ARTICLE IN PRESS

International Journal of Impact Engineering 31 (2005) 861876

Finite element analysis of steel beam to column connections subjected to blast loads
Tapan Sabuwala, Daniel Linzell*, Theodor Krauthammer
Department of Civil and Environmental Engineering, The Pennsylvania State University, State College, PA 16802, USA Received 28 August 2003; received in revised form 20 April 2004; accepted 23 April 2004 Available online 26 June 2004

Abstract The behavior of fully restrained steel connections subjected to blast loads was examined using nite element analysis. Two connections that were tested as part of the AISC Northridge Moment Connection Test Program (Report for AISC, 1994) were studied using ABAQUS. Models were validated by comparing numerical results against AISC Program experimental data. Validated models were then subjected to simulated blast loads and their efciency against those blast loads was veried based on criteria specied in TM5-1300 (Department of the Army, Structures to resist the effects of accidental explosions, 1990). Adequacy of TM5-1300 criteria was investigated and critical zones in the connection details were identied. Based on the results of the study, recommendations for modications to TM5-1300 criteria were made and the effectiveness of the chosen connection details under blast loads was summarized. The results showed that the TM5-1300 criteria for steel connections subjected to blast loads are inadequate. Also the unreinforced (pre-Northridge) connection detail performed poorly under blast loads with excessive deections and above yield stresses in the connection region while the reinforced connection detail showed improved resistance against blast loads and this connection may be an option when detailing steel framed connections to resist blast loads. r 2004 Elsevier Ltd. All rights reserved.
Keywords: Blast loads; Finite element modeling; Steel connections

1. Introduction The study of steel connections subjected to dynamic loads was initiated in the 1960s by Popov [1] wherein tests were conducted to study the cyclic behavior of steel moment-resisting
*Corresponding author. Fax: +1-814-863-7304. E-mail address: dlinzell@engr.psu.edu (D. Linzell). 0734-743X/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijimpeng.2004.04.013

ARTICLE IN PRESS
862 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

connections. Since these early studies, investigations have generally focused on their behavior under cyclic loads, such as those generated during an earthquake. However, since September 11th, there is rising concern in the United States over the safety of building structures subjected to blast loads. When a structural steel frame is subjected to blast loads, the beam-to-column connections, which are responsible for load transfer between different members within the frame, play a major role in structural response. Thus, a better understanding of the behavior of structural steel connections under blast loads is of prime importance. However, few studies have been conducted which analyze the interaction of blast loads and structural components of building structures. Case studies based on past attack on buildings subjected to blast loads have been presented such as by Caldwell [2] which focus on the pattern and severity of blast damage sustained by the structure. Most of these studies however take a macro view of the situation and analyze the effect of blast loads on the buildings as a whole instead of identifying the behavior individual structural components of the building structure under such loads. No experimental studies have been reported that analyze the behavior of steel connections under blast loads. Only one theoretical study [3] and one numerical study [4] investigating steel connection behavior under blast loads has been reported. While some general publications, as that published by Conrath et al. [5], dealing with the design of structural systems to resist blast loadings exist, there are only a limited number of code documents that exist related to blast design. The principal code currently used for the design of structures in the United States to resist blast loads is TM5-1300, Structures to resist the effects of accidental explosions [6]. This document provides guidelines for the safe design of structural elements subjected to short duration dynamic loads (i.e. blast loads) and criteria contained within TM5-1300 are oriented toward industrial building applications common to ammunition manufacturing and storage facilities, (i.e., relatively low, single-story, multi-bay structures). The approach presented in TM5-1300 is centered on the response of structures and structural elements that are idealized as equivalent lumped-mass single degree of freedom systems. While general criteria for proportioning low-rise framed military structures to resist blast loads are provided, this publication does not provide specic design guidelines or performance criteria for steel connections under blast loads. Hence the performance of a steel connection has to be judged based on the performance of the steel frame. However, the adequacy and effectiveness of criteria given in TM5-1300 for steel frames is not well understood due to limited research. Hence, given the absence of specic criteria governing connection behavior under blast loads, this study aimed at verifying the adequacy of criteria presented in TM5-1300 for steel frames containing select connection details. Moreover due to the lack of research and specic design guidelines for blastloaded steel connections, existing connection details proven to be effective against dynamic loads were selected in order to obtain an initial understanding of their effectiveness and behavior under blast loads which can then be used as a stepping stone for future studies in this area.

