TMM4160 Fracture Mechanics Abaqus project report

Joakim Johnsen - [email protected]

3rd

October 2010

Preface
Is it

This is, as you know, a preliminary version of my project on Abaqus in TMT4160 Fracture Mechanics. I am very eager to get as much feedback as possible, especially concerning the quantity of the work that is done up til now. satisfying, too little or too extensive? In the start I experienced a lot of trouble installing Abaqus, but by taking my computer with me to the school it was resolved. Abaqus surely is a massive program, and I suspect it will take years to master it. But I must say that it gives a lot of possibilites, and already now I can see the future benets by using this program in my work.

The images in the report is in high resolution (300 dpi), so to look closer, just zoom in. I hope that this as an intermediate project report meets your expectations.

Joakim Johnsen NTNU, Trondheim.

Contents

1 Preface 2 Building a model in Abaqus CAE 3 Theoretical background for various parameters
3.1 3.3 The stress intensity factor The

. . . . . . . . . . . . . . . . . . . .

2 3 4
4 5 6

3.2

J -integral

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CTOD/CMOD

4 Tension model
4.1 4.1.1 4.1.2 4.1.3 4.2 4.2.1 4.2.2 4.2.3

Elastic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . Load-displacement and stress-strain Estimating . . . . . . . . . . . .

7
7 7 8 9 10 10 10 10

KI

. . . . . . . . . . . . . . . . . . . . . . . .

Computing the

J -integral

. . . . . . . . . . . . . . . . . .

Elastic-plastic material . . . . . . . . . . . . . . . . . . . . . . . . Stress ( ) in front of crack tip . . . . . . . . . . . . . . . . Load vs. displacement . . . . . . . . . . . . . . . . . . . . Calculation

. . . . . . . . . . . . . . . . . . . . . . . . .

5 Bending model 6 Source list

11 11

Building a model in Abaqus CAE

Model building in Abaqus is a ow like process, taking you through dierent steps, assigning geometry and properties of the material as you go. The process is presented beneath in numbered sequence, giving a short presentation of each step.

#1 Part #2 Property #3 Assembly #4 Step #5 Interaction #6 Load #7 Mesh

Here you dene the geometry of your part, and slicing it (if necessary) into partitions. Giving you the possibility to assign dierent material attributes to each partition in the next step (#2 Property) Step for creating the material and assigning them to sections (made in #1 Part). A good thing about Abaqus is that you can create numerous materials, and assign them to dierent sections interacting with each other. So that we in our case can get elastic-plastic behaviour in some parts, and linear elastic in other. In this module an instance of the part (or parts) is made, this makes it possible to assemble parts - with dierent local coordinate axis, to instances relative to each other in a global coordinate system. Here you can divide your analysis into steps. I.e. an initial step for which boundary conditions are assigned, and an analysis step for displacements (or loads). Self explaining, module where interaction between instances is quantied. Loading and boundary conditions for each step. See #4 Step. Here you choose what kind and how many elements to be used, and then apply the mesh to the dierent regions.

This is the way to establish a model in Abaqus, after this the Job is dened, generating selected outputs in #4 Step. An important thing to notice when It is therefore imperative using Abaqus, is that you never encounter units. pressure unit (MPa

to be consequent in your choice of units. For example making sure that your

= N/mm2 )

matches your length unit (mm).

3
3.1

Theoretical background for various parameters

The stress intensity factor

Figure 1: Denition of the coordinate axis ahead of crack tip. Starting with the equation Williams used

2 (2 ), where is an Aires stress function, +1 = r f () to show that the stress elds could be expressed by a singular term C1 / r plus higher order terms. By then introducing complex functions and , = (z + ) Westergaard's solution could be used to nd
an example of the remaining integration constant, which was renamed the stress intensity factor

KI .

Westergaard gave us:

y = x = xy = K is denoted KI , KII , KIII .

KI 2r K I 2r KI 2r

3 cos ( 2 ) 1 + sin 2 sin 2 cos ( 2 ) 1 sin 2 sin 32 3 cos ( 2 ) sin 2 sin 2

(1)

with a subscript corresponding to the mode of loading, i.e.

Figure 2: The modes of loading, and how they can be applied.

As you can see from the equations in (1) the crack. Knowing

KI

scales the stress eld in front of

we can determine the stress eld. From the analysis of

K we can compute whether or not the crack-growth will initiate, by comparing KI to the critical stress intensity factor KIc , measured in lab experiments.
3.2
The

The

J -integral
is, similar to

J -integral

K,

a factor scaling the stress eld in front of Hutchinson (1963) and

the crack tip (assuming linear elastic behaviour). eld. curve (

Rice and Rosengren (1968) (HRR) derived the link between

and the stress

For an elastic-plastic material with a Ramberg-Osgood stress-strain

= /0 + (/0 )n ),

proportional stresses and no unloading, HRR

basically used the same procedure as Williams, and obtained:

ij = 0
where:

EJ 2I r 0 n

1 n+1

fij (, n)

(2)

fij : In : n: 0 :
. The

dimensionless functions of material parameter (n

and

n.

function of hardening and stress state.

=1

elastic material).

0 = E
is

0 , yield stress
Meaning that the value is independent of

J -integral

path independent.

the path you integrate over (see gure 3).

