Thin-Walled Structures 3 (1985) 145-162

Analysis and Design Guidance for the Lateral Stiffness
of Bellows Expansion Joints

N. W. Snedden
UEN/131, Shell UK Expro, Shell-MexHouse, Strand, London, UK


Thin-walled bellows expansion joints are frequently applied in piping
systems for absorbing thermal and mechanical movement. There have
been several failures in service due to the lateral buckling of bellows under
pressure, the most notable case being the Flixborough disaster in 1974.
The problem of bellows squirming, as the phenomenon is more
commonly termed, was first investigated by the Dutch engineer J. A.
Haringx in 1952. In the intervening years there has been little additional
research carried out and consequently there is a paucity of experimental
data and practical theories on the subject. The present paper offers
guidance to avoid bellows squirming and provides the design engineer
with simple procedures for evaluating the stability of a pressurised
bellows subject to either small or large lateral displacements. Formulae
are also presented in order to determine the strength of bellows supports
which limit and control the amount of bellows movement in service.


E1 Equivalent bending stiffness.
F Lateral force.
h Wall thickness.
Ka Elastic axial stiffness.
Ke Elastic lateral stiffness.
C o n v o l u t e d length.
M Bending moment.
n N u m b e r of convolutions, coefficient.
Thin-Walled Structures 0263-82311851503.30© Elsevier Appfied Science Publishers Ltd,
England, 1985. Printed in Great Britain

there are a variety of convolution shapes: S-shaped. y Cartesian coordinates. Such joints are manu- factured in a variety of ways. It has long been established from research and practice that under certain conditions a bellows under internal pressure can become unstable and 'squirm' laterally (see Fig. h y ~ a u ! i c and pneumatic tube forming are among the most common r n a n u f a ¢ ~ g techniques employed. The latter configuration is by far the most popular. W.-~ = bellows mean radius. P Equivalent axial load. universal. and U-shaped (see Fig. 1 INTRODUCTION Bellows expansion joints are widely used in piping systems to absorb differential thermal expansion whilst containing the system pressure and flow..#e relationship for the elastic buckling pressure. gimbal. .. E 1 = bellows equivalent bending stiffness* . C-shaped. a Angle of tilt (from vertical) of convolution..onal and nuclear power plants. Snedden p Internal pressure. roller formi. x. and heating and cooling systems. etc. -~ Mean radius. 0 Angle of rotation of convolution. chemical plants. Bellows are constructed from relatively thin gauge material (normally stainless steel) and are convoluted in order to provide the necessary flexibility needed to absorb mechanical and thermal movements expected in service.146 N. There are many categories of bellows joints: for example. A Equivalent axial displacement of convolution. They are installed in oil refineries. 2). p' Elastic buckling pressure.7~¢2 (1) where p' = internal buckling pressure. 8 Lateral deflection. eom'cnP. viz: "n'EI p' . 1). The phenomenon of squirming was first demonstrated by Haringx in 19521 who showed that pressure buckling of a bellows was analogous to buekting of an Ederstnrt~: F I e ~ a ~k~t. pressure-balanced. *Haringx'sformula for bendingstiffnessis givenin Appendix 1.g. hinged. and ~e = bellows convoluted length. In addition.

2. l~g. Typical bellows convolutions. Squirmed bellows. . 1.Lateral su'ffness of bellows expansion joints 147 S-shaped g'L U-shaped C-shaped ~g.

Snedden It was the investigation into the cause of buckling of bellows which also prompted Haringx to study the problem of elastic stability. bellows tASME Code Classes are related to SafetyClassesdefinedin ANS-5l.e. He showed that the bending characteristics of the system varied considerably with changes in pressure and pivot-point location. a bellows which is only permitted to rotate about a fixed pivot-point. The principal conclusion of his work was that. design. The analytical results of this work were compared with experimental data from bellows tests conducted by Fitzgibbon. of thin-walled cylinders subjected to internal pressure and to explain the buckling of drill pipes.5 in dia. 1/N18. published by a group of US and Canadian companies. At the present time there is no British Standard for the design of bellows expansion joints. i.9 . There is.) under external pressure was relatively good. 2 Newland 3extended Haringx's theory to the more complicated case of a universal expansion joint. recommends standards relating to the manu- facture. Design rules for bellows are also given in Section III (Nuclear Power Plant Components) of the ASME Boiler and Pressure Vessel Cod& where it is stated that bellows expansion joints are allowable in nuclear piping systems conforming to ASME Class 2t and Class 3 standards and in metal containment vessels. for the large bellows (13 in dia. safety and installation of bellows expansion joints for application in piping systems. 5The agreement found between theory and experiment for small bellows (4. by providing a correctly designed supporting structure.e.148 N. the critical buckling pressure could be increased by up to four times the value for the same system with no supports. According to subarticle NB-3649.2-1973. Seide 4 also investigated the effect of pressure on the stability of a hinged bellows. a code of practice for the installation of metallic bellows: BS61296 and the Standards of the Expansion Joint Manufacturers Association 7 (commonly referred to as the E J M A standard). however. W. In all of the aforementioned papers only elastic instability was considered. i.) under internal pressure correlation was poor and believed to be due to friction between the convolution plies. However. two bellows interconnected by a straight length of pipe. This latter document. and that instability of the bellows could occur at both internal and external pressure.1.

