Performance-Based Seismic Bridge Design (2013)

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

12 CHAPTER THREE REVIEW OF INDUSTRY PRACTICE A literature review of the current seismic design standard of practice for the bridge, building, and waterfront/marine industries was conducted. This chapter provides the results. BRIDGE INDUSTRY PRACTICE Current AASHTO Practice Two seismic design methods are codified as minimum stan- dards and permitted by AASHTO. One is a force-based method that is embedded into the AASHTO LRFD Bridge Design Specifications (AASHTO 2012), referred to herein as the AASHTO LRFD, and the other is a displacement-based method that is the basis of the AASHTO Guide Specifica- tions for Seismic Bridge Design (AASHTO 2011), referred to as the AASHTO SGS. The LRFD gives these two meth- ods equal weight, thus permitting the displacement-based method of the SGS to be used in lieu of the force-based method, even though the displacement-based method is outlined only in a guide specification. Both methods use a single-level earthquake input, a 1,000-year return period ground motion. The force-based method has its roots in the improved design procedures that followed the 1971 San Fernando earthquake in southern California. Caltrans and AASHTO quickly updated their design procedures, and in 1981 the Applied Technology Council (ATC) published ATC-6, Seis- mic Design Guidelines for Highway Bridges (ATC 1981). AASHTO adopted this document as a guide specification in 1983, and it was formally adopted into the Standard Speci- fications for Highway Bridges in 1991 following the Loma Prieta earthquake. These design provisions became the basis for the seismic provisions included in the AASHTO LRFD. These force-based provisions were modified over the years as improvements were identified; however, the provisions remain largely as they were formatted in the ATC-6 document. The AASHTO seismic design provisions seek to pro- duce a structure that can resist more common smaller earth- quakes without significant damage and to resist larger, rare earthquakes without collapse. However, in the larger event the damage may be severe enough that repair of the struc- ture is not feasible; the objective is simply to prevent loss of life. While the design approach generally seeks to deliver these performance objectives, there is no direct quantitative check of multilevel earthquake loading, nor is there a direct linkage between the design parameters checked and actual damage states. From the perspective of performance objectives, the two specifications differ in that only the AASHTO LRFD addresses design of more important structures. The AAS- HTO SGS has its origins in part in the Caltrans Seismic Design Criteria (SDC) (2006a) for ordinary standard or conventional bridges. In the case of a bridge with a higher importance being designed with the AASHTO SGS, proj- ect-specific criteria would need to be developed, which is the approach that Caltrans uses for such bridges (Caltrans 2010b). The AASHTO LRFD defines three operational clas- sifications of bridges: Other, Essential, and Critical. The AASHTO LRFD commentary describes Essential bridges as those that should be open to emergency vehicles immedi- ately after a 1,000-year event. Critical bridges must remain open to all traffic after the design event and be open to emer- gency vehicles after a 2,500-year event. However, such per- formance is not directly assessed. The force-based method, as implemented in the AAS- HTO LRFD, is built around the capacity design process that has its origins in New Zealand in the late 1960s. The pro- cess was credited to John Hollings by Robert Park in his interview with Reitherman (2006). John Hollings (Park and Paulay 1975) summarized the process thus: In the capacity design of earthquake-resistant structures, energy-dissipating elements of mechanisms are chosen and suitably detailed, and other structural elements are provided with sufficient reserve strength capacity, to ensure that the chosen energy-dissipating mechanisms are maintained at near full strength throughout the deformations that may occur. In the force-based procedure an elastic analysis of the bridge is performed under the requisite earthquake load- ing and internal forces are determined. Forces in elements that are those chosen for energy dissipation—typically col- umns—are reduced by a response modification factor, R, and then combined with concurrent nonseismic forces to gener- ate the design forces. These forces would typically be in the form of column moments at selected plastic hinging locations. The reinforcement for these locations is chosen to match the

13 required design moments; then these locations are prescrip- tively detailed to be adequately ductile. The remainder of the bridge, including foundations, superstructure, bearings, abut- ments, and the nonyielding portion of the columns, is designed to be able to withstand the maximum possible forces—known as overstrength forces—that the plastic mechanism would ever be capable of generating. This process essentially satis- fies the capacity design objective, although a direct check of the actual expected response, inclusive of yielding effects and demand displacements, is never made, and a direct check of ductility capacity is, likewise, never made. This process was developed to be expedient for design using elastic analysis tools, and is further discussed in chapter five. The aforementioned R-factors are in effect a measure of the ductility capacity of the structural system: Large R-fac- tors imply that the system has a high displacement ductil- ity capacity and small R-factors imply low displacement ductility capacity. One difficulty of this method is that a single R-factor cannot provide a reliable method of damage or performance control under certain structural configura- tions. For example, two reinforced concrete columns that differ only in their height will have two different displace- ment ductility capacities (and therefore R-factors); the lon- ger column will have a lower ductility capacity owing to the increased influence of elastic deformations to the overall dis- placement (i.e., the ratio of plastic hinge length to the overall column length reduces with an increase in column length). Changes in behavior of this kind are best captured using dis- placement-based methodologies, such as those adopted into the AASHTO SGS. The displacement-based method in the AASHTO SGS focuses the designer’s attention on checking the system deformation capacity rather than selecting the precise resistance of the yielding or energy dissipating elements. This method is based heavily on the Caltrans practice for conventional bridges (Caltrans 2006a). The design process then becomes one of checking a trial design, rather than a linear progression of steps to calculate required internal forces in the structure. The process still follows the capac- ity design overall methodology, in that locations for dam- age are selected; these locations then are detailed to deliver adequate displacement or ductility capacity, and that capac- ity is directly checked. In the displacement-design process, the effect of confinement steel, for example, is directly included in the calculation of displacement capacity. Thus, the designer has some direct control over the amount of ductility or deformation capacity that will be provided ver- sus the amount of ductility that is required. The elements that are not part of the energy-dissipating mechanism are subsequently designed to be adequately strong under the maximum expected actions of the plastic mechanism. In principle, this step is identical to the one for force design. In application, the process differs primarily by the material strength factors that are used. In the AASHTO SGS displacement-based design process for the high seismic areas—Seismic Design Category D (SDC D)—the deformation capacity is controlled by limit- ing the maximum amount of tensile strain in the reinforce- ment steel and the maximum concrete compressive strain. The design method links element strains to member curva- ture, then to member rotations, and finally to member and systems displacements. Figure 3 shows the relationship between global and local deformations and damage, where a cantilever reinforced con- crete column is subjected to an inertial lateral force (F) at the FIGURE 3 Cantilever column deformations and limit states.

