National Academies Press: OpenBook
« Previous: Chapter 2 - Research Approach
Page 18
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 18
Page 19
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 19
Page 20
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 20
Page 21
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 21
Page 22
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 22
Page 23
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 23
Page 24
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 24
Page 25
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 25
Page 26
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 26
Page 27
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 27
Page 28
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 28
Page 29
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 29
Page 30
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 30
Page 31
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 31
Page 32
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 32
Page 33
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 33
Page 34
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 34
Page 35
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 35
Page 36
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 36
Page 37
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 37
Page 38
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 38
Page 39
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 39
Page 40
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 40
Page 41
Suggested Citation:"Chapter 3 - Findings and Applications." National Academies of Sciences, Engineering, and Medicine. 2011. LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems. Washington, DC: The National Academies Press. doi: 10.17226/14497.
×
Page 41

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.

18 As part of Tasks 1, 2 and 6 for this project, performance data were collected and archived from sites located in the northeastern, mid-Atlantic, southeastern, southwestern, and western United States consistent with details of the existing inventory. Data are included from 170 sites located through- out the United States and Europe. Table 10 updates the sum- mary similar to Berkovitz (1999) of statewide practice and MSE corrosion monitoring programs that have been imple- mented by State DOTs. These programs have produced data that have been archived into the performance database. In general, the database is self-contained yet structured such that it can be ported to other existing databases. The database is formatted using Microsoft Access, which is linked to a geo- graphic information system (GIS) (ArcView) platform to pro- vide visual and spatial recognition of data. The organization and structure of the various tables and data fields are updated, as necessary, to accommodate different types of information that are identified from available data sets. For example, obser- vations of reinforcement performance and condition may be based on NDT, direct physical measurement, or visual obser- vations, and these data types are archived accordingly. Drop down lists and check boxes are implemented to facilitate min- ing or querying of the database. Separate databases have been developed for MSE reinforcements (Type I), and for ground anchors, rock bolts, and soil nails (Type II). The Type II data- base is similar to the Type I database, but some data fields are different to address corrosion protection measures, different subsurface conditions, and types of monitoring techniques that are specific to the Type II systems. Information within the shell of the database is distributed amongst seven distinct tables comprising a total of 150 data fields. The tables are divided into categories of information similar to those employed in other databases that are based on the FHWA Bridge Management Inventory. The database includes the following tables: • Project, • Walls/Structure, • Reinforcements, • Backfill/Subsurface, • Observation Points, • NDT Results, and • Direct Observations. Microsoft Access data forms were created to facilitate data entry and examples of these forms are included in Appen- dix C. Tables are related with a one-to-many relationship using “project number” as a key parameter. Other relation- ships may also be created, but currently all other tables are considered to be a sub-form to the project form, which serves as the master form. Thus, a project may have a number of walls or backfills. A wall may have numerous observation points, and a number of observations, including NDT or direct phys- ical observations, may be associated with each observation point. For example, the project in Las Vegas, Nevada, described by Fishman et al. (2006) includes three walls; Wall #1 has 15 monitoring stations, Wall #2 has six, and Wall #3 has four. Each monitoring station includes two in-service reinforce- ments wired for monitoring (NDT) and at least two coupons: one plain and one galvanized. Also, direct physical measure- ment of section loss is performed on 18 samples retrieved from six of the monitoring stations (i.e., three reinforcements exhumed from six of the stations). These data are all organized into separate tables that are linked to the Las Vegas, Nevada, entry from the project table. Relationships are also defined between backfill, wall, reinforcements, monitoring stations, and results tables. Each project is associated with a point that is displayed on a map within ArcView as shown in Appendix C. ArcView- mapped points are also linked to the Microsoft Access tables so pertinent information for each project can be displayed next to each point when selected by the user. In this way, the geographic distribution of performance data, as well as specific attributes for each site can be displayed within a GIS platform. Thus, the user may associate the C H A P T E R 3 Findings and Applications

Table 10. Summary of state DOT MSE wall corrosion assessment programs. 19 data with geographic location and view all of the perfor- mance data and pertinent information associated with that location. Type I—Measured Corrosion Rates Consistent with the data needs for reliability analysis and calibration of strength reduction factors for LRFD, the fol- lowing studies were performed: 1. Collection of data on comparison of LPR and weight loss measurements using data available from the existing liter- ature augmented with additional data collected from this project. These data are useful to discern any bias with respect to LPR measurements that should be considered in the reliability analysis. 2. Study of the relationship between corrosion rate and resis- tivity of reinforced fill materials. Resistivity is known to have a significant impact on corrosivity, however, data State Description References California Have been installing inspection elem ents with new construction since 1987, and have been perform ing tensile strength tests on extracted elements. Som e electr oc hem ical testing of in-service reinforcements and coupons has also been performed. LPR and EIS tests were performed on inspection elements at selected sites as part of NC HRP Project 24-28 and results co mp ared with direct physical observations on extracted elements. Jackura et al. (1987), Elias (1990), Coats et al. (1990), Coats et al. (2003- Draft Report) Florida Program focused on evaluating the im pact of saltwater intrusion, including laboratory testing and field studies. Coupons were installed and reinforcements were wired for electrochemical testing and corrosion m onitoring at 10 MSE walls. Monitoring has continued since 1996. Sagues et al. (1998, and 2000), Berke and Saques (2009) Georgia Began evaluating MSE walls in 1979 in response to observations of poor perform ance at one site located in a very aggressive marine environ me nt incorporating an early application of MSE technology. Exhum ed reinforcem ent sa mp les for visual exam ination and laboratory testing. Some in situ corrosion monitoring of in-service reinforcem ents and coupons at 12 selected sites using electrochem ical test techni ques was also performed. McGee (1985), Deaver (1992) Kentucky Developed an inventory and perform ance database for MSE walls. Perfor me d corrosion m onitoring including electroche mica l testing of in-service reinforcem ents and coupons at five selected sites. Beckham et al. (2005) Nevada Condition assessments and corrosion monitoring of three walls at a site with aggressive reinforced fill and site conditions. Exhumed reinforcements for visual examination and laboratory testing; perform ed electrochemical testing on in-service reinforcements and coupons. A total of 12 monitoring stations were dispersed throughout the site providing a very good sample distribution. Fishman et al. (2006) New York Screened inventory and established priorities for condition assessment and corrosion monitoring based on suspect reinforced fills. Two walls with reinforced fill known to meet department specifications for MSE construction are also included in program as a basis for comparison. Corrosion monitoring uses electrochemical tests on coupons and in-service reinforcements. Wheeler (1999, 2000, 2001, 2002a and 2002b) North Carolina Initiated a corrosion evaluation program for MSE structures in 1992. Screened inventory and six walls were selected for electrochemical testing including measurement of half-cell potential and LPR. This initial study included in-service reinforcements, but coupons were not in stalled. Subsequent to the initial study, NCDOT has installed coupons and wired in-service reinforcements for measurement of half-cell potential on MS E walls and em bankments constructed since 1992. LPR testing was also perform ed at approximately 30 sites in cooperation with NCHRP Project 24-28. Medford (1999) Ohio Concerned about the im pact of their highway and bridge deicing program s on the service life of metal reinforcements. Performed laboratory testing on samples of reinforced fill but did not sample reinforcem ents or ma ke in situ corrosion rate m easure me nts. Tim me rm an (1990) Oregon Preliminary study including 1) a review of methods for estimating and measuring deterioration of structural reinforcing elements, 2) a selected history of design specifications and utilization of metallic reinforcements, and 3) listing of MSE walls that can be identified in the ODOT system . Raeburn et al. (2008) Note: EIS = electrochemical impedance spectroscopy.

20 comparing measured corrosion rates to resistivity mea- surements include a lot of scatter. Some of the scatter may be due to spatial and temporal differences between mea- surement of corrosion rate and sampling and testing of reinforced fill materials. However, the study is useful to demarcate threshold levels of resistivity wherein corrosion rates may be significantly affected and to define ranges within which particular metal loss models may apply. 3. Study of the effect of climate/region on measured corrosion rates considering data from different geographic regions associated with different climates, and construction and maintenance practices. The purpose of this study is to fur- ther evaluate if data should be partitioned into regions for the purpose of reliability analysis. 4. Partitioning the data into sites that incorporate reinforced fill materials meeting AASHTO requirements, and consid- ering metal loss or corrosion rate as a function of time. The purpose of this study is to evaluate the robustness of available metal loss models and the probability of exceed- ing metal loss rates used in design. 5. Observation of trends for marginal fills that do not meet AASHTO criteria for reinforced fills. The purpose of this study is to make recommendations on the appropriate parameters for modeling metal loss and the reliability of metal loss estimates for a selected range of resistivity; for example, between 1,000 Ω-cm and 3,000 Ω-cm. Detailed results from these studies are included in Appen- dix D. Data are grouped by quality of reinforced fill, age of sample, and reinforcement type. Figures 7 and 8 summarize the statistics (mean and COV) from these data groups for gal- vanized and plain steel reinforcements, respectively. Rein- forcement type does not appear to have a significant impact on corrosion rates, but lower COVs are realized when data are partitioned into groups defined by reinforcement type. The best results in terms of lower COV are from galvanized reinforcements between 2 and 16 years old, where the COVs range between approximately 30% and 60%. Higher COVs are realized for younger reinforcements (<2 years old) and reinforcements that are older than 16 years. This may be due to variations in the time it takes for the zinc surface to become passivated for younger reinforcements, and the variation of remaining zinc on the surface of older reinforcements. Data are more scattered (i.e., have higher COVs) considering fill materials that do not meet AASHTO requirements (ρmin < 3,000 Ω-cm), and this is may be because, although a low value of ρmin is indicative of the potential for higher corrosion rates, this potential may not be realized if the moisture con- tent is kept low, and moisture content and degree of satura- tion exhibit significant variability. More scatter is evident for plain steel reinforcements. This may be due to the tendency for galvanized surfaces to undergo more uniform corrosion compared to plain steel; also, steel may be more sensitive to changes in environment over the range of conditions for which measurements were obtained. Bias of LPR Measurements Figure 9 depicts observations of corrosion and metal loss with respect to age of the reinforcements for fill conditions meeting the AASHTO criteria described in Table 3. Observa- tions included in Figure 9 are via LPR measurements from sites located in the northeastern, mid-Atlantic, southeastern, Figure 7. Summary of statistics for galvanized reinforcements (*  limited samples).

southwestern, and western United States, and from weight- loss measurements from reinforcements that were exhumed from sites in Europe (Darbin et al., 1988). Since LPR mea- surements render corrosion rate at an instant in time, these data must be extrapolated to estimate metal loss. Metal loss is computed as the product of the measured corrosion times the age of the reinforcement, adjusted for higher corrosion rates assumed to occur during the first 2 years of service. Except for younger reinforcements that are less than 2 years old, it is assumed that 30 μm of zinc per side is lost during the first 2 years, and the measured corrosion rate is considered to be constant thereafter. This assumption is less significant con- sidering older reinforcements. Figure 9 includes approximately 404 data points from LPR measurements and 50 weight-loss measurements. Weight- loss and LPR measurements are not from the same samples, and the samples are from different sites. However, all fills meet electrochemical requirements similar to AASHTO. These data are useful to demonstrate that metal loss extrapolated from LPR measurements are in the same range as those observed directly via weight-loss measurements. Metal losses com- puted from LPR appear to be equal to or higher than those from weight-loss measurements. Thus, the methodology of using LPR measurements to estimate metal loss appears to be conservative (at least for the range of corrosion rates depicted in Figure 9). 21 Figure 8. Summary of statistics for plain steel reinforcements (NA indicates data are not available). 0 50 100 150 200 0 5 10 15 20 25 30 Age of Element, years M et al L os s, μm /si de LPR Measurements Weight Loss Measurements AASHTO Model Figure 9. Comparison of LPR and weight loss measurements for galvanized elements in fill materials that meet AASHTO criteria.

