Int. J. Pres. Ves.

& Piping 46 (1991) 95-111

Stress Analysis of an Autofrettaged Thick-Walled Pressure Vessel Containing an External Groove

S. K. K o h & R. I. S t e p h e n s Mechanical Engineering Department, The University of Iowa, Iowa City, Iowa 52242, USA (Received 31 May 1990; revised version received 1 September 1990; accepted 17 September 1990)

ABSTRACT Stress analysis of an autofrettaged thick-walled pressure vessel containing an external groove and subjected to internal pressure is described. From the equivalence of thermal and autofrettage residual stress fields, the autofrettage residual stress distributions of the external grooved thick-walled pressure vessel were simulated using an equivalent thermal loading. Thermal stresses due to the simulated thermal loadings for various degrees of autofrettage overstrain level were computed using finite element methods. Very high stress concentration factors due to internal pressure and autofrettage loadings (3.9-4-65) were obtained at the external groove root that contained a sharp root radius. This sharp root radius location was considered as the most vulnerable location for fatigue crack formation and growth. Experimental measurement of residual stresses for a fully autofrettaged thick-walled pressure vessel using an equivalent saw cut method resulted in very close agreement with the theoretical autofrettage residual stress distributions. The stress analysis results indicate that lower autofrettage overstrain and a groove geometry change is desirable for enhanced durability.

INTRODUCTION If a sufficiently high internal pressure is applied to a cylindrical thick-walled pressure vessel, plastic deformation will begin at the inside diameter and extend through the cylinder wall as the pressure is increased. D u e to nonuniform elastic recovery after release of the 95 Int. J. Pres. Ves. & Piping 0308-0161/91/$03.50 1991 Elsevier Science Publishers Ltd, England. Printed in Northern Ireland

96

S. K. Koh, R. L Stephens

internal pressure, the outer portion of the cylinder attempts to resume it original size, but the inner portion of the cylinder, which was deformed a greater amount, tries to remain permanently deformed. This process results in a compressive tangential residual stress near the inside diameter that varies logarithmically to tension through the plastically deformed region of the cylinder. The process of producing residual stresses by means of plastic deformation of the pressure vessel is known as autofrettage.1 Compressive tangential residual stresses at the inside diameter due to the autofrettage process increase the strength and retard fatigue crack formation and growth under the pulsating internal pressure loading. It has been shown that autofrettage significantly increases the fatigue life of internally pressurized thick-walled pressure vessels in which fatigue failures originate at the inside surface. 2'3 However, the tangential tensile residual stresses at the oustide surface are of great concern when discontinuities such as grooves and keyways are introduced at the outside diameter. These external structural discontinuities cause a redistribution of the autofrettage residual stresses 4 along with high stress concentrations at the discontinuities and have resulted in external fatigue failures of pressure vessels. 5 In this paper, analytical stress distributions of an autofrettaged thick-walled pressure vessel with an external groove subjected to internal pressure were obtained. Autofrettage residual stress distributions for 50, 75 and 100% overstrain, where percentage overstrain is defined as the percentage of wall thickness subjected to plastic deformation during the application of autofrettage pressure, were simulated using a thermal loading analogy6 and finite element methods. Residual stresses in a fully autofrettaged smooth thick-walled pressure vessel were measured experimentally using an equivalent cut method and compared to theoretical solutions.

T H E O R E T I C A L STRESS DISTRIBUTIONS The elastic stress distributions of a smooth thick-walled pressure vessel resulting from an internal pressure, P~, are given by the well known Lam6 equations, 7 Pia2 ( 1 - ~ )
O r -b2 _ a 2

(la)

PiaZ (1 ~-~) O~-b~_--a2 +

(lb)

Analysis of an auto~rettaged thick-walled pressure vessel

97

where or and oo denote the radial and tangential stresses and a and b are the inside and outside radii of the pressure vessel respectively. Theoretical stress distributions of a partially autofrettaged smooth thick-walled pressure vessel based on the assumptions of the Tresca yield criterion, elastic-perfectly plastic material behavior, plane strain conditions, and no reverse yielding nor Bauschinger effect are given as, 8

