You are on page 1of 19

ADDED NOTES CN4 PLASTICITY ANALYSIS

111

3 2 1

111

2 1

Von Mises yield surface (VM)

Tresca yield surface (T)

3 VM T r
111

2 o VM - o 2 T o 1

1
r= 2 3 o

- o

3 = 0

- plane projection plane = plane through origin normal to 111 T = hexagon and VM = circle (radius r)

Plane stress (3 = 0) T = polygon and VM = ellipse

Figure 1. Yield surfaces plotted in stress space.

-1-

ADDED NOTES CN4 The Tresca and Von Mises yield surfaces are shown in various forms in Fig. 1. The surfaces are plotted in stress space. Stress space is the 1 2 3 coordinate system where each axis represents the value of a principal stress, but not its direction, ie., stress-space axes are not principal axes directions. Points plotted in stress space represent the principal stresses for a given []. If we change [], then the principal stresses will change, and the underlying principal axes directions would also change. We ignore the principal axes directions for plotting stress-space points, but understand that as we move to different points in stress space, there is a change in principal axes directions for each corresponding stress matrix []. The issue of principal axes directions does arise later in developing the stress strain equations for plasticity, and in that context we will assume that the principal axes directions for a particular set of stresses (a point in stress space) and the strains that they produce are the same. We can identify points in stress space as vectors, = (1 2 3). The vector stress (which is not the same as stress vector S used for finding normal and shear stress components) is a useful representation of a stress state [] in terms of its principal stress values. As will be discussed, vector analysis using the stressspace representation is a relatively straightforward means to develop the plasticity equations. Thge equations can also be developed using the full tensor (matrix) representation for stress and strain. In that case, the yield surfaces would be hyper-surfaces in a 6-dimensional space (3 normal stress axes and 3 shear stress axes)! The yield surfaces plotted in stress space in Fig. 1 are various graphical representations of the Tresca and Von Mises yield criterions. In terms of principal stresses, the Tresca criterion is
max max = o

The Von Mises criterion is


eff = 1 2

(1 2 )

+ (1 3 ) + ( 2 3 ) = o
2 2

eff is defined as the effective stress. o is the tensile yield stress of the material. The yield stress will change if the material microstructure is changed, or if the material plastically deforms (strain hardening the plastic flow curve). o is the only material parameter that appears in these yield criterions. Once o is specified, yielding can be predicted for all possible combinations of applied stresses. In this sense, o plays the role of a structure parameter similar to a thermodynamic state variable. -Plane Projection for Yield Surfaces The -plane projection for a VM yield surface in stress space is shown in perspective view in Fig. 2. In order for plastic yielding to initiate, the prinicipal stress state = ( 1 2 3) has to touch the yield surface. Equivalently, the projection dev of onto the -plane must touch the VM circle (radius = r). dev is called the deviatoric component of the stress state . Let n be a unit vector along the normal direction N = (111) to the -plane. Defining the mean hydrostatic

-2-

ADDED NOTES CN4 pressure to be m = ( 1 + 2 + 3)/3, the vector geometry in Fig. 2 gives the following relation for dev ,
1 1 n= (111) = N 3 3 + 2 + 3 1 + 2 + 3 ( n)n = 1 (111) = 1 (111) = mN 3 3 3 dev = ( n)n = mN
dev 1 = 1 m

dev = 2 m 2 dev = 3 m 3 2 VM criterion : dev = r = o = 3

( ) + ( ) + ( )
dev 2 1 dev 2 2

dev 2 2

Figure 2. plane projection for the Von-Mises yield surface (see also Fig. 1).
You can show by substitution of the components that dev = r is identical to eff = o. You can view the deviatoric stress dev to be the given stress state minus its hydrostatic component m N , ie., dev = m N . Since m N is the component of that lies along the hydrostatic

-3-

ADDED NOTES CN4 axis (111), it cannot contribute to yielding; all yield surfaces for volume conserving plastic deformation (V = 0) are open along the (111) direction (Von Mises, Tresca and others). Normality Rule and Levy-Mises Equations The Normality Rule is central to the development of the Levy-Mises plastic strain equations. These equations are widely using for engineering plasticity analysis. The basis of the Normality Rule and the derivation of the Levy-Mises equations is outlined in this section1.