For the purpose of this comparison it is assumed that all samples in Figure 9 are still coated with zinc. Thus, for the AASHTO model, the corrosion rate remains constant after 2 years (4 μm/yr). The AASHTO model appears to be a good upper limit for metal loss throughout the experience period and most of the data points lie well below the envelope described by the AASHTO model (note that many of the data points in Figure 9 overlap one another). Many of these data represent metal loss that is less than half of what is computed with the AASHTO model. This is consistent with the analysis of metal loss and corrosion rate measurements reported by Gladstone et al. (2006). Appendix B includes data for which LPR measurements are directly compared with visual observations. Much of these data are from elements extracted during fieldwork for Task 6 per- formed in cooperation with Caltrans. These data demonstrate that the ratio of maximum metal loss (i.e., loss of tensile strength) to average corrosion rate or metal loss from LPR measurements ranges from 1.2 to 4.8 with a mean of 2.4. This factor appears to be inversely proportional to severity of cor- rosion and tends to range between 2 and 3 when more severe loss of cross section is observed. For galvanized elements, corrosion rates via LPR correlate best with the percentage of zinc remaining on the surface. When more than 70% of the surface is covered by zinc, corrosion rates measured via LPR reflect the rate of zinc loss. However, there may be instances in which localized corrosion of steel may not be reflected in the LPR measurement of corrosion rate. This is more of an issue at sites with relatively poor or marginal qual- ity fill materials where metal loss is less uniform and localized loss of zinc is observed. In general, corrosion rates from LPR measurements are consistent with observations of maximum metal loss considering a factor between 2 and 3 relates the aver- age to the maximum metal loss. This is consistent with the fac- tor of 2 commonly used to relate loss of tensile strength to uniform corrosion losses, as discussed by Elias (1990). Trends Data were analyzed to identify trends from corrosion rate measurements with respect to spatial and temporal variations, climate, environment (marine vs. non-marine), and fill char- acteristics described in terms of electrochemical parameters (ρmin, pH, Cl−, SO4) and organics content. Details of results from data analysis and identification of trends are described in Appendix D and in the interim report that was submitted for the project in April 2007. Spatial Variations Consideration is given to elevation (top vs. bottom) and distance from the wall face (front vs. back). One would expect to observe increased corrosion near the top of the wall and near the wall face due to the potential for infiltration of storm water and relatively higher levels of oxygen within the fill at these locations. However, the majority of the data as described in the interim report indicates that location does not have a significant effect on measured corrosion rates. Data from one site in New York exhibits higher corrosion rates for samples located near the face compared to the backside of the rein- forced fill. Data from several sites in California, where inspec- tion elements were placed along three rows at vertical spacing of 10 feet, suggest that increased corrosion activity may occur near the top of the walls. Given the limited amount of data and lack of a clear trend, spatial variability is considered to be random for the purpose of the reliability analysis and calibra- tion of resistance factor. Temporal Variations The effect of time on corrosion rates is apparent in the data. For galvanized reinforcement and fill materials that meet AASHTO requirements for electrochemical parame- ters, on average, lower corrosion rates are realized from sam- ples with ages between 2 and 16 years compared to those that are younger than 2, or older than 16 years. This is due to the attenuation of corrosion rate with respect to time, and the possibility that higher corrosion rates prevail as zinc is con- sumed from galvanized samples. Although the upper bound of corrosion rate measurements for galvanized reinforce- ments less than 2 years old is close to 15 μm/yr, which is the rate included in the AASHTO model for young (<2 years old) galvanized steel reinforcements, the mean of the mea- surements in this time frame is only about twice as high as measurements obtained after 2 years of service. Higher cor- rosion rates measured after 16 years of service may be due to zinc loss and exposure of base steel; however, the measured corrosion rates are much lower than those measured for plain black steel. Corrosion rates for plain steel attenuate with respect to time, but not as rapidly as those for galvanized elements. This is consistent with corrosion rate models that are based on Equation (1). The Darbin model, Equation (2), applies an exponent of 0.65 to the time factor to describe metal loss of galvanized reinforcements and Elias (1990) applies an expo- nent of 0.8 in Equation (3) to describe metal loss of plain steel elements. A higher exponent reflects a lower attenuation of corrosion rate with respect to time. These temporal variations were considered in the reliability analysis and calibration of resistance factor. Corrosion rates do not necessarily attenuate when fill materials are of marginal quality (i.e., do not meet AASHTO criteria), indicating that a less favorable environment (e.g., high in chlorides) interferes with the formation and suste- nance of a passive film layer. 22

Climate Four regions within the United States were considered including the northeastern, southeastern, high plains and western states. These regions are distinguished by differences in climate, availability of suitable fill materials, use of deicing salts, and prevalence of reinforcement type. For galvanized reinforcements no significant differences were observed. Mean corrosion rates ranged between approximately 1 μm/yr and 2 μm/yr, with slightly higher means observed for the north- eastern and western regions. Because there does not appear to be a significant effect of climate on measured corrosion rates, measurements from different regions are combined to evalu- ate the effects of backfill character, time, and reinforcement type on corrosion rates and observations of metal loss for gal- vanized reinforcements. Thus, data from all the regions were used to generate statistics for galvanized reinforcements used in the reliability analysis and calibration of resistance factor. More significant variations were observed relative to cor- rosion rates for plain steel reinforcements. Mean corrosion rates for plain steel ranged between approximately 3 μm/yr and 20 μm/yr, with much higher corrosion rates observed for the western region. However, climate may not be the only fac- tor, as the western states tend to use fill materials with less favorable electrochemical parameters (higher salt contents) compared to other regions including the Northeast and South- east. These different fill conditions were considered for the reliability analysis and calibration of resistance factors. Thus, the statistics for these different climates were considered sep- arately for plain steel reinforcements. Environmental Conditions (Marine vs. Non-marine) Data from coastal/marine environments, that come from locations near the coast, but are not submerged or in direct contact with saltwater, were separated from non-marine environments. Marine environments did not have a signifi- cant impact on the performance of galvanized reinforcements, however, there was a significant effect observed for plain steel reinforcements. These data demonstrate that plain steel rein- forcements should not be used in marine environments. Effects on corrosion rate from use of deicing salts were not apparent in the data. Effect of Backfill Character The electrochemical properties of reinforced fill have a profound effect on corrosion rates, and resistivity appears to have the strongest influence, although a few data from sites with low pH (pH<4) also exhibit very high corrosion rates. Figure 10 depicts observed corrosion rates from galvanized reinforcements versus measurements of resistivity from sam- ples of fill that are most often taken from stockpiles prior to construction. Figure 10 incorporates 489 data points from 53 sites distributed amongst the states of California, Florida, Kentucky, North Carolina, Nevada, and New York. Reinforce- ment ages range from 1 to 30 years with an average of 13. Therefore, data points in Figure 10 generally depict corrosion rates for zinc, particularly for ρ > 3,000 Ω-cm. Figure 10 depicts scatter that is significantly higher consider- ing lower levels of fill resistivity. This may be due to the variabil- ity of fill conditions at sites that are characterized as having lower quality fill, uncertainty regarding the correlation between sources of samples and fill placed during construction, and the possibility that zinc may be consumed in less than 10 years when ρ < 3,000 Ω-cm. On average, observations from sites with fill resistivities less than 3,000 Ω-cm are approximately an order of magnitude higher than observations from sites with fill resistivity greater than 3,000 Ω-cm. Observations from sites with fill resistivities between 3,000 and 10,000 Ω-cm have average corrosion rates slightly higher than those associated with resistivity greater than 10,000 Ω-cm. Based on these data a power law was regressed to achieve the “best fit” with the data rendering the following equation, which 23 CR = 1400ρ-0.75 R2 = 0.4644 0 10 20 30 40 50 100 1000 10000 100000 ρ (Ω-cm) CR (μ m /y r) Figure 10. Observed corrosion rates versus fill resistivity for galvanized reinforcements (CR  corrosion rate).

