International Journal of Pressure Vessels and Piping 77 (2000) 751±759

www.elsevier.com/locate/ijpvp

Plastic-deformation analysis in tube bending

N.C. Tang*
Quad Engineering, Toronto, Ontario, Canada
Received 29 September 1999; accepted 29 September 2000

Abstract
Plastic-deformation theory is employed to investigate the plastic deformation in pipe and tube bending. The major contribution of this
paper is that it provides solutions to seven common tube-bending questions. In this paper, some practical formulae are developed to explain
the phenomena in tube bending and their magnitudes are also derived. These are: (1) stresses in the bend; (2) wall thickness change; (3)
shrinking rate at the tube section; (4) deviation of neutral axis; (5) feed preparation length of the bend; (6) bending moment, and (7) ¯attening.
An experimental sample was also tested to illustrate that the results of the formulae are very similar to the experimental results. q 2001
Elsevier Science Ltd. All rights reserved.
Keywords: Plastic deformation; Tube bending; Elastic±plastic analysis

1. Introduction mostly coincident with those of the original equations. See

Figs. 3, 6, 8 and 9. The ®nal results provided a good agree-
Over the last decade, several tube-bending phenomena, ment with experimental values. It is successfully demon-
such as stress, neutral axis deviation, ¯attening and bending strated that use of plastic-deformation theory is the unique
moment, have been widely researched [1±4]. However, way to explain the entire tube-bending question at once.
most of the studies are based on only a single topic. To
date, no research has been done on all the phenomena
comprising tube bending, even though these are associated 2. Longitudinal and circumferential stresses
with each other and occur simultaneously.
Bending a straight tube to form an elbow is a process of Under an external bending moment, a longitudinal inter-
metal plastic ¯ow. All of the dimension in the elbow nal force dpx is induced on both sides of the elements as
changes as the plastic ¯ow rules; this is consistent with shown in Figs. 1 and 2
the plastic-deformation theory. Thus, plastic-deformation dpx ˆ s x r daT …1†
theory is the best theory to assess tube-bending topics.
The derivation of relationship between longitudinal and where, r is the average radius of the tube cross-section, s x,
circumferential stress (Eq. (8)) provides a ®rst and impor- the longitudinal stress, and T, the wall thickness of the tube.
tant step. The stress values (Eqs. (10)±(14)) can then be Their centripetal composite force dv is
solved using the maximum-shear-stress theory. Conse-
q
quently, the relationship of stress and strain is illustrated dv ˆ 2Dp x sin
in Eq. (15) via the deformation theory of plastic ¯ow. As 2
a result, these six equations can form a whole composite where, q is the bending angle (see Figs. 1 and 2). For the
group of kernel formulae, and can be used to solve the small angles here take
problems in tube bending.
Some equations in this paper have to be simpli®ed in q
2 sin < sin q < q
order to derive the solutions, which makes one easier to 2
understand the rule of dimension change and provides Therefore
convenience of use. The curves of simpli®ed equations are
q
dv ˆ 2Dpx sin ˆ s x rDaT q …2†
2
* Address for correspondence: 32, Ferne-Rachel Road, Thornhill,
Ontario, Canada L4J 6J4. The circumferential force dpc is produced by dv, which
0308-0161/00/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0308-016 1(00)00061-2
752 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759

