National Information Service for Earthquake Engineering
University of California, Berkeley
Department of Civil and Environmental Engineering
University of California, Berkeley
One of the important overall surprises of the Northridge earthquake of January 17, 1994, was the widespread and unanticipated brittle fractures in welded steel beam to column connections. The economy, versatility and presupposed plastic deformation capacity of welded steel moment-resisting frame (WSMF) building led to their common usage in Los Angeles and elsewhere in the U.S. No casualties or complete collapses occurred during the Northridge earthquake as a result of these connection failures, and WSMF buildings in areas of moderate shaking were not damaged at all. However, a wide spectrum of brittle connection damage did occur, ranging from minor cracking observable only by nondestructive testing to completely severed columns.
This paper reviews the performance of steel buildings during the Northridge earthquake and the implications for design practice. Some of the results of studies undertaken as part of a project initiated by U.S. Federal Emergency Management Agency (FEMA) to reduce the earthquake hazards posed by steel moment-resisting frame buildings. The objective of this project is to develop and verify reliable and cost-effective methods for the inspection, evaluation, repair, and rehabilitation of existing steel frame buildings and for the construction of new ones.
Every earthquake provides new lessons for the earthquake engineering profession. The widespread damage to welded steel moment resisting frame systems was one of the major overall lessons of the Northridge earthquake. The brittle nature of the fractures detected in numerous welded steel beam to column connections, essentially invalidated historic design approaches and code provisions based on "ductile" structural response.
The most commonly observed damage occurred in or near the welded joint of a girder bottom flange to the supporting column flange; complete brittle fractures occurred in many cases. Damage was so severe in some buildings that all of the moment resisting connections at one or more floors failed, or significant permanent lateral displacements occurred. In one case, damage was so severe the building was demolished, and several buildings were evacuated.
Thus far, more than 150 damaged buildings have been identified, including hospitals and other health care facilities, government, civic and private offices, cultural and educational facilities, residential structures, and commercial and industrial buildings. Damage occurred in new as well as old buildings; in tall as well as in short structures. While inadequate workmanship was believed to play a role in the damage observed in some structures, most damaged buildings are believed to be constructed consistent with modern codes and standards of practice, The effect of these observations has been a loss of confidence in the procedures used in the past to design and construct welded connections in steel moment frames, and a concern that existing structures incorporating these connections may not be sufficiently safe.
A particularly disconcerting aspect of this damage is that it often occurred without accompanying distress to architectural finishes and cladding. As a result, reconnaissance reports immediately following the Northridge earthquake often cited the apparent excellent behavior of steel frame buildings. However, severe damage found in buildings under construction at the time of the earthquake, and detailed investigations of WSMF buildings which suffered increasing amounts of damage during aftershocks, quickly identified the true performance.
Current professional judgment is that the historic practices used for the design and construction of WSMF connections do not provide adequate reliability and safety, and should not continue to be used in the construction of new buildings intended to resist earthquake ground shaking through inelastic behavior. As a consequence, pre-qualified connection details and design methods contained in the major U.S. building codes have been rescinded, and emergency code provisions stipulate that new designs be substantiated by testing or test-verified calculations. Several fundamental questions must be answered in order to develop effective and economical design procedures and construction standards, and to restore public and professional confidence in this form of construction. These questions include:
Answering these questions involves consideration of many complex technical, professional and economic issues including metallurgy, welding, fracture mechanics, connection behavior, system performance, and practices related to design, fabrication, erection and inspection. Unfortunately, current knowledge is inadequate.
A coordinated, problem-focused program of research, investigation and professional development has begun under FEMA sponsorship to develop reliable, practical and cost-effective guidelines and standards of practice related to steel moment-resisting frame buildings for:
1. the identification, inspection and rehabilitation of existing at-risk buildings prior to a damaging earthquake,
2. the identification, inspection, and repair or upgrading of damaged buildings following an earthquake, and
3. the design and construction of new buildings.
This program is being managed by the SAC Joint Venture comprised of the Structural Engineers Association of California (SEAOC), Applied Technology Council (ATC) and California Universities for Research in Earthquake Engineering (CUREE). However, all aspects of the program are conducted with active involvement of design and construction experts, researchers and others from throughout the U.S.