Figure 3. The plastic flow curve obtained using tensile tests. Consider the plastic flow curve obtained from a tensile test on a given material as shown in Fig. 3. The yield stress at the starting point, a, is oi. Due to work hardening the yield stress o increases with plastic strain according to the plastic-flow-curve relation, o = o (pl ) . In order to produce a stable plastic strain increment d, the stress must be raised by an amount d. Mechanical stability requires that the work increment 1/2 dd be positive, or simply that dd > 0. The stability requirement is rather subtle because plastic deformation is a dissipative process and the work done by the loading mechanism cannot be recovered (as it can for elastic deformation). The work is dissipated primarily as heat, but there is also work input to produce new defects such as dislocations (the work-hardening effect). Fig. 4 is a representation of the stability criterion for a dissipative system. Imagine a ball sliding (not rolling) on the surfaces shown where the friction force f = fo. Case i is stable, case ii is neutral and case iii is unstable. In all cases the work dw = fodx (total area under the f-x curves) is dissipated when the ball slides dx > 0 (here, you must apply the force fo to make the ball move in all cases as opposed to the friction-free system where the ball rolls). The second-order incremental work 1/2 dfdx (shaded areas), or just dfdx, is an indicator of stability. df is the force increment needed to slide the ball quasi-statically along the path (no inertial effects). In case i dfdx > 0 because the ball moves uphill (stable system). In case ii dfdx = 0 because there is no hill
1

For further details see Fundamentals of Metal Forming, R. H. Wagoner and J.L. Chenot, chapter 7, Wiley (1997) ISBN 0-471-57004-4. Also see Mechanical Metallurgy, G. E. Dieter, McGraw-Hill (1986) for background.

-4-

ADDED NOTES CN4 (neutral system). In case iii dfdx < 0 because the ball moves downhill (unstable system). In terms of the plastic flow curve, case i is the analog of a work hardening material (most engineering metals), case ii is the analog of no work hardening (perfectly plastic material) and case iii is the analog of work softening. Work softening will occur in some materials and it causes unstable deformation modes such as Luders Bands in mild steels or necking in amorphous polymers (ignore possible stabilization effects due to localized strain rate increases during deformation). For engineering plasticity, we consider only stable deformation modes so that dd > 0.

case i (stable)

case ii (neutral)

case iii (unstable)

Figure 4. Stability diagrams for dissipative systems (friction forces). The Normality Rule follows by extending the previous ideas to a 3D stress state. We use principal stresses so that applied stresses [] are represented as vector in stress space. a Furthermore, we will plot the incremental principal plastic strain d due to the stress in the same stress-space coordinates used for plotting (a scale factor would be needed to get the proper units). For this to be justified, the underlying principal axes directions for and d must be the same. This is discussed in more detail in last section of these notes.
The dot product d d is the vector form of dd. The basic assumptions needed for deriving the Levy Mises plastic strain equations are:

(a) (b) (c)

d d > 0 (stable deformation) d is a unique function of the stress state at yielding

Any parameters required in the plastic strain equations can depend only on the microstructual state of the material (characterized here by o)