2. Experimental program Connections selected for the current study were tested as part of the AISC Northridge Moment Connection Test Program. The study was conducted by Engelhardt et al. [7] immediately after the Northridge earthquake to provide insight into causes of steel connection failures and to provide

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 863

Fig. 1. Experimental setup (Engelhardt et al. [7]).

preliminary guidelines regarding retrotting any connections to increase their effectiveness against induced seismic loads. The behavior of eight fully restrained connection details was examined experimentally at full-scale under cyclic loads (Fig. 1). Based on tests of unreinforced, pre-Northridge shear tab connections, modications were made to the connection details to attempt to improve their efciency against seismic loads. It was demonstrated by Engelhardt et al. [7] that the cover plate retrotted connection detail in Fig. 2b demonstrated improved performance against cyclic loads by providing higher values of beam plastic rotation, energy dissipation and ductile failure modes when compared to the original connection details. Connections selected for the current study included an unreinforced connection (Fig. 2a) and the reinforced connection (Fig. 2b). Analytically examining these two connections permitted studying the effects of the additional structural elements used for the reinforced connection on blast load behavior.

3. Analytical approach The general purpose ABAQUS nite element code was selected for the current study. Information from previous studies indicated that, for high rate dynamic loads, ABAQUS performed better than other commercially available nite element codes as per a study conducted by Krauthammer [8]. Finite element models were validated by comparing results to experimental data from the AISC Northridge Test Program. Due to small time durations and high pressure loads required for the current study, the nite element models were created using 8-noded continuum (brick) elements with reduced integration (C3D8R). Wedge elements (C3D6) were used to model curved regions of the beam, which

ARTICLE IN PRESS
864 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

Fig. 2. (a) Unreinforced connection (Engelhardt et al. [7]). (b) Reinforced connection (Engelhardt et al. [7]).

included weld access holes at the top and bottom of the web, and the welds. An isometric view of the nite element models is shown in Fig. 3. Numerical models for certain connection components are shown in Fig. 4.

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 865

Fig. 3. Numerical study models: (a) Numerical model for validation studies (unreinforced). (b) Numerical model for blast studies (unreinforced).

Elasto-plastic material properties with isotropic hardening were selected to simulate material behavior of all components in the nite element model except the welds, which were assigned an elasto-plastic material model with perfect plasticity as a brittle material model for steel was unavailable in ABAQUS Explicit.

ARTICLE IN PRESS
866 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

Fig. 4. Overview of components of numerical model.

Member constituent relationships for model validation were based on test data from the AISC Northridge Test Program. Since design and analysis procedures presented in TM5-1300 are based on nominal material properties, blast studies that followed the validation process utilized nominal properties from the Structural Welding Code [9] for the welds and from other sources [10] for

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 Table 1 Connection component yield stresses and tensile strengths for FE models under blast loads Connection component Beam Column Shear tabs Bolts Nuts Weld Cover plates Yield stress (w/o dynamic increase factor) (MPa) 248.2 344.7 248.2 586.0 586.0 482.6 248.2 Yield stress (w/o dynamic increase factor) (MPa) 352.3 424.7 352.3 721.9 721.9 594.3 352.3 Ultimate tensile strength (MPa) 399.9 482.6 399.9 722.0 722.0 594.6 399.9 867

