Professional Documents
Culture Documents
5.1. Introduction
Crack initiation and propagation is one of the common failure modes for tubular joints subjected to tensile loads. However, simulation using finite elements based on continuum mechanics formulation does not represent the effect of crack initiation and growth, which violates the continuity and integrity of material and geometry.
Four numerical methods conventionally used to simulate crack initiation and propagation include: discrete crack model, fracture mechanics, smeared crack model and continuum damage mechanics. In the discrete crack model, crack develops only along existing element boundaries and the crack growth depends on the mesh size and orientation (Cofer and Will, 1992). For the fracture mechanics approach, the crack size and orientation has to be assumed or known a-prior. In the smeared crack model, the material stiffness decreases to zero when fracture is detected. The continuum damage mechanics - 82 -
Gurson Model for Ductile Fracture approach introduces a damage variable to reflect the amount of damage in the structure throughout the loading history. The constitutive relationship is modified using this damage variable (Lemaitre 1985).
The present study adopts the void growth and nucleation approach which was established by Gurson (1975). The Gurson model simulates the plastic yield behavior of a porous, or void containing, material. Under plastic deformation, the material strain hardens, and voids nucleate and grow, and subsequently lead to deformation localization and fracture. Thomason (1990) reports that all engineering metals and alloys contain inclusions and second-phase particles, to a greater or lesser extent, and this leads to void nucleation and growth .
This chapter begins with a description of the Gurson model formulation. The next section discusses the benchmark study on the classical bar-necking problem. The subsequent section presents the Gurson model simulation in tubular joints, followed by a sensitivity study on the Gurson model parameters. The findings in this chapter are summarized by Qian et al. (2005a).
(5.1)
- 83 -
Gurson Model for Ductile Fracture in which, e refers to the effective stress, m denotes the hydrostatic pressure, f defines the void volume fraction and y is the material yield stress. The yield criterion in Eq. 5.1 becomes von Mises yield criterion when f = 0. For q1 = q2 = q3 = 1.0, the yield function bears the same form as the original Gurson model. Tvergaard (1981) reports that q1 = 1.5, q2 = 1.0 and q3 = q12 = 2.25 better represents materials subjected to plain-strain condition. Instead of f in the original Gurson model, q1f represents the void volume fraction in Eq. 5.1, and magnifies the void volume fraction by a factor of q1 (Thomason, 1990).
The change of void volume arises from two sources: growth of existing voids and nucleation of new voids (Tvergaard, 1981):
df = df growth + df nucleation
(5.2) (5.3)
df nucleation = Av d p
fN sN 1 p N 2 exp 2 2 sN
where Av =
(5.4)
The plastic nucleation strain follows a normal distribution. The nucleated void volume depends on the mean plastic nucleation strain, N, standard deviation, sN, and the void volume fraction of the nucleating particles, fN (Tvergaard, 1981). These three parameters depend on material properties and vary with different materials. Void nucleation initiates once yielding occurs for high strength steel FeE 690 (Arndt and Dahl 1997). fN refers to the ratio of the void volume in the nucleating particles over the entire volume of the material and is thus less than the total void volume fraction of the material, which is normally less than 10% (Thomason, 1990). Otsuka et al. (1987) reports that ductile crack
- 84 -
Gurson Model for Ductile Fracture initiation occurs when the void volume fraction reaches about 4-6% for bending mode. The Gurson model is an upper-bound theory (Gurson, 1977) and does not include the plastic limit load failure of the inter-void matrix (Thomason, 1990).
In this chapter, the load-deformation curves for tubular joints follow the notations below.
Curve name (CT) FE model with crack-front simulation Curve name_f Curve name_m Curve name_c fine mesh is employed medium mesh is employed coarse mesh is employed
Figure 5.1 describes the details of the FE mesh and geometry, with the material and geometric property obtained from ABAQUS benchmarks manual (2001). Figure 5.1(c) indicates the Gurson model parameters used in the analysis, with qi (i =1, 2 or 3) - 85 -
Gurson Model for Ductile Fracture parameters modified by Tvergaard (1981). Figure 5.1(c) indicates little difference between 2D and 3D models.
R=1
P (unit)
Rotating axis
0.5l0 = 4
0.0 0.0
0.2
(unit)
0.4
0.6
0.8
Fig. 5.1 3D solid model; (b) 2D axi-symmetric model; and (c) Load-deformation curves; for the bar necking problem.