14 lae are shown in Figure 4 for SDC B and Figure 5 for SDC C. The averaging process used is evident in the figures by the SDC line’s position relative to the neighboring lines. An estimate of the displacement capacity of the column is cal- culated from the column aspect ratio. The graphs represent cantilever columns with fixity at one end only. Other con- figurations are handled by decomposing the columns into equivalent cantilevers joined at the inflection points. FIGURE 4 SGS Seismic Design Category B displacement capacity (after Imbsen 2006). FIGURE 5 SGS Seismic Design Category C displacement capacity (after Imbsen 2006). For SDC B, this method produces a displacement capac- ity that is a conservative estimate of the spalling limit state. It is conservative because the implicit capacity is less than the average spalling limit state owing to the slightly conservative nature of the analytical spalling data. Like- wise, for SDC C the displacement capacity lies between spalling and the attainment of a displacement ductility of four. These estimates are meant to be easy to calculate and are conservative as shown. The limits are also linked to the minimum confinement reinforcement required in columns in SDCs B and C. These categories each have lower required transverse steel contents, leading to lower expected ductility capacity. Therefore, the maximum per- mitted ductility demand must also be kept low. The idea behind this approach was to ease congestion of reinforce- ment, provided that the lower capacities were adequate for the anticipated demands. If designers wish to expend more effort or if the implicit capacities are too low, they may use the more rigorous displacement capacity calculation center of gravity of the substructure, resulting in the deformed shape shown. Local deformations (strains) are related to the global displacement (∆) through the curvature distribution along the height of the column (often idealized as shown, tak- ing advantage of the plastic hinge length, Lp). Finally, using the strain limits, displacement limits can be determined as indicated on the force-deflection (pushover) response. This calculation is not made in the force-based method. Because of this key difference, it is logical that the displace- ment-based process of the AASHTO SGS is the appropriate method into which to incorporate performance-based design. Although the AASHTO LRFD force-based method attempts to differentiate ordinary, critical, and essential bridges, giv- ing the impression of accommodating different performance objectives, the method is ill suited for a performance-based process specifically because deformation adequacy is not directly checked at the earthquake demand level. That said, at present the AASHTO SGS displacement-based method addresses only ordinary bridges and does not provide crite- ria or guidance for more important structures. This is a key gap that easily could be closed in the near future. It is also important to recognize that performance-based design procedures are possible because a capacity-design process is used. The selection of damage-tolerant elements and their subsequent design to accommodate earthquake demands permits these elements to be designed to respond with more or less damage, depending on the performance objectives desired. Such a process is predicated on the ability to relate engineering demand parameters—strain, rotations, and so on—to damage states and then operational perfor- mance. Thus, the capacity design method is a key compo- nent of the PBSD process. Although both AASHTO design methods are based on capacity design, the AASHTO SGS displacement-based method is better suited for extension into performance-based design. The AASHTO SGS method could be converted to a nominally performance-based approach by using concrete and reinforcing steel strain limits that are correlated to specific damage states—for instance, spalling or bar buckling. The AASHTO SGS method uses implicit formulae to calculate the displacement capacity of reinforced concrete columns for the intermediate Seismic Design Categories (SDCs) B and C. The formulae were derived using data from Berry and Eberhard’s (2003) database of column damage, whereby statistics for experimental tests of columns were developed using the two damage states, spalling of the cover concrete and buckling of reinforcement (discussed in fur- ther detail in chapter six). The experimental spalling data are averaged with the AASHTO analytical spalling limit state to calculate the displacement capacity for SDC B, and the experimental spalling data are averaged with the analytical data corresponding to attainment of a displacement ductility of four for SDC C. The data used for the implicit formu-

15 • ASL 1: 0–15 years • ASL 2: 16–50 years • ASL 3: >50 years. The PLs range from PL0 to PL3, and the correspond- ing expected postdesign earthquake damage levels are as follows: • PL0: No minimum—No minimum level of perfor- mance is specified. • PL1: Life safety—Significant damage is sustained, service is significantly disrupted, but life safety is preserved. The bridge may need to be replaced after a larger earthquake. • PL2: Operational—Damage sustained is minimal and service for emergency vehicles should be available after inspection and clearance of debris. Bridges should be repairable with or without traffic flow restrictions. • PL3: Fully operational—Damage sustained is negli- gible and full service is available for all vehicles after inspection and debris clearance. Damage is repairable without interrupting traffic. Importance is set as either standard or essential, where essential is defined as a bridge that (1) provides for second- ary life safety, such as emergency response vehicle use, (2) would create a major economic impact, (3) is formally defined in an emergency response plan as critical, or (4) is a critical link in security and/or defense road network. Table 1 presents the minimum performance levels, determined by combining seismic hazard level, impor- tance, and ASL. TABLE 1 MINIMUM PERFORMANCE LEVELS FOR RETROFITTED BRIDGES Earthquake Ground Motion Bridge Importance and Service Life Category Standard Essential ASL 1 ASL 2 ASL 3 ASL 1 ASL 2 ASL 3 Lower-Level Ground Motion 50%/75 years (approx. 