various components of the connection details. In addition, nominal yield stresses were increased using dynamic increase factors as required by TM5-1300 to account for the inuence of high strain rates from the blast loads on the mechanical properties of steel. These increase factors were 1.12 and 1.29 for Grade 50 and Grade 36 steels, respectively. Material properties used for the nite element model for the blast load phase of the study are illustrated in Table 1. Surface-to-surface contact interaction capabilities available in ABAQUS were used to account for the various forces generated between interacting parts of the model. A tied contact formulation was used for the welds and a small sliding formulation was used for the other interacting parts such as boltsbolt holes, bolttabs, tabsbeam, and cover platesbeam. For validation, the loads and boundary conditions applied to the numerical model replicated the experimental setup. The column for the numerical model was xed at the ends and at the base as in the experimental setup (Fig. 3). The beam was cantilevered from the column. Cyclic loads corresponding to those used during the experimental tests were applied to the free end of the beam. The primary response quantity used for model calibration was the displacement timehistory of the beam tip. Predicted and measured displacement time-histories as obtained from the numerical model and corresponding experimental values for the selected connection details are illustrated in Figs. 5 and 6, respectively. Maximum differences between the numerical and experimental results are also shown. Figs. 5 and 6 indicate that good numerical prediction of the behavior existed for both connections. The peak differences between numerical and experimental tip displacements were 9.9% for the unreinforced connection and 6.1% for the reinforced connection. The main aim of this study was to analyze the behavior of the selected connection details under blast loads and benchmark their performance against criteria from TM5-1300. As per TM5-1300, for blast-resistant design only the peak response, from the rst cycle, of the structure is important. This rst response cycle is minimally affected by damping in the system and damping effects are subsequently neglected in the theoretical procedure given in TM5-1300 for evaluating blast load response [6]. Thus, damping was not included in the numerical models. Analyses were carried out over a time duration which would produce one cycle of structural response. Peak displacements

ARTICLE IN PRESS
868 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

Fig. 5. Numerical vs. experimental beam tip displacements (unreinforced connection).

and rotations of structural members were subsequently judged based on their response during this rst cycle. Blast pressures were applied as uniformly distributed loads to the inner faces of the beam and column anges (Fig. 3) and are summarized in Table 2.

4. Theoretical/empirical load development and response prediction Simulated blast pressures were generated using procedures outlined in TM5-1300, in conjunction with the SHOCK and FRANG computer codes [11] and [12] respectively and loading functions corresponding to these blast pressures were applied to the numerical models. Theoretical estimates of the behavior of the studied steel connection details under these blast loads were obtained. The procedure outlined in TM5-1300 to estimate an explosive charge size provides blast pressure for walls of containment structures or cubicles due to an internal or external explosion. Hence, it was necessary to consider the connection details as part of a hypothetical room within which the explosion occurred. Considering this, a simple one-story steel frame structure as shown in Fig. 7 was employed as the theoretical room model. One side of the room was considered to be a frangible panel having a surface weight of 0.96 MPa [6] that would provide realistic venting of the explosive gases.

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 869

Fig. 6. Numerical vs. experimental beam tip displacements (reinforced connection).

Table 2 Numerical blast pressures Specimen Unreinf. Member Beam Load (MPa) 3.2 2.0 0.0 8.6 1.5 0.0 3.2 2.0 0.0 8.8 1.5 0.0 Time (ms) 0.0 3.4 48.2 0.0 1.8 48.2 0.0 3.3 48.1 0.0 1.8 48.1

Column

Reinf.

Beam

Column

Only part of the representative room was considered for the numerical model by taking advantage of symmetry conditions with planes of symmetry located at mid-height of the column and mid-length of the beam (Fig. 7).

ARTICLE IN PRESS
870 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

Fig. 7. Theoretical room used for blast study.

Fig. 8. Theoretical model failure mechanisms.