Eliminating q T, we get
ds c …R 1 r cos a† ˆ s x r sin a da
Therefore, circumferential stress can be expressed as
sin a da
ds c ˆ s x r …6†
R 1 r cos a
Integrating the above, the circumferential stress can be
obtained as
Za sin a da R 1 r cos a
sc ˆ sxr ˆ s x ln
0 R 1 r cos a R1r
Introducing a geometry parameter of the tube: k ˆ R=2r; the
range of ªkº in this paper should be 1.4±30
So s c can be written as
2kr 1 r cos a 2k 1 cos a
s c ˆ s x ln ˆ s x ln …7†
2kr 1 r 2k 1 1
To make subsequent derivation easier, the stress function
can be simpli®ed as (see Fig. 3)
2k 1 cos a 1 2 cos a
Fig. 1. Stresses shown in 3D. ln <2
2k 1 1 2k 1 1
Therefore, the relationship between longitudinal and
acts along the circumferential direction. dpc is given by circumferential stress is expressed as
dpc ˆ dv sin a ˆ s x r daT q sin a …3† 1 2 cos a
s c ˆ 2s x : …8†
2k 1 1
It can be expressed in terms of the circumferential stress
2.1. The stress in the outer semi-circle
dpc ˆ ds c Ac ˆ ds c …R 1 y†qT …5†
In the outer semi-circle, the longitudinal stress s x is
where, R is the bending radius of the tube, y the vertical tensile, and the circumferential stress s c is compressive.
distance from the tube center to the element, and s c, the Since the wall thickness is much smaller than the radius
circumferential stress. of the tube, the radial stress s r can be neglected. By the
Since y ˆ r cos a; we have maximum-shear-stress theory, the relation between the
dp c ˆ ds c …R 1 r cos a†qT …5† two stresses is given by
us x u 1 us c u ˆ s s …9†
Let Eq. (3) equal Eq. (5)
where s s is the yield strength of the tube material. Substi-
ds c …R 1 r cos a†qT ˆ s x r daT q sin a tuting Eq. (8) into Eq. (9), the longitudinal stress is

Fig. 2. Stresses shown in two views.

 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759 753

Fig. 3. Comparison of derived and simpli®ed of stress function; Eqs. (7) and (8).

expressed as expressed as
2k 1 1 2k 1 1
sx ˆ ss …10† s x ˆ 2s s …13†
2k 1 2 2 cos a 2k 1 cos a
and substituting Eq. (8) into Eq. (10), the circumferential
stress is expressed as and substituting Eq. (8) into Eq. (13), the circumferential
stress is expressed as
1 2 cos a
s c ˆ 2s s : …11†
2k 1 2 2 cos a 1 2 cos a
s c ˆ 2s s …14†
2k 1 cos a
2.2. The stress in the inner semi-circle
The stress distribution in the outer and inner semi-circles is
In the inner semi-circle, both s c and s x are compressive. shown in Fig. 4.
By the maximum-shear-stress theory, the relation between
them is given by
us x u 2 us c u ˆ s s …12† 3. Wall thickness change

Substituting Eq. (8) into Eq. (12), the longitudinal stress is From the deformation theory of plastic ¯ow, strains and

Fig. 4. Longitudinal and circumferential stress in the tube section.

754 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759

Fig. 5. Wall thickness change.

stresses can be expressed as wall thickness change for the outer semi-circle can be
1 obtained as
dx ˆ …s 2 m…s c 1 s r ††
D x 2k 1 cos a cos a
dro ˆ 2 …17†
1 4k 1 3 2 cos a 2k
dc ˆ …s c 2 m…s r 1 s x †† …15†
D Hence, the wall thickness in the outer semi-circle is
 
1 2k 1 cos a cos a
dr ˆ …s 2 m…s x 1 s c †† to ˆ …1 2 dro †T ˆ 1 2 T …18†
D r 4k 1 3 2 cos a 2k
where d ˆ Dl=l is the deformation of unit length, D the where T is the original thickness of the tube wall. The smal-
plastic ¯ow modulus of the tube material, and m the Poisson lest thickness is located at the outermost point. Substituting
ratio (for plastic deformation m ˆ 0:5†: From Eq. (15), we a ˆ 0 into Eq. (18), it is expressed as
have  
2k 1 1
1 dr dx tom ˆ …1 2 dro †T ˆ 1 2 T: …19†
ˆ 2k…4k 1 2†
D sx 1 sc ˆ sc 1 sr
sr 2 sx 2
2 2
3.2. Inner thickening
2s 2 s x 2 s c
dr ˆ dx r
2s x 2 s c 2 s r Substituting Eqs. (13) and (14) into Eq. (16), the rate of
inner wall thickness change is obtained as
Since the wall thickness is smaller, compared to the tube
radius, the radial stress s r is negligible. Let s r ˆ 0 and 2k 1 2 2 cos a cos a
dri ˆ 2 …20†
y r cos a cos a 4k 1 1 1 cos a 2k
dx ˆ ˆ ˆ
R R 2k Therefore, the wall thickness in the inner semi-circle is
the rate of wall thickness change is  
2k 1 2 2 cos a cos a
2s x 2 s c cos a ti ˆ …1 1 dri †T ˆ 1 2 T …21†
dr ˆ : …16† 4k 1 1 1 cos a 2k
2s x 2 s c 2k
The largest thickness is located at the innermost point.
Substituting a ˆ 1808 into Eq. (21), it is expressed as
3.1. Outer thinning  
2k 1 3
tim …1 1 dri †T ˆ 1 1 T …22†
Substituting Eqs. (10) and (11) into Eq. (16), the rate of 3k2
 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759 755

Fig. 6. Comparison of derived and simpli®ed of diameter shrinking; Eqs. (29) and (30). Both curves are coincident.