-5-

ADDED NOTES CN4

These assumptions are not self evident, nor can they be rigorously derived from physical theory. However, without them we cannot develop a useful set of plastic strain equations. You can view these assumptions in the spirit of thermodynamics. Stable systems are best suited for a deterministic analysis, responses (strains) should be a unique function of the driving forces (stresses), and the parameters in the governing phenomenological equations should only depend on state variables (o). The validity of the plastic strain equations derived must be confirmed by experiment, just as with the laws of thermodynamics. Assumptions (a) and (b) lead to the Normality Rule. Consider a general yield surface in stress space as shown in Fig. 5 where is a stress state onthe yield surface, ie., at the current yield point. In order to produce a plastic strain increment d , the stress must increase by d because of the stability requirement (a). This can be viewed as an expansion of the yield surface because o will increase to o + d o due to work hardening. The stress increment d must lie on the outer side of the tangent plane to the yield surface at the point, as is shown in Fig. 5. If it were inside the tangent plane, it would lie inside the original yield surface where the deformation must be elastic. The normal to the yield surface at is denoted by n . The stability requirement d d = d d cos > 0 means that the angle 90.
The requirement (b) that d be uniquely determined by the stress state atyielding, means that the incremental stress increase d along some loading path through the point can have an arbitrary direction as shown in the inset to Fig. 5. There is only one direction for d that satisfies . If d both of the requirements (a) and (b). That is the direction normal to the yield surface at were not along n you could choose d such that > 90 as is obvious from the inset. Hence, the direction of d must be normal to the yield surface. This is the Normality Rule. A more detailed examination of the geometry in Fig. 5 will also show that the yield surface must be convex outward at every point.

Figure 5. Geometry showing the incremental strain and stress changes at yielding.
-6-

ADDED NOTES CN4

The Levy -Mises equations can be readily obtained from the Normality Rule in conjunction with the Von-Mises yield criterion. Fig. 6 shows a -plane projection for the VM yield surface. As dev discussed earlier, yielding occurs when the deviatoric stress vector touches the VM circle as shown. The Normality Rule requires that d be normal to the yield surface, hence it must be normal to the VM circle. This means that d lies along a radial direction in the -plane, and so dev and d must be coaxial vectors. This would not be the case if we used the Tresca yield surface because the circle in Fig. 6 would then be a flat-sided hexagon (see Fig. 1) and the d vector would not be radial.

dev Figure 6. VM circle in the -plane. and d must be coaxial because of the Normality Rule.

Using the VM yield surface in Fig. 6 we have,


d = dev

(1)

where is a scaling parameter. We will need to evaluate . To do this, take the magnitude d = dev so that,
d = dev

(2)

In order to proceed, we need to invoke assumption (c). This requires that the parameter depend only on the microstructural state of the material, which in our case is characterized by the tensile yield stress o used for the yield criterions. Since a given yield surface is defined by the value of o, must be the same for any point on the VM circle in Fig. 6. This means that we can apply Eq. (2) to any convenient point on the VM circle and use it as the reference stress state for evaluating . We will use a tensile test for the reference state.

-7-

ADDED NOTES CN4 Consider a tensile specimen being plastically strained with the loading direction along the 1axis,

1 = o 2 = 3 = 0 d 1 = d pl 1 1 d 2 = d 3 = d1 = d pl 2 2

The strain increments d2 and d 3 follow from symmetry, d2 = d 3, and the constant volume requirement for plastic strain, d1 + d2 + d3 = 0. Thus, for a tensile test we have,

d = d 12 + d 2 2 + d 3 2 = m =

3 d 2 pl

1 + 2 + 3 1 o = = 3 3 3 2 1 dev = 1 m = o 3 1 dev 2 = 2 m = o 3 1 3 dev = 3 m = o 3 dev dev 2 dev 2 dev 2 = (1 ) + ( 2 ) + ( 3 ) =

2 3 o

The last relation must also follow from dev = r . Substituting into Eq. (2) and using the VM

criterion in the form eff = o , we derive the required value for ,


d 3 d pl 3 d pl = dev = = 2 o 2 eff

(3)

As a final step, we give the form of the plastic-strain equations for any principal stress combination 1, 2 and 3. These are known as the Levy-Mises (incremental) plastic strain equations. Substituting from Eq. (3), the components of the vector strain d in Eq. (1) can be written in terms of the principal stresses 1, 2 and 3 as,