Theoretical values for the pressures were determined based on the plastic moment capacity of the beam and column for each chosen connection detail. TM5-1300 provides different methods for calculating the plastic moment capacity from structural members of a frame based on the type of anticipated collapse mechanism. The present study considered two typical failure mechanisms as shown in Fig. 8.

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 871

Fig. 9. Unreinforced connection side wall load.

A representative pressure time-history for the connection details is graphically depicted in Fig. 9. For the numerical model, an effective pressure time-history as depicted by solid line in Fig. 9 is used. This time-history shows shock and gas pressures acting on a side wall of the hypothetical room. These pressures are then transferred to the column considering the tributary area of the side wall. A similar procedure is used for the roof and the pressures are then transferred to the beam using the effective tributary area. These transferred loads on the beam and the column are distributed over their anges and these resultant pressures as applied to the numerical model are summarized in Table 2.

5. Performance evaluation In addition to estimating blast load response using ABAQUS, additional estimates were obtained using TM5-1300 procedures for comparative purposes. Response was characterized in terms of the maximum deection at mid-span, Xm, and the corresponding rotational deformation at the member end, y. Design charts from TM5-1300 that related dynamic properties of the structural elements, such as the natural period of vibration (TN), resistance (R), and deection (X), to those of the blast overpressures that consisted of the load (P) and time duration (T) were used. A ductility ratio, m, associated with the ratios T/TN and P/Ru, can be obtained from these design charts. Table 3 summarizes values of maximum deection, Xm, ductility ratio, m and rotational deformation, y, for structural elements of the studied connection details along with the safety limit for rotation as specied in TM5-1300. These theoretical estimations indicated that the representative room could withstand the loads from the explosive charge according to the TM5-1300 limiting criteria.

ARTICLE IN PRESS
872 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

Table 3 Blast load response from TM5-1300 Connection detail Unreinf. Member Beam Column Beam Column Max. deection Xm (mm) 15.7 72.1 17.3 69.9 Ductility ratio (m) 2.0 1.7 2.2 1.8 Rotation y (deg) 0.3 1.3 0.3 1.3 Rotational limit ylimit (deg) 2>0.3 2>1.3 2>0.33 2>1.3

Reinf.

Table 4 Predicted displacements and rotations Spec. Member Response quantity Rotation (y) Displacement (X) Rotation (y) Displacement (X) Rotation (y) Displacement (X) Rotation (y) Displacement (X) Peak numerical value 0.2 9.8 mm 0.1 4.2 mm Peak theoretical value (TM51300) 0.3 15.7 mm 1.3 72.1 mm % Difference 38.3 38.3 94 94 Limiting criteria (deg) 2>0.3 2>1.3

Unreinf.

Beam

Column

Reinf.

Beam

0.1 7.7 mm 0.1 5.0 mm

0.3 17.2 mm 1.3 69.8 mm

55.5 55.5 92.8 92.8

2>0.3 2>0.3

Column

6. Results and discussion 6.1. Unreinforced connection Beam response was evaluated based on end rotation, end displacement, and Von Mises stresses. Numerical model displacements and rotations for both the reinforced and unreinforced connections under the prescribed blast pressures are summarized and compared to theoretical predictions from TM5-1300 in Table 4. As shown in the table, numerical model beam rotation for the unreinforced connection was well within deformation criteria specied by TM5-1300 and hence it is considered as safe under the applied blast loads based upon these criteria. However, TM5-1300 predicted values are 38% higher than those predicted by the numerical model and hence the beam was over-designed for the applied blast loads according to TM5-1300.