100 years) PL0 PL3 PL3 PL0 PL3 PL3 Lower Upper Ground Motion 7%/75 years (approx. 1,000 years) PL0 PL1 PL1 PL0 PL1 PL2 Source: FHWA (2006). These performance criteria are then combined with appropriate assessment techniques to determine whether ret- rofit is required. From that point, a retrofit strategy is selected (if required), then approaches to satisfy that strategy are developed and retrofit measures are defined to provide the method for SDC D and, by doing so, calculate a larger dis- placement capacity; however, more confinement steel will typically be required. This is a simple example of how data that relate engi- neering parameters to damage are currently used. A true PBSD methodology will require more data and correlation of this nature. In the case of probabilistically based PBSD, the statistical dispersion of the data around a central ten- dency will also be required. This will be discussed further later in this synthesis. Much of what is discussed in this synthesis is related to reinforced concrete (RC) construction following the Type 1 (ductile substructure with essentially elastic super- structure) design strategy as defined in the AASHTO SGS. This type of construction is the most common, and the bulk of laboratory testing and design methodologies apply to RC construction. Ultimately, the PBSD tech- niques developed for RC will have to also be developed for other types of construction, including Type 1 struc- tures with steel columns or concrete-filled-tube columns. AASHTO Type 2 structures, those with an essentially elastic substructure and ductile steel superstructures (e.g., ductile cross frames or diaphragms) have seen relatively little research attention in the context of PBSD, certainly constituting an area of research need. AASHTO Type 3 structures, those with an elastic superstructure and sub- structure with a fusing mechanism (seismic isolation and supplemental damping devices) between the two, are well researched, with many publications specifically devoted to their analysis and design. Chapter six briefly discusses Type 2 and 3 structures in terms of their ability to reduce structural damage. The two AASHTO seismic design specifications, LRFD and SGS, are becoming increasingly difficult to maintain because it is challenging to maintain parity of treatment between them as new information is added. It is likely that a choice will need to be made in the not-to-distant future regarding keeping both design methodologies—force-based and displacement-based—or dropping the force-based pro- cedures. These maintenance challenges will likely increase if PBSD elements, whether mandatory or optional, are adopted into the specifications. FHWA Retrofitting Manual The FHWA Seismic Retrofitting Manual for Highway Struc- tures: Part 1—Bridges (2006) is essentially a performance- based guideline, which uses a multiple-level approach to performance criteria. It defines two seismic hazard levels, 100-year and 1,000-year return periods, and uses antici- pated service life (ASL), along with importance, to catego- rize suggested performance levels (PLs). The ASL values are as follows:

16 selected approach. The strategy is the overall plan for retro- fitting the bridge and may include several approaches made up of different measures. Example strategies may include do nothing; partial retrofit of the superstructure; or full ret- rofit of the superstructure, substructure, and foundations. Approaches may include such things as strengthening, force limitation, or response modification. Measures are physical modifications to the bridge, such as column jacketing. While the Retrofitting Manual includes performance- based objectives for different levels of importance, service life, and ground motions, it does not address in detail the linkage between the performance levels, PL0 to PL3, and damage limit states. For reinforced concrete, for example, the following limit states are quantified in terms of curvature: • Unconfined concrete compression failure • Confined concrete compression failure • Buckling of longitudinal bars • Fracture of longitudinal reinforcement • Low-cycle fatigue of longitudinal reinforcement • Lap-splice failure • Shear failure • Joint failure. Using these damage limit states and the descriptions of the required performance or service levels following an earthquake, an engineer can establish criteria that would deliver the required performance. For example, to meet the PL3 performance level of fully operational, the limiting per- formance levels would be to prevent— • Unconfined concrete compression failure • Shear degradation • Yielding of the longitudinal bars • Joint failure • Lap-splice failure. The more serious longitudinal bar damage states, such as buckling, fracture, and low-cycle fatigue, would not be an issue if yield of the bars were prevented. The relative lack of detailed procedures to link perfor- mance, damage states, and finally operation reflects the state of practice in 2006, when the manual was published. How- ever, the damage states are overtly stated in terms of physical damage, not simply strain limits or curvature limits in a table. These physical damage states may be compared with those of Berry and Eberhard’s database, discussed in the previous section, which were used for the AASHTO implicit displace- ment equations. However, in the Retrofitting Manual, mate- rial properties and curvature limits reflect the performance and behavior of older materials, and as such the limiting val- ues may be considerably lower than those associated with new construction. These damage states also correspond to the Caltrans visual damage guidelines described in chapter six. Linking these observations together, one may conclude that the way forward into PBSD would be to develop specifi- cations that combine these physical damage states with ana- lytical methods of structural analysis on the one hand, and performance and loss estimates on the other hand. An alternative rating method called the Seismic Rating Method Using Expected Damage is briefly outlined in the retrofit manual. The method provides a concise overview of the process of using the National Bridge Inventory database, standard bridge fragility functions, (the concept of fragil- ity functions is