The Steel Program is divided into two major phases. The first phase focused on the development of Interim Guidelines [I] for the inspection, evaluation, repair, modification and construction of welded steel structures. This phase was supported by limited amounts of laboratory and field testing, as well as focused investigations. Major efforts to identify and verify reliable and cost-effective long term solutions and to develop seismic design criteria for steel frame structures are contained in the second phase.
The backbone of the Steel Program is the development of specific design advisories, guidelines and other criteria for design, inspection, evaluation, repair, modification and rehabilitation of steel moment frame structures. The Interim Guidelines  developed in Phase I were written by ten experts from a variety of disciplines. The Guidelines were subjected to extensive review by engineers, researchers, building regulators and other public officials, and representatives from the steel and construction industries. The scope of the Interim Guidelines covers welding procedures, quality assurance, post-earthquake actions, and new construction. Specific chapters cover: (a) welding and metallurgy; (b) quality control and assurance; (c) visual inspection; (d) non-destructive testing; (e) classification and implications of damage; (f) post-earthquake evaluation; (g) post-earthquake inspection; (h) post-earthquake repair and modification; and (i) new construction.
Some of the results of the preliminary investigations and tests carried out to support the development of the Interim Guidelines and the planning of Phase 2 are described below. Information on activities being undertaken in Phase 2 is presented at the end of the paper.
Four types of surveys were used to assess the damage to steel frame buildings caused by the Northridge earthquake. In the first, in-depth interviews  were conducted with design engineers, building inspectors, contractors and building officials. A number of important difficulties were detected in identifying damaged buildings and in inspecting and repairing them.
In the second type of survey, a brief questionnaire was sent to more than two hundred, randomly selected owners of steel buildings to assess their awareness of problems occurring in steel buildings, whether their building had been inspected by an engineer, and the state of damage, if any, found in their building. This preliminary survey was used to estimate the overall scope of damage to steel buildings and to help identify geographic areas where steel buildings were damaged. Based on results from this survey  and other more detailed information on ground motion intensity and structural damage, the Interim Guidelines recommended detailed inspection of steel buildings be conducted where peak ground motions exceeded 0.2g.
A third level of survey was carried out by engineers who had evaluated damaged steel frames [4,5]. Detailed information was obtained on 89 buildings regarding the types and locations of damage observed and the structural configuration, materials and detailing. This third survey was supplemented by even more detailed surveys of damage in 12 buildings selected for dynamic analyses. Precise comparison of predicted and observed damage was possible for these buildings.
Results were used to improve methods to select buildings for inspection, and to identify joints within a suspect building that should be inspected. For example, on average 70% of the floors of buildings surveyed had serious damage to at least one welded joint. Only 25% of the connections were found without damage. About 20% of the building frames had more than 40% of their connections damaged; in some instances, all connections at one or more floors were damaged.
Survey results also were used to assess methods for predicting damage. For instance, the data show a correlation between damage and the area supported by a welded connection. This suggests that redundancy contributes to improved response, but other factors may be involved. Ground motion intensity was also found to correlate with damage, but limited data at high peak acceleration values make precise interpretation difficult.
Similar interpretations showed that damage in low rise structures was more or less uniformly distributed over height, whereas tall buildings exhibited greater damage in the upper half. Results also show that damage tends to congregate. Thus, finding a severely damaged connection as part of an inspection should trigger inspection of other nearby connections.
Twelve buildings damaged by the Northridge earthquake were selected for detailed analysis by consultants using elastic and nonlinear analysis programs. To support this effort, detailed investigations  of ground motion characteristics during the Northridge earthquake were conducted. This consisted of gathering available strong motion records in or near the subject buildings. In addition, a fault dislocation model was formulated, verified with available records and used to generate time history estimates at the sites of the case study buildings and elsewhere.
Buildings studied had heights from 2 to 17 stories, and were located from Santa Clarita, north of the epicenter, to Santa Monica, to the south. The analyses were intended to help identify the causes of the damage, as well as the ability of analytical methods and modeling assumptions to predict damage. For this reason, heavily damaged buildings were excluded. In two cases, buildings without damage were included. In four buildings, recordings of response during the earthquake were available and in three other cases, ambient vibration tests were performed.