is limited to galvanized reinforcements that are less than 20 years old: Table 11 is a summary of corrosion rates computed with Equation (16) for selected resistivities. The corrosion rate computed at ρ = 3,000 Ω-cm is consistent with the AASHTO model for corrosion of zinc after 2 years in service (i.e., 4 μm/yr) and the corrosion rates computed at ρ = 10,000 Ω-cm and 20,000 Ω-cm are consistent with the statistics presented in Figure 7. It also appears that corrosion rates for galvanized reinforce- ments are not necessarily lower than plain steel considering fill materials with low ρmin. This is not surprising because it is well known that zinc does not perform better than steel for all environments. Reliability analyses and resistance factor calibrations were performed on data groups according to selected ranges of fill resistivity, including 1,000 Ω-cm < ρmin < 3,000 Ω-cm; 3,000 Ω-cm ≤ ρmin < 10,000 Ω-cm; and ρmin ≥ 10,000 Ω-cm. Due to the relatively high variability, marginal fills (with 1,000 Ω-cm < ρmin < 3,000 Ω-cm) should be used with extreme caution. Considerably more effort is needed to sample and test these materials to reliably characterize them and select appropri- ate corrosion rates for use in design (Elias et al., 2009). Use of marginal material is not recommended, but guidance is devel- oped to demonstrate the issues and level of effort required to properly manage the risk that is involved when used. Walls with fill material closer to the 3,000 Ω-cm range may become more prevalent depending on whether or not recommenda- tions from NCHRP Project 24-22 are adopted in practice. Metal Loss Models and Reliability AASHTO Model—Galvanized Reinforcements Figures 11(a) and 11(b) compare corrosion rates measured via the LPR technique to the AASHTO metal loss model (see CR ≈ −1400 160 75ρ . ( ) Table 3). Figure 11(a) includes 150 data points and Figure 11(b) includes 257 data points documenting the performances of galvanized reinforcements within good and high quality fills, respectively. The highest rates of corrosion are observed from elements that are equal to or less than 2 years old, but these higher rates are less than half of the rate of 15 μm/yr included in the AASHTO model. The mean corrosion rate during the first 2 years is approximately twice the mean corrosion rate measured from elements older than 2 years. When reinforcements are greater than 2 years old, the means of the observed corrosion rates are less than half of those antic- ipated on the basis of the AASHTO model. Due to the low rate of zinc loss, most of the observations reflect corrosion rates prior to depletion of the zinc coating. However, zinc may have been depleted when corrosion rates, observed from elements more than 16 years old, are greater than 4 μm/yr. This applies to two points each for Figures 11(a) and 11(b) where the aver- age rates of steel loss for good and high quality fill may be inferred as approximately 6 μm/yr and 4 μm/yr, respectively. Considering a factor of 2 to relate observations of average (uniform) metal loss to tensile strength suggests that steel losses of 12 μm/yr and 8 μm/yr may be used to model corro- sion rates for steel after zinc has been depleted from galva- nized reinforcements, considering good and high quality fills, respectively. The AASHTO model is used to compute the nominal metal loss and corresponding sacrificial steel for the calibration of resistance factors when considering galvanized elements in both good and high quality fills. A Monte Carlo analysis was performed to assess the probability that metal loss in excess of the nominal amount may occur (pf). This analysis uses the means and standard deviations of the observations as described in Figure 7. A lognormal distribution was also assumed to describe the variations in measurements, and the validity of this assumption is verified as described in Appendix E. Because the majority of observations reflect corrosion rates for zinc, these measurements are best suited of estimating zinc life. Results from the Monte Carlo analysis render a 99% prob- ability that zinc coating with an initial thickness of 86 μm will last 15 years considering good quality fill, and 32 years considering high quality fill. Thus, good quality fill supports zinc life similar to 16 years as predicted by the AASHTO model for zi = 86μm, and the zinc life appears to be twice as long with high quality fill. The increased zinc life for high qual- ity fill is due to the lower observed corrosion rates evident in Figure 11(b). Steel loss, X, is assumed to commence subsequent to zinc depletion. The mean steel loss is assumed to occur at a rate of 12 μm/yr with a COV of approximately 0.66. Table 12 pre- sents reinforcement ages corresponding to pf equal to 0.01 and 0.05, and the probability that the sacrificial steel will not be consumed for the intended design life (75 or 100 years). 24 ρ ( Ω -cm) CR ( μ m/yr) 1,000 7.9 3,000 3.5 10,000 1.4 20,000 0.8 Table 11. Computed corrosion rates for galvanized reinforcements at selected resistivities.

Based on the results in Table 12 it appears that the proba- bility of sacrificial steel being consumed within design lives of 75 or 100 years is approximately 10% with good quality fill and 1.5% with respect to high quality fill (i.e., probabilities of 90% and 98.5% that design lives may be exceeded with good or high quality fills, respectively). Plain Steel Reinforcements Figures 12(a) and 12(b) compare corrosion rates measured via the LPR technique to the Elias and Stuttgart metal loss models proposed for design. Figure 12(a) includes 53 data points and Figure 12(b) includes 70 data points documenting the performances of plain steel reinforcements within good and high quality fills, respectively. Compared to data for galva- nized reinforcements, the data for plain steel reinforcements are less ambiguous because only the presence of one metal type along these surfaces needs to be considered; whereas either zinc, steel or both may be present along the surfaces of galvanized reinforcements. Measured corrosion rates plotted in Figures 12(a) and 12(b) were multiplied by a factor of 2 to reflect higher rates of localized corrosion inherent to the behavior of buried steel elements. Significant attenuation of mean observed corrosion rates with respect to time is not 25 0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 Age (Years) CR (μ m /y r) Strip Grid AASHTO Mean Strip Grid AASHTO Mean 0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 Age (Years) CR (μ m /y r) Figure 11(b). Corrosion rates vs. time and comparison with the AASHTO model for galvanized elements within high quality fill. Fill Quality t de sign (years ) X ( μ m) p f = 0.01 (years ) p f = 0.05 (years ) p f @ t de sign 75 708 54 69 0.075 Good 100 1,008 65 84 0.116 75 708 75 102 0.010 High 100 1,008 86 118 0.022 Table 12. Occurrence of sacrificial steel consumption for galvanized reinforcements. Figure 11(a). Corrosion rates vs. time and comparison with the AASHTO model for galvanized elements within good quality fill.

observed. However, more scatter is evident in these data com- pared to galvanized reinforcements [Figures 11(a) and 11(b) compared to Figures 12(a) and 12(b)]. The Elias model described by Equation (3), and the Stuttgart model as described in Appendix A, are considered for design of plain steel reinforcements and the calibration of resist- ance factors considering good and high quality fill conditions, respectively. Given the nonlinear form of Equation (3) (Elias model) the differences between the mean of the observed corrosion rates and the Elias model, depicted in Figure 12(a), are inversely proportional to age/design life. Considering a design life of 50 years, the Elias model renders a mean cor- rosion rate (averaged over 50 years) of 37 μm/yr, compared to the observed mean of 25 μm/yr based on measurements obtained from reinforcements with ages spanning 20 years. The mean of observed corrosion rates from reinforcements within high quality fill is similar to the Stuttgart model (12 μm/yr) for plain steel reinforcements that are older than 2 years. Higher rates are used in the Stuttgart model for the first 2 years of ser- vice; however, this is not very important considering a service life of 75 years. A Monte Carlo analysis was performed to assess the prob- ability that metal loss in excess of the nominal amount may occur (pf). This analysis uses the means and standard devia- tions of the observations as described in Figure 8. A lognor- mal distribution was also assumed to describe the variations in measurements and the validity of this assumption is veri- fied as described in Appendix E. Design lives of 50 and 75 years are considered for plain steel reinforcements within good and high quality fills, respectively. Given the uncertainty associated with variations of observed performance, and lack of data from reinforcements older than 20 years, estimations of sacrificial steel requirements for longer service lives are considered dubious. Table 13 summarizes results from the Monte Carlo simulations of service life. Due to the higher variance inherent to the observed performances, probabilities of exceeding estimated metal losses are higher for plain steel reinforcements than for galvanized reinforcements (Table 12 compared to Table 13). This will be reflected in relatively lower calibrated resistance factors used to achieve the same overall probability that MSE designs will meet the intended service life. 26 0 20 40 60 80 100 120 140 CR (μ m /y r) Data Elias mean 0 5 10 15 20 Age (Years) Data Stuttgart mean 0 20 40 60 80 100 120 140 CR (μ m /y r) 0 5 10 15 20 Age (Years) Figure 12(b). Corrosion rates vs. time and comparison with the Stuttgart model for plain steel elements within high quality fill. Figure 12(a). Corrosion rates vs. time and comparison with the Elias model for plain steel elements within good quality fill.

Marginal Quality Fill Figure 13 compares corrosion rates measured via the LPR technique to the metal loss model proposed for design [Jackura et al. (1987) and described subsequently with Equation 17(a)]. These figures include approximately 200 data points docu- menting the performances of galvanized reinforcements within marginal quality fill. Performance data were obtained from 11 sites distributed amongst California, Nevada, New York, and North Carolina. Much higher scatter is evident in these data compared to corrosion rates observed from good and high quality fills. Higher scatter may be attributed to uncertainties with respect to fill resistivity. Samples, corresponding to these 11 sites, were collected from different locations or sources (stockpiles) but the destinations of these fills relative to spe- cific locations within MSE wall constructions are unknown. Furthermore, characteristics including salt content are not homogeneous and can vary spatially with corresponding vari- ations in the related resistivity. Often results from five to 10 resistivity measurements are available and used to represent fill conditions for a particular site. These measurements depict a range with some measure- ments above 3,000 Ω-cm, and some below 1,000 Ω-cm. This is significant because resistivities neighboring 1,000 Ω-cm appear to be a threshold, and substantially higher corrosion rates are realized at resistivities below this threshold. Thus, although a site may be classified as having marginal quality fill, depending on location within the fill and the actual sources used during construction, there may be locations that have resistivity higher than 3,000 Ω-cm, or less than 1,000 Ω-cm. This is reflected in large scatter in the data as depicted in Fig- ure 13 where measured corrosion rates obtained from a par- ticular site on the same day may vary from less than 4 μm/yr to more than 25 μm/yr. Due to the paucity of data between 2 and 10 years of ser- vice, statistics are generated for the first 2 years of service (μ = 2.4 μm/yr and σ = 1.6 μm/yr) and after 10 years of service (μ = 4.6 μm/yr and σ = 6.3 μm/yr). These statistics demonstrate that the corrosion rates for marginal quality fill are approxi- mately two to three times higher than those observed from good quality fills. A Monte Carlo simulation was performed to estimate zinc life assuming a lognormal distribution of corrosion rates. There are no data from reinforcements between the ages of 2 and 10 years so the statistics from reinforcements less than or equal to 2 years old are assumed to apply until the reinforce- ments have been in service for 10 years. Results from the Monte Carlo analysis render a 99% probability that zinc coating with an initial thickness of 86 μm will last 10 years considering marginal quality fill. This compares with 16 years and 32 years for galvanized reinforcements within good and high quality fills, respectively. Thus, the use of marginal quality fills appears to have a significant effect on zinc life, and zinc life is approx- imately 60% of that expected with good quality fills. 27 Fill Quality t de sign (years) X (μm) pf = 0.01 (years) pf = 0.05 (years) pf @ tdesign Good 50 1,829 25 35 0.16 High 75 1,036 20 33 0.31 Table 13. Occurrence of sacrificial steel consumption for plain steel reinforcements. 0 10 20 30 40 50 60 70 CR (μ m /y r) Strip Grid Model Mean 0 5 10 15 20 25 30 Age (Years) Figure 13. Corrosion rates vs. time and comparison with the Jackura model for galvanized elements within marginal quality fill.

