Anda di halaman 1dari 11

Application of Finite Element Analysis to the Fracture Strength of Steel MRF Connection

Te-Wei Fan Department of Civil Engineering and Engineering Informatics Chung Hua University No.707, Sec. 2, Wufu Rd., Hsinchu City 300, Taiwan (R.O.C.) Tel.: +886-3-5186114 Fax.: +886-3-5372188 Email: dfan@chu.edu.tw

Abstract
During the inspection for moment-resist frame connection assembled by welding after the Northridge earthquake in 1994 California, many brittle failures were found. Most failures occurred in weld zone between beam and column flange connections, among which the most severe damage was found to be in the weld between the lower beam flange and column flange, and these failures were caused by cracks initiated at the root of weld. The main reason for the crack initiation is not plastic deformation as expected, and the plastic hinge is not located at the weak beam far from the connection as it was designed to be. Evaluation results indicated that the major failure mechanism for the brittle failure behavior of MRF(Moment-Resist Frame) connection is caused by the residual stress induced by welding. Brittle fracture strength of MRF connection is influenced by the following welding related factors: welding residual stress, HAZ(Heat Affected Zone), welding parameter, and weld defect. This paper use finite element method to simulate welding residual stress, and apply them into 3D fracture models. With fracture parameters calculated in finite element analysis, information about fracture strength of MRF connection is acquired.

Keywords: residual stress, moment-resist frame, fracture strength

1.

Introduction
Structural damages had been discovered in

tle fracture at weld area, which is affirmed as the major failure mechanism. Besides, plastic hinge

moment-resist frames which were inspected after the 1994 earthquake in Northridge, California. Most of these damages occurred at flanges welded area in beam and column connections, especially at the weld of lower flange of beams. Connection failures caused by crack-like defects were found to be at the weld root. Reason for crack growth is

was not found on beams with position which should be far from connection. Reasons for the

brittle failure of connection include: low toughness, defects in weld, triaxial residual stress state caused by highly restrained condition, dynamic load, and so on. In order to enable the structural design to

tend to the perfection, Efforts need to be taken involve material science, connection design, welding

not the anticipated plastic deformation, but the brit-

procedure, and construction quality control. One feasible aspect is to analyze the effect of welding residual stress on the brittle fracture of MRF connections. In their investigation, Brust, Dong, and Zhong [1] argue that the welding residual stress has decisive influence on the fracture process of highly restrained connections, and the connection design detail could be the most noticeable restrain condition. Welding residual stress induced by connection design is significant with the result that plastic deformation is unable to adequately develop, and eventually result in brittle fracture. Other investigation finished by Kaufmann and Fisher[2] found that many crack-like defect with various dimensions exist at the weld interface between beam and column connections. Inspections made on frac-

finite element process which contains a simplified MRF connection model with related welding procedure.The idea of this step is to simulate welding procedure when MRF connection is constructed, and derive welding residual stress distribution in beam and column flange connection area.The residual stress results will then be imposed on the model of the following analysis which includes a 3-D connection model.Assume an existing crack at the root of the weld could provide fracture mechanics calculation to analysis the fracture behavior of the MRF connection. Finite element method is widely used in various engineering fields. Structures constructed by welding like steel structure buildings, reactors of nuclear power plants, shipbuilding, etc. abidingly confront the challenge of welding related problems. Welding procedure simulation by finite element analysis is studied by many researchers to investigate debatable issue which is welding related in many structures. Simulation conditions may in-

ture surface revealed that inelastic deformation did not occur, and it was concluded that brittle fracture was likely to take place at the beam and column connections. Resistance to earthquake force of

moment-resist frame is mainly dependent on connection strength, for this reason the discovery above exhibits gravity of this issue. If the brittle fracture strength of the connection of moment-resist frame is smaller than plastic moment capacity, then the failure of that connection is brittle. Factors that affect the brittle fracture

cludes welding procedure details like connection details, number and shape of welding layers, heat input, interpass temperature, cooldown condition, etc. The main purpose is to derive precisely distributed residual stress arise from welding which are attests to be the major cause of failures occurred in structure connections. This paper use finite element method to discuss the fracture behavior of MRF connection with the presence of welding residual stress.

strength of connections include: weld defect, residual stress, HAZ, welding parameter. The conception of this research is to subsume welding parameter in the finite element analysis through heat transfer and thermal stress calculation. With the residual stress results, it is combined with the linear-elastic fracture mechanics in finite element analysis, and then information for the fracture strength of connection could be derived. To exert residual stress influence on the connection for fracture analysis, it needs to establish a

2. Objective
The first task is to set up a finite element procedure with a simplified MRF connection 2D model which is used to simulate welding process. The

objective of this task is to derive welding residual stress distribution in beam and column flange connection area when MRF connection is assembled

with welding. In the second task, the residual stress results will then be imposed on a symmetric 3-D MRF connection model, which contains a crack assumption at the root of the weld, finite element analysis with fracture mechanics calculation is accessed to discuss the fracture behavior of the MRF connection.