defined in the Damage Prediction section of chapter six. Briefly, a fragility function relates the like- lihood or probability of attaining a specified damage state to an EDP such as drift, where, for example, first yield of a column can be related probabilistically to drift.), and repair cost data to calculate a ranking, R. These data are combined with estimates of indirect losses, network redundancy, non- seismic deficiencies, remaining useful life, and other issues to determine an overall priority for retrofit. This method fol- lows the four basic steps of probabilistic PBSD to determine the retrofit priority. Although the expected damage method provides an overview of the process, significant data and methodology remain to be developed before the method can be applied with the same level of precision as the more conventional deterministic techniques, which are outlined in detail in the Retrofitting Manual. However, those conventional tech- niques are well developed only through the first two steps of PBSD, seismic hazard analysis and structural analysis. The remaining two steps, damage prediction and loss pre- diction, require more development before they mature, and this development will likely be a focus area in the coming years in earthquake engineering research. BUILDINGS INDUSTRY PRACTICE Overview Buildings seismic design practice evolved similarly to bridge design practice in that life safety has been the pri- mary minimum goal of both design methodologies. Like the AASHTO bridge seismic design procedure, the build- ing codes sought indirectly to provide structures that could resist smaller, more frequent earthquakes with little or no damage and larger, rare earthquakes with significant dam- age, but without loss of life—hence the name, life safety. For example, The International Building Code (2009) defines as its purpose “to establish the minimum requirements to safe- guard the public health, safety and general welfare.…” How- ever, there is no direct check that the life safety performance objective is met for a code-compliant design. Instead, the design must comply with given design parameters and pre- scriptive detailing requirements. If these requirements are

17 met, adequate seismic performance is implied in much the same way as in the AASHTO LRFD. The implication is that if the structure meets the code, then life safety is reasonably assured. However, this approach, while simple, can result in some structures performing better than others under earth- quake loading, even though the structures were designed to the same code. The need to seismically rehabilitate existing buildings has led to the development of “first-generation guidelines” that have PBSD as their core objective. In recognition that exist- ing buildings contained elements that did not conform to new building design requirements, a need existed for alternative means of setting criteria for rehabilitation that departed from criteria for new buildings. Often, existing building elements cannot be made to perform to the strength, stiffness, and ductility levels expected of new buildings, so different crite- ria were needed to guide such rehabilitation. The Structural Engineers Association of California (SEAOC), with FEMA funding, produced its Vision 2000 report, Performance Based Seismic Engineering of Buildings, in 1995 with the goal of defining both rehabilitation and new building seis- mic design criteria. It was the first such document to define multiple discrete levels of earthquake design. These levels are shown in Figure 6. Multiple performance levels were defined, along with multiple earthquake design levels. Then, performance objectives were defined as groups of combined earthquake and performance levels. For example, the Safety Critical Performance Objective was a combination of fully operational performance in the rare earthquake and opera- tional performance in the rare event. At about the same time, FEMA funded the development of national guidelines for the rehabilitation of buildings, which led to the publication of FEMA 273, NEHRP Guidelines for the Seismic Rehabilitation of Buildings (1996). FEMA 273, along with its commentary, FEMA 274, used a similar mul- tiple design level approach. For more important structures, the criteria became more rigorous. The FEMA documents used slightly different earthquake hazard levels, but the con- cept was the same as that first presented by SEAOC. In 2000, FEMA 273 and FEMA 274 were revised into FEMA 356, Prestandard and Commentary for the Seismic Rehabilitation of Buildings (2000f), which represented the second genera- tion of PBSD guidelines for existing buildings. FEMA 356 was subsequently adopted in 2006 as ASCE 41-06, Seismic Rehabilitation of Existing Buildings, which is still the stan- dard for developing rehabilitation designs for buildings. In the ASCE 41-06 approach, several earthquake levels are assessed, depending on the overall performance objec- tive that is selected (see Table 2 for the performance levels and Table 3 for the definitions of the performance levels and their associated damage states). The acceptance criteria are also provided on an element-by-element basis, depending on the desired performance level that is being checked. Thus, more restrictive limits are provided for each element type and for each performance level. The ASCE 41 methodology permits both force- and displacement-based assessments to be made. The force-based approach uses m-factors, which are essentially element-based R-factors. The displacement- based approach uses deformation limits, such as element rotations for moment frames and shear walls and element displacements for bracing elements. While both force- and displacement-based approaches are used, the displacement- based method is preferred and is required in some cases, depending on structural regularity, desired performance objectives, and other parameters. Structural actions are checked at the element level, and each primary lateral force resisting element is classified as either force-controlled or deformation-controlled. When checking force-controlled (brittle) elements, the nominal resistance is used to form a design resistance similar to FIGURE 6 Performance objectives for buildings (SEAOC Vision 2000).