Due to the high scatter inherent to these data and uncer- tainties with respect to fill properties, conservative assump- tions are made regarding zinc life and corrosion rates of base steel subsequent to zinc deletion. For the purpose of estimat- ing service life, zinc life is assumed to be constant and equal to 10 years (with zi = 86 μm) and the observed corrosion rate of steel subsequent to zinc depletion is taken as 32 μm/yr with standard deviation of 21 μm/yr and a lognormal distri- bution. The corrosion rate of steel is based on observations from galvanized reinforcements after 8 years of service with the lower corrosion rates (i.e., < 4 μm/yr) culled from the data as described in Appendix E. The model used to compute nominal sacrificial steel require- ments for design is similar to the recommendations described by Jackura et al. (1987) for “neutral” fill with ρ > 1,000 Ω-cm and salt contents limited as described in Table 2 (Caltrans- Interim model). This model assumes that steel is exposed on the surface of galvanized reinforcements (zi = 86 μm) after 10 years of service and that the base steel will corrode at an average rate of 28 μm/yr subsequent to zinc depletion. Two different metal loss models are studied to illustrate how this affects the reliability of service life estimates. The first model (Model I) is from Jackura et al. (1987) for “neu- tral” fill and the second model (Model II) is a similar form, but with double the corrosion rate for steel as follows: A Monte Carlo analysis was performed to assess the prob- ability that metal loss in excess of the nominal amount may occur (pf). This analysis uses the means and standard devia- tions of the observations as described in the preceding para- graphs and as depicted in Figure 7. A lognormal distribution was also assumed to describe the variations in measurements and the validity of this assumption is verified as described in Appendix E. Design lives of 50 years are considered for galva- nized steel reinforcements within marginal quality fills. Given the uncertainty associated with variations of observed per- formance, and lack of data from reinforcements older than 25 years, estimations of sacrificial steel requirements for longer service lives are considered dubious. Table 14 summarizes Model II m m : (X t years year designμ μ( ) = −( ) ×10 56 17b) Model I m m : (X t years year designμ μ( ) = −( ) ×10 28 17a) results from the Monte Carlo simulations of service life when nominal sacrificial steel requirements are estimated with Models I and II. Table 14 shows that the probabilities of sacrificial steel con- sumption are significantly affected by the nominally com- puted sacrificial steel requirements (i.e., the amount of sacrificial steel estimated for a 50-year design life according to Model I or Model II). In principal, different resistance factors computed with different nominal models should offset the differences in nominal sacrificial steel requirements, render- ing similar design as long as the COVs of the different bias distributions are also similar. Resistance factors will be cali- brated in the next section that will render the probability that reinforcements will be overstressed during their design life to be similar, independent of the metal loss model that is selected (i.e., resistance factors may be calibrated for each model to render the same pf). The effect of the different models on steel requirements is illustrated in the design example presented in Appendix F. Calibration of Resistance Factors Galvanized Reinforcements Data included in Appendix D include observations from galvanized reinforcements and coupons, and from plain steel (i.e., not galvanized) elements. In-service reinforcements and coupons are placed in the same fill conditions but have very different dimensions, and coupons may be placed at both front and back locations with respect to the wall face. Data from in-service reinforcements and coupons were compared, and, on the basis of this comparison, the decision was made to include them in one data set, thus enhancing the quantity of data within each partition. Metal loss is considered in the resistance factor calibration where the bias of remaining strength (i.e., ratio of measure- ments to nominal value used in design) is computed as Equa- tion (15). The nominal remaining strength used in design and in the denominator of Equation (15) is computed as described in Equations (10) through (14) with values of rz1, rz2, and rs from the metal loss model recommended by AASHTO for assessing metal loss of galvanized reinforcements, and described in Tables 2 and 3. Since the oldest MSE walls are approximately 40 years old, direct measurements of remaining strength after 28 Design Model t de sign (years ) X ( μ m) p f = 0.01 (years ) p f = 0.05 (years ) p f @ tdesign Model I 50 1,120 18 24 0.44 Model II 50 2,240 28 40 0.11 Table 14. Occurrence of sacrificial steel consumption for galvanized steel reinforcements in marginal quality fill.

a service life of 75 or 100 years are not available. Therefore cor- rosion rate measurements must be extrapolated to estimate “measurements” of remaining strength used in the numera- tor of Equation (15). The extrapolation also employs equa- tions similar to Equations (10) through (14), but with corrosion rates rz1, rz2, and rs from the observed performance of reinforcements during service. This approximation is con- sidered conservative due to the likely attenuation of corrosion rate with respect to time. The corrosion rates used to extrap- olate metal loss are considered constants over prescribed time intervals, and are higher than those expected to prevail at the end of service. Figures 14 and 15 are examples of inputs and intermediate results from the calibration exercise. Tables 15 to 18 summa- rize the final results from the calibration. The following list describes the steps involved in the calibration process and gen- eration of resistance factors using the Monte Carlo Technique: a) Generate statistics from observations for corrosion rates including the mean (μ), standard deviation (σ), and the shape of the probability density function (pdf). It is impor- tant to select the correct shape of the pdf to represent the data. Probability plots similar to the one depicted in Fig- ure 14 are used to check the match between the empirical 29 -3 -2 -1 0 1 2 3 0.10 1.00 10.00 Corrosion Rate, μm/yr Z Figure 14. Probability plot depicting lognormal distribution for data describing corrosion rates of galvanized reinforcements within good quality fill. 0 0.2 0.4 0.6 0.8 1 0 0.5 1.51 2 2.5 Bias Cu m ul at iv e Pr ob ab ili ty Data Weibull Figure 15. Typical plot showing Weibull distribution for bias; galvanized strip reinforcement with S  4 mm, tf 75 years;   1.5,   1.6 corresponding to mean  1.35 and standard deviation  0.42. f x e x( )⎛⎝⎜⎜ ⎞ ⎠⎟⎟ ⎛ ⎝⎜ ⎞ ⎠⎟     1

Good Quality Fill Good quality fill meets AASHTO requirements for electro- chemical and mechanical properties, and has ρmin in the range of 3,000 Ω-cm and 10,000 Ω-cm. The statistics for reinforce- ments that are between 2 and 16 years, shown in Figure 7, are considered representative of the life of the zinc coating. Thus, the corrosion rate for zinc is assumed to be constant with respect to time with a mean rate of 1.7 μm/yr (rz1 and rz2) and standard deviation of 1.09 μm/yr. The distribution is mod- eled as lognormal based on the probability plot depicted in Figure 14. The data shown in Figure 14 plot close to a straight line with a coefficient of correlation, R2 = 0.96. Probability plots, similar to Figure 14, depicting the distributions used for other corrosion rate measurements described in this report are included in Appendix E. Given the average rate of zinc loss (1.7 μm/yr), and since measurements were made on reinforcements that are less 30 Reinforcement Type Design Life Reinforcement Thickness Simple/Coherent 4 mm 0.45/0.35 5 mm 0.45/0.35 75 years 6 mm 0.50/0.40 4 mm 0.25/0.15 5 mm 0.30/0.20 Strip 100 years 6 mm 0.25/0.20 φ R z i ( μ m) 86 1.35 0.42 0.35 150 1.54 0.26 0.65 λ μ σ φ Reinforcement Type Design Life Reinforcement Thickness/Size Simple/Coherent 4 mm 0.70/0.65 5 mm 0.65/0.55 75 years 6 mm 0.65/0.55 4 mm 0.55/0.50 5 mm 0.60/0.50 Strip 100 years 6 mm 0.65/0.50 W7 0.60/0.50 W9 0.60/0.50 W11 0.60/0.50 75 years W14 0.55/0.50 W7 0.55/0.45 W9 0.55/0.45 W11 0.55/0.45 Grid 100 years W14 0.55/0.45 φ Reinforcement Type Design Life Reinforcement Thickness/Size Simple/Coherent 4 mm 0.85/0.70 5 mm 0.75/0.65 75 years 6 mm 0.70/0.60 4 mm 1.0/0.85 5 mm 0.85/0.70 Strip 100 years 6 mm 0.75/0.65 W7 0.75/0.65 W9 0.70/0.60 W11 0.65/0.55 75 years W14 0.65/0.55 W7 0.90/0.75 W9 0.80/0.70 W11 0.80/0.65 Grid 100 years W14 0.75/0.60 φ Table 15. Summary of  for conservative steel model. Table 16. Effect of zi on computed ; S  4 mm, tf  75 years. Table 17. Summary of  for the zinc residual steel model. Table 18. Summary of  considering high quality fill. data frequencies and the theoretical pdf. Probability grids similar to Figure 14 are generated for each variable used to describe corrosion rates and metal loss. In most cases log- normal distributions were found to fit well with the observed corrosion rates. b) Extrapolate metal loss to the end of the selected service life using the statistics of observed corrosion rates and corre- sponding assumptions regarding the trends of corrosion rates with respect to time. c) Compute the remaining tensile strength, Trem, and the sta- tistics of the resistance bias, λR, via Monte Carlo simula- tions. The distribution of λR is modeled with a pdf. The bias distributions were modeled with either normal, log- normal, or Weibull distributions. Probability plots simi- lar to the one depicted in Figure 15 are prepared to check the match between the empirical data frequencies and the theoretical pdf. d) Compute β and corresponding pf for an assumed value of φ. e) Iterate on φ to converge to the desired target reliability index, βT. Tables 15 to 18 summarize the resistance factors, φ, that converge to βT for the different cases considered (e.g., galvanized reinforcements in good or high quality fill). As shown in Figure 7 and discussed in the previous section, the statistics of corrosion rate measurements are different for fill materials that are considered good enough to meet AASHTO electrochemical requirements (good fill), and those that exceed AASHTO requirements by a wide margin (high quality fill). Therefore, resistance factors are calibrated with respect to fill quality (i.e., good fill and high quality fill).