Yang and Brust [6] used VFT software to develop a welding simulation technique which is applicable in shipbuilding and repair. They built a 3D model with moving arc to calculate weld heat input and further to predict the distortion of ship structure when welding is exerted. Bonn et al. [7] explored the welding residual stress distribution of piping circumferential weld by simulation and experiment to understand the influence on residual stress field from pipe pressure. Vinas et al. [8] used finite element method to simulate residual stress arise from welding repair. An important perspective is advanced showing that not only high residual stress should be measured, but also low residual stress area, to confirm the fading of residual stress from weld.

3. Anterior Achievements
Brust, Dong, and Zhong [1] propounded their theory that the fracture process is decisively influenced by welding residual stress for highly restrained connections. Due to restrained condition which is conspicuously related to connection design, high welding residual stresses are arises and lead to brittle failure without developing plastic deformation. Kaufmann and Fisher[2] took a similar view through surveying many failed connection welds with crack-like defects of different sizes, and there are evidences showing that no inelastic deformation exists. Joh and Chen [3] performed finite element analysis by ABAQUS in welding simulation for connection flange weld of steel structures. In their research linear elastic fracture mechanics was used to study brittle fracture mechanism of the welds in pre-north ridge connection design. A method to predict the brittle fracture strength of steel structure connection is addressed. Zhang and Dong [4] presented their efforts on welding procedure simulation by finite element method, and fracture mechanics is also incorporated with residual stress results to present the impact on structure safety. Applications of this technique in other engineering fields like Varfolomeyev and Siegele [5] used ABAQUS to simulate welding process of core shroud welding in reactor of nuclear power plants. Stress intensity factor is calculated for postulated crack to evaluate the effect upon integrity of reactor structure.

Figure 1. Simplified 2D model for beam and column flange connection

4. Welding Simulation And Factors


This paper use finite element software ANSYS [9] to carry out concepts that stated in the

previous sections. Couple-field analysis was applied, which comprises two modules:transient heat transfer analysis and transient thermal-mechanical analysis. Phase one is the transient heat transfer analysis to derive temperature history of welding process. A simplified 2D model of beam and column flanges connected by a weld with several weld pass is established as shown in figure 1. It needs to incorporate temperature-dependent material prop-

erties as shown in table 1.

Table 1. Temperature-dependent thermal and mechanical properties Youngs Temperature() Modulus (x103 MPa) 20 100 200 300 360 420 500 650 750 900 1000 198.1 193.2 185.5 177.1 173.6 168.7 159.6 148.4 134.4 105.7 63.7 Thermal Expansion(1/) 12.0 12.6 13.4 13.8 14.1 14.3 14.5 15.0 15.2 15.2 15.2 Thermal Conductivity(W/m) 53.33 50.67 47.33 44 42.19 39.48 37.33 33.24 30.04 27.3 27.3 Specific Heat(J/Kg) 430.59 471.39 522.39 573.39 596.81 632.97 675.39 954.28 576.06 579.62 579.62

(a) layer 1

(b) layer 2

(c) layer 3

(d) layer 4

(e) layer 5

(f) layer 6

Figure 2. Thermal stress variations when welding processed from layer 1(a) to layer 6(f) Based on engineering experience, layers of welding process can be set up as follows: nine welding layers are to be deployed into the weld groove between beam and column flanges, these nine welding layers are simulated to be welded sequentially by the element kill and birth option. Interpass temperature is set to be 255 for last welding layer to cooldown to, before next welding layer shows up by giving element birth. When a welding layer being deployed by giving element birth, its temperature is set to be 1600 as the melting temperature with a hold time. The hold time can be set to be 12 or 24 seconds depending on welding method used. To study how layer 6 are shown in figure 2, and the welding residual stress is shown in figure 3 which will be induced to the next phase of fracture mechanics analysis.