Tvergaard (1981) reports that the qi parameters depend on the strain-hardening properties. Faleskog et al. (1998) propose a relationship between strain-hardening and qi values. According to Bessen et al. (2001), q2 = 1.15 provides the best fit for round bar and plane strain specimens. Therefore, three values of q2 are selected: q2 = 1.0, 1.15 and 2.0. In
- 86 -
Gurson Model for Ductile Fracture addition, comparison is carried out between the Tvergaard model and the original Gurson model. A relatively large value of q1 = 2.0 is also included for comparison.
N=0.30 sN=0.10
q2=1.0 q2=2.0
N=0.30 sN=0.10
3.0
1.5
0.0 0.0
0.2
0.4 (unit)
0.6
0.8
0.0 0.0
0.2
(unit)
0.4
0.6
0.8
(a) Effect of q1 Fig. 5.2 (a) Effect of q1; and (b) Effect of q2; on the tensile bar.
(b) Effect of q2
Figure 5.2 compares the effect of qi on the 3D tensile bar model. The effect of qi is noticeable only after the strength reduction becomes significant. Increasing qi either magnifies the effect of void volume fraction f or hydrostatic pressure p, which causes more severe reductions in the tensile strength. In addition, variation of qi has no observable effect on the initiation of strength reduction in the tensile bar.
Gurson Model for Ductile Fracture compared using the Gurson model. Both joint specimens experience brace axial loads. Table 5.1 lists the geometric parameters of these two joints. For both joints, the initial crack locates at the chord saddle.
Table 5.1 Geometry of tubular joints with and without initial crack
Reference HSE (1999) Sander & Yura (1986) Kang et al. (1998c) Wang et al. (2000)
5.4.1.1. T-Joint
Zerbst et al. (2002a) report four T-joint tests with the same joint parameters ( = 0.51, = 7.46, = 1.0, = 10.2) and different pre-crack geometry. The crack is introduced as a surface notch at the saddle point. The crack length, 2c, is 46.5 mm for all four joints. The T-joint with the crack depth, a = 9.94 mm, is selected. Figure 5.3(a) shows the schematic configuration of a T-joint, with the crack front geometry. The numerical analysis includes four different models. Figure 5.3(b) shows crack-front modeling of the surface notch, while Fig. 5.3(c) (e) shows three continuous FE models which ignore the notch (i.e. an intact model), with different mesh density, since the Gurson model depends on values of plastic strains. The fine mesh utilizes forty elements around the quarter brace-chord intersection curve in order to ensure the element aspect ratio to be close to 1:1:1.
- 88 -
Gurson Model for Ductile Fracture The boundary-loading condition of the T-joint assumes pin supports to the chord ends and applies tensile loads along the brace axis. Figure 5.4 compares the load-deformation curves of the numerical analyses and test results. In Fig. 5.4(a), the numerical analysis includes the void growth without the void nucleation property. The initial porosity f0 imposes a critical effect on the joint strength. However, the initial porosity can be improved during material production process. In spite of the large initial porosity (0.05) for the crack-front model, no load reduction occurs in the Gurson model with only the void growth process. Figure 5.4(b) compares joints with and without incorporation of the void nucleation process. Material softening becomes significant with void nucleation under large plastic deformations, and consequently causes a slight load reduction.
Crack tip 2c a t0
t1 d1
Surface notch
Fig. 5.3 (a) T-joint configuration; (b) Crack-front modeling; (c) Fine mesh; (d) Medium mesh; and (e) Coarse mesh; for pre-cracked T-joint.
- 89 -
P/fyt02
P/fyt02
0 0.00
0.05
/d0
0.10
0.15
0 0.00
0.05
/d0
0.10
0.15
(a) f0
Fig. 5.4 Comparison of numerical results for the effect of: (a) f0; and (b) Void nucleation.