than 30 years old, very few measurements are available to describe the corrosion of steel after zinc has been consumed from a galvanized reinforcement. Two different assumptions are applied as described by Elias (1990) that either (1) con- sider the base steel to corrode at the same rate as plain black steel (i.e., not galvanized) or (2) assume that the base steel will corrode at a rate similar to that prevailing as zinc is finally consumed (i.e., corrosion rate does not change abruptly after zinc is consumed). In addition, “measured” corrosion rates for steel were multiplied by 2 to render loss of tensile strength from LPR measurements. A conservative model for steel consumption assumes that the base steel corrodes at the same rate as plain steel (i.e., not galvanized) after the sacrificial zinc layer is consumed. Most of the data used for corrosion rates of plain steel embedded in fill materials meeting current AASHTO guidelines are from plain steel coupons installed at MSE sites located in California, New York, and Florida. The statistics of this data set render a mean corrosion rate and standard deviation of 27 μm/yr and 18 μm/yr, respectively; and the distribution can be approxi- mated as lognormal. A resistance bias is computed for different sizes of strip- type reinforcements (4 mm, 5 mm, and 6 mm) and both 75- and 100-year service lives. The bias tends to decrease with respect to increase in reinforcement size, and is higher con- sidering longer service life. The mean resistance bias, λR, ranges between 1.2 and 1.5 with COV approximately 40% and a distribution that is approximated as a Weibull distri- bution (Vardeman, 1994). Figure 15 is a typical plot showing the distribution of the computed bias. Resistance factors are calibrated using the computed sta- tistics for resistance bias and load bias from the literature. Table 15 summarizes the results of the resistance factor cal- ibration applicable to the conservative steel loss estimate. The resistance factors do not vary significantly with respect to reinforcement size but are lower when considering longer service life. Resistance factors of approximately 0.45 and 0.25 apply to 75- and 100-year service lives, respectively. Resis- tance factors calibrated using the coherent gravity model are slightly lower. The efficiency factor is approximately 0.38 for a design life of 75 years and 0.2 for a design life of 100 years. Additional calibrations were performed to investigate the effect of initial zinc thickness on the computed resistance fac- tors. Table 16 compares results obtained with zi equal to 86 μm and 150 μm per side. The comparison considers 4-mm-thick galvanized strip reinforcements, a design life of 75 years, and the same statistics for metal loss (i.e., zinc and steel) as described for galvanized reinforcements with the conserva- tive steel model. The load bias used in the calibration cor- responds to the coherent gravity method. In each case the computed resistance bias has a Weibull distribution simi- lar to that shown in Figure 15. The computed resistance fac- tor for zi = 150 μm is 0.65 and compared to the case with zi = 86 μm, this result is closer to the current AASHTO specifica- tions (φ = 0.75). In this case zi has a significant effect on the computed φ, which demonstrates that zinc thickness is an important variable to include in resistance factor calibrations. However, data on initial zinc thickness are needed to prop- erly characterize the inherent variation and to incorporate the statistics into a reliability analysis. Use of zi = 86 μm corre- sponds to the minimum requirement and is a conservative approach to modeling initial zinc thickness. The zinc residual model for steel consumption considers that the corrosion rate of the base steel is affected by the pres- ence of zinc residuals. Zinc residuals passivate the steel sur- face and include a zinc oxide film layer adhered to the metal surface and zinc oxides within the pore spaces of the sur- rounding fill. There are very few measurements describing corrosion rates of base steel after zinc has been consumed. A few observations may be applicable from the data set col- lected in Europe (Darbin et al., 1988) wherein zinc is con- sumed relatively rapidly (i.e., within a few years) and from measurements made on walls in the United States that are approaching 30 years of age. Rapid zinc consumption from some of the earlier sites in Europe is due to a relatively thin zinc coating (zi = 30 μm) and moderately corrosive reinforced fill materials. A review of these data renders corrosion rates for steel that are close to 12 μm/yr. This is the metal loss model recommended by AASHTO and is adopted as a basis for comparison with calibrations performed by extrapolating measured corrosion rates with the conservative steel model. Similar to other data sets, a COV of 60% and a lognormal dis- tribution is used to describe the variation. The calibration was performed for both strip- and grid-type reinforcements. The mean of the resistance bias is approxi- mately 1.4 with COV approximately 20% and a distribution that is approximately normal. Table 17 is a summary of the resistance factors calibrated with metal loss measurements extrapolated with the zinc residual model for steel loss. These resistance factors are significantly higher than those obtained with the conservative steel model (Table 15) and are in the range of 0.60 to 0.70 for strip-type reinforcements, and 0.50 to 0.60 for grids. The efficiency ratio for this case is approxi- mately 0.5 and represents an improvement compared to the case in which metal loss measurements are extrapolated via the conservative steel model. High Quality Reinforced Fill (ρ > 10,000 Ω-cm) High quality reinforced fills have ρmin > 10,000 Ω-cm and corrosion rates corresponding to these conditions were observed from sites in Florida (Sagues et al., 1998; Berke and Sagues, 2009) and North Carolina. These data render mean and standard deviation of corrosion rates for the zinc coating 31

of 0.8 μm/yr and 0.5 μm/yr, respectively, for strip-type rein- forcements, and 0.5 μm/yr and 0.2 μm/yr, respectively for grid-type reinforcements. Corrosion rates observed from plain steel coupons older than 16 years correspond to mean and standard deviation values of 11.5 μm/yr and 9.4 μm/yr, and these parameters are used to represent the loss of base steel subsequent to depletion of the zinc coating for this case. Both of these distributions are modeled as lognormal. The mean of the corresponding resistance bias is computed as ranging from 1.4 to 2.0 with COV approximately 10%. The bias distribution is approximately normal considering a 75-year service life, but is better represented by a Weibull distribution considering a 100-year service life. Table 18 is a summary of the resistance factors calibrated with metal loss measurements from sites with high quality reinforced fill. These resistance factors are equal to or higher than those currently specified by AASHTO (see Table 7). The efficiency ratio for this case is approximately 0.5. Changing the initial zinc thickness from 86 μm to 150 μm per side did not have a dramatic effect on the computed resistance factors compared to the case with good quality backfill and the conservative steel model. This is because the resistance factors computed with zi = 86 μm considering high quality fill are closer to one. The resistance factors summarized in Tables 17 and 18 correspond to the target reliability index (βT = 2.3 corre- sponding to pf = 0.01). However, other values of φ corre- sponding to different levels of reliability are also of interest. Table 19 compares the relationship between resistance fac- tors and reliability in terms of β and pf, for different scenar- ios involving good or high quality fill, and strip or grid-type reinforcements. Table 19 considers typical galvanized strip reinforcements with S = 4 mm and grids with W11 longitu- dinal wires. On the basis of data depicted in Table 19, alterna- tive approaches may be contemplated for selecting resistance factors for design rather than calibrating to achieve a target reliability index. For example, consider applying the resistance factors as they stand in the current version of the AASHTO specifica- tions (AASHTO, 2009), where φ = 0.75 and 0.65 for strip- and grid-type reinforcements, respectively. Thus, for designs with strip-type reinforcements the probability (pf) that stress in excess of yield will occur before the end of the intended design life is 0.005 and 0.015, respectively for construction employing high quality and good quality fill. Similarly, for designs with grid-type reinforcements, pf would correspond to 0.008 and 0.018. Thus, MSE walls designed in accordance with current AASHTO specifications, and constructed with high quality fills, have a more favorable pf compared to the target of 0.01. Based on the statistics of the current inventory described by AMSE (2006), this exceptional performance applies to approximately 80% of MSE walls in the existing inventory. The remaining 20%, constructed with good qual- ity fill, are associated with a lower level of performance, and a pf that is nearly twice the target valued of 0.01. Verification of Monte Carlo Analysis Results from the Monte Carlo simulations used to cali- brate resistance factors are verified via comparison incorpo- rating alternative formulations for computing resistance bias and closed-form solutions for reliability index. Although the closed-form solutions are limited to particular distributions of the bias variables, they render estimates for comparison and illustrate the effect of incorporating more realistic distri- butions via Monte Carlo simulations. In general, the verifica- tion study is performed as follows: Step 1. Select a design life and compute the distribution of metal loss using the service life model described by Sagues et al. (2000). Step 2. Compute the bias of the remaining cross-sectional area, λAc, as the ratio of remaining cross section based 32 Table 19. Comparison of relationship between  and  for different fill quality. Strip Reinforcements ( S = 4 mm) Grid Reinforcements (W11) High Quality Fill Good Quality Fill High Quality Fill Good Quality Fill p f p f p f p f 0.55 3.12 0.001 2.73 0.003 2.99 0.0014 2.49 0.006 0.60 2.97 0.0015 2.64 0.004 2.63 0.004 2.24 0.012 0.65 2.93 0.002 2.42 0.008 2.41 0.008 2.09 0.018 0.70 2.72 0.003 2.30 0.010 2.13 0.016 1.93 0.026 0.75 2.56 0.005 2.17 0.015 2.11 0.017 1.79 0.036 0.80 2.48 0.006 2.04 0.020 1.96 0.025 1.69 0.045 0.85 2.27 0.011 1.91 0.028 1.80 0.035 1.52 0.063 0.90 2.19 0.014 1.80 0.036 1.67 0.047 1.39 0.082 φ calibrated for target reliability index βT = 2.3 corresponding to pf = 0.01. φ based on current AASHTO specifications (AASHTO, 2009). φ β β β β

on metal loss measurements to the remaining cross section based on nominal metal loss used in design. Step 3. Compute the bias of the remaining tensile strength, λR, as the product of the random variables including the bias of remaining cross section determined in Step 2, and the bias for yield strength. Step 4. Compute β as a function of resistance factor using the bias of remaining tensile strength determined in Step 3 and available closed-form solutions. Sagues Formulation Sagues et al. (2000) formulated a probabilistic deterioration model for service life forecasting of galvanized soil reinforce- ments. This formulation is based on the following assump- tions, which are the same as those employed in the Monte Carlo simulations: • The distribution of corrosion loss over all elements in the structure mirrors the overall distribution of corrosion measured in the field; • During the early life of the structure, the corrosion rate dis- tribution reflects that of the galvanized elements; • Loss of base steel is initiated after the zinc coating is consumed; • The highest rate of metal loss takes place in the region of maximum reinforcement stress, and the service life of a given element is over when the sacrificial steel in the high- est stressed region is consumed; and • Corrosion rates are constant with time. The resulting formula to compute the probability that metal loss, X, exceeds a given threshold, X′, is given by where P is probability; X is loss of steel defined by tf, zi, rz, σz, rs, σs; X ′ is a given amount of steel loss; tf is service life; zi is the initial zinc thickness; rz is the mean zinc corrosion rate; σz is the standard deviation of zinc corrosion rate; rs is the mean steel corrosion rate; σs is the standard deviation of steel corrosion rate; r0 = zi/tf and is the lowest rate of zinc corrosion for which base steel will be consumed within tf; fz(rz) is the pdf representing zinc corrosion rates, rz; and P X X t z r r f r F X t z f i z z s s z z s f > ′[ ] = ( ) − ′( ) − , , , ,,σ σ 1 i z z r r dr( )( )( )∞∫ 0 18( ) Fs is the cumulative density function (cdf) representing steel corrosion rates. Equation (18) was programmed into an Excel spreadsheet and the integration performed numerically. The numerical integration was performed in increments between 0.1 and 0.01 times r0. In most cases convergence to within E-06 was achieved within 100 increments. The numerical integration was performed for a range of X and the corresponding prob- abilities of exceedance computed for a given service life. For each value of X the bias of the remaining cross section (strip- type reinforcement) is computed as wherein the AASHTO metal loss model (Eq. 5) is used in the denominator to compute nominal remaining cross section. A mean and standard deviation were determined from the dis- tribution of the computed bias to describe the variation of λAc. The bias of the remaining tensile strength was then com- puted as: where λFy is the bias of the yield stress. Assuming that λFy and λAc are uncorrelated and their statistics are known, the mean and standard deviation of λR could be computed using well known relationships between functions of random variables as described by Baecher and Christian (2003). Closed-Form Solutions for Reliability Index For a specific limit state and a single load source, the reliabil- ity index (β) and the resistance factor (φ) can be related using the following formula (Allen et al., 2005), which assumes that the load and resistance bias both have normal distributions: where β = reliability index (dimensionless), γQ = load factor (dimensionless), φR = resistance factor (dimensionless), λQ = mean of load bias (dimensionless), λR = mean of resistance bias (dimensionless), COVQ = coefficient of variation of load bias (dimension- less), and COVR = coefficient of variation of resistance bias (dimen- sionless). β γ φ λ λ γ φ λ = ⎛ ⎝⎜ ⎞ ⎠⎟ − ⎛ ⎝⎜ ⎞ ⎠⎟ ⎛ ⎝⎜ ⎞ ⎠⎟ + Q R R Q R Q R RCOV CO 2 V 2 Q Qλ( ) ( )21 λ λ λR Ac Fy= × ( )20 λAc f S X S t C = − ×( ) − × × −( )[ ] 2 2 12 19( ) 33