-8-

ADDED NOTES CN4

d 1 = 1 dev = [ 1 m ] = d 2 = 2 dev = [ 2 m ] = d 3 = 3 dev

2 1 d pl 1 1 2 ( 2 + 3 ) = 1 2 ( 2 + 3 ) 3 eff
(4)

2 1 1 d 2 ( 1 + 3 ) = pl 2 (1 + 3 ) eff 3 2 2 2 1 1 d = [3 m ] = 3 (1 + 2 ) = pl 3 (1 + 2 ) eff 3 2 2

In order to get the total strains 1, 2 and 3, the ds in Eq. (4) must be integrated over the loading path. For many problems there is a simplification that avoids the actual integration of the ds in Eq. (4). The simplification requires a radial loading path, which is defined as one such that,

1 = c1 f(p); 2 = c 2 f(p); 3 = c 3 f(p)


c1, c2 and c3 are constants, including zero, and f(p) is an arbitrary function of a load-path parameter p. The latter defines the principal stresses in parametric form. The definition of p depends on the particular problem at hand (for example, p would be the internal pressure that increases stresses in thin-wall pressure vessels). For radial loading, the stress ratios in Eq. (4) are independent of the parameter p. For example, the stress ratio term for d1 in Eq. (4) is
1 1 1 2 ( 2 + 3 ) = 2 eff 1 c1 2 (c 2 + c 3 )f(p)

(c1 c 2 )

+ (c1 c 3 ) + (c 2 c 3 ) f(p)
2 2

Hence, f(p) cancels out in the above expression. Integrating d1 over the loading path gives,

1 =

path

dp dp =

d1

pl dpl 1 1 1 1 2 ( 2 + 3 ) dp dp = 1 2 ( 2 + 3 ) path eff eff

For radial loading, we can replace dpl in Eq. (4) by the total strain pl . The Levy-Mises incremental strain equations reduce to the Hencky total plastic strain equations,

1 = 2 =

pl eff

1 1 1 1 2 ( 2 + 3 ) = 1 2 ( 2 + 3 ) 1 1 1 2 2 (1 + 3 ) = 2 2 (1 + 3 ) 1 1 1 3 2 (1 + 2 ) = 3 2 ( 1 + 2 )
(5)

pl eff 3 = pl eff

A plastic modulus can be defined as,

-9-

ADDED NOTES CN4


= eff pl

The plastic strain term pl in Eq. (5), as well as d pl in Eq. (4), is inconvenient because we are trying to solve for the plastic strains from the stresses. In order to apply Eq. (5) to real problems it is necessary to have the plastic modulus in a more computationally useable form. The plastic flow curve, o(pl) vs pl, measured in a tensile test is the basic input data needed for this purpose. The power-law representation, o ( pl ) = A n = eff , is often convenient (but not the only pl possible choice recall that a power law does not include the initial yield stress oi in Fig. 3). A power-law gives

eff pl

A n An = eff = 1 eff eff n 1

(6)

Eq. (6) eliminates the awkward pl term in Eq. (5). All quantities now involve known material parameters, A and n, and the stresses 1, 2 and 3. The Hencky equations appear similar to Hookes law for elastic deformation. They are only superficially the same because the plastic modulus is not a constant, whereas the elastic modulus E is constant. Furthermore, we use the plasticity equations for loading beyond the initial yield stress. Unloading is elastic and the unloading strains are the elastic strains determined using Hookes law and the current values of the stresses. Example Problem Given: A thin-wall spherical pressure vessel under internal pressure p. The initial radius and thickness are ro and bo. The power-law form of the plastic flow curve obtained in a tensile test is o ( pl ) = A n = eff . pl Find: The p vs r relation for plastic expansion, and the critical radius, rc, at p = pmax. From elementary solid mechanics, the principal stresses for a thin-wall spherical pressure vessel are2

1 = 2 = =

pr (circumferential stress); 3 0 (thickness stress) 2b

The loading is radial for this case. That is, even though both r and b are as yet unknown functions of p, the stresses are all proportional to the same function, f(p), given by

1 = 2 =

1 pr f(p); 3 = 0f(p); f(p) = 2 b

Mechanics of Materials, J. M. Gere and S. P. Timoshenko, Brooks/Cole (1984) ISBN 0-534-03099-8.