Figure 3: Integration path

The proof (a rather long one) of path independance comes from the expression

I=

du wn1 T dx ds, where w (= 0 d ) is strain energy density, n1 surface 1 normal in x-direction, and T is traction on surface. From gure 3 we see that I = JAB + JBC + JCD + JDA , given that I = 0 (path independent) and that n1 = Tk = 0 on path BC & DA, we obtain JAB = JCD = JDC . This is very helpful, because calculation JCD (which can be very close to the crack tip) can be problematic, but calculation JAB (which can be taken far from crack tip) is
both rather easy and accurate.

3.3

CTOD/CMOD

CTOD (Crack Tip Opening Displacement) is dened in more than one way, in Anderson two denitions are used (see gure 4a).

(a) Displacement at original crack tip and at intersection of a

(b) Hinge model

90

vertex with crack anks.

Figure 4: Two denitions of CTOD and hinge model One way of estimating CTOD and CMOD (Crack Mouth Opening Displacement) is by using the hinge model (gure 4b), for plastic material behaviour. By using that the triangles are similar we obtain this formula:

=
where

r(W a)V r(W a) + a (0, 1)).

(3)

is the rotational dimensionless factor (r

It is quite clear that

this approach only works when a plastic hinge exists, and that elastic deformation has to be added. This can be done by adding the strip-yield model, which gives

2 el = KI /(Y S E ).

Summarizing the plastic and elastic contributions gives:

= el + p =

2 KI rp (W a)Vp + Y S E rp (W a) + a

(4)

Tension model

(a) Geometry of plate

(b) Deformed geometry

(c) Old mesh

(d) New mesh

Figure 5

Half of this plate (symmetry) was modelled in Abaqus 6.9.2 CAE, the results of the analysis is presented in gure 5b, and processed throughout this chapter.

4.1

4.1.1 Load-displacement and stress-strain

Elastic material

(a) Load-displacement curve

(b) Stress-strain curve

Figure 6: Some plots of material response (absolute value of stress & load)

One immediate observation is that the plots in gure 6 matches my choice of a linear elastic material. Another observation is that the stresses measured reaches a value above 1200MPa, which is an un-realistic value as the yield strength fy (300, 600) for most common steel components. It is also worth noticing that a ner mesh don't aect the read-outs concerning load-displacement (gure 6a) nor stress-strain (as a consequence of load-displacement).

Computing eective Young's modulus are using plain strain, meaning 188200MPa, is a bit too low. that

E yields; E = = 188 200MPa. We E E = 1 2 . So the E we found, equal to

4.1.2 Estimating K

KI was estimated from the dataplot of the stresses near the crack tip (r KI with respect to KI yields: KI = 11 2r . The (0, 4mm)). Solving 11 = 2r method of tting was used to ensure a best possible approximation of KI ,
results of the curve tting can be viewed in gure 7. In this gure the eect of mesh renement can be viewed, seeing that making the mesh ner results in lower stresses and thus lower fracture toughness (KI

ij ).

Figure 7: Estimation of stress intensity factor

KI

old mesh new mesh

KI (MPa mm) 15848MPa mm 6800MPa mm

4.1.3 Computing the J -integral

Abaqus can be used to compute the

J -integral

by dening the location of the

crack on the part, and then asking for an ODB History Output of the J-integral for that crack. I used 10 contour integrals, which resulted in this plot:

Figure 8: Plot of Abaqus J-integral calculation

We see that the tendency is a J-value of about plot.

1250Nmm/mm2 ,

except for one

By following the tendency the deviating plot was discarded, ending up

with consistent data. Figure 8 also demonstrates the eect of mesh renement, which in this case is approximately none. Both for old and new mesh we obtain

J = 1247kJ/m2 .

An easy way to check if this number is way o or not, is to

J and K for an elastic material, J = EI , here 200000MPa E = 10.32 = 220 000MPa KI = 1247 220000 = 16 555MPa mm. We see that the relationship formula between K and J is not always correct, for example K varies with regards to mesh renement, something J does not.
use the relationship between This is in coherence with gure 9.

K2

Figure 9: Plastic (Hutchinson, Rice, Rosengren (HRR)) and elastic zone (Linear Elastic Fracture Mechanics (LEFM))

4.2

Elastic-plastic material

Now we replaced the elastic material in 4 with an elastic-plastic material, with properties given in #1.1. That resulted in a rather big change in the deformation of the model, as depicted in gure 10.

Figure 10: Deformation of tension model with elasto-plastic material

4.2.1 Stress () in front of crack tip

It was desirable to nd out how the change in material parameters would aect the stress in front of the crack tip. By dening a path along the crack ligament and outputting stress-data along this path we got:

Figure 11: Stresses in front of crack tip

This is very dierent from the stress-plot obtained in gure 7, and we see that it would be impossible to make

ij = KI / 2r

t uniquely to this, i.e. we need

to theories to describe it suciently, HRR and LEFM, as seen in gure 9.

4.2.2 Load vs. displacement 4.2.3 Calculation J

Same procedure as in section 4.1.3, by using Abaqus' built-in contour integrals, gave these plots:

10

(a)

(b)

Figure 12: J-integral and CMOD/CTOD for EP1.

Resulting in

J = 820kJ/m2 .

The CMOD and CTOD were found by creating sets

for the proper nodes (# 1.2), and then creating history output for transversal displacement in x-direction. Analysis gave:

= 0.75mm

and

V = 1.04mm

after

1s.

Both in case of

and CTOD/CMOD mesh renement's aect on read-outs

are incremental.

5 6

Bending model Source list

Zhiliang Zhang, Dept. of Structural Engineering. 1.1 Table 1, page 6 1.2 Section 5.3, page 35

# 1 Manual for the computational excercise in TMM4160 Fracture Mechanics

11