Inspection and Management of Bridges with Fracture-Critical Details (2005)

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.

39 OVERVIEW OF FATIGUE Fatigue is considered a serviceability limit state for bridges. This is because fatigue cracks do not typically compromise structural integrity and are more of a maintenance issue. However, as was recognized by the Task Committee on Redundancy of Flexural Systems (A1), fatigue is the most common cause of reported damage in steel bridges. The fatigue design and assessment procedures outlined in this appendix are included in the AASHTO specifications for bridges (A2). As a result, steel bridges that have been built in the last two decades have not and should not have any signifi- cant problems with fatigue and fracture (A3). However, bridges designed before the modern specifications will continue to be susceptible to the development of fatigue cracks and to fracture. Detailing rules are perhaps the most important part of the fatigue and fracture design and assessment procedures. These rules are intended to avoid notches and other stress concentrations, as well as the use of details known to be very fatigue sensitive. They also often result in details that have improved resistance against brittle fracture as well as fatigue. Modern steel bridges are also detailed in a way that appears much cleaner than those built before the 1970s. There are fewer connections and attachments in modern bridges and the connections use more fatigue-resistant details, such as high-strength bolted joints. In bridges there are usually a large number of cycles of sig- nificant live load and fatigue will almost always precede frac- ture. Therefore, controlling fatigue is practically more impor- tant than controlling fracture. Usually, the only measures taken in design that are primarily intended to ensure fracture resis- tance are those to specify materials with minimum specified toughness values [such as a Charpy V-notch (CVN) test re- quirement]. As explained in chapter three, toughness is speci- fied so that the structure is resistant to brittle fracture despite manufacturing defects, fatigue cracks, and/or unanticipated loading. However, these material specifications are less impor- tant for bridges than the S–N curves and detailing rules. Nominal Stress S–N Curves The established method for fatigue design and assessment of steel bridges in the United States is the nominal stress approach. The nominal stress approach is based on S–N curves, where S is the nominal stress range and N is the number of cycles until the appearance of a visible crack. Details are designed based on APPENDIX A Background Discussion on Fatigue, Fracture, Nondestructive Evaluation, and Repair and Retrofit the nominal stress range in the connecting members rather than the local “concentrated” stress at the detail. The nominal stress is usually obtained from standard design equations for bending and axial stress and does not include the effect of stress con- centrations of welds and attachments. AASHTO (A2) has seven S–N curves corresponding to seven categories of weld details (A through E’), as shown in Figure A1. The fatigue design procedure is based on associating the weld detail under consideration with a specific category. The effects of the welds and other stress concentrations, includ- ing the typical defects and residual stresses, are reflected in the ordinate of the S–N curves for the various detail cate- gories. Consequently, the variability of fatigue life data at a particular stress range is typically about a factor of 10. The AASHTO S–N curves in Figure A1 are also used throughout North America for a variety of other welded structures, including the American Institute for Steel Con- struction Manual of Steel Construction (A4), the American Railway Engineering and Maintenance-of-Way Association Manual for Railway Engineering (A5), the American Weld- ing Society (AWS D1.1) Structural Welding Code (A6), and the Canadian Standards Association (CSA S16-2001) Limit States Design of Steel Structures (A7). The S–N curves can be represented by the following power law relationship: where N is the number of stress cycles, S is the nominal stress range, and A is a constant particular to the detail category, as given in Table A1. In the nominal stress range approach, each detail category also has a constant-amplitude fatigue limit (CAFL), which is given in Table A1. The CAFL is the stress range below which no fatigue cracks occurred in tests conducted with constant- amplitude loading. The approach to designing and assessing bridges for fatigue is empirical and is based on tests of full-scale mem- bers with welded or bolted details. Such tests indicate that: • The strength and type of steel have only a negligible effect on the fatigue resistance expected for a particular detail (A8–A10). • The welding process also does not typically have an effect on the fatigue resistance (A11,A12). N A S = 3 1( )

• The primary effect of constant-amplitude fatigue loading can be accounted for in the live-load stress range (A8– A10); that is, the mean stress is not significant. (The rea- son that the dead load has little effect is that, locally, there are very high residual stresses.) It is worth noting that when information about a specific crack is available, a fracture mechanics crack growth rate analysis should be used to calculate remaining life (A13– A16). However, in the design stage, without specific initial crack size data, the fracture mechanics approach is not any more accurate than the S–N curve approach (A17). Therefore, there will be no further discussion of the fracture mechanics crack growth analysis. Effect of Corrosion on Fatigue Resistance Many bridge owners are concerned about the influence of cor- rosion on the fatigue and fracture performance of both frac- 40 ture-critical and non-fracture-critical bridges (FCBs and non- FCBs). However, the concern is greater with respect to frac- ture-critical structures, as indicated by several responses to the survey and as discussed in chapter three. The full-scale fatigue experiments upon which the fatigue rules are based were carried out in moist air and therefore reflect some degree of environmental effect or corrosion fatigue. Hence, the lower-bound S–N curves in Figure A1 can be used for the design of details with a mildly corrosive envi- ronment (such as the environment for bridges, even if salt or other corrosive chemicals are used for deicing) or provided with suitable corrosion protection (galvanizing, other coat- ing, or cathodic protection). Some design codes for offshore structures have reduced fatigue life by approximately a fac- tor of two when details are exposed to seawater (A18,A19). At the relatively high stress ranges at which most acceler- ated tests are conducted, the effect of seawater is clearly detri- mental. However, there is evidence that the effect of corrosion in seawater is not so severe for long-life, variable-amplitude fatigue of welded details. Full-scale fatigue experiments in sea- water at realistic service stress ranges do not show significantly lower fatigue lives, provided that corrosion is not so severe that it causes pitting or significant section loss (A20). At relatively low stress ranges near the CAFL typical of service loading, it appears that the build-up of corrosion product in the crack may actually increase crack closure and retard crack growth, at least enough to offset the increase that would otherwise occur owing to the environmental effect. The fatigue lives seem to be more significantly affected by the stress concentration at the toe of welds than by the corrosive environment. AASHTO Curves 1000 100 10 100 10 104 105 106 107 108 Number of Cycles St re ss R an ge , M Pa St re ss R an ge , ks i FIGURE A1 Lower-bound S–N curves for seven primary fatigue categories from AASHTO specifications. Dotted lines are the constant-amplitude fatigue limits and indicate detail category. AASHTO Category CAFL (MPa) A 165 B 110 Bí 83 C 69 D 48 E 31 E’ Coefficient A (MPa3) 81.9 x 1011 39.3 x 1011 20.0 x 1011 14.4 x 1011 7.21 x 1011 3.61 x 1011 1.28 x 1011 18 CAFL = constant-amplitude fatigue limit. TABLE A1 PARAMETERS FOR S–N CURVES