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 Table 5 Observed stress summary Specimen Unreinf. Component Beam Column Stress concentration region Lower weld access hole Upper and lower connection point with shear tabs Around bolt holes Bolt shank Bottom groove weld Beam ange at end of top cover plate Lower connection point with shear tabs Around bolt holes Bolt shank Bottom groove weld Tip of top cover plate Maximum stress (MPa) 365.0 124.1 Dynamic yield stress (MPa) 352.3 424.7 Comment Local yielding 873

Shear tabs Bolts Welds Reinf. Beam Column Shear tabs Bolts Welds Cover plates

248.2 262.0 587.4 379.2 151.7 248.2 262.0 358.5 358.5

352.3 722.0 594.3 352.3 424.7 352.3 721.9 594.3 352.3

Near failure Local yielding Local yielding

Stress contour plots of the numerical model indicated that the highest stresses were generated near the weld access hole at the bottom of the beam web due to the reduced area of the web at that location. The observed peak stress in this region was 364 MPa, which was marginally higher than the dynamic yield stress for the beam (352.2 MPa), indicating initiation of local yielding at this location. Column response under simulated blast loads was elastic with low stresses and deformations. Table 5, which summarizes stresses for the reinforced and unreinforced connections under the prescribed blast pressures, indicates that under the simulated blast loads, stresses in the bottom groove welds between the beam and column ange reached a maximum value of 587.4 MPa, which was marginally smaller than the weld dynamic yield stress of 594.6 MPa. These results showed that localized weld failure may occur at these locations. 6.2. Reinforced connection Table 4 indicates that predicted values for beam response in terms of the tip rotation and displacement were overestimated for the reinforced connection. Moreover, differences between predicted and theoretical response quantities for the beam were 55% higher than corresponding differences for the unreinforced connection. Increased discrepancies between TM5-1300 predictions and those from the ABAQUS model were caused by additional stiffness provided to the reinforced connection from the cover plates that were not addressed by TM5-1300 criteria. Thus, as expected, the reinforced connection was more efcient than the unreinforced connection for limiting deections of the beam and the column under blast loads.

ARTICLE IN PRESS
874 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

Table 5 indicates that stresses generated in the beam for this connection were lower when compared to those in the unreinforced connection. They were concentrated in the beam ange at the end of the top cover plate with the maximum stress being 379.2 MPa, which indicated the formation of a yielding zone at that point. This behavior corresponds with the design philosophy of the reinforced connection for seismic loads whereby cover plates were utilized to position plastic hinges in the beam away from the connection zone. Beam stresses in the connection region ranged between 103.4 and 124.1 MPa, which was approximately 65% lower than the peak stress observed for the unreinforced connection. As in the unreinforced connection, the column remained elastic. One of the key problems identied with the unreinforced pre-Northridge connections was the concentration of stresses at the groove welds. The cover plates added to the reinforced connection aimed at reducing this stress concentration for induced seismic loads. Response of the reinforced connection under simulated blast loads indicated that adding cover plates to reduce weld stress concentration can be effectively used for blast loads. Reinforced connection groove weld stress concentrations were signicantly reduced, with peak stresses (358.5 MPa) being 33% less than peak stresses observed in the unreinforced connection (Table 5). Moreover, stresses on the other weld components were also reduced to a maximum of 255.1 MPa. The cover plates were subjected to high stresses at their tips under blast loads with maximum 358.5 MPa at the top cover plate tip, which indicated local yielding. Lower stresses occurred for the bottom cover plate. The reason for the difference in stress was caused by the blast load application procedure, which initially deformed the beam upwards with high pressures. The top cover plate attempted to resist this motion leading to stress concentrations at its tip.

7. Conclusions Based on the results obtained from the numerical models the following conclusions were made regarding connection performance and adequacy of the TM5-1300 criteria. 1. According to the TM5-1300 criteria, structural members were over-designed for the blast loads as peak values of displacements and rotations from the numerical models were lower than the limiting criteria given in TM5-1300. 2. The reinforced connection performed better than the unreinforced connection under blast loads, exhibiting lower displacements, rotations and stresses. 3. The bottom groove weld was an area of concern for the unreinforced connection as it was subjected to high stresses exceeding its dynamic yield stress. This behavior matches observations made during testing of this connection detail under cyclic loads for the AISC Northridge Test Program [7] where the connection detail failed due to localized fracture of the bottom groove welds. 4. Flange cover plates added to the reinforced connection were efcient in reducing the stress concentrations on the groove welds. 5. The formation of plastic hinges in the reinforced connection would occur away from the connection zone at the end of the cover plate as indicated by initiation of yielding of the beam ange at this point.