Orr= oryS(b2~a2( 1 - ~b2~rp2- b2 - ] I_ 2-~S

+[

lnaj 2b 2

e]
In for for a---r<--p (2a)

(
and

b2~"p2
a

a2 [p2- b2
~-~ In

~]}

=Oy~ 1 - ~ ] [ ~ + ~ l

p<-r<-b

b ][p

+1_ 2b 2 =or~ 1 + ~ ~-~+b-y-~_a2 ~-2

In In

for for

a~r~p

(2b)

p<-r<-b

where the radius p denotes the radius of the elastic-plastic boundary during the autofrettage process. If the yon Mises yield criterion were used for the derivation of the autofrettage residual stresses, 9 then the same equations would be obtained, except that (2/X/3)oys would replace Oys in eqn (2). Theoretical autofrettage residual stresses obtained above by elasticplastic analysis can also be determined using an equivalent thermal loading.I If there is a steady state heat flow such that Ta and TO are the temperatures at r = a and r = p of the pressure vessel respectively, then the temperature distribution of the smooth pressure vessel becomes T~-Tp In-p T = In (p/a) r =T O for for

a<-r<--p p<-r<-b

(3)

In this thermal loading, the radius p corresponds to the elastic-plastic radius in the elastic-plastic analysis of the autofrettage residual stress distribution. The resulting thermal stress distributions due to thermal loading given in eqn (3) are the same as those given in eqn (2), if the

98

S. K. Koh, R. 1. Stephens

yield strength, ays, in eqn (2) is replaced by the following relation,

fie ( T a -- To)

2(1 - v) In (p/a) - oys

(4)

where tr is the coefficient of thermal expansion. Therefore, if an equivalent temperature given in eqn (4) were imposed on the thickwalled pressure vessel, thermal stress distributions would be theoretically equal to the autofrettage residual stress distributions. In eqn (2) for the autofrettage residual stress distributions, the compressive tangential residual stress occurs at the bore region and varies logarithmically to tension through the wall. Therefore, radial cutting through the thickness of an autofrettaged pressure vessel causes an opening of the pressure vessel due to a relaxation of the residual stresses analogous to the form of a pure bending moment. The theoretical residual stresses for a 100% overstrain can be obtained by solving an elastic problem of a curved beam subjected to a pure bending moment, M, necessary to join the spilt opening angle y as shown in Fig. 1, resulting in, 1
= yE b

a,

~-~[-ln r

a2 b2 b2:fi(1-~)lnb]

(5)

oo=

l-In;

b2_a 2 l + F ) ln

Similarity between eqn (5) and the theoretical residual stresses for a 10(1% overstrain, i.e. P = b in eqn (2) yields
~ - ~ = Oys

rE

or

8~

= Oys

(6)

I ~ . 1.

Curved beam subjected to pure bending moment.

Analysis of an autofrettaged thick-walled pressure vessel

99

depending on the Tresca or von Mises yield criteria assumed in the elastic-plastic stress analysiLs, respectively.

