Beruflich Dokumente
Kultur Dokumente
412 Vol. 68, MAY 2001 Copyright 2001 by ASME Transactions of the ASME
material points can not leave, due to the axisymmetry. By virtue of these particular conditions, the directional derivative
The starting point is the kinematical equation derived in the of curl v along may be put in the form
Appendix for the time derivative of the curvature of the material
lines of a continuous medium in motion. In this equation, the time 1 curl v e
b e grad t (3)
rate of the curvature at a given point of a material line is ex- 2 s s s
pressed as a function of the strain rate tensor field, the velocity
field, and the curvature and the torsion of the material line at and Eq. 1 becomes
the given point 1 tn
grad t t . (4)
1 tn 1 1 curl v 2 s
tb b t . (1)
2 s 2 2 s Furthermore, the gradient of may be expressed as a function of
The amounts involved in Eq. 1 are the components t , tn , and the strain rate field by applying the formulas for this operator in
tb of the strain rate tensor normal and shear strain rates in the cylindrical coordinates and then eliminating the derivatives of the
tangent t, principal normal n and binormal b directions of the velocity components through the expressions which relate the
line, the directional derivatives of these components in the tangent strain rates to the velocity field, also in cylindrical coordinates.
direction of the line ( / s), and the curl of the velocity field v. Denoting the radial and axial unit vectors of the cylindrical coor-
For a material line coplanar with the axis of revolution in an dinates by er and ez , the result is
axisymmetrical motion of the continuous medium Fig. 1, Eq. 1
can be simplified. First, since the material line remains plane, its
torsion vanishes. Further, if the axis of revolution is chosen as
grad z 2 r r
r 1 rz
e
2 z
r z
1 rz z
e . (5)
the z-axis of a cylindrical coordinate system r, ,z and the princi- On applying Eq. 4 to the deformed notch profile Fig. 2, the
pal normal is the tangent rotated 90 degrees counterclockwise shear strain rate tn remains null along , and as a consequence
Fig. 1, the binormal b coincides with the negative of the unit its derivative in the tangent direction vanishes. This is due to the
base vector e . Finally, the axisymmetry of the velocity field v absence of shear stresses acting on the notch profile, which im-
A form of this equation involving still fewer variables can be
obtained. According to the axisymmetry, a principal direction of R 0 1 R 0 R 0 a 0 1
2 . (19)
the strain rate tensor is e , and the other two, eI and eII , are R a0 a
contained in the axial plane. By applying the Mohrs circle pro- Therefore, the notch root radius R is determined by three values
cedure, the strain rate rz can be expressed as a function of the dependent on the displacement field of the minimum cross section
strain rates I , II in the directions eI and eII and of the angle of the notch: the radius a of the section and the values and
between the radial direction and eI , so that taken at the notch root by the radial stretch ratio and its derivative
rz I II sin 2 (8) along the r-axis. Equation 19 is a kinematical relationship that
allows the notch root radius to be found from the single function
rz I II of the distance to the center of the cross section, to which that
sen 2 2 I II cos 2 . (9) displacement field reduces.
z z z
Since at the root of the notch the axial and radial directions are
principal directions of stress and strain rate, the conditions to be 3 Application to Bridgmans Solution
particularized in Eq. 4 for point A are
The first specific hypothesis of Bridgmans analysis determines
the displacement field at the minimum cross section of the tensile
c I z II r . neck. Therefore, when Bridgmans solution is applied to a tensile
2 z
notched specimen, the result of particularizing Eq. 19 for this
Thus, for the root of the notch, Eq. 9 and subsequently Eq. 7 displacement field is being assumed as valid for the specimen
become and it provides a means to check such application of Bridgmans
solution.
rz According to experimental evidence, Bridgman assumed the
2 z r c (10)
z axial strain rate z to be uniform across the minimum cross sec-
tion, which leads to the uniformity of the overall strain rate and
z strain states, including the standard equivalent plastic strain p .