In the case of lognormal distributions for load and resis- tance bias: For a given load factor and known load and resistance sta- tistics, Equations (21) and (22) are satisfied for selected val- ues of resistance factors, rendering related pairs of reliability indices and resistance factors. From the computed pairs of β versus φR, resistance factors can be selected corresponding to the targeted level of reliability. Table 20 presents selected results and compares resistance factors computed via Monte Carlo simulations to those computed via the Sagues service life model and closed-form solutions for β. The load bias used in these analyses refers to the coherent gravity method and is a lognormal distribution (D’Appolonia, 2007). A Weibull or normal distribution is used to describe the variation of λR. Results are presented for two cases: (1) where the probability density function (pdf) for λR used in the Monte Carlo simula- tion is normal, and (2) where the pdf is described with a Weibull function. Because the closed-form solutions only consider probability density functions to be normal or lognormal this comparison demonstrates the importance of properly captur- ing the distribution of the pdf in the analysis, and the need for numerical simulations (i.e., Monte Carlo simulations). Table 20 demonstrates that when λR is normally distributed, the compar- isons between the Monte Carlo simulations and the closed- form solutions are very good. Note that since the distribution of load bias is lognormal, the closed-form solution, assuming a normal distribution for both λR and λQ, does not always give the best results, even when λR is normally distributed. Alternatively, when λR is described with the Weibull func- tion, the closed-form solutions, which consider the distri- β γ λ φ λ = +( ) +( )⎡⎣⎢ ⎤ ⎦⎥ + ln ln Q R R Q Q R1 1 1 2 2COV COV COVQ R2 21 22( ) +( )[ ]COV ( ) bution to either be normal or lognormal, do not necessarily compare very well with the results from the more robust Monte Carlo simulations. Therefore, numerical analyses (e.g., Monte Carlo simulations) are necessary to properly model the distribution of λR. Plain Steel Reinforcements Resistance factor calibrations are performed considering the use of plain steel (i.e., not galvanized) reinforcements. The purpose of these calibrations is to define design parame- ters that are appropriate for plain steel reinforcements and demonstrate the advantage of using galvanized reinforce- ments. Since the use of plain steel reinforcements has been limited, most of the data from the performance of plain steel reinforcements are from coupons placed at sites where galva- nized in-service reinforcements are employed. The few exam- ples where plain steel reinforcements have been used in the United States are for grid-type reinforcements, as plain steel strip-type reinforcements are not readily available. Therefore, the calibration is performed for grid-type reinforcements. The calibration is performed considering reinforced fill quality that meets AASHTO criteria for electrochemical properties, and both good and high quality fill are considered. Good Quality Fill Based on the summary of statistics from corrosion rate mea- surements depicted in Figure 8, a mean corrosion rate and standard deviation of 25 μm/yr and 14 μm/yr, respectively, represent the statistics for plain steel grid-type reinforcements within good quality fill, and the distribution can be approxi- mated as lognormal. The resistance bias is computed for dif- ferent sizes of grid-type reinforcements (W7, W9, W11, and W14). The AASHTO metal loss model only applies to galva- 34 Table 20. Summary of comparison between closed-form solutions and Monte Carlo simulations.1 Closed-Form Monte Carlo Normal Lognormal Reinforced Fill Quality w/Details of Steel Loss Model Life (yrs) Thickness (mm) pdf R Good w/Zinc Residuals 75 4 Normal 0.60 0.55 0.60 Good w/Conservative Steel 75 15 Normal 0.45 0.50 0.50 High Quality Fill 75 5 Normal 0.65 0.55 0.65 Good w/Conservative Steel 75 4 Weibull 0.35 NA 2 0.25 Good w/Conservative Steel 75 6 Weibull 0.40 0.10 0.35 High Quality Fill 100 4 Weibull 0.85 NA 2 0.55 1 Coherent gravity method applied to galvanized strip type reinforcements with assumed initial zinc thickness of 86 μ m. T =2.3 was used to compute . 2 NA m eans a result is not available because β > 2.3 could not be achieved using the closed-form solution. β φ λ φ φ φ

nized reinforcements, therefore, the nominal metal loss model used in the denominator of Equation (15) is based on data col- lected by the National Bureau of Standards for plain steel in fill materials similar to those typically used in the construction of MSE and described by Eq. (3). The analysis is limited to a 50-year service life since the sacrificial steel requirements considering 75- and 100-year service lives are considered to be impractical. Thus, a shorter service life is considered appropriate when using plain steel as opposed to galvanized reinforcements. The mean of the resistance bias, λR, tends to decrease with respect to increase in reinforcement size and ranges between 1.4 and 1.9 with COV between 30% and 40%, and a distribution that is approximately normal. Table 21 summarizes the results of the resistance factor cal- ibration. The resistance factors tend to increase with respect to reinforcement size and are approximately 0.1 to 0.15 lower than those computed for galvanized reinforcements with the conservative steel model and longer service lives as depicted in Table 15. The efficiency ratio (φ/λR) for this case is approx- imately 0.2, which is also lower than the efficiency ratio com- puted for galvanized reinforcements High Quality Fill (ρ > 10,000 Ω-cm) Based on the summary of statistics from corrosion rate measurements depicted in Figure 8, a mean corrosion rate and standard deviation of 12 μm/yr and 9.6 μm/yr, respectively, represent the statistics for plain steel grid-type reinforcements within high quality fill, and the distribution can be approxi- mated as lognormal. The resistance bias is computed for differ- ent sizes of grid-type reinforcements (W7, W9, W11, and W14). For this case, the nominal metal loss model used in the denominator of Eq. (15) is based on the Caltrans-Select model (Jackura et al., 1987) described in Table 2, corresponding to rs = 13 μm/yr. Given the more favorable sacrificial steel require- ments compared to the previous case, the analysis considers service lives of 75 years. The mean of the resistance bias, λR, tends to decrease with respect to increase in reinforcement size and ranges between 1.1 and 1.2 with COV between 30% and 35%, and a distribution that is approximately normal. Table 22 summarizes the results of the resistance factor calibration. The resistance factors tend to increase with respect to reinforcement size and range between 0.25 and 0.35. The efficiency ratio (φ/λR) for this case is approxi- mately 0.25. Marginal Fill Quality Resistance factors are calibrated considering the use of fill that does not meet AASHTO criteria for electrochemical parameters as described in Table 3. Fill with pH in the range of five to seven, but with ρmin between 1,000 Ω-cm and 3,000 Ω-cm is referred to as marginal quality fill. This calibration is performed considering the use of galvanized reinforce- ments and a 50-year service life. Based on the analysis of the observed corrosion rates for marginal fill, and the paucity of data for reinforcements less than 10 years old, extrapolations of metal loss assume that the zinc coating will survive 10 years. Corrosion rate measure- ments are available from six sites located in California that appear to reflect corrosion rates of base steel subsequent to depletion of the zinc coating. A mean corrosion rate and stan- dard deviation of 32 μm/yr and 21 μm/yr, respectively, and a lognormal distribution are used to describe the statistics of these measurements. These statistics appear to be conserva- tive compared to corrosion rates observed from plain steel elements that are more than 10 years old at the time of mea- surement as described in Appendix E. Computations of resistance bias and corresponding calibra- tions of resistance factors are performed considering nominal requirements for sacrificial steel computed with Models I and II as described by Equations (17a) and (17b). These calcula- tions consider zi = 86 μm, a design life (tdesign) equal to 50 years, and grid reinforcements with W20 size longitudinal wires. Results from these computations are presented in Table 23 in terms of the statistics of the resistance bias and corresponding calibrations of resistance factors. As expected, the bias associated with Model I is less than Model II, but the COVs are nearly the same. Due to the differ- ences in the bias, the resistance factor calibrated for Model I is also less than that associated with Model II. However, because the COVs of the bias are similar the calibrated resistance fac- tors render the same design efficiency, φ/λR, for each case. Sim- ilar design efficiencies result in similar design details for a given MSE geometry, load case, design life, and so on. An example problem is presented in Appendix F that demonstrates that 35 Reinforcement Type Design Life Reinforcemen t Thickness/Size Simple/Coherent W7 0.25/0.20 W9 0.30/0.25 W11 0.35/0.25 Grid 50 years W14 0.40/0.35 φ Table 21. Summary of  considering plain steel reinforcements and good quality fill material. Reinforcement Type Design Life Reinforcemen t Thickness/Size φ Simple/Coherent W7 0.20/0.20 W9 0.30/0.20 W11 0.35/0.25 Grid 75 years W14 0.35/0.30 Table 22. Summary of  considering plain steel reinforcements and high quality fill material.