STRESS DISTRIBUTIONS OF AN E X T E R N A L G R O O V E D T H I C K - W A L L E D P R E S S U R E VESSEL In many practical cases, external discontinuities such as keyway grooves, holes and cracks can be present in a thick-walled pressure vessel. These discontinuities cause a redistribution of residual stresses that may be difficult to determine using elastic-plastic analysis. It was shown by Pu & Hussain 4 that an active thermal loading can be used to produce thermal elastic stresses equivalent to autofrettage residual stresses in a partially autofrettaged thick-walled pressure vessel with external notches or cracks. In this paper, autofrettage residual stresses in a thick-walled pressure vessel with an external groove were found using a finite element method and thermal loading analogy. Linear elastic stress analysis of a thick-walled pressure vessel with an external groove using the finite element method was performed. The vessel was subjected to an internal pressure of 386 MPa and thermal loading for the simulation of autofrettage. An internal pressure of 386 MPa is the typical operating pressure of the actual pressure vessel. Configuration of the pressure vessel with a rectangular groove at the outside surface is shown in Fig. 2. The pressure vessel has an inside radius, a, of 85 mm and an outside radius, b, of 142 mm. Wall thickness ratio, defined as b / a , is 1.67. Groove root radius, R, is 1.5 mm. Three autofrettage loading conditions, 100, 75 and 50% overstrain, were analysed. In the thermal loading analogy, the equivalent temperature distribution given by eqn (4) was imposed, depending on percentage overstrain conditions. Due to the symmetry of the pressure vessel, as shown in Fig. 2, only half a segment of pressure vessel was used for the finite element analysis. The ANSYS finite element program was employed to perform the analysis. 11 A typical finite element mesh used to find solutions is shown in Fig. 3. Boundary conditions are shown in Fig. 3(a) and a refined mesh near the groove root is shown in Fig. 3(b). A uniform pressure of 386 MPa was applied on the inside surface for the internal pressure loading. For thermal loading, the temperature distribution due to steady state heat flow was first analysed using the finite element program, and then the nodal temperature results were transferred as an input for the thermal stress calculations. Plane strain conditions with ez = 0, where z is the longitudinal direction, was used for all loading

100

S. K. Koh, R. I. Stephens

\\

\
A'-A . . . . . . .

\
~

F i g . 2

Thick-walled p r e s s u r e vessel with a n e x t e r n a l g r o o v e (a = 85, b = 142, d = 10, h = 25, W = 57, R = 1-5, unit: m m ) .

(a) Generated mesh

(b) Refined mesh near thc root of groove Fig 3.

Finite e l e m e n t m e s h used for stress analysis.

Analysis of an autofrettaged thick-walled pressure vessel

101

g~
3 .(]1 0 " 10 . . . . 20 . . 30

Number of Elements (a) 4 node isoparametlic quadrilateral element

4.0 ~ '//l~t,..,.~ d " " " I . . . .

r~
o

ill. . . . . . . . . . . . . .
10 20 Number of

30

Elements

(b) 6 node isoparametric Uiangul~ element

K t

3.5

,01
. 7 i 8 . i 9 . , 10 . , 11 , I 12

.....
, i 13 , i 14 , 15

3.0

Number of Elements (c) 8 node isoparametric quadrilateral element

Fig. 4. Convergenceof theoretical stress concentration factor, K,, due to internal pressure loading for three different finite element types.

cases in the analysis of the thick-walled pressure vessel. For the finite element modeling, three different types of two-dimensional plane strain solid elements were used: a four-node isoparametric quadrilateral element, a six-node isoparametric triangular element, and an eightnode isoparametric quadrilateral element. For each element type, convergence was checked in order to ensure correct results by avoiding possible input errors and inherent errors in some element types. Convergence of the theoretical stress concentration factor, Kt, due to internal pressure loading for each element type is shown in Fig. 4. In the final meshes, the number of elements around one quarter of the groove root circle for the four-node quadrilateral element, six-node triangular element, and eight-node quadrilateral element were 28, 20 and 12 elements, resulting in element sizes of 0-06, 0.12 and 0.15 mm, respectively.

102

S. K. Koh, R. 1. Stephens

RESULTS A N D DISCUSSION The tangential stress distribution along the planes A - A ' , B - B ' , and C - C ' , as designated in Fig. 2, due to internal pressure only are plotted in Fig. 5. Here data points represent the finite element solutions and the solid lines represent interpolation curves of the finite element solutions. Plane A - A ' is the radial plane opposite the external groove, and planes B - B ' and C - C ' are the radial planes through the groove root and through the center of the groove, respectively. In Fig. 5, a significant stress concentration exists along the plane B - B ' , open triangles, but the stress distribution along the plane A - A ' , open squares, was not influenced by the existence of an external groove. The difference between the finite element results and theoretical solutions was less than 3%, which was considered as an excellent agreement. It must be noted that tangential stresses near the groove root were not equal to the maximum principal stresses, since the tangential stress direction differed from the maximum principal stress direction by an angle less than 5 due to the presence of the groove. But the difference between the maximum principal stress and the tangential stress was less than 1% near the root region due to the small difference in angle. Therefore, the maximum principal stresses were used near the groove root in Fig. 5. The theoretical stress concentration factor, Kt, defined as
g t = Oma-'-"~x ON (7)

