumented with strain gages to correlate the laboratory data to the measured field strains. The strain gages were installed at the junction where the liner meets the outer corrugation, as this corresponded to the location of gages installed in the field study. See Figure 2-95 for the test specimen dimensions and Figure 2-96 for a photograph of an instrumented specimen and strain gage location. Based on the data from the field study, all the measured strains in the pipe wall were compressive, a condition that will not lead to cracking in polyethylene. This is consistent Figure 2-94. Removal of junction specimens for laboratory fatigue test.

68 C B A A = 6.3 mm (0.25 in.) B = 33 mm (1.3 in.) C = 114 mm (4.5 in.) Figure 2-95. Dimensions of test specimen for cyclical fatigue tests. with other live load studies that have been conducted, includ- ing the Plastics Pipe Institute’s live load railroad study at the Transportation Technology Center (39) and the Minnesota Department of Transportation’s live load study of the per- formance of corrugated HDPE pipe with shallow fills (42). However, even though tensile strains were not measured in the field study, it was decided to load the test specimens in tension anyway to evaluate a worst-case scenario because it is feasible that a poor installation or an impingement (e.g., from a rock or other object in the backfill material) could result in localized tensile strains in the pipe wall. The test specimens were cycled at 5 Hz (0.2 second per load cycle) on an MTS universal testing machine; this fre- quency was determined to be the maximum the specimens could be cycled without heating the specimens and adversely affecting the results (see Appendix I). Note that for a typi- cal SEPTA railcar traveling at 80 km/h (50 mph), the average dynamic loading rate over the pipe is approximately 1.5 Hz (6 peaks observed in 4 seconds) for a given train, though the time between two adjacent axles on a truck is approximately 0.1 second. The laboratory test was accelerated in that the specimens were continually cycled, while the field pipes expe- rienced dynamic loading for only 8 to 12 seconds every hour, the rest of the time experiencing only static loads. The procedures outlined in ASTM D7791, “Standard Test Method for Uniaxial Fatigue Properties of Plastics” (43), were roughly followed for the fatigue test; however, because the test specimens were taken from the junction of the inner wall and outer corrugation of the pipe, they did not have a constant cross section within the gage length as required. Therefore, Figure 2-96. Instrumented test specimens.

69 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 1000 1500 2000 2500 3000 3500 4000 4500 5000 G ri p D is pl ac em en t (in .) St ra in (m ic ro st ra in ) Time (s) Strain Displacement 50,400 50,401 50,402 50,403 50,404 50,405 Figure 2-97. Dynamic liner strain and grip displacement measured in laboratory fatigue test for specimen from Pipe 2 (49% PCR). the stresses and strains throughout the gage length of the test specimens were variable. Regardless, this was inconse- quential since the specimens were instrumented with strain gages at the junction of the liner and corrugation (the same location where strains were measured in the field) and the fatigue test parameters were based on the strain mea- surements at this location. Note that this location was not the thinnest section within the gage length of the specimen, so it did not represent the location of maximum stress (or strain) in the specimen. The test was strain-controlled rather than stress-controlled, which was appropriate based on the type of loading experienced in the field. The test specimens were pre-loaded to a strain of 2400 microstrain (tension), then cycled to a fixed grip displace- ment of +/− 0.05 mm (0.002 in.). This resulted in peak-peak measured dynamic strains of 1200 microstrain (tension) at the junction of the liner and corrugation. Note this was three times the magnitude of the maximum dynamic strain measured in the field study for Pipe 2 and four times the maximum dynamic strain measured for Pipe 1. Additionally, the testing was con- ducted in a worst-case tensile condition as opposed to the compressive strains measured in the field. Figure 2-75 showed that the maximum peak-peak dynamic strains observed at this location of the pipes in the field were −311 microstrain (compression) occurring at the 0° (crown) location for Pipe 1 and −404 microstrain (compression) occurring at the 270° (springline) location for Pipe 2. Figure 2-97 shows a trace of the grip displacement and dynamic strains measured at the liner for Pipe 2 (49% PCR) during a typical 5 second interval. Two specimens from Pipe 2 (49% PCR) and one specimen from Pipe 1 (virgin) were tested for up to 10 million cycles. No fatigue-related cracking or permanent deformation was observed on any of the test specimens. Based on these results, it was concluded that neither pipes manufactured with vir- gin materials nor pipes manufactured with PCR materials are prone to fatigue-related failures resulting from live loads from commuter railroad and highway applications. Table 2-15 summarizes the fatigue test data. Pipe ID PCR Content Grip Displacement, mm (in.) Peak-Peak Strain, microstrain Maximum Measured Peak- Peak Field Strain, microstrain Tested Cycles, millions Results 1 1-1 0% +/− 0.05 (0.002) 1200 (tension) 311 (compression) 1.0 No failures 2 2-1 49% +/− 0.05 (0.002) 1200 (tension) 404 (compression) 6.4 No failures 2-2 49% +/− 0.05 (0.002) 1200 (tension) 404 (compression) 10.0 No failures − − − Table 2-15. Summary of laboratory fatigue test data for Pipes 1 and 2.

70 2.2.6 AASHTO Material Specification and Design Methodology Proposals Based on the results from the laboratory and field stud- ies, proposals were developed for AASHTO design meth- odology and material specification requirements for pipes manufactured with recycled materials. They are summa- rized in Chapter 3. Proposed draft changes to AASHTO M 294 for the inclusion of pipes manufactured with recy- cled materials are shown in Appendix J, and proposed changes to the AASHTO design specification are shown in Appendix K. Additionally, an AASHTO Standard Practice was developed for determining the service life of corru- gated HDPE pipes manufactured with PCR materials (see Appendix L).