ARTICLE IN PRESS
T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876 875

6. Failure of the unreinforced connection would appear to occur in the connection zone as indicated by local yielding of the region near the bottom weld access hole and failure of the bottom groove welds. These conclusions appear to indicate that criteria presented in TM5-1300 used to judge the adequacy of a steel frame based purely on rotations of the structural members is not adequate and should be revised. The studied connection details satised these criteria but were shown to be subjected to high localized stresses that were indicative of failure. Thus, it would be advantageous to incorporate a strength based criterion into the TM5-1300 document that would augment existing serviceability criteria to evaluate steel frame performance under blast loads. In addition, any strength criteria added to TM5-1300 should account for contributions to the connection strength and stiffness, and the subsequent increase in blast resistance, which results from the addition of cover plates and column stiffeners to the connection region. It is also recommended that unreinforced connection details similar to those included in this study should not be used for blast-resistant structures as they show undesirable failures within the connection zone near the column face. The studied reinforced connection details, which included cover plates for the beam anges and stiffeners for the column web, showed improved performance under blast loads with reduced stresses in the connection region and indication of plastic hinge formation in the beam away from the connection.

Acknowledgements This work was supported by The US Army Corps of Engineers through a project with the Protective Technology Center (PTC) at Penn State University entitled Protective Technology Research, Development and Implementation in Support of DoD Force Protection Needs. The assistance of PTC personnel and personnel at the US Army Corps of Engineers ERDC is gratefully acknowledged. In addition, data and background information provided by Dr. Michael D. Engelhardt at the University of Texas was invaluable to this investigation.

References
[1] Popov EP. Seismic moment connections for moment-resisting steel frames. Report No. UCB/EERC-83/02, Earthquake Engineering Research Center, Berkeley, CA; 1983. [2] Caldwell T. Bomb blast damage to a concrete-framed ofce building Ceylinco House - Columbo, Sri Lanka. Proceedings of Structures Congress, New Orleans, LA; 1999. [3] Krauthammer T. Structural concrete and steel connections for blast resistant design. Int J Impact Eng 1999; 22(910):887910. [4] Krauthammer T, Oh GJ. Blast resistant structural concrete and steel connections. Int J Impact Eng 1999;22: 887910. [5] Conrath EJ, Krauthammer T, Marchand KA, Mlakar PF. Structural design for physical security. State of the practice report, American Society of Civil Engineers. 1999. [6] Department of the Army. Structures to resist the effects of accidental explosions, TM5-1300. 1990. [7] Engelhardt MD, Sabol TA, Aboutaha RS, Frank KH. AISC Northridge moment connection test program. Report for AISC, 1994.

ARTICLE IN PRESS
876 T. Sabuwala et al. / International Journal of Impact Engineering 31 (2005) 861876

[8] Krauthammer T, Lim J, Oh GJ. Findings from three computer code validations with precision impact test data. Proceedings of the 29th Department of Defense Explosive Safety Seminar, New Orleans, LA; 2000. p. 1820. [9] American Welding Society. Structural Welding Code for Steel/ANSI/AWS Dl.194. 1994. [10] Salmon GC, Johnson EJ. Steel structures: design and behavior, emphasizing load and resistance factor design. New York: HarperCollins College Publishers; 1994. [11] Wager P, Connett J. SHOCK Users Manual, Naval Engineering Lab, Port Hueneme, CA; 1989. [12] Wager P, Connett J. FRANG Users Manual, Naval Engineering Lab, Port Hueneme, CA; 1989.