Northridge Earthquake Case Study

One of the major aftermaths of the Northridge Earthquake of January 17, 1994, was the widespread connection damage that posed a major question regarding the behavior of field-welded, field-bolted moment frame connections, also known as Pre-Northridge connections.

Before the Northridge Earthquake, Steel Moment Resisting Frames (SMRFs) were believed to have ductile behavior that would achieve high-cycle fatigue. As a result, fatigue was not considered to be a failure mode for these connections during a seismic event.

After the Northridge Earthquake and the widespread connection failure in steel moment frame buildings, it was concluded that many connections failed at what appears to be relatively few cycles. Observations after the Northridge Earthquake

indicated that these connections failed at both a relatively low stress level and at only a few cycles of vibration.
Appearance of the cracks which in most cases started from the weld at the bottom flange of the beam and on some occasions propagated into the column flange or even column web (as in the case of the two-story building discussed later in this document in which cracks expanded all the way through the column web), undermined the
1
confidence in the ductile behavior of the steel moment frames [37]. According to a report by Youssef et al. in 1995, brittle fracture in and around the groove weld connecting the beam flanges to the column flange was observed in more than 150 steel moment frame buildings after the Northridge Earthquake [37].
As a result of these failures, many researchers tried to gain a better understanding of the causes of damage observed in the connections of the steel moment frames. Due to the complexity of the problem, the SAC steel project was initiated by FEMA as a joint venture between Structural Engineers Association of California (SEAOC), Applied Technology Council (ATC), and Consortium of Universities for Research in Earthquake Engineering (CUREE). Since the response of the structures is often dominated by the first mode, the SAC project was never focused on the potential effects of higher modes. Consequently, all the performed SAC tests represented the first mode type of motions.
The majority of published results of this nationwide project ([10], [11], [12], [13], [14], [15], [16], [17], [18], [19], and [20]) concentrated on local connection defects that potentially initiated the observed cracks. For instance, the existence of the weld access hole (web cope hole) and discontinuity of the bottom flange weld were shown to be the cause of porosity and slag inclusions in the weld at the center of the beam and potentially one of the main reasons for crack initiation. Also, leaving the backup bar after the beam flange to column flange full penetration welds were completed
2
(which was the common practice in Pre-Northridge connections) showed to be a source of porosity and slag inclusions and eventually contributed to the initiation of cracks at the weld area [4].
Although remarkable research under the SAC project was performed to address the above issues, damage to some buildings could not be reconciled by use of these failure mechanisms. This led to renewed interest in studying the effects of low-cycle fatigue combined with the higher modes of vibration that can increase the cumulative fatigue at critical connections, and as a result, potentially create the observed connection failures.
Through an investigation of the role of the higher modes in the fatigue damage, the current study focuses on the contribution of higher modes of vibration to the damage observed in steel moment connections during the Northridge Earthquake. In other words, the contribution of higher mode motions to the stress histories at the connection beams and columns is investigated, as are cyclic fatigue type damages. In essence, this study shows that a large number of cycles at higher frequency and at significant but lower stress levels than the primary mode could be a major cause of increasing the cumulative fatigue at the connections of steel moment frames, potentially creating connection failures similar to those observed in the moment connections during the Northridge Earthquake.
3
As a part of this study, two buildings damaged by the Northridge Earthquake have been investigated.

These buildings were two and ten stories, respectively, and used steel moment frames as the lateral load resisting system in both directions.

Chapter two of this document describes the definition of fatigue failure and summarizes the research done in the area of fatigue behavior of steel moment connections and the concept of low-cycle fatigue.

In chapter three of the current document, the analytical case studies are explained and the investigated buildings are described. Also, the observed damage in these buildings is studied.

Chapter four contains a series of linear and non-linear time-history analyses and includes a very thorough analytical study on the stress histories at the critical locations of the buildings. Furthermore, the contribution of each mode of vibration to total stress is investigated.

Chapter five establishes a comprehensive fatigue analysis procedure, which was developed using the Palmgren-Miner method. In addition, low-cycle fatigue behavior of Pre-Northridge connections are studied in this chapter, and S-N curves established for the high-cycle fatigue range are extended to the low-cycle region using the limited test results that are available. Fatigue analyses are performed