Figure 5.5(a) compares three fN values on the crack-front model. The Gurson model depends significantly on the fN value. A large value of fN corresponds to a less ductile material, and the strength reduction initiates at a very small deformation level. The comparison between different fN values shows that the difference between fN = 0.10 and 0.20 is much more dramatic than the difference between fN = 0.04 and 0.10. Figure 5.5(a) incorporates the load-deformation curves for the three meshes of the intact model. For all three meshes, the Gurson parameters assume the following values: q1 = 1.5, q2 = 1.15 N = 0.10, sN = 0.05 and fN = 0.10. The crack-front model develops from the medium mesh model. This implies that the mesh size further away from the crack-front is similar to the medium mesh. Similar load-deformation characteristics are observed between the crackfront model and the medium mesh for the same material parameter fN = 0.10. This indicates that the mesh refinement in the crack-front may affect the local stress evaluation around the crack front. It does not, however, impose a significant influence on the global - 90 -
Gurson Model for Ductile Fracture joint strength computation since the very refined elements in the crack front do not form a collapse mechanism along the joint load path. Variation of the material parameters, fN, may exert a comparatively stronger effect on the joint strength. On the other hand, with the same material parameters, the effect of void growth and nucleation becomes most significant in the fine mesh which causes the largest plastic strain and most pronounced strength reduction. Figure 5.5(b) compares the load-CTOD (Crack-Tip Opening Displacement) curves. The difference between the numerical analysis and the test increases as the joint deformation increase, since the current FE model does not simulate crack propagations.
Test FE
d0=298.5mm
P/fyt02
2000 fN = 0.04
0.10
0.15
1.2
(a) Effect of fN
(b) CTOD
Fig. 5.5 Comparison of numerical results for: (a) Effect of fN; and (b) CTOD.
5.4.1.2. X-Joint
The HSE report (1999) describes a series of X-joint tests with and without cracks. One Xjoint with through-thickness crack at the saddle position is selected in this study. The - 91 -
Gurson Model for Ductile Fracture crack length extends about 15% of the brace-chord intersection perimeter. The throughthickness crack is simulated in two ways. In the normal fine FE mesh as indicated in Fig. 5.6(a), eighty elements are utilized to simulate half of the brace-chord intersection, with the elements corresponding to the crack removed. The element aspect ratio in the bracechord intersection area is around 1:1:1. In the other mesh scheme, the crack-front is simulated in the refined model.
Figure 5.6(b) compares the numerical crack-front model and test results. The tested joint experiences strength reductions as crack propagates. Without the Gursons algorithm, the FE crack-front model sustains increasing strength within the prescribed deformation. Different fN values with q1 = 1.57, q2 = 0.97 (as suggested by Faleskog, et al., 1998) and initial porosity equal to 0.01 are investigated. The Gurson model analysis with a smaller fN (0.04) value terminates due to the numerical convergence at a deformation level corresponding to the load reduction in the test. With a larger value of fN (0.20), the analysis terminates at an even earlier deformation level due to very large deformation of the crack-tip elements. This proves the inappropriateness of a large fN value. Table 5.2 lists the joint strength for the two joint specimens with crack-front simulations. The strength definition follows the plastic limit load approach presented in Chapter Four. In cases where the FE analysis terminates earlier than the plastic limit load approach, the joint strength refers to the load level at the end of the analysis.
Figure 5.6(c) presents the results of the normal fine FE models. With the absence of the Gurson model, the crack propagation is not captured, and the joint strength prediction can be un-conservative if a large deformation limit defines the joint strength. Both the crack- 92 -
Gurson Model for Ductile Fracture front model and normal fine mesh (with the same material properties) show similar ultimate strength levels compared to the test results. This implies that the simulation of the crack-front singularity is not necessary in the strength analysis with the Gurson model.
15 P/fyt02 P/fyt02
15
10
termination of analysis
=0.48 =15.1 =0.5 =9.1
10
termination of analysis
=0.48 =15.1 =0.5 =9.1
0.06
0.08
0.10
Fig. 5.6 (a) Crack-front mesh and fine mesh; (b) Crack-front FE results; and (c) Fine mesh FE results; for X_HSE joints.
Both cracked joint models demonstrate the importance of the Gurson model parameters. Although the crack in the T-joint does not cause strength reductions, the inclusion of a large fN (0.20) magnifies the effect of crack and leads to under-estimations of the joint - 93 -
Gurson Model for Ductile Fracture strength. For the X-joint, a large fN (0.20) value terminates the numerical analysis at a small deformation level with the joint strength lower than the test results, as shown in Table 5.2. For both joints, the stress evaluation within the crack-front region differs highly from the normal FE mesh without the crack-front simulation. However, the small crack-front region does not form a critical failure in the joint load-path. As a result, only marginal strength difference exists between the crack-front model and normal FE model.