The analysis results (see, for example, Refs. 7, 8 and 9) indicate that the case study buildings were very strong in comparison with the design forces incorporated in building codes. In many of the buildings the estimated response spectrum were nearly double those considered in current building codes (assuming elastic response, Rw = 1). Elastic analysis results showed that the most heavily damaged buildings were only stressed 2 to 3 times their capacities; in several cases, damaged buildings were predicted to remain essentially in the elastic range of response, suggesting that the buildings were 4 to 8 times stronger than required by code. The main reason for this over-strength appears to be the use of large-sized members to satisfy stringent code drift requirements.
Comparisons of damage survey data with results of elastic analyses of the buildings (using recorded and simulated Northridge earthquake records developed for the building sites ) show relatively poor correlation. Analyses suggest that the most heavily stressed joints are most likely to be damaged; however, the precise location and severity of damage was not reliably predicted by conventional elastic dynamic analyses. The 60% most highly stressed connections in a structure (relative to their capacities) have roughly equal chance of being damaged. Areas of low computed stress were also subject to damage. Thus, analysis may not be a good way of assessing the particular joints to inspect, though it may indicate floors that should be inspected. The reasons for differences between observed and computed behavior include the effects of initial defects and poor workmanship, and the limitations of current analytical methods and models. For instance, inclusion of slabs and panel zones had an important effect.
Elastic and inelastic dynamic analyses indicate that higher mode effects are important. As a result, equivalent lateral static force methods in the elastic range, and nonlinear push-over analyses have limited value for longer period structures. Similarly, the predicted distribution of damage in longer period structures is very sensitive to the ground motion considered. High velocity pulses in the ground motion records also resulted in substantial higher mode response, even for short structures. Additional analytical investigations were used to assess effects of structural modeling, member fracture, and structural configuration. These are reported in Ref. 10. Hypothetical buildings (especially shorter ones) subjected to severe shaking representative of the lightly populated areas of north of San Fernando Valley were most susceptible to collapse or severe damage.
Most design calculations are based on an assumption that plane sections remain plane during deformation. However, review of experimental data and results of finite element analyses suggest that this is far from true, with high local bending and shear deformations being induced in beam and column flanges. This is especially pronounced when plastic shearing deformations occur in the panel zone. Results demonstrated that these panel zone deformations were often very large. In such cases, the distribution of shear stress over the depth of the beam's web is not uniform, often concentrating the majority of the shear force in the highly stressed beam flanges. Compounding this situation is the fact that actual material properties are not uniform, and vary randomly from member to member and systematically with loading direction, section size, and welding procedures. Normal member to member variation of material properties may result in members stronger than the connecting weld, or a column that is weaker than the supported beam. As a result, the joint may have negligible inelastic deformation capacity, regardless of workmanship.
A total of 37 full size beam-to-column connections were tested as part of the Phase I investigation [9 and 11]. Twelve specimens were constructed in utilizing pre-Northridge details, half of the specimens had W36xl5O beams and half had W3Ox99 beams. Fourteen-inch wide-flange sections were used as columns in both cases. Dual certified (fy>50 ksi) steel was used. Slabs were not included. All specimens exhibited brittle appearing fractures; some fractured without any plastic deformation, while others deformed to a plastic rotation of 0.02 prior to fracturing.
The damaged specimens were repaired or upgraded. Repair consisted simply of rewelding the connections using high notch tough FCAW procedures; backing bars were removed, the root pass of the CJP weld on the beam flange to column flange connection was air-arc gouged and repaired with a fillet weld. This is the prevalent practice in repairs of damaged buildings in the Los Angeles area. Test results indicated that the repaired specimens, constructed with careful quality control, were able to retain their pre-damage strength and stiffness. Plastic rotation capacities were not significantly different from those achieved in the first tests. Thus, improved workmanship and materials did not significantly improve the inelastic performance of these details.
Some of the specimens were upgraded in an attempt to improve their plastic deformation capacity; inclined haunches were applied to one or both sides of the beam at its connection to the column. This detail moves the plastic hinge away from the face of the column. These tests supplemented earlier tests at the University of Texas  and elsewhere (see Refs. 1 and 13) which utilized trapezoidal and rectangular-shaped cover plates, vertical fins, or side plates. Results for triangular haunches indicate that they are able to increase the plastic deformation capacity of the connection to a plastic rotation of at least 0.03. Inconsistent results have been obtained with cover plates.