-10-

ADDED NOTES CN4

We can the use the Hencky equations to find the circumferential strain 1 (= 2),
1 = ln 2r r 1 1 = ln = 1 ( 2 + 3 ) = 2 2ro ro 2

Evaluate eff and use the power-law flow curve for the plastic modulus ,

eff =

1 (1 2 )2 + ( 1 3 )2 + (2 3 )2 = 2

1 n

1 1 n eff

1 n

1 1 n

Substituting into the Hencky relation for 1 gives,


r 1 n 1 pr n ln = = ro 2 A 2 2bA
1 1

The thickness b can be converted to its initial value bo by using the constant volume condition for plastic deformation in a thin-walled cylinder,

V = 4r b = Vo = 4ro b o
2 2

therefore, r b = ro b o
2 2

Eliminating b gives,
r 1 pr3 n ln = 2 ro 2 2ro bo A
1

Finally, we can solve for p explicitly in terms of r,


2 n+1 ro2 bo A r p= ln r3 ro
n

A plot of p vs r will have a maximum pmax at r = rc. This is an instability point since p decreases for r increasing beyond rc (the sphere will burst - assuming that the pressure source is stiff, i.e., as r increases rapidly p will not decrease). We can find the value of rc by taking the derivative,
2 dp 2 n+1 ro2 b o A r 2n +1 ro bo A r = 3 ln + n ln dr r4 r3 ro ro n n1

1 =0 r

-11-

ADDED NOTES CN4

Hence,
c = ln rc n = ro 3
n 3

or, rc = ro e

This is similar to necking in a tensile test where the axial load reaches a maximum during deformation. As an exercise, do the plastic strain analysis for a tensile test using the Hencky equations. The loading parameter is the applied force F. Principal Axes When deriving the Levy-Mises equations in the previous section, we assumed that the underlying dev principal axes for the stresses, or , and the corresponding strain, d , are the same. If this were not the case, we could not connect the strains to the stresses with a common set of axes; it would not make sense to have normal strains in different directions from the corresponding normal stresses. For isotropic elastic deformation, the principal axes for stress and strain are the same. In a similar fashion, we can justify the assumption that the principal axes for stress and plastic strain coincide using the general tensor (matrix) form for the deviatoric stresses and the requisite plasticity analysis3.
dev 11 [ dev ] = dev 21 dev 31 dev 12 dev 22 dev 11 12 13 1 0 0 13 dev 23 = [] m [I] = 21 22 23 m0 1 0 31 32 33 0 0 1 dev 33

dev 32

Using the tensor forms of the VM yield criterion and the Normality Rule, the Levy-Mises equations are given in matrix form as
dev 11 d11 d12 d13 [d] = d21 d22 d23 = [ dev ] = dev 21 dev d31 d32 d33 31 dev 12 dev 22 dev 32 dev 13 dev 23 dev 33

To demonstrate that the principal axes for the stresses and strains coincide, let n = [n] be a principal axis direction for the principal stress p corresponding to any stress state []. Using matrix representation, we require that [n] and p be eigenvectors and eigenvalues of [], hence

[][n] = p [n]
3

Advanced Mechanics of Materials, H. Ford and J. Alexander, Wiley (1977); Mechanical Behavior of Materials, F. McClintok and A. Argon, Addison-Wesley ((1966).