was calculated at both the root and the center of the groove. Nominal stress, os, was defined as the stress occurring at the outside radius for

D 5 o 4

A-A'

B-B'
C-C

f 0.0

i 0.6

i 0.8

, 1 0

0.2

0.4

(r-a)/W Tangential stress distributions due to internal pressure loading along A - A ' , B - B ' , and C - C ' in a thick-walled pressure vessel.

Analysis of an autofrettaged thick-walled pressure vessel

103

the reduced wall thickness. This definition of nominal stress is analogous to a net nominal stress. 12 Therefore, from eqn (lb) the nominal stress due to internal pressure P~ is given as, 2Pia 2
oN = b 2 - a 2 (8)

where bo is the outside radius of the thick-walled pressure vessel with reduced thickness, i.e. bo = b - d . Using the typical operating internal pressure of 386 MPa, theoretical stress concentration factors, K, of 3.93 1.72 were obtained at the root and at the center of the groove, respectively. Nominal stress, maximum stresses and theoretical stress concentration factors due to internal pressure loading only are listed in Table 1. Due to the sharp radius of the groove root, the maximum principal stress at the groove root was much higher than at the center of the groove, as shown in Table 1. However, a considerable stress concentration at the center of the groove also exists. In order to determine the autofrettage residual stresses using the thermal loading analogy, a temperature distribution due to the steady state heat flow must be found. The ANSYS finite element program was used to first determine the temperature distribution. Since the thermal stresses result from the temperature gradient, the temperature, To, can be assigned as zero for convenience. The initial temperature input for steady state heat flow was calculated from eqn (3) using the yield strength Oys, of 1170 MPa, Young's modulus, E of 210 GPa, thermal expansion coefficient, tr, of 6.8 x 10-6/C, and Poisson's ratio, v, of 0.29. The initial temperature, Ta, at the inside radius r = a resulted in 589, 467 and 332 C for 100, 75 and 50% overstrain cases, respectively. After performing the temperature analysis using the ANSYS program, the finite element analysis was again performed to obtain the thermal stresses from the calculated nodal temperature results. These thermal stresses are equivalent to the autofrettage residual stresses as already mentioned.
TABLE 1

Nominal Stresses, and Finite Element Solutions of Maximum Principal Stresses and Theoretical Stress Concentration Factors for Internal Pressure Loading
Loading condition om~@B' (Mea) am,~@C' (MPa) oN (MPa)
@B'

K,
@C'

Internal pressurea a p~= 386 MPa.

2147

938

547

3-93

1.72

104
2

S. K. Koh, R. 1. Stephens

o
B-B'

o C-C
-1 i i i I

0.0

0.2

0.4

0.6

0. B

1.0

(r-a)/W Fig. 6. Tangential stress distributions due to 100% overstrain loading along A - A ' , B - B ' , and C - C ' in a thick-walled pressure vessel.

Finite element stress distributions for each overstrain condition along the planes A - A ' , B - B ' , and C - C ' are shown in Figs 6-8, where the data points represent the finite element solutions and the solid lines represent the interpolation curves of the finite element solutions. A very high stress concentration at the groove root can be clearly noticed for each autofrettage loading case. Finite element solutions of the tangential residual stresses along plane A - A ' for each overstrain level resulted in very close agreement with the theoretical solutions in eqn (2b) derived for the smooth thick-walled pressure vessel. The differences between these finite element solutions and the theoretical solutions were less than 3%. Just as in the case of internal pressure loading, nominal stress in the autoffettage loading was defined using a thick-walled pressure vessel with reduced wall thickness in order to be consistent with the definitions for all loading conditions. From eqn (2b), using a reduced thickness, the tangential stress at the outside radius,

oA-A' B-B' o C-C

Fig. 7.