Simple weldment specimens were tested to assess various weld procedures, initial defects, repair methods and loading rates . These results clearly demonstrate the importance of quality welding and the greater reliability that can be achieved with high notch-tough weld consumables. Tests were also conducted on a few details appropriate for new construction. These specimens utilized the same steel materials as for the previous beam to column tests, but utilized high notch-tough weld wire. In addition, they included reinforcement of the end region to shift the plastic hinge region away from the face of the column. These specimens utilized horizontal cover plates and horizontal haunches. Generally superior behavior was obtained in these tests .
The Phase 1 Interim Guidelines provide the best answers within the current state-of-knowledge on what to do about welded steel moment frames. However, Phase I has also demonstrated the limitations of current knowledge. The substantial damage, including collapse, of many modern steel frame buildings in Kobe, and increasing reports of damage in the San Francisco Bay Area apparently due to the Loma Prieta earthquake, has heightened the appreciation worldwide of the need for developing reliable, practical and cost-effective solutions to this problem.
The Phase 2 effort addresses these solutions through eleven inter-related tasks spanning over 48 months. The detailed Work Plan for Phase 2 has been finalized through the efforts of the Technical Advisory Panels and the SAC management team, working in conjunction with FEMA and a Project Oversight Committee. Brief summaries of some of the technical investigation areas being undertaken to develop improved Seismic Design Criteria are presented below.
Performance of Steel Frame Buildings during Past Earthquakes - Various investigations are being undertaken to assess the performance of steel moment frame buildings in past earthquakes. In addition to the Northridge earthquake studies, information is being gathered related to the Kobe, Landers/Big Bear, Loma Prieta, Whittier Narrows, and other earthquakes. Results will be interpreted to help assess damage screening and inspection criteria, identify details and other structural features associated with the presence or absence of damage, evaluate the accuracy of analytical methods, and assess the economic, social and other impacts of damage.
Materials and Fracture - This task examines the mechanical properties of steel materials in commercially-available structural members, including new materials just coming on the market. It also identifies the influence of various factors on the behavior of simple, fracture critical welded joints such as the orientation, history and rate of loading, the strength and notch toughness of base materials, joint restraint, and local details. Material characteristics required to develop proposed connection details will also be identified.
Joining and Nondestructive Testing - A variety of investigations undertaken in coordination with investigations related to Materials and Fracture and Connection Performance. These aim at understanding the factors (e.g., welding consumables, procedures, and the relative strengths of the weld metal and base metal) that control behavior, establishing the sensitivity of ultrasonic testing techniques, assessing promising new NDE procedures, and developing criteria for welding and inspection. Bolted and partially restrained joints are also studied.
Connection Performance - Detailed finite element and other analyses are being utilized to devise methods for predicting the deformation and strength capacities of connections and to develop simplified analytical methods suitable for design practice. These analyses will be closely coordinated with the connection test program. Tests will be conducted initially to assess parameters that control the behavior of promising new details as well as of commonly used pre Northridge and current designs; later tests will be used to validate the details and design methods to be incorporated in the Seismic Design Criteria. Tests will include single and double sided beam to column connections, with and without slabs. In addition to welded steel moment connections, simple, bolted and partially restrained connections will be studied.
System Performance - Focused investigations are underway to assess the effect of various structural and ground motion parameters on global and local demands. Hypothetical buildings having 3, 9 and 20 stories, located in regions of relatively high, moderate and low seismicity, are used as the basis of these studies. Different computer programs and modeling approaches are being utilized to study the effect on seismic demands of ground motion intensity and dynamic characteristics, and structural configuration, proportioning and modeling, as well as of the deterioration of hysteretic characteristics due to local buckling, brittle fracture, and so on. Also, the safety and reliability of steel moment-resisting frame systems will be evaluated considering the possible occurrence of brittle fractures. Potential benefits of alternative framing systems having partially restrained, bolted or energy dissipative connections are being investigated.
Performance Prediction and Evaluation - Results of the investigations on seismic demands are being synthesized and interpreted along with results of studies on the capacities of various details and connections to achieve a consistent set of performance-based design and analysis procedures for steel moment frame structures. These are directed at the evaluation of existing steel buildings as well as the design of new ones. Analysis and modeling simplifications suitable for design are being assessed, and special procedures for regions of lower seismicity are being examined.