Based on the above equations, the thicknesses of the inner Substituting Eqs. (13) and (14) into Eq. (24), we have
and outer semi-circles is shown in Fig. 5.  
22…1 2 cos a† 2k 1 1
1 s
cos a
 2k 1 cos a 2k 1 cos a s
4. Circumferential shrinking dci ˆ 2
2k 22…2k 1 1† 1 2 cos a
1 s
2k 1 cos a 2k 1 cos a s
The shrinking of the diameter is caused by the circumfer-
ential stress on the tube cross-section. From the deformation cos a 2k 2 1 1 2cos a
ˆ : …26†
theory of plastic ¯ow, the relationship between strains and 2k 4k 1 1 1 cos a
stresses, which also use Eq. (15), can be derived as follows.
From Eq. (15), take s x and s c, hence
4.1. Average shrinking rate
1 dx dc
ˆ
D sc 1 sr ˆ sx 1 sr The shrinking rates for outer and inner semi-circles are,
sx 2 sc 2
2 2 respectively, given by the following equations:
i.e. 2 Zp=2 cos a 2k 1 3 2 2cos a
dco ˆ da …27†
2s 2 s x 2 s r p 0 2k 4k 1 1 1 cos a
dc ˆ dx c …23†
2s x 2 s c 2 s r
2 Zp cos a 2k 2 1 1 2cos a
Since the wall thickness is small, s r can be neglected. Thus, dci ˆ da …28†
p p=2 2k 4k 1 1 1 cos a
2s c 2 s x
d c ˆ dx …24† Therefore, the average shrinking rate is
2s x 2 s c
For the outer semi-circle: dco 1 dci
dc ˆ …29†
y r cos a cos a 2
dxo ˆ ˆ ˆ
R 2kr 2k Eqs. (27)±(29) given above are too complicated to be used
Substituting Eqs. (10) and (11) into Eq. (24), we have in an engineering context. The following simpli®ed
  equation is therefore introduced (see Fig. 6 for a comparison
22…1 2 cos a† 2k 1 1 of d c with d n):
2 s
cos a 2k 1 2 2 cos a
 2k 1 2 2 cos a  s
dco ˆ 0:16
2k 2…2k 1 1† 1 2 cos a dn ˆ …30†
1 s k 1 0:8
2k 1 2 2 cos a 2k 1 2 2 cos a s
cos a 2k 1 3 2 2cos a The diameter ªdº and perimeter ªCº after bending are,
ˆ2 respectively, expressed by the following equations:
2k 4k 1 3 2 cos a
…25†  
0:16
ds ˆ …1 2 dn †d ˆ 1 2 d …31†
For the inner semi-circle: k 1 0:8
cos a
dxi ˆ 2 C ˆ …1 2 dn †pd: …32†
2k
756 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759

Fig. 7. Experimental sample.

5. Neutral axis deviation Here

Z908  2k 1 1

Since longitudinal stress in the outer half is smaller than
in the inner half (Eqs. (10) and (13)), and the outer wall is 0 2k 1 2 2 cos a
thinner than the inner wall for balance of the moment of the  
2k 1 cos a cos a
internal force, the neutral axis should move to its inner side  12 da ˆ No
as shown in Fig. 7. 4k 1 3 2 cos a 2k
The summation of all the differential forces Po and Pi Substituting Eqs. (13) and (21) into Eq. (35), we have
acting on the neutral axis must be in equilibrium, i.e.
Z1808  2k 1 1 
Po …0:636r 1 E† ˆ Pi …0:636r 2 E† or Pi ˆ s s rT
908 2k 1 cos a
Pi 2 Po …33†  
E ˆ 0:636r 2k 1 2 2 cos a cos a
Pi 1 Po  12 da
4k 1 1 1 cos a 2k
where, constant 0.636 is the distance from the center of ˆ s s rTNi …37†
gravity of the half-section to the x-axis; Po and Pi are the
summation of all differential force acting on the centroid of Here
the outer and inner parts of the walls, respectively.
Z1808  2k 1 1