0.4 08 1.o (r-a)/W Tangential stress distributions due to 75% overstrain loading along A - A ' , B - B ' , and C - C ' in a thick-walled pressure vessel.

-~ .o

o.o

0.2

i o.~

Analysis of an autofrettaged thick-walled pressure vessel

0.6

105

0.3

1
o

0.0

-0.3

C-C
I

-0.6 0.0

o.2

0.4

0.6

o.B

1.0

(r-a)/W

Fig. 8. Tangential stress distributions due to 50% overstrain loading along A - A ' ,
B - B ' , and C - C ' in a thick-walled pressure vessel.

i.e. r = bo = b - d, was defined as the nominal stress by

~2 2a 2 r~2 ,. b2 o

[bo

b o - a [ 2bo

(9)

The elastic-plastic radii for 75 and 50% overstrain cases were not changed; however, for the 100% overstrain case it was redefined as the outside radius bo, since the original outside radius b does not exist in the pressure vessel with a reduced wall thickness. Nominal stress, oN, defined for each loading case, is listed in Table 2. As shown in Figs 6-8, the magnitude of compressive tangential residual stress at the inside radius increases as the percentage overstrain increases. This implies that for both yielding and fatigue inside surface problems, a higher internal operating pressure can be applied to a smooth thick-walled pressure vessel that has been subjected to a larger overstrain. It was reported that a 100% overstrain condition resulted in the maximum fatigue life for a smooth thick-walled pressure vessel without any external grooves, where cracks grew from the inside surface. 2 However, it can be observed that the maximum tensile stress
TABLE 2 Nominal Stresses, and Finite Element Solutions of Maximum Principal Stresses and Theoretical Stress Concentration Factors for Autofrettage Loading

Loading condition

o~o~@B'

mo~@C r

oN

K,
@B' @C' 1-88 1.72 1-76

(Mea)
Percentage overstrain 100 75 50 2044 1462 692

(MPa)
825 633 299

(MPa)
440 368 170 4-65 3.97 4-07

106

S. K. Koh, R. L Stephens

at the external groove root increases as the percentage overstrain increases, meaning that longer fatigue crack formation and growth life at the external groove root will occur for a small percentage overstrain under the same pulsating internal pressure. Maximum principal stresses, o . . . . at the locations B' and C' along with the nominal stress, oN, and stress concentration factors, Kt, are shown in Table 2 for each overstrain loading case. Significant differences in the maximum principal stresses for different percentage overstrain are shown in Table 2. At the groove root, i.e. location B', the maximum principal stress in a 100% overstrain case was 40 and 195% higher than in 75 and 50% overstrain cases, respectively. However, essentially no differences in Kt between 75 and 50% overstrain cases for both the center and root of the groove can be seen in Table 2. Also, theoretical stress concentration factors at points B' and C' for the 100% overstrain case were only 16 and 8% higher than for 75 and 50% overstrain cases, respectively. The small differences in Kt, compared to the large differences in the m a x i m u m principal stress, between percentage overstrain cases at both the center and the root of the groove were due to the similar difference in the ratio between maximum and nominal stresses. A maximum principal stress contour plot near the groove root due to 100% overstrain loading is shown in Fig. 9. Similar stress contours near the groove root due to 75 and 50% overstrain loadings, and internal pressure loading were obtained. Maximum principal stress contours from all loading conditions showed significant stress concentration at the sharp root of the groove. Figure 10 shows the m a x i m u m principal stresses along the groove root surface. Regardless of the loading

Figo 9 Maximum principal stress contours near the groove root due to 100%

overstrain loading.

Analysis of an autoffettaged thick-walled pressure vessel

2.0 . , . , . , . , . , . .