41 Severely corroded members may be evaluated as Category E details (A21), regardless of their original category (unless of course they were Category E′ or worse to start with). How- ever, pitting or significant section loss from severe corrosion can also lower that fatigue strength (A21,A22). Loading for Fatigue Design It is important to note that the load producing fatigue cracking is comprised of the entire load spectrum crossing the bridge. The load that produces a fracture, on the other hand, is typi- cally the largest load from the load spectrum. Bridges experi- ence what is known as long-life, variable-amplitude loading [i.e., very large numbers of random amplitude cycles greater than the number of cycles associated with the CAFL (A23)]. If the percentage of stress ranges exceeding the CAFL is greater than approximately 0.01%, the history of N variable stress ranges can be converted to N cycles of an effective stress range that can then be used just like a constant-amplitude stress range in S–N curve analysis. Typically, Miner’s rule (A24) is used to calculate an effective stress range from a histogram of variable stress ranges. Theoretically, this effective constant- amplitude stress range results in approximately the same fatigue damage for a given number of cycles as the same num- ber of cycles of the variable-amplitude service history. If the stress ranges are counted in discrete “bins,” as in a histogram, the effective stress range, SRe (A23), can be calculated as: where αi = number of stress cycles with stress range in the bin with average value Sri divided by the total number of stress cycles (N). In the AASHTO specifications (A2), the stress range from the fatigue design truck (i.e., the HS15) represents the effec- tive stress range. No additional safety factor is used for Miner’s rule, because it is relatively accurate for truck loading on bridges. For large numbers of cycles, the AASHTO speci- fication has another check that involves comparing the stress range from the fatigue design truck with one-half of the CAFL. This check is actually intended to compare the CAFL with a stress range that is twice that produced by the fatigue design truck (i.e., dividing the resistance by two is the same as multi- plying the load by two). Although somewhat confusing in application, the intent is to ensure that almost all of the stress ranges should be below the CAFL, but that occasionally the stress range can exceed the CAFL with no significant effect. Fatigue Life Prediction Methodology for Existing Bridges The 1990 AASHTO Guide Specifications for the Fatigue Evaluation of Existing Steel Bridges (A25) can be used for the fatigue evaluation of steel bridges. The purpose of the guide S Si ri i Re ( )= ( )( )∑ α 3 1 3 2 is to provide procedures for calculating the remaining fatigue life of existing steel bridges using concepts of probabilistic limit states. The fatigue evaluation procedures in the guide were adopted from and are identical to the proposed proce- dures presented in NCHRP Report 299: Fatigue Evaluation Procedures for Steel Bridges (A26). The effects of repeated loading on the fatigue life of an existing bridge are defined in terms of the remaining life of the structure. This means that the effect of exceeding the allowable fatigue stress is a reduc- tion of the remaining fatigue life of the structure rather than immediate failure. There are two levels for which the remaining fatigue life can be calculated, the remaining safe life and the remaining mean life. The remaining safe life provides a much higher level of safety. The safe life has an exceedence probability of 97.7% for redundant members (the same probability inherent in using the basic design S–N curves). The mean remaining life, however, is the best estimate of the actual remaining fatigue life of the detail under consideration. The mean life has an exceedence probability of 50%. The safe life is used for design and for a first screening analysis. The impact of reaching a safe life equal to zero should be relatively minimal; it may mean having to repair just a small percentage of the details on a bridge. In this case, the calculated mean life may be used to determine when the cracking will be so pervasive that half the details will be cracking, requiring extensive retrofitting or possibly replacement of the bridge. If the remaining life that is esti- mated is not satisfactory, there are four options: (1) recalcu- late the fatigue life more accurately (possibly using load test- ing), (2) restrict truck traffic on the bridge, (3) repair or modify the detail, and/or (4) perform more frequent inspections of the detail. The procedures in NCHRP Report 299 (A26) are the best available procedures for predicting the number of years before substantial fatigue cracking may occur at a detail. It is difficult to accurately predict all of the variables that the bridge will experience over its remaining life, such as past, present, and future truck volumes and stress ranges, which can also vary with changing truck weights over time. Although calculations are by far the most common and favored methods of estimating remaining fatigue life in a bridge, they are also the least accurate. Field instrumenta- tion and monitoring has consistently demonstrated that mea- sured in-service stress range histograms result in the most accurate estimates of remaining life. Analysis methods rely on approximate load models and simplified structural analy- sis models, both of which are typically conservative. Mea- surements made in the field reflect actual site conditions and traffic patterns when recorded over a sufficient period of time. Because the stress range is directly measured at the detail in question, there is no error introduced through ana- lytical simplifications.

Because in nearly all cases field measurements indicate that in-service stresses are less than predicted, costly retrofits can be eliminated or reduced in number. The resulting sav- ings will almost always far exceed the cost of the additional efforts associated with the field instrumentation. Despite this potential for savings, most agencies do not use this approach, as will be discussed in the review of the survey results. OVERVIEW OF FRACTURE Fracture may be defined as rupture in tension or rapid exten- sion of a crack, leading to gross deformation, loss of function or serviceability, or complete separation of the component (A4,A14,A27). Because the scope of this synthesis is limited to practical information, there are many interesting aspects of fracture that are not discussed. However, there are several good texts that can serve as a starting point for more in-depth studies (A14,A15,A27). It is important to rapidly assess the potential for fracture whenever there is a crack in any tension element (tension member or tension flange of flexural member). However, the occurrence of fatigue cracks does not necessarily mean that the structure is in danger (A13,A27). In some redundant structures, a fatigue crack may stop propagating with no intervention at all as a result of redistribution of stresses (A28). Usually, however, a fatigue crack will propagate and eventually cause a fracture if not repaired in a timely manner (A13,A15,A27). The development of a fatigue crack in a frac- ture-critical member (FCM) should, in most cases, warrant closure of one or more lanes, posting the bridge for cars only, or result in complete closure of the bridge until repairs can be made. Fracture is the rupture in tension or rapid extension of a crack leading to gross deformation, loss of function or serviceability, or complete separation of the component. Details that have good fatigue resistance, Category C and better, are usually also optimized for resistance to fracture. Detailing rules to avoid fracture are very similar to the common sense rules to avoid fatigue. For example, inter- secting welds should always be avoided owing to the proba- bility of defects and excessive constraint. Intersecting welds, or even welds of too close proximity, have caused brittle fractures [e.g., the Hoan Bridge in Wisconsin (A29) and the SR-422 bridge in Pennsylvania (A30,A31)]. Weld backing bars must usually be removed to achieve the needed resis- tance to both fatigue and fracture. Stress concentrations such as reentrant corners should be avoided and instead transition radii that are ground smooth and flush should be provided. This commonly occurs at copes in floorbeams at connections designed for vertical shear only. The ends of butt welds are always a potential location of defects. It is important to use run-out tabs and to later grind the ends of the weld flush or to a radius. Fillet weld termina- tions should not be ground, for this will expose a very thin ligament near the weld root that will tear easily. 42 Fracture Assessment Procedures A fracture assessment is made using fracture mechanics prin- ciples (A13–15,A32,A33). In the absence of other established procedures, an often used reference is the 1999 British Stan- dard, BS 7910, “Guide on Methods for Assessing the Acceptability of Flaws in Metallic Structures” (A32). There is presently no comparable U.S. standard; however, BS 7910 is used by some U.S. industries (primarily the oil and gas industries). A fracture assessment requires some knowledge of the fracture toughness of the steel and/or weld metal, which may be estimated from the Charpy energy or CVN (A15,A33). Unlike fatigue, the susceptibility to fracture is strongly dependent on the type of material and even the particular heat of the steel or lot of weld metal (A14,A15,A34,A35). The CVN may be obtained from mill reports for the steel and from the certifications for the weld metal, if good records exist. Lacking such specific information, for bridges built since 1974 when minimum Charpy requirements were first included in the ASTM A709 specification for bridge steels, the CVN for the steel plate and weld metal may be assumed to be at least as large as the minimum specified values. For older structures, or for cases when the assumed minimum values are not sufficient, it may be necessary to drill a core and get a sample of the steel, then make and test some Charpy specimens. A minimum of three specimens should be tested from each plate or structural shape. The fracture assessment will not be discussed further because it is documented else- where and is not the emphasis of this report. Causes for Fracture Other Than Fatigue As mentioned previously, fracture in bridges is almost always a result of fatigue cracking. However, if a detail and material are particularly susceptible to fracture, failure will usually occur as the loads are applied for the first time during or soon after construction. An example is the fracture that occurred when the Caltrans Workers Memorial Bridge (previously known as the Bryte Bend Bridge) was under construction in 1970. The fracture, shown in Figure A2, was attributed to use of low-toughness A514 steel (A36) at a point where a trans- verse bracing member was welded along the edge of the pri- mary tension flange of a tub girder, creating a stress concen- tration at the reentrant corner. A514 steel, marketed under the name of T1 steel, is quenched and tempered with a minimum specified yield stress of 690 MPa. It is also possible that fracture can occur in service directly without apparent fatigue crack growth. For example, at poor details that are highly constrained, such as the intersection point of two or three welds, fracture may occur in service directly from small crack-like weld discontinuities, such as the fractures that originated at shelf plate details in the Hoan Bridge in Mil- waukee in December 2000 (A29) (Figure A3) or the SR-422 failure (A30,A31). Repair and retrofit of these shelf-plate details