-12-

ADDED NOTES CN4

Now consider the deviatoric stress [ dev ]

[ ][n] = [][n] [I][n] = [n] [n] = (


dev m p m

m )[n]

Thus, [n] is also a principal axis direction for [ dev ] and the principal stress is p m. Finally, consider the incremental strain [d] in conjunction with the Levy Mises relations

[d][n] = [ dev ][n] = ( p m )[n]


Thus, [n] is also a principal axis direction for [d] . The principal incremental strain is (p m). Based on this, the principal axes for the stresses and plastic strain increment are seen to coincide. We are therefore justified in plotting the stresses and strain increment together at any given point in stress space, as was done in Fig. 5. GENERAL APPROACH TO PLASTICITY ANALYSIS We used vector geometry and the Normality Rule to derive the form of the incremental plasticity equations, the Levy Mises equations. Vector geometry with the -plane representation for the Von Mises yield criterion make this a very simple derivation because the strain increment vector must be coaxial with the deviatoric vector stress direction, ie., normal to the Von Mises circle in the -plane. For other versions of the yield criterion in stress space, the yield surfaces have the form f(1, 2, 3) = c where c is a constant and f is a function of the principal stresses. The projection on the -plane would no longer be a circle. In order to obtain the normal direction to the yield surface, one can use the fact that the vector gradient f of a scalar function f(1, 2, 3) is a vector along the normal to the surface given by f(1, 2, 3) = c. By definition, the vector gradient is
f f f f f = = i 1, 2 , 3

and the unit vector normal to the yield surface would be

f n= f
The derivation of the Levy Mises equations using this approach will be illustrated next. The deviatoric part of the vector stress = (1, 2 , 3 ) is defined as

-13-

ADDED NOTES CN4


dev dev = m (111) = (1 m, 2 m, 3 m ) = (1 , dev , dev ) 2 3 m =
dev 1

dev 2 dev 3

1 + 2 + 3 3 2 1 = 1 ( 2 + 3 ) 3 2 2 1 = 2 (1 + 3 ) 3 2 2 1 = 3 (1 + 2 ) 3 2

Although we dont have to use the following facts (which are derived previously), these will save considerable algebraic manipulation in the analysis. Recall that dev is the projection of onto the -plane, and the Von Mises yield criterion eff = o can be stated in terms of the radius r of the Von Mises circle,
2 2 dev = r = o = eff 3 3

The equation for the Von Mises yield surface in stress space is
f (1, 2 , 3 ) = eff = 1 2

( 1 2 )

+ ( 1 3 ) + ( 2 3 ) = 0
2 2

o is a constant for a given yield surface and it is the tensile yield stress of the material in its current microstructural state. This is a material state parameter that enters into the yield criteria for engineering metals and alloys. The partial derivatives required for f are,
3 1 dev f eff 1 1 = = 1 2 ( 2 + 3 ) = 2 1 1 1 eff eff 3 1 dev f 1 1 = eff = 2 2 (1 + 3 ) = 2 2 2 2 eff eff 3 1 dev f 1 1 = eff = 3 2 (1 + 2 ) = 2 2 3 3 eff eff

Hence,

3 1 dev 3 1 dev f = = 2 eff 2 o 3 1 dev 3 1 3 f = = r= 2 o 2 o 2

The unit normal vector to the Von Mises yield surface is

-14-

ADDED NOTES CN4

f 3 1 dev 3 1 dev n= = = 2 o 2 eff f


This verifies that the normal to the Von Mises yield surface is coaxial with the deviatoric component of the vector stress , as we showed using the -plane geometry. The Normality Rule is used to determine the direction of the strain increment due to . This requires that the vector d = (d1,d2 ,d3 ) be along the normal to the yield surface, hence,
3 1 dev d = n = 2 eff d = n =

For a fixed microstructural state of the material, ie., a yield surface corresponding to a fixed o value, the scaling parameter can be evaluated using any stress loading path. For a tensile test with 3 as the loading axis, d3 = dpl and d1 = d2 = - d3 /2. This gives,
2 = d = d1 + d2 + d2 = 2 3 3 dpl 2

Substituting for produces the final form of the Levy Mises equations,
3 dpl dev d = 2 eff 3 dpl dev dpl 1 d1 = 1 = 1 2 ( 2 + 3 ) 2 eff eff 3 dpl dev dpl 1 d2 = 2 = 2 2 (1 + 3 ) 2 eff eff d3 = 3 dpl dev dpl 1 3 = 3 2 (1 + 2 ) 2 eff eff