Table 5.2 T- and X_HSE joint strength for varying fN with crack-front simulation
FE results Joint fN (kN) 0.04 T-joint 0.10 0.20 0.04 X_HSE 0.10 0.20
a
FE/Test
Test (kN)
2170
0.91
2397
2589*
1.33
1945
* Strength taken at the end of analysis Strength corresponding to first crack (with chord bending effect incorporated if present) Ultimate strength (with chord bending effect incorporated if present)
b
Table 5.2 compares the current FE predictions against ISO (2001) mean strength formulation and FE computation from the reported study (Zerbst et al., 2002b & HSE, 1999). The ISO joint strength corresponding to crack initiation under-estimates the X_HSE joint strength by about 20%. Over-conservatism is observed for X_HSE joint using ISO ultimate strength formulation. On the other hand, both ISO strength predictions over-estimate the T-joint strength. The FE prediction presented by Zerbst et al. (2002b) shows a lower joint strength when plasticity occurs, although the joint strength does not - 94 -
Gurson Model for Ductile Fracture indicate any sign of strength reduction. For X_HSE joint, the FE results reported by the researchers (HSE, 1999) shows increasing joint strength with increasing joint deformation, since crack effect is not incorporated into the FE modeling.
In the selected tests (except for X2), cracks initiate in the loading history, which leads to reductions in the joint capacity. Since the crack locates in the chord wall, the Gurson model property is assigned to the welds and chord material. The default Gurson model parameters are: q1 = 1.5, q2 = 1.15, q3 = 2.25; and the material parameters are: N = 0.10, sN = 0.05, fN = 0.10. The values for N , sN and fN are selected within the nominal range empirically, since they are not reported in the respective tests. Section 5.5 investigates the effect of these parameters, through a detailed sensitivity study.
FE_f FE_f(G)
0 0.00
0.02
0.04
/d0
0.06
0.08
0.10
0.04
/d0
0.08
0.12
(a) X1
(b) X2
Fig. 5.7 Comparison of test and numerical results for: (a) X1; and (b) X2.
Figure 5.7 compares the X-joint behavior with and without the Gursons algorithm. The two duplicated tests for X1 fail by fracture failure. The tests stop once the crack penetrates through the chord wall, although no sign of load reduction exists. Similar trend among different meshes is observed as that of the T-joint. Hence, Fig. 5.7 shows only results from fine mesh, where substantial strength reduction occurs. For both X1 and X2, the predicted peak loads remain similar to that in the tests, although the corresponding deformation levels are larger than that in the tests. In X1 tests, the fracture failure is defined as the through-thickness penetration of the crack in the chord wall, which has not mobilized the full joint capacity. In X2 tests, no sign of crack failure exists at the end of the tests. The first crack in all four tests does not limit the joint capacity as shown in Fig. 5.7. This implies that the joints offer a significant amount of reserve strength beyond the crack initiation. It is the propagation of crack that leads to the loss of joint capacity.
- 96 -
Kang et al. (1998c) reported the effect of chord pre-load on the moment capacity of Xjoints. The OPB test without the presence of chord load is selected as the reference in the current study. In Kangs test, pure moment is applied through a four-point load mechanism on the brace, as illustrated in Fig. 5.8(a). In the FE simulation, pure moment is applied. The measurement of brace rotation is taken consistently within the pure moment loaded brace corresponding to the test. The element size is refined in the braceto-chord intersection such that the element aspect ratio is around 1:1:1.
d0=169mm
F/2
F/2
Mopb/fyd1t02
F/2 Fx/2
F/2
N=0.10,
fN=0.10, sN=0.05
0 0.0
0.3
Fig. 5.8 (a) Load application in Kangs test; and (b) Comparison of FE and test; for OPB loaded X-joint.
- 97 -
Gurson Model for Ductile Fracture Figure 5.8(b) compares the test and numerical results. Without the Gurson model, the Xjoint sustains increasing load with further deformation. Although the load reduction is not captured by the Gursons algorithm, the presence of voids softens the material and reduces the joint strength when plasticity occurs and propagates in the joint. This weakens the joint strength without introducing an abrupt load reduction. Fracture failure occurs far beyond Yuras deformation limit, which is a deformation level that practical joint may not be able to achieve (Zettlemoyer, 1988).