43 are discussed in special directives from FHWA and are not dis- cussed further in this report. Subcritical crack propagation in bridge elements may also occur by stress corrosion cracking (SCC), although this is a concern only for very-high-strength steels; that is, steel with yield strength much greater than 100 ksi. SCC involves elec- trochemical dissolution of metal along active sites under the influence of tensile stress (A14,A15). As hydrogen is liber- ated in the process, this failure mechanism may also involve hydrogen cracking. SCC can be distinguished from fatigue cracking by examining the fracture surface under a light microscope. SCC has occurred in prestressing cables (A37) and in A490 bolts that were out of specification (A38). Specification of Charpy V-Notch for Fracture Resistance Fracture behavior depends strongly on the type and strength level of the steel or filler metal. The fracture resistance of each type of steel or weld metal varies significantly from heat to heat, or from lot to lot. Although fracture toughness can be measured directly in fracture mechanics tests (A14–A16), the usual practice is to characterize the toughness of steel in terms of the impact energy absorbed by a CVN specimen (A15). Because the Charpy test is relatively easy to do, it will likely continue to be the measure of toughness used in steel specifications. Because it is not directly related to the fracture toughness, CVN energy is often referred to as notch toughness. The notch toughness is still very useful, however, because it can often be correlated to the fracture toughness and then used in a fracture mechanics assessment (A15,A32,A33). Figure A4 shows a plot of the CVN energy of A709 Grade 50 (350 MPa yield strength) structural steel at varying temperatures. These results are typical for ordinary hot-rolled structural steel. The fracture limit state includes phenomena ranging from the brittle fracture of low-toughness materials at service load levels to ductile tensile rupture of a component. The transi- tion between these phenomena depends on temperature, as reflected by the variation of CVN with temperature as shown in Figure A4. The transition is a result of changes in the underlying microstructural fracture mode. Brittle fracture on the so-called lower shelf in Figure A4 is associated with cleavage of individual grains on select crystallographic planes. Brittle fracture may be analyzed with linear-elastic fracture mechanics theory because the plastic zone at the crack tip is very small. At the high end of the temperature range, the so-called upper shelf, ductile frac- ture is associated with the initiation, growth, and coalescence of microstructural voids, a process requiring much energy. The net section of plates or shapes fully yields and then rup- tures with large slanted shear lips on the fracture surface. Transition-range fracture occurs at temperatures between the lower and upper shelves and is associated with a mixture of cleavage and shear fracture. Large variability in toughness at constant temperature and large changes of temperature are typical of transition-range fractures. AASHTO specifications for bridge steel and weld filler metal require minimum CVN values at specific temperatures FIGURE A2 Brittle fracture of flange of tub girder of Bryte Bend Bridge in Sacramento, California, while under construction. FIGURE A3 Fractured girder of the Hoan Bridge in Milwaukee (top) and view of critical shelf plate detail featuring intersecting welds.

(A25). As shown in Figure A4, the typical lower-shelf CVN is about 10 J. Therefore, when a minimum CVN of 20 J or more is specified at some temperature, the most important result of such a specification is that the lower shelf of the Charpy curve will start at a temperature lower than the specified tem- perature. This indicates that the lower shelf of a structure loaded statically or at intermediate strain rates such as traffic loading on a bridge is even lower, a phenomena known as the temperature shift (A15). Because of the temperature shift, the temperature at which the CVN requirement is specified may be greater than the lowest anticipated service temperature. If the material is not on the lower shelf at service tempera- ture, brittle fracture will not occur as long as large cracks do not develop. It essentially does not matter what the specified CVN value is as long as it is at least 20 J. Usually, an average from three tests of 34 J (25 ft-lbs) or 27 J (20 ft-lbs) is speci- fied at a particular temperature. The greater the value of the average CVN requirement, the more certain that the material 44 is well above the lower shelf; however, there may be a greater premium to be paid with diminishing increases in certainty. Table A2 lists, as an example, values of impact energy presently required for base metal in FCMs for the three tem- perature zones into which the United States is divided. [For nonfracture-critical members (non-FCMs), typically only 20 J is required at the same temperatures.] The complete table is found in the LRFD Specifications (A2). For example, 50-mm thick flange plates of grade 345W steel for an FCB to be built at a site where the lowest anticipated service temperature (LAST) is −15°C must have the minimum impact energy for Zone 1, which is 34 J at 21°C. History of Development of the Fracture Critical Plan The fracture toughness requirements are based on a correla- tion between the fracture toughness in terms of the stress- TEST TEMPERATURE, °F A588 19mm Plate Charpy Data 0 50 100 150 200 250-50-100-150 0 10 30 50 70 20 40 60 90 25 50 75 100-25-50-75 TEST TEMPERATURE, °C CV N, J ou le s CV N, ft -lb FIGURE A4 Charpy energy transition curve for A709 Grade 50 (350 MPa yield strength) structural steel. ASTM A709 Steel Grade Plate Thickness (mm) and Joining Methoda Zone 1 LAST = −18°C (Joules at °C) Zone 2 LAST = −34°C (Joules at °C) Zone 3 LAST = −51°C (Joules at °C) 250F Up to 100 M & W 34 at 21 34 at 4 34 at −12 Up to 50 M & W 34 at 21 34 at 4 34 at −12 Over 50 to 100 M 34 at 21 34 at 4 34 at −12 345F, 345WF Over 50 to 100 W 41 at 21 41 at 4 41 at −12 HPS-485WF Up to 100 M & W 48 at −23 48 at −23 48 at −23 Up to 65 M & W 48 at −1 48 at −18 48 at −34 Over 65 to 100 M 48 at −1 48 at −18 48 at −34 690F, 690WF Over 65 to 100 W 68 at −1 68 at −18 Not permitted aM = mechanically fastened; W = welded. TABLE A2 MINIMUM CHARPY IMPACT TEST REQUIREMENTS FOR FRACTURE-CRITICAL MEMBERS

45 intensity factor and the CVN, including a “temperature shift.” The temperature shift accounts for the differences in strain rate between the Charpy impact test and the traffic loads. For example, the time it takes to reach the yield strain in bridges from an overloaded truck is about one second ver- sus one millisecond in the Charpy test (A15). Several inde- pendent studies verified the temperature shift (A39–A41). For example, several girders were tested at the U.S. Steel Research Laboratory that contained fatigue cracks initiated at the ends of cover plate details (A41). A more extensive study was done at Lehigh University that examined an increased number of specimens fabricated from grade 250, 345, and 690 steels (A39). In both studies, the fatigue cracks were grown at temperatures and loading rates that closely resemble those encountered in bridges, and the cracked girders were sub- jected to periodic overloads until fracture occurred. The temperature shift was the subject of some disagreement in the development of notch toughness requirements in the early 1970s. The prevailing opinion in the debate was that ordinary hot-rolled mild structural steels had sufficient notch toughness. The specified CVN values are the typical minimum values that could be consistently achieved. Having the speci- fication was very important to screen out unusually brittle materials. Roberts and Krishna (A39) and Roberts et al. (A40), among others, argued that the fracture-critical specification should require a level of dynamic toughness sufficient to arrest pop-in-type fractures occurring from local brittle zones. Hartbower (A42) proposed an alternate fracture control plan that called for the CVN test to be done at the same tempera- tures as the LAST for the bridge service site; that is, without exploiting the benefit of the temperature shift. This would mean that CVN would be required at −34°C for Zone II (most of the country). To achieve notch toughness at these lower temperatures, the steel would require additional processing, such as normalization. This increased processing would sig- nificantly increase the cost of the steel, which is a major part of total bridge costs. Ultimately, the temperature shift concept was retained, largely out of economic necessity, and the fracture control plan placed a strong emphasis on defect control to prevent the pop-in-type fracture events. Therefore, the CVN require- ments represented a compromise that allowed the continued use of hot-rolled steel rather than normalized steel. The good fracture-resistance track record of mild steel shows that requiring normalized steel would have been unnec- essary. Experience in service and in full-scale experiments has verified the temperature shift and the adequacy of the notch- toughness requirements for mild steels. Some premature cracks and fractures occurred in tied arches fabricated with A514 steel. A514 steel is quenched and tempered steel with a mini- mum specified yield stress of 100 ksi. This steel is much dif- ferent than mild steel and should be avoided for FCMs. Frac- tures that occurred during construction in the box girders of the Bryte Bend Bridge in California also occurred in A514 steel. Consequently, higher CVN requirements are implemented for the higher strength steel, as shown in Table A2. Again a compromise was reached, which resulted in the slightly increased CVN for FCMs at the same temperature shift. At the same time, more stringent fabrication rules were applied for FCMs to control the possibility of initial defects. Most importantly, the stress range for fatigue design was arbitrarily dropped by about one category, which resulted in a decrease in the allowable stress range of approximately 25% in most cases. A 25% decrease in the stress range should essentially double the fatigue life. It is important to realize that this decrease in the allowable stress range for FCMs in the Standard Specifications has been dropped in the most recent LRFD Specifications (A2). The Bridge Welding Code D1.5 calls for higher impact energies in weld metal than in base metal, a reasonable requirement given that the welds create stress raisers and high-tensile residual stresses. Given that the cost of filler metal is relatively small in comparison with the overall cost of materials, it is well worth specifying high-toughness filler metal to mitigate the effects of high-tensile stresses induced by welding. For FCMs, 34 J is required at −29°C for weld metal, regardless of the temperature zone. For non-FCMs, weld metal is required that can give 27 J at −29°C for Zone 3, and at −18°C for Zones 1 and 2. The somewhat higher toughness requirements shown in Table A1 for thick plates that are to be welded is the result of the higher degree of constraint and plane–strain behavior at temperatures higher than would be the case for thinner plates. This is more of a problem in welded members, because weld- ing increases the potential for initial flaws and defects that can initiate brittle fracture. Another feature of the specifica- tion requires higher toughness for steels with yield strength greater than 345 MPa. The AASHTO material toughness specifications provide the minimum level of toughness required to sufficiently min- imize the risk of brittle fracture. The existing AASHTO frac- ture control plan has generally done a good job of preventing fracture failure in bridge structures since design using its pro- visions. A few brittle fractures have still occurred without noticeable fatigue crack growth in previously designed bridges, however, although the steel met the modern CVN toughness requirements (i.e., those required by the fracture control plan or general AASHTO toughness requirements). In most such cases, the fracture can be traced back to one or more aspects of the fracture control plan related to controlling the size of defects. If all aspects of the fracture control plan are properly implemented, the risk of brittle fracture is minimized by the current specifications and fracture will be a rare event in ser- vice. However, in the ensuing decades, very few FCBs were