Application of the Levy Mises equation requires introduction of a plastic flow curve, eff = o = o (pl ) . Inverting this relation allows the plastic strain increment dpl to be expressed

in terms of the stresses 1, 2 and 3. The incremental strains can then be integrated over a stressloading load path to obtain total strains 1, 2 and 3. For radial loading paths, the integration is trivial and the Hencky total plastic strain equations are easily obtained. The development here, was done using principal stresses and principal strains. This allows a simple visualization of the stresses and strains using 3D stress-space vectors along with relatively straightforward vector analysis. However, if the underlying principal axes for stresses and strains they produce are not the same, it does not make physical sense to plot vector stresses and vector strains in the same 3D stress-space coordinates. The principal axes for stress and
-15-

ADDED NOTES CN4 strain need not be the same for some formulations of the yield criteria. In this case the full tensor representations for stresses ij and the corresponding strain increments dij must be used. The yield surface, f(ij) = c, now represents a 5D hyper-surface in 6D stress space (ij and dij are symmetric tensors so six independent stress-space axes are needed to plot tensor components). The normal to the yield surface is given by the generalized vector gradient,

f f = ij
The normal to the yield surface is in the direction of f . The Normality Rule still applies and requires that the strain increment dij must be along the normal direction, ie., along the direction specified by f . Putting this together, and using a tensile test to evaluate the scaling parameter, the full tensor plastic strain equations for the specified yield criteria can be derived.

As an example, the full tensor equation for the Von Mises yield surface is eff = f(ij) = o, where
eff = 1 2

11

2 2 22 ) + (11 33 ) + ( 22 33 ) + 3(12 + 2 + 13 + 2 + 2 + 2 ) 21 31 23 32 2 2 2

The deviatoric component of the tensor stress ij is defined as

dev = ij ij

11 + 22 + 33 ij = ij m ij 3

Differentiating eff and combining terms gives the vector gradient in the form

f 3 1 dev = eff = ij ij ij 2 eff


After some manipulations, you can show that
dev = dev dev = tr [ dev ][ dev ] = ij ij ij f 3 1 dev 3 = ij = ij 2 eff 2

2 eff 3

The Normality Rule requires that dij be along the normal to the yield surface, hence,

-16-

ADDED NOTES CN4

dij = n =

f ij f ij

3 1 dev ij 2 eff

Note that the tensor Levy Mises equations have the same form as the vector Levy Mises equations. A tensile test is used to evaluate the scaling parameter . In this case d33 = dpl, d11 = d22 = - d33/2 and d13 = d23 = d12 = 0.
dij = dijdij = tr ([d][d]) = 3 dpl = 2

Substituting for , the tensor form of the tensor Levy Mises equations becomes

dij =

3 dpl dev 3 dpl ij = [ ij m ij] 2 eff 2 eff

Yield Criterion for Slip: Schmids Law ( = c) Plastic deformation due to slip is an interesting case to consider. For a given stress state ij, the principal axes for the strain increments dij due to different slip systems activated by ij, will not be the same, nor will they in general be equal to the principal axes for the stress. Heres a case where we need to use tensor components if we want to represent the stresses and strains on the same set of stress axes (a 6D stress space). You will see during the following steps that a Normality Rule is derived for plastic strain due to slip. This is in fact one of the arguments that is used to justify the Normality Rule. It is valid for slip, and slip is the basis of plastic strains therefore the Normality Rule should be a good assumption for all engineering plasticity analysis (this is not a bulletproof justification, but it is satisfying that the Normality Rule holds apriori for slip). The plastic strain increment dij due to the shear strain increment d on any slip system is given by

dij =

d d (nisj + njsi ) = 2 N ij 2

n and s are unit vectors defining the slip plane normal and slip directions, respectively. Slip (plastic strain) occurs when the resolved shear stress on the slip plane, in the slip direction, reaches a critical value . This is Schmids Law, = c. Schmids Law is a yield criterion for slip, and c is a yield stress that plays the same role as o does in the Von Mises and Tresca criterions. Using the transformations based on the stress (traction) vector S, the resolved shear stress due ij is