5.4.2.3. K-Joint
The K-joint test reported by Wang, et al. (2000) is selected as the reference. No initial crack is introduced in the joint test. Similar to the study of X- and T-joints, three FE schemes are adopted to verify the effect of mesh density on the joint response. In the fine mesh, four layers of elements are employed in the gap region of the chord. Sixteen elements are utilized along the 30 mm gap length. The element aspect ratio in the gap and the brace-to-chord intersection region is maintained around 1:1:1. Figure 5.9(a) shows the boundary conditions of the K-joint.
Figure 5.9(b) shows the load-deformation curves for K-joints. The under-prediction of the ultimate joint strength may be caused by the un-reported material strain-hardening. In spite of the initiation and propagation of the crack in the K-joint, no load reduction exists in the compression brace behavior. With the Gurson model, a slight load reduction is observed at a relatively large deformation level, due to the reduction in the tensile brace resistance initiated by void growth and nucleation processes. Conventional FE analysis without the Gurson model does not indicate a reduction in the strength. The first crack - 98 -
Gurson Model for Ductile Fracture reported in the test occurs much earlier than the ultimate strength. However, the joint strength continues to increase with the presence of crack.
d0=217.4mm
30 Psin/fyt02
20 First crack 10
N=0.10, fN=0.10, sN=0.05
x y z
0 0.00
0.04
/d0
0.08
0.12
Fig. 5.9 (a) Boundary condition; and (b) Comparison of FE and test results; for the K-joint
5.4.3. Discussion
Table 5.3 lists the joint strength obtained from the FE analyses (using the Gurson model with N = 0.10, sN = 0.05 and fN = 0.10) and the test results for the intact joints. There are three strength definition adopted for the FE results (obtained from the fine mesh): the plastic limit load approach; Lus deformation limit (Lu et al., 1994) and 15% plastic strain (Dexter and Lee, 1999a). For the T-joint, the joint strength at the end of the test is recorded. Among the X- and K-joints, the peak loads in the tests correspond to fracture failure of the joint except for X2, for which, no crack develops. Table 5.3 compares the current FE results against the ISO mean strength, which shows very conservative
- 99 -
Gurson Model for Ductile Fracture predictions for X-joints under brace tension. For X-joints under OPB and K-joint under brace compression, over-estimation of the ISO formulation is observed.
Plastic load Joint Test FE X1 (kN) 2248 X2 (kN) 397 2154 392 10.4 210 FE/Test
15% strain FE 2016 334 10.9 206 FE/Test (kN) 1054a 2194 200 302
b a
ISO ISO/test
18.2 277
225
Table 5.4 Displacement at 15% plastic strain for three joint types
Coarse/Fine
Lus deformation limit shows similar strength predictions as the plastic limit load approach. Both approaches correspond to a state earlier than the fracture failure in the tests. The plastic strain limit approach defines the joint strength corresponding to the first attainment of 15% plastic strain at the element integration point around the brace-chord intersection. The non-dimensional joint displacement /d0 corresponding to the 15% plastic strain for different mesh schemes is tabulated in Table 5.4. The displacement in
- 100 -
Gurson Model for Ductile Fracture the coarse mesh increases up to 3.3 times that of the fine mesh, which implies significant size dependence of the plastic-strain criterion.
Load path remains crucial in the void growth and nucleation process. Gursons approach to simulate the effect of ductile fracture is based on the assumption of continuum mechanics, with the yield condition and plastic flow rule modified by the void volume fraction f. Stress redistribution is mobilized once material softens due to the propagation of voids, with the adjacent material mobilized, contributing to the joint strength. It therefore, delays the formation of the complete collapse of the tensile load path. Consequently, load reductions predicted by the Gurson model correspond to larger displacements as compared to those obtained from the tests for most of the joints studied.
For X-joints with high ratio (X1), a very small portion of the chord wall between the two braces forms the critical load path. For small joints (X2), the chord material between the two braces constitutes a relatively larger volume. Stress redistribution from the highly stressed area to other regions help to release the high stress in the critical load path, which in turn delays the effect of the void growth and nucleation. As shown by the FE results in Fig. 5.7, there is an obvious strength reduction in X1 at around the deformation level of /d0 = 0.08. However, there is only slight strength reduction for X2, when the deformation level exceeds /d0 = 0.10. This shows that the effect of crack is more significant for X-joints with high than low ratios, which is consistent with the observation reported in the Health and Safety Executive (HSE) report (HSE, 1999).