5. Linear-Elastic Fracture Mechanics


The fracture criterion for brittle failure could be defined by using energy criterion or stress intensity factor. This paper apply energy criterion for eva-

luating the fracture strength of connection, by which energy release rate(G) calculation is needed and the critical energy release rate(Gc) must be defined. Gc can be treated as a measure of material ductility, by which an assumed crack will grow when energy release rate G is equal to Gc. Though the connection remains its load bearing capability when an assumed crack starts to grow, but the loading at that moment still be considered as the brittle fracture strength for simplification and conservative assessment. In finite element analysis, J-integral technique can be calculated and convert it to the energy release rate at crack front. Due to the weld area defect often appears to be in different size and type, so a typical type of crack which exists at the edge of beam and column welding fusion line is assumed and the depth of crack is 0.13mm (0.005in). The reason for the crack type

different hold time would affect the residual stress, the hold time can be further set to be 8 or 16 seconds for investigation. Heat convection coefficient

should be set between the surface of steel plate and the contact air which the bulk temperature is 25 as room temperature. Heat transfer analysis and thermal stress analysis could be solved simultaneously or sequentially. When all nine layers have been deposited then the model will cooldown to room temperature by natural convection. As the room temperature is reached by all elements, the thermal stresses at that state would be regarded as the welding residual stress. The 2D beam and column flanges connection model is shown in Figure 1. The welding thermal stress as layer 1 to

and depth assuming is that the anterior research made by Shuey [10] and Popov [11] had proposed a result

of critical energy release rate for this type and depth of crack which existed in certain type of connection. Based on the procedure proposed by Joh [3], the critical energy release rate can be derived in finite element method with referenced experiment result by the following steps: First, establish a finite element model which is equivalent to the connection made for experiment with an assumed crack. Secondly, apply the known brittle failure load from experiment on the finite element model and calculate the energy release rate at crack front using J-integral and conversion. With different test models repeating the steps above, the representative critical energy release rate could be derived.

200 GPa in room temperature.

Due to the me-

chanical and thermal properties of weld are slightly different from base metal, the same material properties of base metal are used for both. Base on the

conclusion of Joh and Chen that the thickness of backing plate at the weld root is not conspicuously influence results of the energy release rate, so the thickness of backing plate is not taken into consideration. The residual stress results acquired from the previous section would be input into the equivalent node of the 3D model by nodal displacements. Phase one of finite element analysis for the 3D model is to increase the load at the free end of beam to derive the elastic stiffness of the model for verification. Phase two is the applying of failure load from experiment to calculate the critical energy release rate. In phase three, analysis of 3D models with and without residual stress inclusion were performed with increasing load until yielding stress reached. Energy release

6. Finite Element Analysis


Before using the experiment results from Popov and Shuey to define the Gc value, an appropriate finite element model is established for verification. As proposed by Joh and Chen, the elastic stiffness derived from load-deformation relationship of connection test result can be used. The specifications of finite element model are as follows:W14257 A572 Grade 50 is used for the column and W36 150 A572 Grade 50 is used for the column. This model is 1/2 symmetric as shown in figure 4 and its mesh in Figure 5 which will also be used later for fracture mechanics calculation. The crack depth is 0.13mm (0.005 in) at the root of the beam and column flange connection weld, and this crack is extended along the whole length of the weld. When meshing the model, quarter-point element technique is applied at the crack front to simulate the singularity of stress field near the crack front. The model is applied with a mechanical load of 873kN (195.15kips) as the failure loading which is obtained from the experiment result of Shuey. The material properties are listed in Table 1 where yield strength is 350 MPa and elastic modulus is

rate variations with increasing load and influence of residual stress on energy release rate will be evaluated and compared for further information.

7. Results and discussion 7.1 Residual stress


The residual stress distribution from finite element analysis of the 2D model is shown in Figure 3. The figure shows that there are two tensile stress zones, the upper area of the weld and the root of the weld. Beyond that area, the residual stress is compressive. The results show good agreement with engineering experience. Results of residual stress distribution in the HAZ of weld from different hold time are showed in Figure 6. The solid line represent stresses with longer hold time and the dashed line represent stresses with shorter hold time. It shows that

longer hold time results in a reduction of the max-

imum stress. However, these changes in the residual stress did not influence the energy release rate for postulated cracks with different depth as shown in Figure 7. In Figure 7 the stress intensity factor are calculated from finite element analysis using fracture mechanics.

nection is closely related to the residual stress scale and distribution. It provides evidence to show that beam and column connection will be getting unstable as the load increases.

7.4 Effect of Crack Depth


The effective fracture toughness value from the experiment made by Shuey is 57.9 kJ/m2

7.2 3D Model Verification


Elastic stiffness acquired from

(330lb/in), which can be defined as the maximum energy release rate when brittle fracture load comes up. To consider the influence of crack depth on

load-deformation relationship of finite element analysis is 23185.6kN/m approximately, which is compared to the value of 23547 kN/m derived by Joh and Chen by finite element analysis, and the value of 25053kN/m derived by Shuey by experiment. The errors are shown to be 1.5% and 7% respectively. The results indicated that the 1/2

the fracture toughness, the finite element model is modified into different crack depth models. The

maximum energy release rate from analysis results are shown in Figure 10. It shows that with increasing loading applied on different crack depth model until the energy release rate reaches the critical value, different fracture loads are acquired, and it also indicated that defining the load when crack growth starts as the brittle fracture strength is reasonable.

symmetric model with crack included is in well simulation condition. The energy release rate converted from J-integral by 2 different paths along crack front is shown in Figure 8. The results indicated that the path-independent J-integral calculation of fracture mechanics can be realizes in this model and the results are reliable. Furthermore, the figure shows that the maximum energy release rate is occurred near the center of the crack, which is the same as evaluation recording from experiment tests.