2k 1 2 2 cos a cos a

Z908 Z908 12 da ˆ N i
Po ˆ s xo dAo ˆ s xo rto da …34† 908 2k 1 cos a 4k 1 1 1 cos a 2k
0 0

Then, Eq. (33) become

Z1808 Z1808
Pi ˆ s xi dAi ˆ s xi rti da …35† Ni 2 No
908 908 E ˆ 0:636r …38†
Ni 1 No
Substituting Eqs. (10) and (18) into Eq. (34), we have
As Eqs. (36)±(38) given above are too complicated for use
Z908  2k 1 1

Po ˆ s s rT in engineering, the following simpli®ed equation is intro-
0 2k 1 2 2 cos a duced (for a comparison of E with En, see Fig. 8):
 
2k 1 cos a cos a 0:42
 12 da ˆ s s rTNo …36† En ˆ r: …39†
4k 1 3 2 cos a 2k k
 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759 757

Fig. 8. Comparison of derived and simpli®ed neutral axis deviation; Eqs. (38) and (39). Both curves are coincident.

6. Feed preparation length and (36), i.e.

 
0:42
The feed preparation length of a bend is shorter than the M ˆ 4Po …0:636r 1 E† ˆ 4s s rTNo 0:636r 1 …41†
geometric centroid axis arc length (Fig. 7). Its in¯uence k
factors are: (1) The sector of constant deviation of neutral Section modulus W is given by
axis. Within this sector all sections are ®nally formed, calcu-
late the length of this sector, which is by the arc length of D4 2 d 4 D2 1 d 2
W ˆ 0:1 ˆ 0:8rT …42†
neutral axis. (2) Transitional sector, it is between the section D D
of constant deviation of neutral axis and the section of Substituting Eq. (42) into Eq. (41), the bending moment
straight tube. This transitional sector behavior of the becomes
cross-section, which from a straight tube gradually varied    
to the ®nal deformation dimensions. These dimensions 5rD 0:42
M ˆ ssW 0:636 1 No ˆ s s WI …43†
include wall thickness, diameter shrinking and neutral axis D2 1 d 2 k
deviation. Therefore, for calculating the length of this where
sector, only half of the deviation value is used. From Fig.    
7 the distribution of the neutral axis can be seen. 5rD 0:42
Iˆ 0:636 1 No …44†
The length of constant deviation sector is D2 1 d2 k

lc ˆ …R 2 E†…v 2 2b† A simpli®ed equation for J is now introduced. For a compar-

ison of J with I, see Fig. 9
The length of the transitional sector is
  0:42
E J ˆ 1:41 1 …45†
lt ˆ R 2 2b k
2
Substituting Eq. (45) into Eq. (43), the bending moment is
The preparation length of the whole bend is given by
 
l ˆ lc 1 lt ˆ Rv 2 E…v 1 b† …40† 0:42
M ˆ s s W 1:41 1 : …46†
k
where b is the angle of the transitional section.
From Fig. 10 it can be seen that
b < 30±358: 8. Flattening

The axial stress in the outer half of the tube section is

7. Bending moment given by Eq. (10) (Fig. 4). For convenience of realization,
let us assume s x ˆ s s ˆ constant: Returning to Eq. (2)
After determining the neutral axis deviation E, it shows (Figs. 1 and 2), the centripetal composite force Dv is
that the bending moment on either side of the neutral axis is given by
equal. The total internal bending moment, M, may be found q
by multiplying the outer quadrant by four, refer to Eqs. (33) Dv ˆ 2Dpx sin ˆ s x rDaT q
2
758 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759

Fig. 9. Comparison of derived and simpli®ed bending moment function; Eqs. (44) and (45). Both curves are coincident.