-17-

ADDED NOTES CN4

= ijn isj 1 ( ijnisj + jinisj) 2 1 1 = ( ijn isj + ijn jsi ) = ij (n isj + n jsi ) 2 2 =

Summation convention is used here. The second line follows since ji = ij (symmetric tensor). The third line follows since i and j are dummy (repeated) subscripts and we can interchange the labels i and j in the jinisj term. The yield surface for = c (in 6D stress space) is then,

1 = f( ij) = ij (n isj + n jsi ) = c 2 The normal direction to the yield surface is given by the generalized (6D) vector gradient,

f 1 1 f = = (n isj + n jsi ) = N ij ij 2 2
Nij is the (6D) vector normal to the yield surface. As you can see from the dij equation for slip systems, the strain increment must always be coaxial with this direction. In other words, a Normality Rule is derived from the basic stress and strain relations for slip systems. In order to use the d ij equations for slip as plasticity equations, it would be necessary to introduce a plastic flow curve = (ji). To get total strains, you would have to add the strains from all of the active slip systems and then integrate the resultant over a load path. One difficulty to be faced is that you must develop a physically acceptable basis for determining the shear increment d that occurs for each active slip system in the average crystal grain. The sum of the strains due to all slip systems must of course be the same as the overall macroscopic strain in the polycrystalline solid. We will not consider this problem further in these notes. Summary For most engineering plasticity analyses, we use an approach based on the Von Mises or Tresca yield criterions. For the record, we note that a third yield criterion known as the maximum reduced stress criterion can also be used. This criterion is expressed as

2 dev max{ 1 m , 2 m , 3 m } = max{ 1 , dev , dev } = o 2 3 3


It can be shown that the Tresca and the reduced stress criterion provide limits between which the acceptable yield criteria for volume-conserving plastic deformation must lie. These are not wide limits. As an exercise, try plotting the reduced stress criterion yield surface along with the Von Mises and Tresca yield surfaces on a -plane projection.

-18-

ADDED NOTES CN4 PRESSURE VESSELS4: Spherical The principal axes 1 and 2 are defined by any two orthogonal great circles around the sphere. The principal axis 3 is normal to the sphere wall. The radius is r, the wall thickness is b and the subscripts o denote initial values. The principal stress 3 is not identically zero, but we ignore it because it is the order of the pressure p which small compared to the other stresses that are magnified by the ratio r/b (r/b >>1 for a thin-wall vessel).

pr 3 0 2b r b 1 = 2 = ln 3 = ro bo 1 = 2 = V = 4r 2 b = Vo = 4ro2 b o

Cylindrical The principal axes 1 and 2 are defined by the cylinder axis and a circle around the cylinder circumference, respectively. The principal axis 3 is normal to the cylinder wall. 3 can again be ignored. The radius is r, the wall thickness is b, the length is L and the subscripts o denote initial values. Two cases are possible. For closed-end vessels there is an axial stress 2 due to the pressure on the end faces. For open-end vessels (pressure is delivered by flexible tubes) there is no axial stress because there is no pressure acting against the end faces to generate axial loads. For both cases, the circumferential stress 2 is usually called the hoop stress.

pr pr 2 = 3 0 2b b L r b 1 = ln 2 = ln 3 = ln Lo ro bo V = 2rLb = Vo = 2ro L o b o 1 =

pr 3 0 b L r b 1 = ln 2 = ln 3 = ln Lo ro bo V = 2rLb = Vo = 2ro L o b o 1 = 0 2 =

For a derivation of the formulas, consult a solid mechanics textbook. For example, Mechanics of Materials, D. Roylance, Wiley (1996) ISBN 0-471-59399-0

-19-

You might also like