constructed, in part because the fabrication costs were signif- icantly greater. High-Performance Steel The recently developed high-performance steels (HPS) are much tougher but also more expensive than the ordinary Grade 250, 345, or 345W steel. The first grade to be fully integrated into the AASHTO specifications is the HPS 485W. Two additional steels are currently being developed, an HPS 690 W Cu-Ni and an HPS 345W. Both of these steels are much tougher than the ordinary Grade 250, 345, and 345W steels. The main reason for developing HPS has been to improve weldability, but a large gain in toughness has been a desirable by-product. Development of HPS in the United States has been done under a joint research and devel- opment program by FHWA, U.S. Navy, and the American Iron and Steel Institute (A43). In contrast to ordinary bridge steel, Grade HPS 485W shows upper shelf behavior at test temperatures down to approxi- mately −20°C. Even at the extremely cold service temperature for Zone 3 (LAST = −51°C), this steel exhibits toughness in the upper transition region; without the benefit of the temperature shift, the CVN energy still exceeds 150 J. Likewise, high dynamic and crack arrest toughness can be expected of this steel. Clearly, HPS 485W steel provides a level of toughness far exceeding the current requirements, which are based on pre- venting brittle fracture. In current research underway at the Turner–Fairbank High- way Research Center of FHWA, the fracture toughness of HPS are being determined as a function of temperature, loading rate, and plate thickness (A44,A45). At the same time, the fracture resistance of fatigue-cracked I-girders fabricated from HPS 485 is being determined. The test girders are cyclically loaded until a crack grows to the desired length; the girder is then cooled to −34°C and subjected to a typical design overload. If the girder does not fracture, the fatigue crack is grown larger and the same overload is reapplied. The HPS 485W steel girder resisted the full design overload until approximately 50% of the tension flange area was lost to fatigue, meaning that the net section of the cracked tension flange had yielded. In contrast, the full-scale tests discussed previously (A39, A41) on girders made of Grade 345 steel fractured when the net section of the cracked tension flange reached a stress of approx- imately 60% of the yield stress, meaning that the fracture was brittle. In contrast, the fracture mode for the Grade HPS 485W indicates a large amount of plastic deformation before failure. The HPS grade steels provide a toughness level that far exceeds the minimum requirements cited in Table A2. Although HPS 485W costs more than ordinary grade 345W steel, the advantages afforded by higher strength more than offset the difference in material costs. This has significantly altered the economic factors that were considered in setting 46 the current AASHTO toughness requirements. The cost pre- mium for higher toughness is lower than it was 30 years ago. Consideration of Increasing Toughness Requirements for Bridge Steel Fortunately, today’s scrap-based steel typically has improved toughness relative to the steel of the 1970s. This is the result of: (1) a reduction in carbon, (2) a reduction in impurities such as phosphorous and sulfur, as well as (3) an increase in alloying elements such as nickel and chromium, which enhance toughness. Nickel and chromium and other beneficial alloys just happen to be in the scrap that is derived primarily from automobiles and other sheet metal, which tends to have greater alloying than structural steel made from iron ore. Because steel today has greater toughness than the steels of the 1970s is perhaps the best argument for increasing the notch toughness requirements for FCMs. This would allow designers to take advantage of the superior toughness char- acteristics of HPS (A45). In other words, because the CVN specifications were essentially just below what was consis- tently attainable in the 1970s, and because the toughness has generally improved over time, it is rational to raise the bar somewhat to ensure that the bridge steel represents the best practice of today. This could probably be done without forc- ing the mills to use any additional processing and therefore would not significantly affect the cost of steel. Even though the argument can be made that the increased toughness is not absolutely necessary, it is always better to have greater tough- ness if there is no cost penalty. If there was no temperature shift, extremely large cracks, greater than 350 mm, can be tolerated in mild steel. If the notch toughness requirements were increased, the increased tough- ness could also be used to offset the decrease in reliability associated with no longer decreasing the stress range for FCMs in the LRFD Specifications (A2). To make it more palatable, this increase could be combined with some loosening of the definition of FCMs that would, for example, allow two-girder bridges that are known to be redundant, as discussed earlier. Research currently underway at FHWA’s Turner–Fairbank Highway Research Center is studying ways of taking advan- tage of the benefits of higher toughness as in most modern plate steel, especially HPS. These benefits include: • Elimination of special in-service inspection require- ments, such as hands-on or arms-length inspection for fracture-critical structures for HPS. • Reduction in the frequency of inspections for the super- structure components of HPS bridges. • Elimination of the penalty for structures with low redun- dancy for HPS. Ultimately, the greatest benefit might be achieved by pro- viding a solid foundation for structural innovation. The cur-

47 rent U.S. practice of providing a high level of structural redun- dancy prevents engineers from considering other structural systems that can result in more efficient, lower-cost bridges. For example, two-girder and tied-arch bridges are rarely built today in the United States, even though experience has shown that in many situations they are very economical. A higher factor of safety against fracture will increase the reli- ability level of low-redundancy systems, thereby reducing this barrier to innovation. Although higher toughness makes bridges more tolerant to longer cracks, it does not significantly increase fatigue life. Fatigue cracks grow according to a power law; therefore, most of the fatigue life is spent growing the crack while it is very small. Additional fracture toughness, greater than the minimum specified values, will allow the crack to grow longer before the member fractures. But, at that late stage, the crack is growing so rapidly that relatively few cycles are needed to reach the end of the life. Ultimately, decisions to specify toughness beyond the minimum required level must be made based on cost. The current AASHTO steel toughness requirements were devel- oped using the cost factors that existed in the 1970s and con- sidering the state of the art in steel production at that time. Modern steel processing practice has made it more econom- ical to produce high-toughness steels such as A709 HPS 485W. In addition, the Grade 250, 345, and 345W steels pro- duced today have typical CVN that far exceed the minimum requirements. The current AASHTO specifications for mater- ial toughness may need to be reevaluated to take maximum advantage of the higher-toughness steels used today. INSPECTION AND NONDESTRUCTIVE TESTING Periodic in-service inspection provides a final safety net to detect cracks before they grow to a critical size. Because of the repetition of details in a bridge, when one crack is found, it is very likely that similar details may also be cracked. Therefore, the first and most urgent step in planning repairs and retrofits is to thoroughly inspect the bridge for other cracks, usually visually but often followed up with nonde- structive evaluation such as magnetic-particle testing (MT) or dye-penetrant testing (PT), especially for FCMs. The focus of inspection for fatigue cracks should be on details similar to the one that cracked on elements of the bridge in tension, with a priority for elements with high live-load stress ranges and FCMs (A46–A48). Elements for which the applied stress remains in compres- sion need not be inspected closely for fatigue cracks, because complete fracture is not possible. Owing to welding residual stresses, a crack can still occur in a structural element that undergoes cyclic loading even if the applied stress remains in compression. However, these cracks will usually arrest as they grow away from the welds as the tensile residual stress field either decreases or is relieved by the cracking (A4). Inspection is primarily performed visually. The survey noted that many agencies inspected non-FCBs from the ground with binoculars, whereas the FCBs were visually inspected hands on. Nondestructive testing (NDT) is used some for FCMs, in particular if there is an indication that is clearly not a crack from the visual inspection. NDT methods presently used in service for bridges include but are not lim- ited to the following: PT, MT, and ultrasonic testing (UT). Radiographic testing and eddy-current testing are usually only used in the shop and therefore are not discussed here. Dye-Penetrant Testing PT is used to detect surface discontinuities only. A penetrat- ing liquid dye, either visible or fluorescent, is placed on the surface of the member and will enter any discontinuities. After a period of time, up to 30 min for extremely fine, tight discontinuities, the excess dye is removed and the area is allowed to dry. A developer is then applied, pulling the resid- ual wet dye from the discontinuities as shown in Figure A5. Penetrant inspection is inexpensive, simple, and easy to learn. However, inspectors need to be properly trained in not- ing the difference between real and false indications that often occur from a slight weld undercut and other disconti- nuities that are not significant. Magnetic Particle Testing MT involves the use of magnetic field lines to determine whether surface or near surface cracks exist by the disruption of the lines. This disruption of lines results from a disconti- nuity in the member; for example, a crack. The material can either be magnetized through direct magnetization or by plac- ing a magnetic field (indirect magnetization) on the member. Once the field is established, magnetic particles (typically in the form of a powder) are placed on the inspection surface. Discontinuities are exposed when they are trapped in the leak- age of the magnetic field and the location, shape, and size of a crack can accurately be determined. Figures A6–A8 depict this process and the required equipment. MT can be conducted very quickly and, compared with other NDT methods, it is relatively cost-effective in terms of FIGURE A5 Example of indication from dye penetrant inspection.