Economic, Social and Political Issues - A variety of activities are being undertaken to assess the practicability of the Design Criteria and to assess the potential economic, social and political impacts of their implementation. These activities include trial applications and economic and performance assessments of buildings designed using various procedures and criteria, and identification of other barriers to effective implementation of the final Seismic Design Criteria.
While the Interim Guidelines represent current U.S. thinking on the proper evaluation, inspection and repair of WSMF buildings, there are clearly many uncertainties and unresolved questions. In Phase 2 of the FEMA/SAC Steel Program additional research and testing will more clearly define the parameters controlling the performance of connections and systems. This research will also develop and verify reliable and cost effective procedures for design of new moment frame buildings and for the evaluation and rehabilitation of existing ones.
The information in this paper was prepared under contract (EMW-95-K-4672) to the U.S. Federal Emergency Management Agency as part of its Program to Reduce Earthquake Hazards in Steel Moment Frame Structures. The assistance and support of Michael Mahoney and Robert Hanson, FEMA's Program Officer and Technical Advisor, respectively, are greatly appreciated. The author is indebted to the efforts of numerous individuals and organizations that support or contribute to the Steel Program, especially to Project Directors Ronald Hamburger and James Malley. The findings, conclusions and recommendations in this paper are not intended for use in design or evaluation of structures. They do not represent official policy of FEMA or the SAC Joint Venture.
1. Interim Guidelines: Evaluation, Repair, Modification and Design of Welded Steel Moment Frame Structures, FEMA 267, FEMA, Washington DC, Aug. 1995.
2. Gates, W. and Morden, M., "Lessons from Inspection, Evaluation, Repair and Construction," Surveys and Assessment of Damage to Buildings Affected by the Northridge Earthquake, Report SAC 95-06, SAC Joint Venture, Sacramento, Dec. 1995.
3. Durkin, M.E., "Inspection, Damage and Repair of Steel Frame Buildings Following the Northridge Earthquake," Surveys and Assessment of Damage to Buildings Affected by the Northridge Earthquake, Report SAC 95-06, SAC Joint Venture, Sacramento, Dec. 1995.
4. Bonowitz, D. and Youssef, N., "SAC Survey of Steel Moment Frames Affected by the 1994 Northridge Earthquake," Surveys and Assessment of Damage to Buildings Affected by the Northridge Earthquake, Report SAC 95-06, SAC Joint Venture, Sacramento, Dec. 1995.
5. Youssef, N., Bonowitz, D. and Gross, J., A Survey of Steel Moment Resisting Frame Buildings Affected by the 1994 Northridge Earthquake, NISTR-5625, NIST, Gaithersburg, MD, April 1995.
6. Somerville, P. et al, Characterization of Ground Motion During the Northridge Earthquake, Report SAC 95-03, SAC Joint Venture, Sacramento, Dec. 1995.
7. Krawinkler, H., "Systems Analysis and Behavior," Design Applications Manual and Seminar Speakers Notes, SAC, Report SAC IR-95-03, September 1995.
8. Analytical and Field Investigations of Buildings Affected by the Northridge Earthquake of January 17, 1994, Technical Report SAC 95-04 (parts A and B), SAC Joint Venture, Sacramento, Dec. 1995.
9. Case Studies of Steel Moment Frame Building Performance in the Northridge Earthquake, Report SAC 95-07, SAC Joint Venture, Sacramento, Dec. 1995.
10. Parametric Analytical Investigations of Ground Motion and Structural Response, Report SAC 95-05, SAC Joint Venture, Sacramento, Dec. 1995.
11. Experimental Investigations of Beam to Column Connections, Technical Report SAC 96-01 (Parts A and B), SAC Joint Venture, Sacramento, 1996.
12. Engelhardt, M. and Sabol, T., Testing of Welded Steel Moment Connections in Response to the Northridge Earthquake, Progress Report, AISC Advisory Subcommittee on Special Moment Resisting Frame Research, October 1994.
13. Test Summary Reports, Report SAC 96-02, SAC Joint Venture, Sacramento, July 1996.
14. Kaufman, E. and Fisher, J., "A Study of the Effects of Materials and Welding Factors on Moment Frame Weld Joint Performance Using a Small-scale Tension Specimen," Report SAC 95-08, SAC Joint Venture, Sacramento, CA, Dec. 1995.
Updated December 9, 1997.
Send comments or questions to EERC Library