c r c . (11)
r Indeed, if z is not dependent on r, from the conditions of axi-
symmetry and incompressibility it follows that
An undeformed elemental length dr 0 lying on the r-axis trans-
forms into dr after deformation and remains on the r-axis since 1 1 a
this axis moves along itself due to the axial and mirror symmetry. r p (20)
2 z 2 a
Let be the value of the ratio dr/dr 0 at point A. Hence, at this
point 1 1 a
r p ln (21)
2 z 2 a0
z 1 z 1 z
r (12)
r r0 r0 r a dr a
e r0 (22)
and Eq. 11 transforms into r0 a0 dr 0 a 0
c
c
1 z
r0
(13)
a 0
a
a0
d
dr 0 a0
0. (23)
which, after some rearrangement, can be integrated to yield Particularizing these values of and in Eq. 19, it simplifies
to
z
c ctec 0 (14) R a
r0 , (24)
R0 a0
where the constant has been determined from the undeformed
specimen, for which the curvature of the notch is c 0 , is equal to i.e., the notch root radius and the minimum specimen radius vary
unity and z vanishes. In terms of the notch root radius R and its proportionally during the deformation process. This must be sat-
initial value R 0 c1/R, c 0 1/R 0 , Fig. 2: isfied by the tensile notched specimens for which Bridgmans so-
lution is applicable.
R0 1 R0 z The second specific hypothesis of Bridgmans analysis was as-
(15) sumed in the development of the formula that gives the tensile
R r0
load as a function of the neck geometry and the stress-strain curve
If the elastic compressibility is neglected, incompressibility fol- of the material. The development of the formula is next summa-
lows, and for the points of the r-axis it takes the form rized as an introduction to the hypothesis.
2r 2 1
2R0
a0
ln 1
a0
2R0
Davidenkov r a0
I and II be the principal stresses contained in the axial plane, s I and 1
the distance along the isostatic line tangent to I , R I the curvature Spiridonova R 4R0
radius along the isostatic line tangent to II , and the angle
between the direction of I and the r-axis. The general equation is
Eisenberg
and
Yen
aR a2
r
r
r
1
2
1
R0
2a0
ln 1
a0
R0
I I II I
cos 0, (25)
sI R II r
and for the r-axis where
d r r z r
a0 1 x 2 dx
0, (26) F 1 . (32)
dr r R0 0 a0
f x,
being the curvature radius of the isostatic lines normal to the R0
r-axis at the points of intersection with it. Furthermore, the equal-
With the exception of the factor F(a 0 /R 0 ) henceforth, notch
ity of the stresses r and arises from that of the corresponding
factor this relation between the tensile load P and the radius a is
strain rates, and as a consequence the difference z r is the
that of an unnotched round bar of initial radius a 0 made from the
equivalent stress , which remains uniform in all the section due
same material as the circumferentially notched round specimen.
to isotropic hardening such a property implies a biunivocal rela-
Therefore, these two relations can be predicted one from another
tion H( p ) between this stress and the equivalent plastic strain
by dividing or multiplying the tensile load by the notch factor and
p , so that if one of them is uniform so is the other. Thus Eq.
considering the radius a as that of an unnotched bar or as that of
26 becomes
the minimum cross section of a notched one. In particular, the
d z
dr
0
with H 2 ln a0
a (27)
maximum loads of these two tensile bars are the same except for
this factor, whereupon the load of plastic collapse P m of a circum-
ferentially notched round specimen should be given by
and can be integrated with the boundary condition z at r
a since r 0 at this point to give a0
P m a 20 R m F , (33)
a
R0
dr
z 1 . (28) R m being the tensile strength of the material. Another important
r consequence of Eq. 31 is the possibility of determining the
The tensile load P sustained by the specimen is the resultant force stress-strain curve H( p ) by testing notched round specimens
of the axial stresses z acting on any cross section. Then, P can be and recording the tensile load as a function P P(a) of the radius
expressed as a function of the variables introduced up to now by of the minimum cross section. The function H() would be
integrating the value given by Eq. 28 over the minimum cross P a 0 e x/2
section H x (34)
a0
P 0
a
z 2 xdx
0
a
1
a dr
2 xdx
a 20 e x F
R0
and would be valid for a larger range of strain than that provided
a
2
a 2 1 2
0
a 1
0
r
xdx dr
by a standard tensile test, since the plastic instability of a notched
specimen would not mean the end of the test for the purpose of
determining the stress-strain curve because it does not produce a
geometrical configuration change as radical as the necking of a
1 a r2 smooth specimen.
a 2 1 dr
a2 0 Table 1 shows different functions giving the curvature radius of
the isostatic lines along the r axis according to Eq 30, as well as
1 2
a r r the corresponding notch factors as calculated from Eq. 32.
a 2 1 d . (29)
0 a a
4 Finite Element Solution
The second hypothesis formulated by Bridgman states the
As previously stated, Eqs. 24, 31, and 33 provide support
variation of the curvature radius along the r-axis. Apart
to assess the validity of Bridgmans solution for tensile blunt
from Bridgman, other researchers 3,6 have proposed or
notched specimens by checking them with numerical or experi-
suggested alternative hypotheses, but they all fall under the gen-
mental results. The numerical results were preferred to the experi-
eral formulation
mental ones due to the difficulties of performing the measure-
a f
r R
,
a a
. (30)
ments involved in Eqs. 24 and 31, since there are no simple
methods to measure radii of curvature in a contour or tensile
stress-strain curves beyond the tensile strength.