From Fig. 10 Da1 ˆ Da2 ˆ Dan therefore Dv1 ˆ Dv2 ˆ The section modulus of the ring cross-section is
Dvn ˆ constant: It is likely that for uniform load …Dv=Da†
acting on a half-ring bridge, the worst-case of ¯attening is jT 2 6z
zˆ or jˆ …49†
on the top part. Since ¯attening or ovalization is mostly 6 T2
determined by technology and mold design, there is no
Substituting Eqs. (48) and (49) into Eq. (47), the bending
need to investigate the ovalization deformation curves.
moment of circumferential direction at the top point is
For determining the ¯attening start condition, see Fig. 10
obtained as
MF ˆ …1 2 0:636†rV ˆ 0:364rV …47†
3:43r 2
MF ˆ zs s …50†
From Eq. (2), for a quarter sector we have RT
The critical point where ¯attening starts is when the ring
p
V ˆ ssr Tq bending reaches the elastic limit
2
MF ˆ zs s
From the left side of Fig. 10, take a thin ring of thickness ªj º
where j ˆ Rq: Then i.e.

p j 3:43r 2
V ˆ ssr T …48† ˆ1
2 R RT

Fig. 10. Internal force distribution of ¯attening.

 N.C. Tang / International Journal of Pressure Vessels and Piping 77 (2000) 751±759 759

Therefore, when 4. neutral axis deviation; Eq. (39)

2
3:43r 0:42
R# …51† Eˆ r ˆ 4:3 mm
T k
¯attening should occur without any special control.
5. the feed preparation length of the bend; Eq. (40)
Rv E…v 2 b†
9. Calculation for a sample lˆ 2 ˆ 141:4 2 10:9 ˆ 130:5 mm
57:3 57:3
Based on the above theoretical analysis, a realistic experi-
ment was done. The parameters of the experiment are as 6. bending moment; Eq. (46)
follows:  
0:42
M ˆ s s W 1:41 1 ˆ s s 1812 £ 1:7 kg mm
outside diameter of the tube D ˆ 32 mm; k
wall thickness of the tube T ˆ 3 mm;
bending radius R ˆ 45 mm; and
bending ratio k ˆ R=D ˆ 1:41: 7. ¯attening condition; Eq. (51)

3:43r 2
In the experiment, pure bending method was used, i.e. R# ˆ 240 mm:
non-mandrel and non-booster. Before bending, the outside T
surface of the straight tube is marked by grooves with
an equal spacing of 10 mm. The result after bending is 10. Conclusions
illustrated in Fig. 7. Flattening, buckling and fracturing
did not occur during bending and after. Even for such big The objective of this paper is to provide a prediction of
strain and thinness at the extrados of the elbow, for internal the tube-bending phenomena. The major contribution of this
high pressure testing, the failures occur in the straight paper is that it provides solutions to seven common tube-
portion and not at the elbow. This can be explained by bending questions. The signi®cant illustration can be seen
some associated factors, such as extrados shape and strain using the small bending radius sample. The analysis devel-
hardening. oped here is different from the elastic and elastic±plastic
From the above derivations, a set of useful formulae are analyses, which include Von Karman's higher bending
given below, which yield agreement with the realistic radius ratio assumption. These analyses could not achieve
experiment the strain up to 0.47, see the experimental sample in Fig. 7.
However, for tube bending, various sizes exist in the plastic-
1. the wall thickness at the outermost point; Eq. (19) deformation range. Therefore, the plastic-deformation
  theory employed in this paper is the useful methodology
2k 1 1 to explain tube-bending phenomena.
to ˆ 1 2 T ˆ 2:47 mm
2k…4k 1 2†

2. the wall thickness at the innermost point; Eq. (22) References

  [1] Mackenzie D, Boyle JT. A simple pipe bend element for piping ¯ex-
2k 1 3
ti ˆ 1 1 T ˆ 4:1 mm ibility analysis. Int J Press Ves Piping 1992;51:85±103.
8k2 [2] Munz D, Mattheck C. Cross-section ¯attening of pipe subjected to
bending. Int J Press Ves Piping 1982;10:421±9.
[3] Kitching R. Smooth and mitred pipe bends: the stress analysis of pres-
3. diameter shrinking; Eq. (31) sure vessels and pressure vessel components. International Series of
  Monographs in Mechanical Engineering, vol. 3, chap. 7, 1970. p. 323±
0:16 33.
ds ˆ 1 2 d ˆ 26 £ 0:9276 ˆ 24:12 mm [4] Von Karman T. Uber die Formanderung dunnwandiger Rohre, insbe-
k 1 0:8
sonders federnder Ausleichrohre. Z Vereines Deutsch Ingen
1911;55:1889±95.