18 way column shear is checked in AASHTO seismic design methods. When checking deformation-controlled elements, maximum inelastic deformations for each performance level are specified. For example, a reinforced concrete beam with symmet- ric top and bottom reinforcement, transverse reinforcement that conforms to minimum confinement details, and a low shear demand would have permissible plastic rotation angles of 0.01, 0.02, and 0.025 rad for immediate occupancy, life safety, and collapse prevention, respectively. These checks are similar to those made using the AASHTO SGS, although they are made at the rotation level rather than strain level. ASCE 41-06 also permits two methods of inelastic dis- placement estimation, the coefficient method and the capac- ity spectrum method. Chapter five describes these methods. It has been observed that the FEMA documents and the follow-on standard, ASCE 41-06, have several significant shortcomings: First the procedures do not directly address control of economic losses, one of the most significant decision maker concerns. Also, the procedures are focused on assessing the performance of the individual structural and nonstructural components that comprise a building, as opposed to the building as a whole. Perhaps most significantly, the reliability of the procedures in delivering the design performance has not been characterized (ATC 2003). In fact, many engineers who have worked with and applied the documents believe that they are too conserva- tive and restrictive, and lead to inappropriate engineering analysis and strengthening of structures (Searer et al. 2008). One of the main complaints by Searer and colleagues is that TABLE 2 REHABILITATION OBJECTIVES (ASCE 41-06) Target Building Performance Levels Operational Performance Level (1-A) Immediate Occupancy Performance Level (1-B) Life Safety Performance Level (3-C) Collapse Prevention Performance Level (5-E) E ar th qu ak e H az ar d L ev el 50%/50 year a b c d 20%/50 year e f g h BSE-1 (≈10%/50 year) i j k l BSE-2 (≈2%/50 year) m n o p Notes: 1. Each cell in the above matrix represents a discrete rehabilitation objective. 2. The rehabilitation objectives may be used to represent one of three specific rehabilitation objectives, as follows: Basic safety objective (BSO) k and p k and m, n, or o p and i or j k and p and a, b, e, or f m, n, or o alone Limited objectives k alone p alone c, d, g, h, or l alone TABLE 3 DAMAGE CONTROL AND BUILDING PERFORMANCE LEVELS (EXCERPTED FROM ASCE 41-06) Target Building Performance Levels for Seismic Rehabilitation Collapse Prevention Level (5-E) Life Safety Level (3-C) Immediate Occupancy Level (1-B) Operational Level (1-A) Overall Damage Severe Moderate Light Very Light Structural Little residual stiffness and strength, but load- bearing columns function. Large permanent drifts. Building is near collapse. Some exits blocked. Some residual strength and stiffness left in all stories. Gravity load-bearing ele- ments function. Building may be beyond economical repair. No permanent drift. Structure substantially retains original strength and stiffness. Minor cracking of facades and partitions. Elevators can be restarted. Fire protection operable. No permanent drift. Struc- ture substantially retains original strength and stiff- ness. Minor cracking of facades and partitions. All systems’ important to nor- mal functions are operable. Comparison with Performance Intended for New Buildings Significantly more dam- age and greater risk Somewhat more damage and slightly higher risk Less damage and lower risk Much less damage and lower risk Note that the number/letter designator reflects structural damage by the number and nonstructural damage by the letter. Numbers range from 1 to 5, letters from A to E. Lower numbers and earlier letters in the alphabet are better. The range of structural and nonstructural damage states reflects the range of performance permitted for seismic rehabilitation.

19 performance-based earthquake engineering is not a subject that is suitable for “standardization” or “cookbook-ization.” Their point is that performance-based earthquake engineer- ing, particularly for rehabilitation of existing structures, must rely on first principles, be done on the merits or lack thereof of each structure, and be performed by appropriately trained and qualified engineers. These issues must be taken into account in the development of any new document pur- porting to guide the performance-based engineering process. Following the 1994 Northridge earthquake, FEMA spon- sored the SAC Program to Reduce the Earthquake Hazards of Steel Moment-Frame Structures because an unexpectedly large number of moment frame buildings were damaged in the earthquake. SAC was a joint venture formed by SEAOC, ATC, and the Consortium of Universities for Research in Earthquake Engineering (CUREE. This effort led to the FEMA 350 Recommended Seismic Criteria for New Steel Moment-Frame Buildings (2000b) report and the three asso- ciated reports (FEMA 351, 352, and 353; 2000 c, d, e). The importance of this effort with respect to PBSD is that— these recommended criteria specifically quantified performance in terms of global behavior of buildings, as well as the behavior of individual components, and also incorporated a formal structural reliability framework to characterize the confidence associated with meeting intended performance goals (ATC 2003). The consideration of global or system behavior, plus the addition of the reliability framework, represent a substantial leap forward with respect to PBSD, which directly corre- sponds to the probabilistic framework that was described conceptually in chapter one. Although the framework pro- posed was complex and not ready for adoption into formal design specifications, it did provide a launching point for the next effort, ATC-58, Seismic Performance Assessment of Buildings. The incremental improvements seen from the Vision 2000 and FEMA 273 approaches onward are relevant for the bridge seismic design community because they are indica- tive of the amount of work that must be expended to develop PBSD procedures and methodologies. Recent/Current Efforts In the wake of the 1989 Loma Prieta and 1994 Northridge earthquakes, where little loss of life was incurred, some rela- tively new structures were damaged to the point that they could not be reoccupied, and the ensuing direct and indirect economic losses were surprisingly high [$7 billion in Loma Prieta and $30 billion in Northridge (ATC 2003)]; so high that conversations began about what it meant to meet code. It was clear that the structural and nonstructural damage was not consistent with public expectations of acceptable dam- age in a code-compliant design (FEMA 2006). The development of first- and second-generation guide- lines, as described previously, was undertaken to address the disparity between actual structure performance and code- inferred performance. However, it was also recognized that new design should be kept simple by the use of advanced assessment techniques. The road to PBSD of buildings was going to be longer than some had perhaps first envisioned, and it would take a great deal of effort. To address the way forward, both the Earthquake Engineering Research Center (EERC), which is now part of PEER, and the Earthquake Engineering Research Institute (EERI) drafted action plans for development of PBSD. These plans were published as FEMA 283, Performance-Based Seismic Design of Build- ings—an Action Plan (1996), and FEMA 349, Action Plan for Performance Based Seismic Design (2000a), and were authored by EERC and EERI, respectively. In 2001, FEMA awarded a contract to ATC to