As a consequence, the integral of the right-hand side of Eq. 29 A large number of finite element solutions have been computed
depends only on the ratio a/R, which according to relation Eq. for elastic-plastic circumferentially notched tensile round bars
24 holds constant during the deformation; so that the expression 13,14. In general the results approach Bridgmans solution as
of the tensile load transforms into the sharpness of the notch profile decreases, but the reported de-
tails do not allow Eq. 24 to be checked, so a finite element
a0 a0 a0
P a 2 F a 2 H 2 ln F (31) solution was specifically obtained for this purpose. The length of
R0 a R0 the tensile bar numerically modeled, the radius of the gross
Fig. 3 Notch profile and finite element mesh used for the nu- Fig. 5 Deformed notch profiles at different load levels P m is
merical solution the maximum load
with plastic unloading until practically vanishing. With regard to Eq. 19 with Bridgmans solution, but whereas the first is based
Eq. 24, the notch profile in the neighborhood of the notch root only on the linear displacement distribution assumed at the mini-
has been plotted at different load levels in Fig. 5 from the dis- mum cross section of the bar, the second comes also from the
placements of four nodes lying on it. The corresponding values of curvatures assumed for the isostatic lines along the r-axis. So the
the notch root radius R were determined as that of the circumfer- errors that the deviations from these two assumptions produce on
ence symmetric to the r-axis which passes through the node on the the notch root radius a local effect and on the tensile load a
notch root and the nearest on the notch profile. global effect will be assessed.
The resulting R/R 0 values are plotted in Fig. 6 as a function of For large plastic deformation, the linear displacement distribu-
the minimum area expressed by the ratio a/a 0 . The exact solution tion is a good approximation, as shown by the finite element cal-
for incompressible material provided by Eq. 19 allows the de- culated displacement field Fig. 4. According to the absence of
viation from Eq. 24 to be explained. For small strain, the hoop deviations, in this regime no disagreement should be expected
and radial stretches a/a 0 and hardly differ from unity, so Eq. between the theoretical prediction of the notch root radius given
19 predicts no significant difference between the radii R and R 0 by Eq. 24 and the finite element results; indeed no significant
unless a strain gradient giving rise to a value of not much disparity is detected, since the slight scatter of Fig. 6 is attribut-
below 1/R 0 be produced at the notch root. As the displacements able to the numerical errors arising from the finite discretization of
plotted in Fig. 4 indicate, this requires a notch profile very much the bar. On the contrary, in the previous elastic and elastic-plastic
sharper than the finite element modeled one, so that the change of regimes roughly up to plastic instability the displacements devi-
curvature cannot precede that of the overall dimensions. Since ate from the linear distribution near the notch root Fig. 4. This
large deformations initiate at the early stages of plastic instability, has no effect on the notch root radius for small strains, as already
Eq. 24 is verified until then simply because there is no geometry stated, and only at the transition to large deformation the begin-
change, even though the radial displacements do not satisfy en- ning of plastic instability, do the data in Fig. 6 indicate some
tirely Eq. 22 Fig. 4. When the large geometry changes do disagreement between the theoretical results and the numerical
occur, the notch profile is deformed according to Eq. 22 as ones.
shown in Fig. 6 by the points clustered around the straight line As far as the tensile load is concerned, Fig. 7 shows good
representing this equation. This is in agreement with the trend of agreement between the values given by the finite element model-
the radial displacements to fulfill Eq. 22 as plastic unloading ing and Eq. 31 in the range covering general and extended yield-
increases Fig. 4. The maximum deviation from Eq. 24 takes ing up to plastic instability. The two results differ from maximum
place just before plastic instability, when strains are already large, load, i.e., when the notch root radius largely changes. The tensile
but the radial displacements still differ from Eq. 22 Fig. 4. load being affected, the discrepancy is a global effect and cannot
Finally, the tensile load obtained by the finite element modeling be due the displacement field at the minimum cross section, since
is plotted in Fig. 7 against the logarithmic strain 2 ln(a0 /a). Equa- in this range these displacements deviate so little from the linear
tion 31 is also plotted in Fig. 7 with Bridgmans notch factor of distribution as to produce no local effect Eq. 24 is satisfied. So,
Table 1. The unit load used in the plots is P 0 a 20 0 , namely, it can only be attributed to the inaccuracy of the curvatures
the area of the minimum cross section multiplied by the stress adopted for the isostatic lines. The alternatives to Bridgmans cur-
constant 0 of the strain hardening curve. vatures given in Table 1 would not improve the agreement, be-
cause they would only modify the notch factor and would produce
a displacement of the curve in Fig. 7 parallel to the ordinate axis.