48 equipment and procedures. In contrast to PT, MT can reveal shallow cracks below the surface, is very accurate, requires less time, and may be more economical after the equipment is obtained. This procedure is favored among many inspectors. Ultrasonic Testing UT inspection is another commonly used NDT method in practice. By using high-frequency sound waves, surface and subsurface discontinuities can be detected. As the sound waves travel through the material and reflect back, the presence and location of any discontinuities, which also cause reflections, in the member can be detected. This information is then dis- played on a cathode ray tube screen for interpretation. Advantages in using this type of test are the ability to detect small internal discontinuities, accuracy, and nearly instantaneous test results. The primary disadvantage to this test is that highly trained and experienced technicians are needed to operate and accurately interpret this type of test results. The international fabrication scanning tour noted that automated UT, providing a permanent record, is often used outside the United States. Because small, possibly innocuous, discontinuities can be detected with UT, acceptance criteria are presented in AWS D1.5. These acceptance criteria are workmanship standards; that is, they represent the typical quality level easily achiev- able by good welders. The AWS D1.5 UT acceptance crite- ria are not based on the effect that the rejectable discontinu- ities might have on the resistance to fatigue and fracture; they are typically more strict than necessary. The AWS standards are for new fabrication and were not intended for existing structures. It is a misapplication to apply these workmanship standards to an evaluation of exist- ing bridges. If there is no impact on fatigue and fracture, an owner will be far more reluctant to take out of service or repair (at owner expense) a structure that is found to have poor workmanship than the owner would be when the com- ponent is still in the fabrication shop and the repairs would be at the fabricator’s expense. Notwithstanding that it is a misapplication, the AWS D1.5 criteria have frequently been used to assess the UT of butt welds in service. On many occasions, an indication with a rejectable decibel (dB) rating will be cored out of a large groove weld for destructive examination and characterization of the actual flaw. Figure A9 shows the results of many of these investigations, as conducted by Dr. Eric Kaufmann at Lehigh University’s Advanced Technology for Large Struc- tural Systems (ATLSS) Center, plotted in terms of the dB rat- ing versus the actual flaw size. Also shown are the D1.5 rejec- tion limits for various thickness butt welds. It can be seen that the AWS criteria are very conservative. For typical plating thickness, the AWS criteria will reliably screen out defects of only a few millimeters in width. FIGURE A6 Magnetic particle testing—Placement of magnetic field. FIGURE A7 Magnetic particle testing—Application of magnetic particles. FIGURE A8 Magnetic particle testing—Indication.

49 Established fracture mechanics principles can be used to define acceptable initial crack sizes that will not propagate to the critical size in the lifetime of a structure (A15,A32). These types of calculations are referred to as “fitness-for-purpose” calculations, indicating that although a component may have rejectable discontinuity, it can be proven that the component is fit for its defined purpose (lifetime and anticipated loading). RETROFIT METHODS This section presents commonly used repair and retrofit tech- niques for fatigue-critical details, as well as retrofit tech- niques to improve the redundancy of fracture-critical bridges (FCBs). A distinction is made between a repair and a retro- fit: a repair is intended to arrest the propagation of a fatigue crack, whereas a retrofit is intended to either (1) upgrade the fatigue resistance and prevent the occurrence of fatigue cracking, or (2) create an alternative load path in the event of fracture to make an FCM into a non-FCM. Issues related to repair and retrofit of fatigue cracks will be discussed first followed by some discussion related to retrofit of FCBs. One of the best sources for information related to retrofit strategies is Fatigue and Fracture in Steel Bridges by John W. Fisher (A13). Although out of print, it remains an excellent source for material on retrofitting fatigue- or fracture-damaged bridges. Repair and Retrofit of Fatigue Cracks Many different methods are used for the repair of fatigue cracks and retrofit of fatigue-prone details. The choice of method depends on the circumstances of the fatigue cracking and may also depend on the availability of certain skills and tools from local contractors who would perform the repairs. Repair and retrofit techniques can be placed in three major categories: (1) surface treatments, (2) repair of through- thickness cracks, and (3) modification of the connection or the global structure to reduce the cause of cracking. Surface Treatments Weld toe surface treatments include grinding, gas tungsten arc or plasma remelting of the weld toe, and impact treat- ments. These techniques can be used as “weld improvement” retrofit methods; that is, for increasing the fatigue strength of uncracked welds. With any of these treatments, the improve- ment in fatigue strength can be attributed to one or a combi- nation of the following: • Improvements in the weld geometry and corresponding reduction in the stress concentration, • Elimination of some of the more severe discontinuities from which the fatigue cracks propagate, or • Reduction of tensile residual stress or the introduction of compressive residual stress (A49,A50). The easiest and lowest cost of these treatments is hammer peening, which is very effective and commonly used. Some of the methods, including hammer peening, can also be used for the repair of shallow surface cracks up to 3 mm deep. A relatively new process, known as Ultrasonic Impact Treatment, has been the subject of several recent studies (A51,A52). The process, developed in Russia, is similar to air hammer peening, but applies the treatment at a very high fre- quency, up to 35 kHz. This technique uses sound waves to excite a peening device that introduces compression into the steel at the toe of fillet welds. Fatigue testing done on treated and untreated specimens cut from the plate girder concluded that, for the conditions tested, Ultrasonic Impact Treatment altered the performance of a Category C detail by imparting to it the fatigue strength of a Category B detail. Further research is being conducted to address the effect Ultrasonic Impact Treatment will have, if any, on the fatigue threshold of details and how different fatigue stress ranges, welding processes, and quality control affect the results. Once either treatment is applied to the welds that have already been in service, the remaining fatigue life is at least as good as the life of the original detail when it was new. In other words, there is no remaining effect of prior fatigue load- ing cycles. In most cases, these treatments result in fatigue strength of the treated detail that is at least one fatigue “cat- egory” greater than the original detail; that is, the next great- est S–N curve in the AASHTO set of S–N curves can be used to predict the residual life of the repaired detail. Surface treat- ments only affect the weld toes; therefore, fatigue cracks may still develop from the weld roots. Hole Drilling Hole drilling is perhaps the most widely used repair method for fatigue cracks or retrofit method for fatigue-critical details. It is often used as a temporary measure to arrest a propagating crack, followed eventually by more extensive repairs. It is rare that any repair scheme such as repair welding or modifying 10 5 0 -5 -10 -15 0 2 4 6 8 10 12 Reject for t < 18 mm (70o class A) Reject for t < 64 mm Reject for t < 203 mm Defect Rating (db) Measured Defect Width (2a) (mm) FIGURE A9 Actual defect size from destructive examinations of cores from welds with UT indications of various dB rating. (Data from Lehigh University.)