the assembly had collapsed by a combination of bellows squirming and jack-knifing of the mitred pipe joints. Despite the wealth of advice available in the abovementioned codes and standards. tragically. the entry on bellows squirming is limited to a few cautionary paragraphs on the phenomenon and a formula for evaluating the elastic buckling pressure of a bellows. Lateral stiffne~ of bellows expansion joints 149 greater than 1 in nominal diameter are not permissible in Class 1 construction. Many reports have been produced explaining the cause of the disaster. there is a paucity of design guidance given on the important problem of lateral instability of bellows. Class 1. liquid metal piping. Reference is made to the fact that design rules for Class 1 components are currently under development. among the more interesting contributions are papers by Newlandll and Webster. testing. Division 1. the ends of which are rigidly supported. The effect of plasticity in the convolutions on the non-linear lateral stiffness of bellows is highlighted. ~°the cause of the explosion was traced to a temporary by-pass assembly consisting of a dog-leg pipe connected by bellows expansion joints to reactor vessels at the ends. design. That there have been several failures of bellows in service due to squirming cannot be wholly attributed to the lack of suitable design guidance. the deaths of 28 people. as a result of gross human error. The ASME Code is also relatively weak in this area. It was the subsequent release of cyclohexane that led to the explosion and. in the present paper limited guidance in the form of simple design formulae is offered for assessing the stability of a pressurised bellows subject to an initial lateral deflection and for evaluating the strength of bellows lateral supports.e. for example. It is the author's intention that the design recommendations given should be of practical use to pipework and project engineers alike. fabrication and instal- lation. the bellows must be capable of satisfying the maximum squirming pressure fixed by the applicable ASME Design Service Limits. there is an interesting and lengthy ASME Code Case (N-290-1. and the Code Case cited earlier states that the bellows design is only acceptable against instability if the defined pressure testing requirements are met. of bellows for service in Section III. In the case of the Flixborough disaster in 1974. . i. In the EJMA standard. 12 In view of the shortage of design information available. approved April 1983) which details rules for materials. Nevertheless. etc. The official inquiry showed that.

there are only a few published theories on the problem of bellows bending stiffness: Haringx's I analysis m a d e use of the theory of unsymmetrical bending of circular plates and was confined to the mathematically more simple case of bellows with rectangular-shaped convolutions.150 N.~2 (6) where p = internal pressure. Assuming then that elementary beam theory is valid. P and F the bending m o m e n t . M. The derivation of this rela- tionship is discussed by Newland 3 and Seide. 14examined the bending flexibility of U-shaped convolutions and produced simple design charts for a range of bellows dimensions. 3) dZy EI-~ = . In addition to the above theories. W. Snedden 2 THEORY 2. . on the bellows. axial and lateral forces. and n = (5) p = p~. H a m a d a e t al.Py .Fx + M (3) The solution of this equation is M Fx y = Acosnx+Bsinnx~ P P (4) w h e r e A and B are unknown constants. the differential equation for the lateral deflection of a bellows is (see Fig. respectively. there is an elementary formula relating bellows bending stiffness and axial stiffness. 4 Seide 4 points out that the circumferential flexural stiffness of a bellows is considerably greater than the meridional stiffness and that for small deflections the bellows cross-section will remain undistorted after loading. 13 However.1 Small deflection analysis There are numerous formulae in the literature for calculating the elastic axial stiffness of a bellows. viz: EI = ~Kar-£ (2) w h e r e Ka = bellows elastic axial stiffness.

.. 3 are dy _ 0 at x = 0 (7a) Y-dx y= 6 atx= ~ (7b) and from equilibrium of the bellows F~ + P6 M .c o s n g ) + 2sinn---------~..2~ ] (11) . Lateralstiffne~ of bellowsexpansionjoints 151 P :/" °'/. .2g ] 6 = n (9) P(1 + cosng This relationship can be re-written as F = Ke6 (10) w h e r e Ke = bellows elastic lateral (shear) stiffness Ke = [ g ( l . Simple beam model for bellows.El I ' :J x J v . 3.c o s n ta) + 2sinngn . Fig. (8) 2 Substituting eqn (4) into eqns (7a) and (7b) and using eqn (8) yields the following solution for the lateral deflection between the ends of the bellows F [ g ( 1 . T h e b o u n d a r y conditions for the model shown in Fig.