at
4
critical locations of the moment frames using the established procedure, and the contribution of higher modes in cumulative fatigue is evaluated.
Finally, chapter six summarizes the results of the study and compares the pattern of cumulative fatigue at critical connections to the observed damage. Also, conclusions of this study are included in this section.
5
CHAPTER 2: BACKGROUND
Although structural steel is an excellent building material that has positive characteristics which make it behave well in many conditions, there are still concerns with its fatigue behavior and possible cyclic fatigue failures of steel components in the scientific community.
The topic of cyclic fatigue has been thoroughly addressed by researchers in the areas of fracture mechanics and material science, resulting in some good publications in these areas. In one of the best books in the field [9], Norman E. Dowling explains the concept and applications of low and high cycle fatigue failures. Most of the examples in this book deal with material steel, which is specifically the focus of the current study. Likewise, Rolfe and Barsom cover the concept of fatigue in a thorough manner in their book [7], which is a classic book on the topic.
Historically, the birth of fracture mechanics goes back to 1920s and studies by Griffith. He studied the fracture behavior of silica glass and focused on the effect of defects in lowering the fracture strength of silica glass. He expressed his theories based on the conservation of energy in a closed system (first law of thermodynamics) [27]. The next steps were taken by Irwin (1948) [24] and Orowan (1945) [28] who worked on the fracture of steel and considered the plastic work done during the fracture.
6
In general, fatigue is a type of fracture failure which occurs when a material is subjected to cyclic or repeated loading [27], [9]. In other words, structural members subjected to cyclic loading may fail at stress levels lower than expected as a result of the fatigue phenomenon. Fatigue failure can be represented in three simplified steps [27]:
1. Crack initiation: the material transition from the virgin condition to the formation of macro crack.
2. Crack propagation: stable growth of the crack after the crack initiation phase.
3. Final fracture: unstable, rapid growth of the crack.
If there are pre-existing defects in the material, the crack initiation step could be eliminated, causing the total fatigue life to decrease. The three steps explained above could be represented in the form of [27]:
Nf =Ni +Np
In the above equation, Nf represents the number of cycles to failure, Ni is the number of cycles for crack initiation, and N p shows the number of cycles for crack
propagation. When the number of cycles to failure is expected to be relatively large
(typically larger than 103 cycles), the concept of high-cycle fatigue is often used to 7
represent the situation. On the other hand, when the number of cycles is not large (typically less than 103 cycles) the condition will be referred to as low-cycle fatigue [27].
During earthquakes, steel moment connections could be subject to low-cycle fatigue. Higher mode effects which create a significant number of stress cycles at the connection (in some cases with relatively high stress levels) need to be investigated. Although demand in the beams and columns of connections could be less than the member strength, cumulative fatigue at the connections could potentially damage them during a seismic event.
Although the low-cyclic fatigue failure has been well researched and documented in the last decades, this issue has not been translated properly into the structural engineering practice and commonly used design manuals [27].
SEAOC Seismology Committee, FEMA 350 task group, strongly recommends that further research to be done on the issue of low-cycle fatigue [38].

The FEMA 350 commentary cites low-cycle fatigue as the main cause of failure in some laboratory connection tests but does not give any information or any possible recommendations on the issue.

8

Some of the SAC task groups addressed the low-cycle fatigue issue in their individual reports. For instance, the work done by Ricles et al. in 2000 [37] has a chapter on low-cycle fatigue with a proposed method for predicting crack initiation and extension over the life cycle of a connection utilizing finite element analysis [37], [38].

Barsom (2000) [6], concludes that fatigue is the failure mechanism of the connection. This report was never distributed to the practicing engineers, as only selected SAC committee members received it.

The report by Krawinkler et al. (1983) [26] cites low-cycle fatigue as the failure mechanism of the Pre-Northridge connections. The concept of “cumulative damage” is discussed in this document. The author indicates that each connection remembers the past events, and these past seismic events consume part of the predictable and quantifiable life of a

connection.
In a follow-up to his 1983 report, Krawinkler introduces the cumulative damage testing criteria method for establishing cyclic life of a connection in the ATC-24 report [5].
Furthermore, in his confidential report to SAC “Development of Loading Histories for Testing of Steel Beam-to-Column Assemblies” in August 2000, Krawinkler again
9
suggests the cumulative damage testing per ATC-24 to determine the performance of the connection.
Bertero and Popov (1965) [8], discusses the low-cycle fatigue as a potential cause of failure in the steel members through a series of tests on beam specimens with large deformations. Their tests shows that fatigue, which in the case they studied was mainly caused by local buckling of flanges, was the failure mode of the specimens. They emphasize that fatigue life of a structural component can not be solely estimated by the fatigue characteristics of the material. Other factors need to be considered for determination of the low-cycle fatigue endurance of a structural member like type and size of a member, states of stress and strain across and along the critical region of a member, and most importantly the magnitude and history of alternating stress and strain.
Popov and Pinkey’s (1969) paper [33], indicates that buckling and the cumulative fatigue associated with it are the main modes of failure for the rolled beams.
In their interesting paper submitted to the 10th World Conference in Earthquake Engineering, Kuwamura and Suzuki (1992) [25] conclude that the Pre-Northridge connection has a finite cyclic life and that low cycle fatigue is the failure mode for this connection.
10
Partridge et al. (2000) [31] show that fatigue is the principal failure mechanism of the Pre-Northridge connection. In this paper, constant cyclic tests were performed on 10 beam-column connections. This paper along with other publications by Partridge, Allen, Richard, and Radau ([30], [32], [34], [35], and [36]) strongly demonstrate that the Pre-Northridge connection failure during the Northridge Earthquake was the result of low-cycle fatigue.
As described in this chapter, the low-cycle fatigue issue has been widely addressed in the literature before and after the Northridge Earthquake; however, the methodology of implementing the fatigue considerations in the state-of-the-art design practices has not been properly developed.