5 Discussion In fact, the curvatures derived by Bridgman seem to be an excel-
The adequacy of Bridgmans solution for axisymmetrical lent solution of the type defined by Eq. 29, since this can be
notched tensile bars will be discussed by comparing the finite derived not only from the circumferences assumed by Bridgman
element solution of Section 4 with the main theoretical predictions to be the family of isostatic lines to which the axis r belongs, but
derived in Section 3, which give the notch root radius Eq. 24 also from other families of curves as ellipses or parabolas. A
and the tensile load Eq. 31 as a function of the specimen mini- plausible explanation of the inaccuracy arises when Eq. 24 is
mum diameter. The two results are a consequence of combining substituted into Eq. 29 and it becomes apparent that the assumed
6 Concluding Remarks
A theoretical development was carried out aimed at deriving a
relationship between the notch root radius of an axisymmetrical
notched tensile bar and the displacement field at the minimum
cross section for large geometry changes of the notch profile and
for elastic-plastic incompressible material. Bridgmans displace-
ment and stress solution for a tensile neck was examined in com-
bination with that relationship with regard to their application to
axisymmetrical blunt notched bars under tension, and this allowed
the entire load-minimum diameter curve of the bar to be pre-
dicted. A finite element solution of a particular case was com- Fig. 8 Vectors concerning the motion of a material line
puted for comparison, and different steps of the analytical devel-
opment were checked, the following conclusions being drawn as
to the effects produced by the inaccuracy of Bridgmans solution can be exchanged. Let q be the vector amount to be differentiated
for the notched bar on the predicted notch root radius and tensile and t the normal strain rate in the direction tangent to ; given
load: no significant error of the former was observed over the that e t is the ratio ds/ds 0 of the deformed and undeformed length
entire range of plastic deformation small and large, but signifi- of an elemental arc of the material line, such a rule would be
cant differences were found in the latter after plastic instability.
Acknowledgments
q
s e s0
t
1 q 1 q t q
t
e s0
t
e s0
The aim of this section is to derive a kinematical equation re-
lating the curvature of a deformed material line to the velocity t t t t
n n n n t n t (A6)
field in a body which is being subjected to deformation. The equa- s s s s
tion is obtained for the general case and particularized for that of
axisymmetry in Section 2 of the main text. A further development of this expression requires an explicit
Let be the curve which represents the deformed material line form to be found of the time derivative of the tangent vector. This
at present time. The position vector of a material point is r, v is can be done by applying Eq. A5 to the position vector r, decom-
the velocity field, s is the distance along , and are the cur- posing subsequently the velocity gradient into the sum of its sym-
vature and the torsion of , and t, n, and b are the vectors of an metric the strain rate tensor and skew part which is equivalent
orthonormal basis in the tangent, principal normal, and binormal to the vector product with 1/2 curl v as first factor, and finally
directions of Fig. 8. All these quantities and their directional introducing the shear strain rates tn and tb for the tangent, prin-
derivatives along are related by the Frenet equations: cipal normal, and binormal directions. All this yields
r
s
t (A1) t
r r
s s
r v
t t t grad v t t t
s s
t 1 1 1 1
n (A2) t curl v t t t tn n tb b curl v t.
s 2 2 2 2
(A7)
n
t b (A3) The differentiation of this expression and the subsequent use of
s
the Frenet equations provide the directional derivative of the time
b derivative of the tangent vector. Then, its projection onto the nor-
n. (A4) mal vector n can be found and substituted into Eq A6, which
s
finally gives:
In order to introduce the velocity field, a rule is required by
which the order of differentiation between the directional deriva- 1 tn 1 1 curl v
tb b t . (A8)
tive and the material time one denoted by a superimposed dot 2 s 2 2 s