the detail does not begin with drilling the crack tips. For retrofit, hole drilling is often used to isolate a detail or to intercept potential cracks before they can propagate far into main elements. By properly tensioning a high-strength bolt in the hole it can be considered a Category B detail. A hole by itself is basically a Category D detail. For repairs, the hole drilling method requires placing a hole at the tip of the crack, essentially blunting the tip of the crack, thus removing the high-stress concentration associ- ated with the sharp tip. However, the hole needs to be a spe- cific diameter to be successful in arresting the crack. Typi- cally, a hole diameter of 4 in. (200 mm) is recommended because this has proven to be effective. When a more refined estimate of the required hole size is necessary, relationships have been developed to define the size of the hole needed to arrest the crack (A53,A54). Appropriate checks on the net section capacity of the member should be made. The fatigue resistance of a hole can also be increased by the cold expansion of the hole, which upsets the material around the perimeter of the hole and introduces beneficial compressive residual stress around the hole. This technique is widely used in aluminum airframes. Once the hole is drilled, a tapered mandrel (also referred to as a drift pin) slightly larger than the hole can be forced through the hole by hitting the pin with a hammer. As the pin passes through the hole, the hole plastically deforms creating the compres- sive field around the hole. This method will not work when the hole used contains a crack. If the hole has a crack enter- ing it, the hole is forced open by the pin as the pin is driven into the hole because the cracked edge is compliant (flexible) and the hole does not provide sufficient constraint to induce the compressive stresses at the edge. However, if the hole is drilled ahead of the crack tip, the hole may be cold expanded. Adding Doubler and Splice Plates Another technique that can be used to repair through-thickness cracks is by the addition of doubler plates, or doublers. Dou- bler plates add material to the cross section to either increase or make up for the cracked cross-sectional area. Doubler plates may be bolted (Figure A10) or welded (Figure A11). From a fatigue-resistance standpoint, bolted doublers are always bet- ter than welded ones, because a high-strength bolted connec- tion can be considered an AASHTO Category B detail, whereas a welded connection will be Category E or worse. It is therefore usually recommended that only bolted doubler plates be used for permanent repair or retrofit on bridges. The philosophy of doublers for fatigue crack repair is to add cross-sectional area, which in turn reduces stress ranges. For instance, if a fatigue crack grows across the full depth of the bridge girder, there are two ways that it can be repaired. First, a vee and weld repair can be specified, but the base metal that is weld repaired will have at best a Category D fatigue resistance (A54–A56). To ensure that the weld repair 50 will have adequate fatigue resistance, doubler plates can be added after the weld repair to decrease the stress range. The one problem with this repair is the alignment of the two sides of the crack before the weld repair. As can be seen in Fig- FIGURE A10 Bolted doubler plate repair. Dotted line represents crack line beneath doubler plate and circle is the hole drilled at crack tip to intercept further growth. FIGURE A11 Welded doubler plate detail. (Note that corners should be rounded.)

51 ure A12, the cracked surface usually develops buckles, making their alignment difficult. In this case, the second option would be to use thicker doublers and bolt them to the girder. The thicker plates add enough cross section assuming the crack will not be weld repaired, and bolting them together then ensures that any buckles can be straightened out. This technique is par- ticularly useful when a full-depth crack forms in a bridge girder. The doubler plates are then meant to make up for the lost cross section from the crack. Doublers are also typically used to restore a section that has been heavily damaged by corrosion. Retrofit of Fracture-Critical Members to Make Them non-Fracture-Critical Members There are a few methods that have been developed that were identified in the literature to improve the redundancy of FCBs. Some of these were illustrated in the main body of the report. Others will be summarized briefly here. Prestressing Strands On some bridges, the addition of prestressing strands or rods has been used to supplement tension members. An example is the Girard Point Bridge that carries I-95 over the Schuylkill River near Philadelphia. The bridge is a double-deck can- tilevered truss that has a very long suspended center span. Stainless steel rods were added to supplement the primary hanger members in the event that the existing hanger truss member were to fracture. The members are preloaded so that the additional rods carry a portion of the dead load and so that if a hanger were to fracture, there would be minimal “snap” as the full load was dynamically transferred to the rods. Sim- ilar systems have been incorporated to other tension members in other trusses such as diagonals and chords. Another related retrofit technique is to string high- strength strands along the bottom face of the bottom flange of a fracture-critical plate girder. The strands are then pre- loaded and a portion of the dead load transferred to the newly added strands. Thus, in the event the existing tension flange fractures, there are additional tension members (strands) with sufficient capacity available to carry both dead and live loads. In addition, depending on the level of post-tensioning applied, the dead-load stresses (as well as live-load stresses) can be reduced in the existing bottom flange. Another approach that used post-tensioning was on the Hazard, Woodhead, Dunlavy, and Mandell Street tied arch bridges in Houston, Texas. In these bridges, the 2 ft × 2 ft tie girder was internally post-tensioned and encased in concrete. (As part of the strategy to improve redundancy the tie was not encased in concrete.) Four post-tensioning strands are in each tie. However, because the tie is encased in concrete, it is inac- cessible for future inspection. These recently built steel tied arch bridges span 224 ft (68 m) over the freeway and carry two lanes of traffic, two bicycle lanes, a utility parapet in each direction, and sidewalks outside of each arch for a total width of 60 ft (18 m). One of these spans is shown in Figure A13. Note the shallow tie girder. Post-tensioning of the arch tie provided redundancy and virtually eliminated tension in the tie, which allayed concerns about the history of problems with tie beams on other tied arch bridges, but necessitated passing the arch tie flanges through the junction with the arch rib. FIGURE A12 Full-girder-depth fatigue crack of Lafayette Street Bridge in St. Paul, Minnesota. FIGURE A13 Post-tension tied arch bridge in Houston, Texas.

Bolted Redundancy Plates Another technique that has been used on some bridges is the addition of bolted redundancy plates. This retrofit con- sists of bolting plates or angles to existing tension mem- bers. The primary function is to provide an additional com- ponent(s) so that if the tension flange of an existing member were to fail at some location along the length, the added component would assume the full dead and live loads of the tension flange. For example, the two-girder approach spans on the Poplar Street Bridge in East St. Louis, Missouri, have been retrofit by bolting thick HPS plates that would take the place of the tension flange along the web just above the tension flange, as shown in Figure 15 in chapter two of this report (page 17). Because the added components are bolted to the existing member, there is no direct path for the fracture to travel into the added component. Hence, the member becomes inter- nally redundant as with a riveted built-up member. For exam- ple, in a riveted tension flange comprised of several plates it has been observed that the cover plate can fully fracture, although the other elements of the member remain and the member continues to take the entire load. When a retrofit plate is added to a bridge with no live load, it also reduces the live-load stress range in the existing member. The technique has also been used in new construction on a large truss bridge that carries SR-33 over the Lehigh River near Easton, Pennsylvania. On this new bridge, redundancy plates were bolted alongside of selected tension chords that were identified to be critical, as shown in Figure 13 in chap- ter two of the report (page 16). The plates were fully spliced at the panel points and nominally connected to the member along the length. An advantage of redundancy plates used in the design is that the components share both dead and live loads throughout the life of the bridge. Although these techniques add internal member redun- dancy, they do not add overall structural load path redun- dancy. In other words, in the unlikely event that the entire lower chord failed only one lower chord remains. Use of Composite Construction The SR-33 Bridge near Easton, Pennsylvania, incorporated an additional measure to increase redundancy. The top chord is fully composite with the concrete deck (see Figure A14). Traditionally, in truss bridges, only the floorbeams and stringers are made composite with the deck. The deck is cast- in-place reinforced concrete supported by steel stringers and transverse floorbeams. It is the only composite truss in the state of Pennsylvania and possibly the United States. The structure is a four-span continuous haunched steel deck truss that is composite with the reinforced concrete deck. The main river span is 181 m and the depth of the trusses varies from 11 m to 22 m. 52 Controlled load testing and long-term monitoring of the bridge confirms that there is excellent load distribution between trusses and that stresses in the upper chord in the negative moment regions are very small owing to the com- posite action (A56). Hence, it is believed that the choice to make the deck composite with the upper chord added sub- stantial redundancy. Installation of Additional Girders This retrofit technique is only applied when a structure is to be widened. In cases where the existing structure is function- ally obsolete and additional or wider lanes are required, an additional exterior girder is sometimes added. If the new girder is adequately attached to the existing girders, full load sharing can be realized. Thus, a two-girder bridge can become a four-girder bridge and be removed from the list of FCBs in an owner’s inventory. It is emphasized that the new girder must be sufficiently attached to the existing girder and deck. This technique was employed on the Pennsylvania Turn- pike near the Valley Forge Interchange northwest of Philadel- phia. In this example, a two-girder bridge was widened by adding parallel girders adjacent to the existing riveted girders. In effect, a two-girder cross section was converted into a four- girder cross section. Although the girder spacings are unequal and the distance between the two original girders may still be considerable, the structure as a system should not be consid- ered fracture critical as long as the connections between the girders and to the concrete deck are adequate. The additional girders can also be used to retrofit cross girders. For example, two girders were used to provide redundancy to existing steel cross girders carrying multi- girder composite spans, as shown in Figure A15. The bridge carries I-95 just north of Philadelphia. The retrofit required Stringer note shear studs Shear studs on top chord FIGURE A14 Shear studs being installed on top chords, floorbeams, and stringers on the SR-33 bridge near Easton, Pennsylvania.