- 101 -
Gurson Model for Ductile Fracture The brace-chord intersection of a T-joint, as compared to an X-joint, is under additional tension due to bending generated by chord end reactions. This complicates the loading on the crack front. The assumed Gurson properties for the intact joint may not be a good representation of the real material property and this may be a possible cause for the significant void growth and nucleation effect in the intact fine mesh with N = 1.0, sN = 0.05 and fN = 0.10.
The gap region between the braces forms the critical load path for K-joint. The gap near the weld toe of the tension brace experiences tensile and shear stresses. However, the brace tension is proportioned by the brace angle , and the tensile stress induced by the local bending may not be of a significant magnitude. In Fig. 5.9, there is a marginal difference between the analysis with and without the Gurson model around the first plateau of the load-deformation curve. However, as plasticity spreads in the joint, the difference becomes more pronounced.
Gurson Model for Ductile Fracture each material parameter, the rest of the parameters remain as constants. As the strain hardening parameter varies, N, sN and fN take the values of 0.10, 0.05 and 0.10 respectively. As N changes, sN and fN equal 0.05 and 0.10, with the reported material strain hardening property included. For the variation of sN, N and fN both take the value of 0.10. With fN varied, N and sN remain at 0.10 and 0.05.
5.5.1. Strain-Hardening
Different strain-hardening relationship causes different levels of plastic strain under the same loading condition. The current study adopts the power hardening law described in Eq. 5.5.
E , = y E y
for ,
N
y
(5.5)
for
>y
The hardening relationships vary the value of N: 5, 10, 20 and (corresponding to elastic-perfectly-plastic property). When the hardening relationship changes, the Youngs modulus remains at 205 GPa. The joint deformation at the load reduction, as tabulated in Table 5.5, reflects the effect of void growth and nucleation.
Similar trends in the three types of joints are observed. Strength reduction occurs at a smaller deformation level with a larger N (lower strain-hardening). The strain-hardening relationship for the reported X-joint test resembles that represented by N = 10. For the Tjoint, N = 10 also approximates closely the reported strain-hardening. For all three joint types, N = 5 corresponds to an un-realistically high strain-hardening, which produces
- 103 -
Gurson Model for Ductile Fracture very high joint strength. Higher strain hardening delays the development of plastic strain, which postpones the void growth and nucleation process.
The joints with elastic perfectly plastic property, which is normally assumed in the design, require special consideration in the design procedure. With zero strain-hardening, the plastic strain attains easily the nucleation strain, which causes dramatic loss of the joint strength as demonstrated in Table 5.5. This does not reflect the real material behavior, for which strain-hardening exists to some extent.
Table 5.5 Peak joint strength and corresponding deformation for various material properties
X1 /d0 0.095 0.051 0.029 0.011 0.097 0.075 0.105 0.105 0.075 0.031 0.075 0.084 Ppeak/fyt02 191 109 82.2 65.0 111 103 123 118 103 79.6 103 108 /d0
T-joint Ppeak/fyt02 17.2 12.6 10.8 9.53 11.1 10.9 12.4 11.8 10.9 10.0 10.9 11.3 /d0 0.117 0.067 0.043 0.032 0.060 0.049 0.119 0.091 0.049 0.033 0.049 0.060
K-joint Ppeak/fyt02 38.0 30.5 28.7 26.1 26.7 26.8 26.9 26.8 26.8 26.7 26.8 26.8 0.064 0.024 0.018 0.015 0.018 0.017 0.018 0.018 0.017 0.018 0.017 0.018
5 10 20
0.0 N (sN=0.05, 0.10 fN=0.10) 0.30 0.04 fN (N=0.10, 0.10 sN=0.05) 0.20
sN 0.05 (N=0.10, fN=0.10) 0.10
Test results
0.049
105
0.103
12.5
28.7
- 104 -
5.5.2. Effect of N
The value of N takes 0.0, 0.1 and 0.3 to assess the effect of N on the Gurson model. The case of N = 0.0 implies immediate void nucleation once material yields. A relatively large N (0.3) corresponds to a very ductile material. The values of sN and fN takes 0.05 and 0.10, respectively, as N is varied. Figure 5.10 illustrates the effect of N on X1 and Tjoint. The ultimate load level does not vary much with the N value, with strength reductions first observed in both joints with N = 0.10.