this is indeed the case. The results from this exercise demon- strate how changing the metal loss model used in design will not effect a change in design if the resistance factors are prop- erly calibrated. The best way to achieve a more efficient design is to improve the COV of the bias. This may be achieved by using models that do a better job of capturing the behavior (e.g., capture trends that may be related to space, time, fill characteristics, and site conditions) and by improving the quality and quantity of performance data. Type II—Condition Assessment For Type II reinforcements, installation details have an effect on the vulnerability of the system to the surrounding environment and corresponding susceptibility to corrosion, and on our ability to probe the elements and interpret data from NDT. Relevant details include steel type, corrosion pro- tection measures, drill hole dimensions, bond length, free/ stressing length, total length, date of installation, level of pre- stress, grout type, and use of couplings. For rock bolts, the grout surrounding the reinforcement is often the only corro- sion protection afforded to the reinforcements. More complex installation details are incorporated into ground anchorages that include elaborate corrosion protection measures, as described by PTI (2004). Construction details, durability of different material components, and workmanship associated with the corrosion protection system affect the service life and durability of ground anchorages. Generally speaking, rock bolts are more susceptible to metal loss from corrosion com- pared to ground anchorages. For these reasons, results from condition assessment and analysis of data relative to rock bolt and ground anchor installations are distinct. Table 24 is a summary of sites with Type II reinforcements that were included in the fieldwork conducted as part of Task 6, and where measurements of corrosion rate and information on the condition of the reinforcements were obtained. Six of the installations described in Table 24 are rock bolts, and three are ground anchorages. Reinforcement age ranges from 8 to 43 years when monitoring was conducted (2007–2008). A variety of site conditions prevail, but in general, the sites provide an environment that is slightly acidic with pH ranging between four and six, and fairly conductive with resistivities less than 10,000 Ω-cm. One exception to this is the National Institute for Occupational Safety and Health (NIOSH) Safety Research Coal Mine (SRCM), which presents a more aggres- sive environment relative to corrosion. Grout type is an especially important detail as the extent and type of grout surrounding an element affects the vulner- ability of the system to corrosion. Portland cement-based grout is alkaline and protects the steel reinforcement by pas- sivating the surface as well as providing a barrier to moisture and oxygen. Half-cell potential measurements, depicted in Figure 16, are useful to assess if the steel surface is passi- vated, or if corrosion is occurring. The alkalinity of the port- land cement grout tends to shift the potential at the surface of a steel reinforcement to a more positive value. A half-cell potential greater than −200 mV relative to a CSE indicates the surface of the steel reinforcement is passivated. Figure 16 depicts the means and ranges of half-cell potential measure- ments from sites listed in Table 24. In general, sites with resin- grouted reinforcements exhibit half-cell potentials less than −200 mV, which on average range between approximately −400 mV and −700 mV. Reinforcements surrounded with portland cement grout exhibit half-cell potentials greater than −200 mV (maximum values). Although there are some notable exceptions, in general, these data serve to demonstrate the effectiveness of portland cement grout to protect steel earth reinforcements. The best demonstration of the effectiveness of portland cement grout to passivate the steel reinforcements is with respect to the dam tie-downs wherein the steel wires are sur- rounded by portland cement grout within a concrete gravity dam. The fully grouted steel bar tendons at the Barron Moun- tain Rock Cut are generally passivated, but there are some elements of the population for which the grout protection appears to be compromised. The degree of protection afforded to the strands behind the anchor plate at the reaction blocks along the I-99 17th Street exit ramp in Altoona, PA, do not appear to be protected by grout and this will be discussed later in this section when describing the integrity of ground anchor installations. The half-cell potentials with respect to the restress- able anchors at the same site in Altoona, PA, are lower because measurements reflect conditions along the surface of the gal- vanized trumpet head, and, in contrast to steel, zinc is not passivated by alkalinity. For resin grout installations, relatively high half-cell poten- tials indicate that corrosion may have occurred, but these 36 Model R 1 R 1 COV R 1 β p f / R I [Eq. (17a)] 1.01 0.29 0.29 0.30 2.37 0.009 0.297 II [Eq. (17b)] 1.63 0.46 0.28 0.50 2.26 0.011 0.301 1 Norm ally distributed. λ λ λφ φσ Table 23. Resistance bias and calibration of  considering construction with marginal quality fill.

measurements are also affected by salt concentrations and moisture content of the surrounding rock mass. Higher salt concentrations and dry conditions tend to shift half-cell potentials to more negative values. This is evident from mea- surements taken at several locations in western New York where monitoring was performed at regular intervals over a 2-year duration (see Appendix C). Figure 17 presents the means and ranges of corrosion rate measured via the LPR technique at some of the sites listed in Table 24. Corrosion rates for resin-grouted rock bolts at the Barron Mountain and Beaucatcher Rock Cuts are relatively low. However, the LPR measurements only reflect corrosion rates in areas that are in direct contact with the surround- ing earth and may not include areas where there is a gap or void separating the steel reinforcement surface from the 37 Site Highway State Reinforcement Type Year Installed Anchorage Type Prestress (kips) Corrosion Protection Comments Barron Mountain Rock Cut I-93 NB NH Rock bolts— Grade 150 prestressing steel rods and Grade 80 smooth steel rods 1974 Polyester resin grout 40 None Grouted in bond zone only—bare stressing length Barron Mountain. Rock Cut I-93 SB NH Rock bolts— Grade 150 prestressing steel rods and Grade 80 smooth steel rods 1974 Polyester resin grout 40 None Grouted in bond zone only—bare stressing length Beaucatcher Rock Cut I-240 W NC Rock bolts— Grade 150 prestressing steel rods 1982 Epoxy resin grout 40 Grout Grouted full- length Safety Research Coalmine (SRCM) NIOSH Pittsburgh Research Laboratory PA Roof bolts— Grade 60 steel rods 1988 Resin grout, expansion shell or slot and Wedge N/A Grout/none Fully-grouted, or nongrouted roof bolts Safety Research Coalmine NIOSH Pittsburgh Research Laboratory PA Roof bolts— Grade 60 steel rods 2000 Resin grout N/A Grout Fully-grouted roof bolts Barron Mtn. Rock Cut I-93 NB NH Rock bolts— Grade 150 prestressing steel rods 1974 Portland cement grout 0 Grout Fully-grouted passive elements New Brunswick, Canada Dams in the Musquash River Basin N/A Dam tie- downs—Grade 270 cold drawn steel wire grouted into rock 1964 Portland cement grout 125 Grout Fully-grouted parallel wire, buttonhead anchorages 17th Street exit ramp I-99 PA Reaction blocks—Grade 270 seven wire strand ground anchors grouted into rock. 1992 Portland cement grout 100 Class I— double corrosion protection system Restressable strands surrounded by grease in trumpet head 17th Street exit ramp I-99 PA Reaction blocks—Grade 270 seven wire strand ground anchors grouted into rock 1992 Portland cement grout 100 Class I— double corrosion protection system Nonrestressable strands surrounded by grout behind the bearing plate NOTE: NIOSH = National Institute for Occupational Safety and Health. Table 24. Summary of sites with Type II reinforcements evaluated during Phase II. -1000 -900 -800 -700 -600 -500 -400 -300 -200 -100 0 N H I-93 (NB) N H I-93 (SB) NC I-240 Bruceton M ine D am Tie-D ow ns N H I-93 (tendons) PA Anchors (NR) PA Anchors (R) E c o rr (m V) Resin Grout Portland Cement Grout Figure 16. Ranges of half-cell potential measurements for Type II reinforcements.

surrounding rockmass (i.e., the electrolyte). Thus, these mea- surements demonstrate that resin grout can effectively protect the surface, assuming the surface is adequately covered. Much higher corrosion rates are evident from the NIOSH SRCM. This correlates well with the harsh environmental conditions at this site that includes pH ranging between 2.5 and 3.5 and sulfate concentrations between 800 ppm and 7000 ppm. Also, roof bolts installed at the SRCM may not have a gap behind the anchor plate similar to the rock cut installations such that LPR measurements reflect corrosion rates near the proximal end where moisture and oxygen are more prevalent. Data in Figure 17 also confirm that the non-restressable anchors tested at the I-99 17th Street Exit ramp in Altoona, PA, may not be adequately protected by portland cement grout behind the anchor plate and localized corrosion is occurring at an average rate of 5 μm/yr. Grout quality and the potential for gaps are indicated from results of sonic echo and ultrasonic testing. These tests indi- cate that gaps often exist behind the bearing plate, even for fully grouted installations (Withiam et al., 2002). Results from sonic echo testing also provide information on remaining pre- stress and this may also correlate with conditions along the bonded or anchorage of the reinforcements. Since these con- ditions may deteriorate with respect to time, these measure- ments are also useful to interpret service life and durability of rock bolts. Specific details for rock bolt and ground anchor installations, data from condition assessment, data interpre- tation, and reliability analysis relative to durability and service life, are described in the following sections. Rock Bolts Sites with rock bolts described in Table 24 do not incorpo- rate corrosion protection measures other than grout, there- fore details of the condition and type of grout are particularly important. Resin Grout Although resin grout may provide a barrier from corro- sion, there is strong evidence to suggest that areas of the sur- face may not be covered and vulnerable to corrosion. Even for fully grouted rock bolts it is likely that a gap exists behind the anchor plate that is vulnerable to corrosion. Prestress- ing tends to cause resin grout to crack, compromising its abil- ity to act as an effective barrier to corrosion. Poor coverage has been observed, both from the results of NDT and direct observations (Fishman et al. 2005; Fishman, 2005). Results from sonic echo testing imply that often the degree of cover- age afforded by the grout is relatively low, or grout quality is poor (voids and cracks exist). This is consistent with Comp- ton and Oyler (2005) who reported that the resin grout only covered approximately 60% of the surface for fully grouted roof bolts that were exhumed for observation. Fishman (2005) also reported incomplete coverage in the bonded zone of grouted end-point anchorages exhumed at the site of the Barron Mountain Rock Cut. Kendorski (2003) estimates the design life of unprotected rock reinforcement systems is approximately 50 years. Results from this study demonstrate that, in instances where the design load of rock reinforcements is based on pullout resistance, the design life may be longer than 50 years, depending on site conditions. Other factors, such as loss of prestress, may also affect the service life of rock bolts with end-point anchorages. Results from sonic echo tests, that were confirmed from lift-off test- ing, indicate that a relatively high proportion (approximately 30%) of resin-grouted bolts with end-point anchorages have lost significant levels of prestress at the Barron Mountain Rock Cut (Fishman, 2004; Fishman, 2005). However, this may not be as much of a problem for fully grouted elements. Loss of prestress may be indicative of poor grout cover along the bonded zone, or due to weathering of rock beneath the bearing plate for end-point anchorages. 38 0 10 20 30 40 50 60 70 NH I-93 (SB) NC I-240 Bruceton Mine PA Anchors (NR) CR (μ m /y r) Figure 17. Ranges of corrosion rate measurements for Type II reinforcements.