107

Pressure

1.5

75%

O.S.

0.5,

0.0[

is

30
A n g l e

4s
o (Degree)

60

75

go

Fig. 10. Maximum principal stress variations along the groove root due to each loading condition.

conditions, the peaks of maximum principal stress occurred at the angle 0 between 15 and 20 from the beginning of the root curvature, and they decreased very rapidly as the angle increased. In order to experimentally determine the autofrettage residual stresses for a 100% overstrain case using the equivalent saw cutting method, rings of 20 m m thickness were taken from a fully autofrettaged thick-walled pressure vessel and were saw-cut through the thickness as shown in Fig. 11. Theoretical stress distributions in the thick-walled pressure vessel due to 100% overstrain are equivalent to the stress distributions in a curved beam due to the bending m o m e n t which is required to put the split ring together. Therefore if an autofrettaged ring were cut through the thickness, then the cut would make a certain opening angle of the split ring due to relaxation of tangential autofrettage residual stresses. As shown in Fig. 11, punch marks were made to measure the split opening angle and a series of strain gages were attached on the opposite side surface of the saw-cut. Combina-

Punch n a ~ ~ r Saw-cut

/ i,, ..... .,

Fig. 11. Saw-cutof autofrettaged ring and strain gage locations.

108
1000

S. K. Koh, R. I. Stephens

5OO

Eq.(3. l~.a)

.(3,18.b1

/ /:'
@
([.
-I000
0 0.2

__'-.--TM ....... ' ' "....

A Expcrimen(a[ Oo(Rosetlc)

0.4

0.6

0.8

(r-a)/W

] ~ , ]l~, Autofrettaged residual stress distributions across the thickness for a 100%
overstrained thick-walled pressure vessel.

tions of uniaxial and 45 rosettes were used across the thickness of the wall. Figure 12 shows the autofrettage residual stress distributions calculated from the relaxed strains by the saw-cut across the thickness of the wall. The data points represent the experimental stress values with opposite signs after the saw-cut, while the lines respresent the theoretical residual stress values prior to the saw-cut. The tangential relaxation stress data, calculated from the uniaxial strain gages using o - - E e , were expected to be in error, which turned out to be small compared to the calculated relaxed stress data from the rosettes using the generalized Hook's law, as shown in Fig. 12. This was attributed to the relatively small magnitude of radial relaxation stress. G o o d agreement of residual stress distributions between the experimental measurements and theoretical calculations can be seen in Fig. 12, especially for the von Mises criterion. This good agreement implied total relaxation of the autofrettage residual stresses after the saw-cut, leaving the split ring supposedly free from residual stresses. Noticeable scatter in the

Analysis of an autofrettaged thick-walled pressure vessel

TABLE 3 Split Opening Angles from the Saw-cut of Autofrettaged Rings

109

Experimental opening angle (o)a

4.98, 4.97, 4.68, 5-05 4-84, 5.04, 5.25 a Average value: 4.97

Theoretical opening angle ()

4-21 (Tresca) 4.86 (yon Mises)

tangential residual stress distributions between experiment and theory can be seen at both the inside and outside surfaces. This was attributed to the assumptions made on the theoretical derivation regarding elastic-perfectly plastic material behavior and yield criteria. The maximum difference in tangential residual stress between experiment and theory was observed at the bore, resulting in approximately 28% for the Tresca criterion and 9% for the von Mises criterion. The distance between the punch marks measured after the saw-cut was used for split angle calculations. The opening angles obtained from the seven rings taken from the autofrettaged pressure vessel are listed in Table 3 along with the average angle. These values can be compared to the theoretical opening angles calculated from eqn (6). Young's modulus, E, of 200 GPa, and yield strength, crys, of 1170 MPa were used for these calculations. The experimental opening angle showed results closer to the theoretical opening angle based on the von Mises yield criterion than to that based on the Tresca yield criterion.