conduct a long-term project to prepare the next generation of PBSD guidelines, a multiyear, multiphase effort known as the ATC-58 Project. The project was estimated at $21 million for two phases in 2004 dollars, but funding was not avail- able for the full scope, and the project was finally funded at about 50% of the original estimated level. The project, which is still in progress, issued a 75% draft in 2011, 10 years after beginning, reflecting that the project is not typical in terms of effort or time to complete. Several features of the ATC-58 project that are relevant to this synthesis were the subject of a workshop that is discussed here, along with the current plan for the ATC-58 project and the project’s technical direction. In 2002, a workshop was held with primary stakehold- ers—building owners, tenants, lending institutions, build- ing regulators, and others who had no formal training in probabilistic risk assessment concepts—to develop methods to communicate earthquake risk (ATC 2002). The “partici- pants confirmed that life losses, direct losses, and indirect economic losses are the primary aspects of earthquake con- cern.” Some participants were keenly interested in being able to quantify the length of time a facility might be out of service and to quantify associated economic losses. Interest- ingly, life safety was not a primary topic in the workshop, and it is believed that was because most participants felt that assurance of life safety was a given in a code-compliant structure. Also of interest was the participants’ preference to consider earthquake effects and losses using a scenario event rather than probabilistic considerations of earthquake losses. Annualized losses were the least favored way to compare data. All participants understood the uncertainties in the prediction of losses from seismic events, and even though there was not a preference for probabilistic treatment, the use of confidence (as in 10% chance of exceedance) was favor- ably received. The current plan for the project is to develop a methodol- ogy that will provide a framework for identifying probable

20 consequences in terms of human losses, direct economic losses, and indirect economic losses. The framework is out- lined in detail in Seismic Performance Assessment of Build- ings Volume 1—Methodology of the 75% draft of ATC 58-1 (2011), and “the companion Volume 2 provides guidance on implementing the technology, including instructions on how to use an electronic Performance Assessment Calcula- tion Tool (PACT) that has been developed to enable practical implementation of the methodology.” The technical direction of the ATC-58 project has the goal of replacing performance levels that are currently used, such as fully functional, immediate occupancy, life safety, and collapse prevention, with probable future earthquake impacts as measures of performance. The following impacts are considered: • Casualties—the number of deaths and injuries of a severity requiring hospitalization • Repair cost—including the cost of repairing or replac- ing damaged buildings and their contents • Repair time—the period of time necessary to conduct repairs or replace damaged contents, building compo- nents, or entire buildings • Unsafe placards—the probability that a building will be deemed unsafe for postearthquake occupancy. Clearly, these impacts would be relevant to decision mak- ers, such as owners of buildings, but the approach also rep- resents a leap beyond the current method of attempting to quantify building performance based on a limited set of ele- ment deformation or force levels. To address the necessary methodology and data require- ments, significant effort has gone into developing tools that knowledgeable designers might eventually use to produce consistent and technically sound performance-based designs. The 75% draft of Volume 2—Implementation Guide outlines this process, provides the PACT calculation package, and describes the probabilistic damage state (or fragility, which is defined in the Damage Prediction section of chapter six) database that supports the process. The National Institute of Standards and Technology Interagency Report (NISTIR) 6389, UNIFORMAT II Elemental Classification for Building Specifications (NIST 1999) specifies unique identification codes for common structural and nonstructural systems and components. The PACT database ties system and component fragilities to those identification codes, thus permitting PACT users to simply input a code that will then invoke the appropri- ate fragility function. Fragility databases are clearly a useful method to ensure consistent application of fragilities across the design community. To fully deploy PBSD, the bridge community will need to develop a similar methodology. Significant progress has been made in the identification of knowledge gaps and the research required for the full imple- mentation of PBSD in buildings, including limitations of the ATC-58 project. This was published in the NIST Grant/ Contract Report (GCR) 09-917-2 report (NIBS 2009), which outlines in detail 37 research topics that are critical for PBSD implementation. Included are tasks related to the determina- tion of data to define fragility relationships for components found in old and new buildings, and to determine the perfor- mance of buildings designed according to prescriptive codes and standards in order to improve building codes and ensure a smooth transition to the widespread use of PBSD in the next decade. As a general observation, the range and complexity of construction types is probably greater for buildings than for bridges. Thus, the PBSD development effort for bridges should be less than that for buildings. ASCE 7-10 The ASCE 7 Standard, Minimum Design Loads for Build- ings and Other Structures, is the primary structural refer- ence governing the seismic design for the International Building Code (IBC). The 2010 version of the ASCE 7 (ASCE 2010) has taken a significant departure from pre- vious editions. The mapped values for seismic ground motions are probabilistic ground motions that are based on uniform risk rather than uniform hazard. This means that a notional, standard, or generic probability of collapse has been used to translate the seismic hazard (a property of a structure’s site and geographic location) to a seismic risk of structural collapse (a property of site, location, structure type, and assumed damage state). An overview of the pro- cess and rationale for the change is given in FEMA P-749 (FEMA 2010) and in Luco et al. (2007). FEMA P-749 also provides an overview of the ongoing development and regu- latory process of the U.S. building codes with respect to seismic design. The previously mapped values were for uniform seismic hazard with a 2% chance of exceedance in 50 years. This hazard was known as the maximum considered earthquake (MCE) and was given as spectral accelerations with uniform probabilities of exceedance as a function of period. The new mapped values are for 1% chance of exceedance in 50 years, risk-adjusted maximum considered earthquake (MCER) spectral accelerations. Stated differently, these mapped val- ues represent ground motions that result in a 1% chance that the structure could collapse in 50 years. The new risk-based maps result in slight reductions in the 1-second spectral accelerations in many areas of the country, including the eastern United States. These reductions range from 0% to 20%. In the more seismically active areas of the country—California, Alaska, and Hawaii—the 1-second spectral accelerations increase slightly, on the order of 0% to 20%. ASCE 7-10 provides maps of these adjustment factors.