53 the addition of new steel support columns to carry the loads from the new cross girders to the footings. The retrofit was installed as a preemptive strategy, because no problems with the cross girders have been observed. Pin and Hanger Retrofits There are two common strategies used to retrofit pin and hanger bridges. Both will be discussed here. It is important to note that although this system improves the redundancy of the given pin and hanger assembly, it does not necessarily add load path redundancy to the entire structure. However, for the approaches that actually remove the pin and hanger and replace the detail with a full moment connection, the case could be made that, if sufficient diaphragms are present, there may be sufficient alternate load paths. Addition of Supplemental “Catcher” Systems This form of retrofit is typically used on pin and hanger sys- tems. In the typical application, an additional group of com- ponents are added to “catch” the suspended girder should the existing pin and hanger system fail. A typical installation of this system is shown in Figure A16. Removal of Pin and Hanger Assembly In this approach, the entire pin and hanger assembly is removed and replaced with a new short section of girder that is attached to existing portions of the girders with full moment splices. The girders are then made continuous for live load and even some proportion of dead load. Field instru- mentation conducted on the bridge in Figure A17 confirmed that after the retrofit, the bridge behaved as a typical contin- uous multispan bridge (A57). The ability of the structure to behave as a continuous mul- tispan bridge, primarily in the negative moment regions, must be adequately checked. During construction, either false work or strong backs are required to ensure that the bridge is stable. The process can be completed with a live load on the bridge. Figure A17 illustrates a two-girder bridge where the pin and hanger were removed and replaced. DEVELOPMENT OF FRACTURE CONTROL PLAN This section reviews the history of the development of the fracture toughness requirements for steel and weld filler metal, which are a central part of the bridge fracture control plan contained in D1.5. The requirements appear to have served their purpose; that is, there have been no catastrophic New steel support columns (typ.) Existing cross girder Newly added cross girders on each side of cross girder FIGURE A15 Additional support girders are used to provide redundancy to steel cross girders on I-95 bridge north of Philadelphia, Pennsylvania. “Catcher” beam FIGURE A16 Catcher system as used on a typical pin and hanger bridge. (Courtesy: Modjeski and Masters, Inc.)

fracture problems with bridges since their implementation. However, many things have changed since the development of these specifications. There are fundamental differences in the steel and the way it is produced. Some of these differ- ences may have some impact on the temperature shift and other assumptions in the original development of the tough- ness requirements. The difference between the fracture control plan provisions and the provisions for non-FCMs elsewhere in AASHTO/ AWS D1.5 is primarily that there are more strict fabrication and shop-inspection requirements to control weld flaws and other crack-like defects. For example, both radiographic test- ing and ultrasonic testing are required on all groove welds for fracture control elements. In addition, CVN requirements for welds and base metal are increased for fracture control ele- ments. The provisions result in an even lower probability of brittle fracture in new FCMs than for typical non-FCMs. REFERENCES A1. Task Committee on Redundancy of Flexural Systems of the ASCE–AASHTO Committee on Flexural Mem- bers of the Committee on Metals of the Structural Division, “State-of-the Art Report on Redundant Bridge Systems,” Journal of Structural Engineering, Vol. 111, No. 12, Dec. 1985. A2. AASHTO LRFD Bridge Design Specifications, 3rd ed., American Association of State Highway and Trans- portation Officials, Washington, D.C., 2004. A3. Fisher, J.W., “The Evolution of Fatigue Resistant Steel Bridges,” 1997 Distinguished Lectureship, Paper No. 971520, 76th Annual Meeting of the Transportation Research Board, Washington, D.C., Jan. 12–16, 1997, pp. 1–22. A4. LRFD Specification for Structural Steel Buildings, Manual of Steel Construction: Load and Resistance 54 Factor Design, 3rd ed., American Institute of Steel Construction, Chicago, Ill., 1999. A5. “Steel Structures,” AREMA Manual for Railway Engi- neering, American Railway Engineering and Mainte- nance of Way Association, Landover, Md., 2002. A6. Structural Welding Code—Steel, ANSI/AWS D1.1- 02, American Welding Society, Miami, Fla., 2002. A7. Limit States Design of Steel Structures, CSA S16- 2001, Canadian Standards Association, Toronto, ON, Canada, 2001. A8. Fisher, J.W., K.H. Frank, M.A. Hirt, and B.M. McNamee, NCHRP Report 102: Effect of Weldments on the Fatigue Strength of Steel Beams, Highway Research Board, National Research Council, Wash- ington, D.C., 1970, 114 pp. A9. Fisher, J.W., P.A. Albrecht, B.T. Yen, D.J. Klinger- man, and B.M. McNamee, NCHRP Report 147: Fatigue Strength of Steel Beams with Welded Stiffen- ers and Attachments, Transportation Research Board, National Research Council, Washington, D.C., 1974, 85 pp. A10. Keating, P.B. and J.W. Fisher, NCHRP Report 286: Evaluation of Fatigue Tests and Design Criteria on Welded Details, Transportation Research Board, National Research Council, Washington, D.C., Sep. 1986, 66 pp. A11. Petershagen, H. and W. Zwick, Fatigue Strength of Butt Welds Made by Different Welding Processes, IIW-Document XIII-1048-82, International Institute of Welding, West Germany, 1982. A12. Petershagen, H., “The Influence of Undercut on the Fatigue Strength of Welds—A Literature Survey,” Welding in the World, Vol. 28, No. 7/8, 1990, pp. 29–36. A13. Fisher, J.W., Fatigue and Fracture in Steel Bridges, John Wiley and Sons, New York, N.Y., 1984, 336 pp. A14. Anderson, T. L., Fracture Mechanics—Fundamentals and Applications, 2nd ed., CRC Press, Boca Raton, Fla., 1995, 88 pp. A15. Barsom, J.M. and S.T. Rolfe, Fracture and Fatigue Control in Structures: Applications of Fracture Mechanics, 3rd ed., American Society for Testing and Materials, West Conshohocken, Pa., 1999. A16. Broek, D., Elementary Fracture Mechanics, 4th ed., Martinis Nijhoff Publishers, Dordrecht, the Nether- lands, 1987. A17. Kober, W., E.J. Dexter, E.J. Kaufmann, B.T. Yen, and J.W. Fisher, “The Effect of Welding Discontinuities on the Variability of Fatigue Life,” Fracture Mechan- ics, Vol. 25, ASTM STP 1220, F. Erdogan and R.J. Hartranft, Eds., American Society for Testing and Materials, Philadelphia, Pa., 1994. A18. Guide for Fatigue Strength Assessment of Tankers, American Bureau of Shipping, New York, N.Y., June 1992. A19. Fatigue Design Guidance for Steel Welded Joints in Offshore Structures, UK Health & Safety Executive (formerly the UK Department of Energy), Her Location of former pin and hanger FIGURE A17 Typical complete girder splice installed to replace a pin and hanger connection in a two-girder bridge.