SUMMARY AND CONCLUSIONS 1. Autofrettage residual stresses were successfully determined by imposing an equivalent thermal loading in a thick-walled pressure vessel with an external groove. 2. Significantly high theoretical stress concentration factors, K,, of 3.93, 4.65, 3.97 and 4.07 were obtained at the external groove root due to internal pressure, 100, 75 and 50% overstrain loadings, respectively, from the finite element analysis. 3. As the percentage overstrain increased, both magnitude of the compressive tangential residual stress at the bore and the tensile tangential residual stress at the external groove root increased. Higher percentage overstrain autofrettage was detrimental to the pressure vessel with discontinuities at the outside surface.

110 4.

s. K. Koh, R. I. Stephens
Finite element solutions of tangential stresses along the thickness opposite to the external groove were not influenced by the existence of an external groove, resulting in excellent agreement between the finite element solutions and theoretical solutions. Experimental measurements of the autofrettage residual stresses using both split opening angles and relaxed strains due to the saw-cut of a fully autofrettaged pressure vessel showed excellent agreement with the theoretical solutions. Peaks of the maximum principal stresses regardless of loading conditions occurred at an angle between 15 and 20 from the beginning of the root curvature, which was considered as the most vulnerable location for pressure vessel failure. The stress analysis results showed a high possibility of failure at the groove root due to the high stress concentration. This indicates that the failure can be avoided or the life of the pressure vessel can be extended by changing the groove shape to alleviate the stress concentration at the root, and by decreasing the percentage overstrain to a certain o p t i m u m level in order to reduce the tensile tangential residual stresses at the outside diameter without sacrificing the beneficial compressive tangential residual stresses at the bore.

5.

6.

7.

ACKNOWLEDGEMENTS The authors would like to thank J. A. Kapp and the US Army Watervliet Arsenal for sponsoring this research.

REFERENCES 1. Franklin, G. J. & Morrison, J. L. M., Autofrettage of cylinders: prediction of pressure/external expansion curves and calculation of residual stresses. Proceedings of Institution of Mechanical Engineers, 174 (35) (1960) 947-74. 2. Davidson, T. E., Eisenstadt, R. & Reiner, A. N., Fatigue characteristic of open-end thick-walled cylinders under cyclic internal pressure. Journal of Basic Engineering (Dec. 1963) 555-65. 3. Morrison, J. L. M., Crossland, B. & Parry, J. C. S., Strength of thick cylinders subjected to repeated internal pressure. Journal of Engineering for Industry (May 1960) 143-53. 4. Pu, S. L. & Hussain, M. A., Residual stress redistribution caused by notches and cracks in a partially autofrettaged tube. Journal of Pressure Vessel Technology, 103, (Nov. 1981) 302-6.

Analysis of an autofrettaged thick-walled pressure vessel

111

5. Kapp, J. A. & Pu, S. L., Fatigue design of thick-walled cylinder considering the OD as a failure site. Pressure Vessel Design, PVP-57, American Society of Mechanical Engineers, 1982. 6. Hussain, M. A., Pu, S. L., Vasilakis, J. D. & O'Hara, P., Simulation of partial autofrettage by thermal loads. Journal of Pressure Vessel Technology, 103 (Nov. 1981) 302-6. 7. Ugural, A. C. & Fenster, S. K., Advanced Strength and Applied Elasticity. Elsevier Publishing Co., New York, 1981. 8. Hill, R., The Mathematical Theory of Plasticity. Oxford University Press, London, 1950. 9. Davidson, T. E. & Kendall, D. P., The design of high pressure containers and associated equipment. In The Mechanical Behavior of Materials under Pressure, ed. H. LI. D. Pugh. Elsevier, Amsterdam, 1970, pp. 54-118. 10. Timoshenko, S. P. & Goodier, J. N., Theory of Elasticity. McGraw-Hill, New York, 1970. 11. Desalvo, G. & Swanson, J. A., ANSYS User's Manual. Swanson Analysis System Inc., June 1985. 12. Peterson, R. E., Stress Concentration Factors. John Wiley, New York, 1974.