21 actual structures designed using the specifications. How- ever, ASCE 43-05, Seismic Design Criteria for Structures, Systems, and Components in Nuclear Facilities (2005a), also uses a risk-targeted seismic hazard, and that standard does include displacement-based methodologies, although they are not outlined in detail. Nonetheless, it is clear that the building profession is incrementally migrating toward probabilistic PBSD. International Code Council Performance Code The International Building Code (IBC), and its adjunct ASCE 7-10, are force-based and prescriptive and, as such, do not contain performance-based provisions. However, the International Code Council (ICC) publishes a performance- based document, International Code Council Performance Code (ICC-PC). The ICC-PC allows the user to achieve vari- ous design solutions and is intended to envelop the single solution obtained using the basic IBC. To that extent, the IBC is considered to provide an acceptable solution that will comply with the ICC-PC, thus making the ICC-PC the higher- level document. This is a model code that must be adopted or otherwise permitted by a jurisdiction before it may be used on a project. The ICC-PC addresses seismic loading in addi- tion to considering performance-based design for other load- ings and performance types (i.e., nonstructural). The ICC-PC uses four performance groups (PGs) and four design events, as shown in Table 4. The performance groups are: • PG I—agricultural, temporary and minor storage facilities • PG II—facilities other than I, III, or IV • PG III—buildings representing substantial hazard to human life (assembly of more than 300 people in one area, schools, health care, power facilities, occupancy of more than 5,000 total, etc.) • PG IV—essential facilities (hospitals, fire, rescue, police, emergency shelters, air traffic control towers, buildings with critical national defense functions, etc.). The reasons that the acceleration levels change are related to (1) how the accelerations change with increasing return period throughout the country, and (2) the consideration of structural fragility or probability of collapse that has a statis- tical distribution. For example, in San Francisco, where the 100-year return period accelerations are almost as high as the 2,500-year (MCE) accelerations, the risk of collapse, when considering the full distribution of capacity (i.e., fragility), is larger than in Memphis, where the 100-year accelerations are significantly smaller than the 2,500-year accelerations (Luco et al. 2007). Although reasons for the changes in spec- tral acceleration levels are apparent in Luco et al., it is not clear whether similar changes to the basic hazard level would be present for the 1,000-year event, which AASHTO uses. The return period for ground motions producing a uniform risk will not correspond to a single value across the country. At sites where earthquakes occur relatively frequently, such as California, the return period generally will be less than at sites where earthquakes occur infrequently. The ASCE 7-10 MCER shaking approach produces shaking values that are approximately, but not exactly, equal to 2,500 years. The change from uniform hazard to uniform risk repre- sents a shift away from simple quantification of earthquake damage in terms of EDP and code-implied performance toward more complete performance-based design that cal- culates chances of DMs occurring. This approach attempts, at a notional level, to combine the first three of the four steps of PBSD: seismic hazard, structural response, and damage prediction. The simplification of using a generic probability of col- lapse may provide a somewhat inaccurate indicator of the actual risk of collapse compared with the risk that would be calculated by a rigorous analysis based on system-spe- cific fragilities. However, the method does pick up general trends related to regional ground motion differences. The ASCE 7-10 design methodology is still predicated on a force- based methodology, and thus, there is no direct rationaliza- tion of the methodology with respect to DM prediction for TABLE 4 ICC-PC MODEL CODE Increasing Level of Performance PERFORMANCE GROUPS PG I PG II PG III PG IV In cr ea si ng M ag ni tu de o f E ve nt M A G N IT U D E O F D E S IG N E V E N T Very Large (Very rare – 2,475-year return period) Severe Severe HigH Moderate Large (Rare – 475-year return period) Severe HigH Moderate Mild Medium (Less frequent – 72-year return period) HigH Moderate Mild Mild Small (Frequent – 25-year return period) Moderate Mild Mild Mild

22 These performance groups are essentially the same as those in the IBC and ASCE 7. The performance group, combined with four design events, is then used to define the maximum level of damage that can be tolerated, as shown in Table 4. Four damage levels are defined in terms of struc- tural damage, nonstructural system damage, occupant haz- ards, overall extent of damage, and hazardous materials: • Mild—No structural damage and the building is safe to occupy. • Moderate—Moderate structural damage, which is repair- able, and some delay to reoccupancy can be expected. • High—Significant damage to structural elements, but no large falling debris occurs; repair is possible, but significant delays in reoccupancy can be expected. • Severe—Substantial structural damage, but all sig- nificant components continue to carry gravity load demands; repair may not be technically possible; the building or facility is not safe to reoccupy, as reoccu- pancy may cause collapse. The owner and the principal design professional (PDP) have the responsibility to develop performance criteria, have them peer reviewed, and gain approval by the code official. The design reports must document required ele- ments such as the goals and objectives of the project, per- formance criteria, bounding conditions (restrictions on use to ensure the desired performance is achieved) and critical design assumptions, system design and operation require- ments, and operational and maintenance requirements. It is