Knowledge of surface area in electrical contact with earth is required to reconcile corrosion rates from LPR measure- ments. This surface area is more difficult to determine for rock bolt installations compared to MSE (i.e., Type I) reinforce- ments. First, the length of the bolt and the grouted length need to be estimated, and then the amount of coverage afforded along the grouted length must be assessed. Installation details for rock bolts are not readily available from construction plans, and details related to bolt length and the length of the bonded zone must be obtained from field notes when available. The length of the bond zone can also be estimated from knowledge of the lock-off load, drill hole diameter, and by estimating the allowable bond stress at the grout/rock interface. How- ever, another utility of sonic echo testing is the confirming or obtaining of missing information about the geometry of the installation. Data from sonic echo tests have been used in this study to verify bolt lengths and bond lengths that may then be used to estimate the surface area of the rock bolts in con- tact with the surrounding rock in order to reconcile corrosion rates from LPR measurements. The rock bolts installed at Barron Mountain were only grouted along the bonded length (i.e., end-point anchor- ages), thus the free/stressing length is more vulnerable to cor- rosion. This is confirmed by direct visual observations of bolts that were exhumed from the site as reported by Fishman (2005), and is evident in the results from sonic echo tests per- formed on more than 50 rock bolts from this site. This is use- ful since direct observations of metal loss from portions of reinforcements that have been retrieved can be compared to corrosion rates measured in situ with LPR. Based on results from sonic echo testing, the bond lengths of rock bolts along the southbound barrel of I-93 at the Barron Mountain Rock- Cut range between 3 and 13 feet. Rock bolts at the Beau- catcher Cut are fully grouted, and results from sonic echo tests indicates that the bolts that were sampled for testing are approximately 10 feet long. LPR measurements depicted in Figure 17 demonstrate cor- rosion rates for sections of rock bolts surrounded by grout are relatively low. Although LPR measurements are useful to assess corrosion along grouted areas, more information is needed to assess corrosion rates for exposed sections (i.e., not surrounded by grout) of the reinforcements. Direct observa- tions of exposed portions of exhumed reinforcements indi- cate that corrosion in these vulnerable areas is much higher than that indicated via LPR measurements. Thus, critical locations that may control design life include the gap behind the anchor plate or other exposed areas. Metal loss of exposed portions of the reinforcement behind the anchor plate, or other areas, may be expressed using the Romanoff equation as where t is time in years. Table 25 is a summary of metal loss measurements obtained from rock reinforcements that have been exhumed from sites located in the United States, Swe- den, Finland, and England. These data are useful to assess the variability associated with the constant A that appears in Equation (23). Based on the data in Table 25, A in Equation (23) has a mean value of 60 μm/yr, a standard deviation of 40 μm/yr, and can be approximated with a lognormal distribution. The data in Table 25 represent site averages. However, data from six rein- forcements retrieved from the Barron Mountain Rock Cut were analyzed and rendered a mean of 66 μm/yr for A. Also, X side A yr side t yr μ μm m⎛⎝⎜ ⎞⎠⎟ = ⎛ ⎝⎜ ⎞ ⎠⎟ ( )0 8 23. ( ) 39 Site Country Type Age (yrs.) X ( μ m) A ( μ m/yr) Barron Mountain Rock Cut, I- 93 NB 1 USA Resin-grouted bar 32 880 55 Barron Mountain Rock Cut, I- 93 SB 2 USA Resin-grouted bar 33 1498 91 State Route 52, Ellenville, NY 3 USA Expansion shell 20 1690 154 Pyhasalmi 4 Finland Split set 1.5 63 46 Pyhasalmi 4 Finland Split set 1 28 28 Hemmaslahti 4 Finland Split set 2 43 25 Hemmaslahti 4 Finland Swellex 1.5 90 65 Kerretti 4 Finland Swellex 2 56 32 FIP Case 4 5 Sweden Ground anchor 26 445 33 Dovenport Royal Dockyard 6 England Ground anchor 22 750 63 1 Fishm an (2005). 2 NCHRP 24-28. 3 Withiam et al. (2002). 4 Lokse (1992). 5 FIP (1986). 6 Weerasinghe and Adams (1997). Table 25. Summary of metal loss measurements from rock bolts and ground anchors.

measurements of metal loss from steel elements embedded in fills with resistivity between 3,000 Ω-cm and 10,000 Ω-cm as discussed in the resistance factor calibrations for Type I rein- forcements render a value for A equal to 54 μm/yr. Further- more the statistics generated from Table 25 are relatively close to the nominal metal loss expressed by Eq. (3), which applies to plain steel buried in a wide range of environments. Thus, the statistics rendered from the data in Table 25 appear to be rea- sonable. The statistical variation of metal loss represented by these parameters can be used to calibrate resistance factors for LRFD similar to that for Type I (MSE) reinforcements. How- ever, there are some notable differences as described in the following example: The example resistance factor calibration considers that Type II reinforcements are not redundant and failure of a sin- gle element could mean that a block of rock is loosened, lead- ing to a system failure. For the purpose of this example a target reliability index, βT, equal to 3.1 corresponding to pf ≈ 0.001 is adopted. This is consistent with past geotechnical design prac- tice for foundations as described by Withiam et al. (1998). The load bias, λQ, used in the resistance factor calibration presumes that rock bolts are actively loaded and prestressed during installation; and prestress is verified via lift-off testing as described by PTI (2004). Thus, the uncertainty relative to the design load is much less compared to Type I reinforce- ments, whereby the loads are passive and transferred to rein- forcements as the system deforms. Design loads are enforced upon the reinforcements during installation, and PTI (2004) recommends that verified lock-off loads be within 5% of design specifications. For the purpose of this example, λR = 1 and COV = 10% are used, and considered to be conservative estimates. The resistance bias is computed as follows: D D X X D D X D i i i ∗ ∗ = − ∗ ∗ < = ∗ ≥ 2 2 2 0 26 for 2 for 2 ( ) A D c  = ∗π 2 4 25( ) λR ult c F A T =   nominal ( )24 For this example, it is assumed that the design capacity of the rock bolt is based on the resistance mobilized along the bonded length and not the structural capacity of the bolt. The structural resistance (tensile strength) remaining at the end of the design life must equal the design load enforced on the system during installation. If the difference between the lock- off load and the original structural capacity of the bolt is high enough, then enough structural capacity may be available at the end of the design life to sustain the design loads even though metal loss may not be explicitly considered during design. A nominal resistance, Tnominal, equal to 40 kips, is used in this example similar to the Barron Mountain and Beaucatcher Rock Cuts described in Table 24. The reinforcements are assumed to be solid bars made from Grade 150 steel that has a guaranteed ultimate tensile strength (GUTS) of 150 ksi. The GUTS is considered as the nominal strength and Fult (the sta- tistical variable for ultimate strength of steel) is considered to have a normal distribution equal to 1.05 times the nominal and COV equal to 0.1 similar to that described by Bounopane et al. (2003) and the statistics used to describe the variation of yield strength for Type I reinforcements. Equation (23) is used to compute metal loss, X, where the parameter A is varied statis- tically according to a lognormal distribution with μA = 60 μm and σA = 40 μm. The calibration is performed considering bolts with a 1-inch initial diameter, Di, and 50-, 75- and 100-year design lives. Table 26 is a summary of the computed bias, resistance fac- tor, and probability of occurrence for each design life. These results indicate that for this example metal loss does not have a significant impact on performance for design lives of 50 and 75 years, but should be taken into account for service lives in excess of 75 years. The results from this example depend on the selected values for Tnominal, Di, and Fult. If these inputs vary then the results depicted in Table 26 do not apply. The pur- pose of this example is to demonstrate the approach and identify the input needed for a complete calibration. More data are needed to assess typical design scenarios before a more complete calibration can be performed. Current AASHTO specifications specify φ = 0.8 relative to rupture resistance for high-strength steel reinforcements. Assuming that metal loss is not considered in design but using the same statistical properties for the remaining vari- ables renders pf < 0.0001. The 75-year case shown in Table 26, 40 R t f (years ) Distribution p f 50 2.36 0.33 Weibull 1.0 < 0.0001 75 2.22 0.42 Weibull 1.0/0.80 ≈0.001/0.0001 100 2.14 0.55 Weibull 0.55 ≈0.001 μ σ φλ Table 26. Results from reliability-based calibration considering the yield limit state for rock bolts.

but with φ = 0.8 renders pf ≈ 0.0001. This implies that the cur- rent AASHTO specifications imply a service life of approxi- mately 75 years for the selected example. Portland Cement Grout Based on results from NDT and direct observations, port- land cement grout quality generally appears to be good for the rock bolts inspected at the sites listed in Table 24 and depicted in Figure 16. Half-cell potential measurements indicate that the presence of a passive film layer protects the rock bolts from corrosion. However, passivation of the steel may be compro- mised by the presence of chlorides or acidic conditions. Chlo- rides may be present along the rock face as a residue from salt spray produced from deicing of the highway. There is some evidence of this from the testing of fully grouted tendons at the Barron Mountain Rock Cut (Fishman, 2004). LPR measure- ments indicate very low average rates of metal loss and sig- nificant metal loss was not observed from several elements that were exhumed for inspection (Fishman, 2005). In gen- eral, for installations where lower half-cell potentials (less than −200 mV) are realized corrosion monitoring should be per- formed to assess the rate of metal loss. Very high rates of corrosion are possible when the passive film layer is locally compromised. Grout cracking may also occur at the beginning of the bonded zone due to the application of prestress. How- ever, the surface is still protected via the alkaline environment until crack widths exceed a minimum value (e.g., 1 mm). Grout and return tubes are used to install portland cement- grouted rock bolts such that the grout quantity can be adjusted as needed when grout is lost from the drill hole. This is in con- trast to resin-grouted bolts for which a fixed quantity of grout is inserted into the drill hole. Thus, the coverage from port- land cement grout is expected to be better when compared to resin-grouted installations. Also, prestressed rock bolts require a stressing length that is often protected by grease and a plastic sheath. However, a gap behind the anchor plate is still possible unless a trumpet head assembly is employed. Thus, the main concern with portland cement-grouted rock bolts is with respect to a gap behind the anchor plate. This may be remedied with a trumpet head assembly filled with grease. Alternatively, metal losses need to be considered using Equation (23), similar to those for resin-grouted installations. Ground Anchors If the ground anchor system is protected with an adequate corrosion protection system [e.g., meeting the requirements of PTI Class I (PTI, 2004)], then corrosion is generally not a prob- lem. Observations from NDT indicate that, generally, grout quality along the bonded zone appears to be good and no defects or anomalies were encountered along the stressing lengths. The main concern for ground anchorages is near the anchor head assembly and the fact that high-strength, pre- stressed steel elements may be vulnerable to hydrogen embrit- tlement and SCC. Hydrogen embrittlement and SCC tend to occur in acidic or chloride-rich environments, and, without proper detailing and workmanship at the anchor head assem- bly, the service life of the elements is severely compromised by these environments. Time to failure for hydrogen embrittle- ment and SCC can be relatively short, and the previously cited models describing rate of metal loss for uniform corrosion or incursion of pit depths are not applicable to assess service life. Measurements of half-cell potential and corrosion rate for ground anchors indicate that portland cement grout does not always serve to passivate steel, and the area behind the bearing plate is vulnerable to corrosion. Use of a trumpet head filled with grease appears to be a more effective measure to protect the reinforcements as it is isolated from the surrounding envi- ronment by virtue of the dielectric properties of the grease. Loss of prestress may also affect service life. Current obser- vations in the database do not include sites where this has been observed to be a problem. However, if the anchorage zone is within soil or rock types that creep, then loss of pre- stress could affect service life. Regular maintenance of lock- off loads could be implemented to achieve a given service life for these conditions. 41

Next: Chapter 4 - Conclusions and Recommendations »
LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s National Cooperative Highway Research Program (NCHRP) Report 675: LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems explores the development of metal loss models for metal-reinforced systems that are compatible with the American Association of State Highway and Transportation Officials' Load and Resistance Factor Design Bridge Design Specifications.

NCHRP Research Results Digest 364: Validation of LRFD Metal Loss and Service-Life Strength Reduction Factors for Metal-Reinforced Systems summarizes the results of research to further validate some key results of a project that resulted in publication of NCHRP Report 675.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!