55 Majesty’s Stationery Office, London, United King- dom, 1984. A20. Roberts, R., et al., Corrosion Fatigue of Bridge Steels, Vols. 1–3, Reports FHWA/RD-86/165, 166, and 167, Federal Highway Administration, Washington, D.C., May 1986, 44, 168. 564 pp. A21. Outt, J.M.M., J.W. Fisher, and B.T. Yen, Fatigue Strength of Weathered and Deteriorated Riveted Members, Report DOT/OST/P-34/85/016, Federal Highway Administration, Department of Transporta- tion, Washington, D.C., Oct. 1984, 138 pp. A22. Albrecht, P. and C. Shabshab, “Fatigue Strength of Weathered Rolled Beam Made of A588 Steel,” Jour- nal of Materials in Civil Engineering, Vol. 6, No. 3, 1994, pp. 407–428. A23. Fisher, J.W., A. Nussbaumer, P.B. Keating, and B.T. Yen, NCHRP Report 354: Resistance of Welded Details Under Variable Amplitude Long-Life Fatigue Loading, Transportation Research Board, National Research Council, Washington, D.C., 1993, 38 pp. A24. Miner, M.A., “Cumulative Damage in Fatigue,” Jour- nal of Applied Mechanics, Vol. 12, 1945, p. A-159. A25. Guide Specifications for the Fatigue Evaluation of Existing Steel Bridges, American Association of State Highway and Transportation Officials, Washington, D.C., 1990. A26. Moses, F., C.G. Schilling, and K.S. Raju, NCHRP Report 299: Fatigue Evaluation Procedures for Steel Bridges, Transportation Research Board, National Research Council, Washington, D.C., 1987, 100 pp. A27. Dexter, R.J. and J.W. Fisher, “Fatigue and Fracture,” In Handbook of Bridge Engineering, W.F. Chen, Ed., CRC Press, Boca Raton, Fla., 1999. A28. Youngberg, C.J., R.J. Dexter, and P.M. Bergson, Fatigue Evaluation of Steel Box-Girder Pier Caps: Bridge 69832, Report MN/RC 2003-18, Minnesota Department of Transportation, St. Paul, Mar. 2004, 67 pp. A29. Wright, W.J., J.W. Fisher, and B. Sivakumar, “Hoan Bridge Failure Investigation,” Federal Highway Administration, Washington, D.C., 2001. A30. Kaufman, E.J., R.J. Connor, and J.W. Fisher, “Failure Analysis of the US 422 Girder Fracture—Final Report,” ATLSS Report No. 04-21, Center for Advanced Technology for Large Structural Systems, Lehigh University, Bethlehem, Pa., Oct. 2004. A31. Connor, R.J., E.J. Kaufmann, J. Jin, and C.H. Davies, “Forensic Investigation of the SR422 Over the Schuylkill River Girder Fracture,” Proceedings of the Twenty-First International Bridge Conference, Pitts- burgh, Pa., June 14–16, 2004. A32. “Guide on Methods for Assessing the Acceptability of Flaws in Metallic Structures,” BS 7910, British Stan- dards Institute, London, 1999. A33. Dexter, R.J. and J.W. Fisher, “Fatigue and Fracture,” In Steel Design Handbook, LRFD Method, A.R. Tam- boli, Ed., McGraw–Hill, New York, N.Y., 1997. A34. Dexter, R.J., W.J. Wright, and J.W. Fisher, “Fatigue and Fracture of Steel Girders,” Journal of Bridge Engineering, Vol. 9, No. 3, May/June 2004, pp. 278–286. A35. Dexter, R.J., “Fatigue and Fracture,” The Structural Engineering Handbook, 2nd ed., E.M. Lui, Ed., CRC Press, Boca Raton, Fla., 2004. A36. “State Cites Defective Steel in Bryte Bend Failure,” Engineering News Record, Vol. 185, No. 8, Aug. 20, 1970. A37. Vehovar, L., “Hydrogen-Assisted Stress-Corrosion of Prestressing Wires in a Motorway Viaduct,” Engineer- ing Failure Analysis, Vol. 5, No. 1, 1998, pp. 21–27. A38. Kulak, G.L., J.W. Fisher, and J.H.A. Struick, Guide to Design Criteria for Bolted and Riveted Joints, 2nd ed., Prentice Hall, Englewood Cliffs, N.J., 1987, 352 pp. A39. Roberts, R. and G.V. Krishna, “Fracture Behavior of A36 Bridge Steels,” Report FHWA-RD-77-156, Fed- eral Highway Administration, Washington, D.C., 1977, 59 pp. A40. Roberts, R., G.R. Irwin, G.V. Krishna, and B.T. Yen, “Fracture Toughness of Bridge Steels—Phase II,” Report FHWA-RD-74-59, Federal Highway Adminis- tration, Washington, D.C., 1974, 418 pp. A41. Schilling, C.G., K.H. Klippstein, J.M. Barsom, S.R. Novak, and G.T. Blake, Low Temperature Tests of Simulated Bridge Members, Report No. 97.021- 001(3), American Iron and Steel Institute, Washing- ton, D.C., 1972. A42. Hartbower, C.E., “Reliability of the AASHTO Tem- perature Shift in Material Toughness Testing,” Struc- tural Engineering Series No. 7, Federal Highway Administration, Washington, D.C., 1979. A43. Wright, W.J., “High Performance Steel—Research to Practice,” Public Roads, Vol. 60, No. 4, 1997, pp. 34–38. A44. Wright, W.J., “Fatigue Strength and Fracture Resis- tance of HPS-485W High Performance Steel—Work in Progress,” Paper No. T103-6, Proceedings, Struc- tural Engineers World Congress, Elsevier, 1998. A45. Wright, W.J., H. Tjiang, J. Hartman, and P. Albrecht, “Fracture Resistance of Modern Bridge Steels,” Pro- ceedings, ASCE Structures Congress, Philadelphia, Pa., 2000. A46. Yen, B.T., T. Huang, L.-Y. Lai, and J.W. Fisher, Manual for Inspecting Bridges for Fatigue Damage Conditions: Final Report, Report FHWA-PA-89-022 + 85-02, Fritz Engineering Laboratory, Bethlehem, Pa., 1990, 174 pp. A47. Demers, C. and J.W. Fisher, Fatigue Cracking of Steel Bridge Structures, Vol. I: A Survey of Localized Crack- ing in Steel Bridges—1981 to 1988, Report FHWA- RD-89-166, also, Vol. II: A Commentary and Guide for Design, Evaluation, and Investigating Cracking, Report FHWA-RD-89-167, Federal Highway Administration, McLean, Va., Mar. 1990. A48. Harland, J.W., R.L. Purvis, D.R. Graber, P. Albrecht, and T.S. Flournoy, Inspection of Fracture Critical

Bridge Members, supplement to the Bridge Inspec- tor’s Training Manual, final report, Report FHWA-IP- 86-26 (PB87-163440), Federal Highway Administra- tion, McLean, Va., Sept. 1986, 232 pp. A49. Braid, J., R. Bell, and D. Militaru, “Fatigue Life of as- Welded, Repaired, and Hammer-Peened Joints in High-Strength Structural Steel,” Welding in the World/Le Soudage le Monde, Vol. 39, No. 5, 1998, pp. 248–261. A50. Harrison, J.D., “Further Techniques for Improving the Fatigue Strength of Welded Joints,” British Welding Journal, Vol. 13, No. 11, 1966. A51. Tryfyakov, V., P. Mikheev, Y. Kudryavtsev, and D. Reznik, “Ultrasonic Impact Peening Treatments of Welds and Its Effect on Fatigue Resistance in Air and Seawater,” 25th Annual Offshore Technology Confer- ence, Houston, Tex., May 3–6, 1993. A52. Roy, S., J. Fisher, and B. Yen, “Fatigue Resistance of Welded Details Enhanced by Ultrasonic Impact Treatment (UIT),” Proceedings of the Eleventh International Offshore and Polar Engineering Con- ference, Stavenger, Norway, June 17–22, 2001, pp. 309–313. 56 A53. Fisher, J., B.M. Barthelemy, D.R. Mertz, and J.A. Edinger, NCHRP Report 227: Fatigue Behavior of Full-Scale Welded Bridge Attachments, Transporta- tion Research Board, National Research Council, Washington, D.C., 1980, 47 pp. A54. Dexter, R., R. Fitzpatrick, and D. St. Peter, “Fatigue Strength and Adequacy of Fatigue Crack Repairs,” Report SSC-425, Ship Structure Committee, Wash- ington, D.C., May 2003. A55. Gregory, E.N., G. Slater, and C.C. Woodley, NCHRP Report 321: Welded Repair of Cracks in Steel Bridge Members, Transportation Research Board, National Research Council, Washington, D.C., 1989, 52 pp. A56. Dexter, R.J. and B.A. Kelly, “Research on Repair and Improvement Methods,” International Conference on Performance of Dynamically Loaded Welded Struc- tures, Proceedings of the IIW 50th Annual Assembly Conference, San Francisco, Calif., July 13–19, 1997, pp. 273–285. A57. Frank, K.H. and C.F. Galambos, “Application of Fracture Mechanics to Analysis of Bridge Failure,” Proceedings, Specialty Conference on Safety and Re- liability of Metal Structures, 1972.