the responsibility of the owner and PDP to develop a plan that will meet the required performance requirements. Pre- scriptive limits are not provided in the ICC-PC, and this approach puts significant responsibility on the PDP to sat- isfy the performance objectives over the life of the facility. This may constitute a significant, practical shortcoming of the performance code approach, as illustrated by the New Zealand “leaky home” case study described in chapter two. MARINE INDUSTRY PRACTICE Deterministic PBSD, using the first two steps of PBSD, has been used extensively in the marine industry for well over a decade, primarily because of the development of the Marine Oil Terminal Engineering and Maintenance Stan- dards (MOTEMS 2011) and the Port of Los Angeles/Port of Long Beach (POLA/POLB 2009) seismic design guidelines. In general, marine PBSD uses a multiperformance-level approach that limits damage and downtime in small to mod- erate earthquakes, and prevents structural collapse and loss of life in large seismic events. The overall design philosophy and many of the provisions and recommendations have been adopted from the bridge industry, with a heavy influence from Priestley et al. (1996). Both the MOTEMS and POLA/ POLB guidelines are summarized. MOTEMS California has developed guidelines for design and mainte- nance for marine oil terminals (MOTs). Included in these provisions (adopted as Chapter 31F of the California Build- ing Code) are multilevel performance-based seismic design requirements. These provisions have become standard prac- tice for many waterfront structures other than MOTs, includ- ing piers and wharves. Seismic performance is characterized using two performance levels. The return period for a Level 1 or Level 2 seismic hazard is defined based on the quantity of oil the terminal processes, the number of transfers per year per berthing system, and the maximum vessel size. However, all new MOTs are classified as high risk, with return periods of 72 years and 475 years, respectively, for Levels 1 and 2. • Performance for Level 1: – Minor or no structural damage – Temporary or no interruption of operation. • Performance for Level 2: – Controlled inelastic structural behavior with repa- rable damage – Prevention of structural collapse – Temporary loss of operations, restorable within months – Prevention of a major oil spill. Performance is quantified using material strain limits and nonlinear static capacity (pushover) curves. Strain limits are material and location specific (i.e., strain limits associ- ated with pile in-ground plastic hinging are more restrictive than strain limits in the pile-to-deck connection). Demand is determined through a two-dimensional (2-D) nonlinear static demand procedure (simplified or refined) based on recom- mendations by Priestley et al. (1996) or the FEMA 440 modi- fications of the ATC-40 Capacity Spectrum Method, which is described in chapter five of this synthesis. Additionally, if the structural configuration is irregular, three dimensional (3-D) linear modal procedures are required. Nonlinear dynamic analyses are optional, and capacity protection is achieved through methods similar to those used in bridge design. Port of Los Angeles/Port of Long Beach The Port of Los Angeles and the Port of Long Beach each have developed their own seismic design guidelines for waterfront structures, with a primary emphasis on marginal wharves (parallel to shore). Both guidelines are similar to each other and to the MOTEMS approach. Seismic perfor- mance is characterized using performance-based procedures with three performance levels, an operating level earth- quake (OLE) with a 72-year return period, a contingency level earthquake (CLE) with a 475-year return period, and a design earthquake (DE), which represents two-thirds of the

23 maximum considered earthquake (MCE). The DE ground motion is associated with meeting the minimum require- ments of the 2007 California Building Code, which invokes ASCE 7-05 (ASCE 2005a). Thus, the DE is meant to be a direct check of life safety consistent with ASCE 7-05 and was added to demonstrate minimum compliance with ASCE 7-05 to the building official. Seismic performance for each performance level is defined by operability, reparability, and safety concerns, as follows: • Performance for the OLE – No interruption of operations – Forces and deformations (including permanent embankment deformations) shall not result in struc- tural damage – All damage shall be cosmetic in nature and located where visually observable and accessible – Repair shall not interfere with wharf operations. • Performance for the CLE – Temporary loss of operations of less than 2 months is acceptable – Forces and deformations (including permanent embankment deformations) may result in con- trolled inelastic behavior and limited permanent deformations – All damage shall be repairable and shall be located where visually observable and accessible for repairs. • Performance for the DE – Forces and deformations (including permanent embankment deformations) shall not result in struc- tural collapse of the wharf, and the wharf shall be able to support the dead load of the structure includ- ing cranes – Life safety shall be maintained. Seismic performance is primarily quantified using strain limits for each level of ground motion. Strain limits are material and location dependent. Structural capacity is generated using nonlinear static analyses. Demand is deter- mined primarily using the substitute structure method of a 2-D segment of the structure. However, if the wharf is considered irregular, 3-D modal response spectra or linear response history analysis may be employed, and nonlinear response history analyses may be used to verify seismic displacement demands. Capacity protection is enforced for all elements except the piles, where a strong deck-weak-pile philosophy is used. As with other first-generation perfor- mance-based standards, seismic hazard analysis and struc- tural analysis are well defined, but damage analysis and loss analysis are not.