N=0.0 (G)
=1.0 =25.5
N=0.0 (G) =0.51 =7.46 N=0.10 (G) =1.0 =10.18 N=0.30 (G) d0=298.5mm
X1(test B) 15
Test
P/fyt02 5
100
10
50
sN=0.05 fN=0.10
0 0.00
0.04
/d0
0.08
0.12
0 0.00
0.04
/d0
0.08
0.12
(a) X1 Fig. 5.10 Effect of N on Gurson model for: (a) X1; and (b) T-joint.
(b) T-joint
Table 5.5 reflects the effects of N on three joint types via the joint peak load and the corresponding deformation. The same trend exists for all three joints. The ultimate strength level does not vary significantly with N. However, the deformation level corresponding to the peak strength becomes relatively larger for X- and T-joints with N = - 105 -
Gurson Model for Ductile Fracture 0.30 than N = 0.10 and 0.0. A small value of N (0.0) does slightly reduce the joint strength once plasticity occurs. This early initiation of void nucleation does not cause a consequently earlier load reduction for all three joints as shown in Table 5.5 and Fig. 5.10.
For all the three joint types, the effect of void nucleation becomes more pronounced in N = 0.10, followed by N = 0.0 and N = 0.30. For X-joints, fracture failure occurs at the deformation level around /d0 = 0.05 in the test. All three values of N cause a strength reduction at a larger deformation level. For T-joint, load reduction does not occur in the test. The value of = 0.30 delays both the initiation of void nucleation and the consequent load reduction, and thus shows a closer correlation with the tested joint. The effect of the N on the tension brace of K-joint is least prominent, due to the fact that the primary failure mode of the K-joint is not fracture failure.
Both Table 5.5 and Fig. 5.11 demonstrate the significant effect of fN. From Eqs. 5.3 and 5.4, the void nucleation rate is directly proportional to fN. Small fN (0.04) delays the development of void volumes and a relatively large fN (0.2) accelerates the void - 106 -
Gurson Model for Ductile Fracture nucleation process, causing premature strength reductions. The joint strength differs more significantly between large fN values (0.1 and 0.2) than between smaller fN values (0.04 and 0.1). As mentioned, however, fN = 0.2 does not describe a realistic material.
fN=0.04 (G) =1.0 =25.5 fN=0.10 (G) =1.0 =17.5 fN=0.20 (G) d0=407.4mm X1(test A) 120
N=0.10
sN=0.05 (G) =1.0 =25.5 sN=0.10 (G) =1.0 =17.5 test A test B 120 d0=407.4mm
X1(test B)
80
0.08
0.12
0 0.00
0.04
/d0
0.08
0.12
(a) fN Fig. 5.11 Effect of (a) fN; and (b) sN; on X1 joint.
(b) sN
The value of sN does not affect the joint behavior significantly. Similar joint strength among different values of sN is observed for all joints, with slight variation in the peak strength and the corresponding displacement. The effect of fN and sN is more pronounced for X- and T-joints, with very little effect on the K-joint behavior.
5.6. Summary
The Gurson model reflects the fracture failure by load reductions in the load-deformation curves of tubular joints. The accuracy of the Gurson model depends primarily on two factors: the mesh size and the material properties. The detailed mesh refinement in the - 107 -
Gurson Model for Ductile Fracture crack front does not form a critical load path, and thus is not necessary in joints with initial cracks. The mesh size should remain sufficiently small near the critical load path to compute accurately the strain value. The current study recommends more than three layers of elements across the chord wall thickness with the element aspect ratio maintained around 1:1:1 near the brace-chord intersection. The Gurson model parameters qi (i = 1, 2 or 3) does not introduce an effect on the tensile bar until the post-peak state. The void nucleation parameters (N, sN and fN) determine the accuracy of the Gurson model, with the effect of fN being most significant. A large value of fN (> 0.10) does not represent the realistic material. Without material data, fN shall vary between 0.0 to 0.10. The value of N hardly affects the ultimate joint strength, while a large value of N (0.30) postpones the void nucleation process. Hence, 0.0 N 0.10 is suggested. sN imposes marginal effect on the joint response. A nominal range of 0.05 to 0.10 is recommended.
The effect of void growth and nucleation on the tubular joint depends also on the load path. Among the three types of joints investigated, K-joint is the least sensitive to variation of the nucleation parameters compared to X- and T-joints.
- 108 -