8.

Conclusions
This paper establishes a simplified MRF

connection model which includes weld layers to simulate the welding residual stress. With the results of residual stress distribution, 3D finite element models containing assumed cracks were analyzed with fracture mechanics to assess the influence of residual stress and postulated crack on the fracture strength of connection. Fracture parameter calculation at crack front after stress analysis is converted to energy release rate to evaluate the fracture behavior. Residual stress distribution obtained from 2D finite element model are compared with different hold time when simulate welding process. The

7.3 Effect of Welding Residual Stress


By increasing the load until the material yield stress is reached, two analyses of 3D models were performed. One with residual stress and another without residual stress included. The results are

shown in Figure 9. The curve represent the model with residual stress possesses higher energy release rate as compared with the one without residual stress included. Furthermore, as the stress increases with

loading, the energy release rate also increases rapidly. It indicated that the brittle fracture strength of con-

results shows that longer hold time setting for welding simulation generate a reduction of the

maximum stress. At the mean time, residual stress changes caused by different hold time simulation for welding did not influence the energy release rate emphatically for a postulated crack. With residual stress distribution introduced into 3D model which contained a postulated crack, fracture parameters were calculated. It is concluded that the maximum energy release rate is occurred near the center of the crack, which is in agreement with related experiment tests. Energy release rate calcu-

plications

for

Structural

Performance.,

Journal of Structural Engineering, Vol. 126, No. 3, 306-315. [5]. Varfolomeyev, I., and Siegele, D. (2002). Analysis of Residual Stresses and Assesment of Postulated Cracks in a Core Shroud Weld., Computational Weld Mechanics, Constraint, and Weld Fracture, PVP-Vol. 434, 33-38. [6]. Yang, Y. P., Brust, F. W., and Kennedy, J. C. (2002). Lump-Pass Welding Simulation Technology Development for Shipbuilding Applications., Computational Weld Mechanics, Constraint, and Weld Fracture, PVP-Vol. 434, 47-54. [7]. Bonn, R., Metzner, K., Kockelmann, H., Roos, E., and Stumpfrock, L. (2002). Temperature and Residual Stress Fields in an Austentic Circumferential Pipe Weld.,

lated from finite element analysis results with residual stress included rise more rapidly with increasing loading than the one with no residual stress included. Thus it can be said that in fracture strength evaluation of MRF connection, residual stress contained in is a necessary factor. In other words, the brittle fracture strength of connection is closely related to the residual stress scale and distribution.

References
[1]. Brust, F. W., Dong, P., and Zhang, J. (1997). Influence of Residual Stresses and Weld Repairs on Pipe Feacture., Approximate Methods in the Design and Analysis of Pressure Vessels and Piping Components,

Computational Weld Mechanics, Constraint, and Weld Fracture, PVP-Vol. 434, 55-62. [8]. Vinas, G., Dauda, T., Moyes, N., and Laird, A. (2002). Application of 3D Finite Element Modelling to Repair Weld Simulation in Industrial Applications., Computational Weld Mechanics, Constraint, and Weld Fracture, PVP-Vol. 434, 39-46. [9]. ANSYS. ANSYS/Advanced Analysis Techniques Guide, Version 6.0, ANSYS, Inc., Canonsburg, PA, USA. [10]. Shuey, D. B., Engelhardt, M. D., and Sabol, T.

PVP-Vol. 347, 173-191. [2]. Kaufman, E. J., and Fisher, J. W. (1995). Fracture Analysis of Failed Moment Frame Weld Joints Produced in Full-scale Laboratory Tests and Buildings Damaged in the Northridge Earthquake. Tech. Rep. SAC 95-08, SAC Joint Venture, Sacramento, Calif. [3]. Joh, C., and Chen, W. F. (1999). Fracture Strength of Welded Flange-Bolted Web Connections, Journal of Structural Engineering, Vol. 125, No. 5, 565-571. [4]. Zhang, J., and Dong, P. (2000). Residual Stresses in Welded Moment Frames and Im-

A. (1996). Testing of Repair Concepts for Damaged Steel Moment Connections. Tech. Rep. SAC 96-01, SAC Joint Venture, Sacramento, Calif. [11]. Popov, E. P., Blonder, M., Stepanov, L., and

Stozidar,

S.

(1996).

Full-scale

Steel

Beam-column Subassemblages, Rep. No. SAC-96-01, Part 2.