-~. In this situation a more elaborate theory is required to predict the bellows behaviour. by truncation. 8 = ?/4) and relatively low internal pressure. ~ . In service. When this happens the convolutions buckle in a local manner and distort . ~ zr2( (16) 2. into eqn (15) finally gives 12El 12P Ke. for e < < (---7 (12) It is also possible to derive this result from Haringx's theory. which is sometimes termed column instability.2 Large deflection analysis The results obtained for the lateral stiffness in eqns (12) and (16) are only relevant to bellows undergoing small elastic deformations. (In addition to the type of squirming discussed here.152 N. in-plane instability can occur if the bellows intemal pressure is high. It has been observed from full-scale buckling experiments carried out by the author 15 that the assumption of undistorted cross-sections is sufficiently accurate for bellows subject to gross shear deformation (e. Snedden Using a Maclaurin series to expand the sine and cosine terms it can be shown. bellows are subject to much larger displacements and are often stressed beyond the elastic limit. W. however. The condition for instability of the bellows is given by Ke = 0 at p =p' (13) and from simple beam theory 12El Ke . respectively. (3 at p = 0 (14) It follows from the work of Haringx that K~- Substituting f o r p a n d p ' from eqns (1) and (6).g. that eqn (11) simplifies to 12EI 6P E1 Ke .

4. Lateral stiffness of bellows expansionjoints 153 such that they are no longer perpendicular to the bellows neutral axis./2 tar ] (19) ~=. each ~ ~idIink2s ~ inghinges Fig. The deformation of a bellows subject to combined internal pressure and lateral loading is asymmetric. 7) By maintaining this assumption it is possible to establish a simple procedure for investigating the non-linear lateral stiffness of a bellows. Thus by analysis of the geometry shown in Fig. the overall shape of the deformed structure possesses 180° rotational symmetry about the mid-point of the neutral axis.nl_ T 1 . 4). Large displacementmodel. In the new model postulated here (see Fig. viz. 4 the equation for the lateral displacement between the ends of the bellows is . simulating the rotational stiffness of a single convolution. The mathematical model which will be utilised is similar to the one considered by Newland 3 where the bellows was represented as a strut made up of a number of rigid links joined by linear spring hinges. 4) the hinges are replaced by non-linear springs which each rotate through an angle defined by ON =f(MN) (17) where Ms = bending moment applied to convolution N (18) and N = convolution number (N = I corresponds to convolution at end of bellows see Fig.

Experiments too are costly and not always feasible to perform. the former option can prove uneconomical if a non-linear analysis of a large model is essential. respectively. finite element theory may be employed or tests conducted. such as shown in Fig.3 Proposed solution There are a number of ways in which the angular stiffness of the convolution can be determined. straining is greatest in the longitudinal fibres furthest away from the neutral axis) III where contact occurs between adjacent convolutions These can be conveniently termed the elastic.w h e r e localised plasticity occurs at the root and crown of the convolution (by simple analogy with a beam subject to pure bending. The unknown quantities which have to be estimated are the slopes and . A n alternative and much simpler approach. The application of such theories to bellows expansion joints is formidable and beyond the scope of the present paper. However. 5. This figure indicates three distinct regions: I--where the rotation of the bellows convolution is linearly proportional to the applied bending moment I I . elasto-plastic and contact regions. would be to describe the moment/rotation characteristic of eqn (17) by means of a tri-linear function. eqn (19) can be readily solved using an iterative procedure by assuming an initial value of 8. W. Provided the elasto-plastic moment/angular deflection relationship of eqn (17) is known for a single convolution. Shell theory is sufficiently advanced to enable the analysis of shells of revolution subject to non-symmetric loading 16 but the governing differential equations are complex and their solution is dependent on the aid of a large digital computer. 2.. based on experimental observation. Snedden where a s = vertical tilt (in radians) of convolution N 1 ~N = Z OM (2o) M-1 and n = number of convolutions in bellows. For example.154 N.

5 first yielding of the material occurs. By using the EJMA design equations 7 or simple beam theory together with the maximum shear stress theory (Tresca) to predict . Lateral stiffness of bellows expansion joints 155 6000 55OO 5O00 4500 [] 4000 3500 3000 N 2500 1 2000 1500 1000 5O0 I J i i i b 0. nEl (21) It may be safely assumed that in region II the deformation of the con- volution is perfectly plastic and that at point 1 in Fig. 05O 0 2 ANGULAR OEFLECTION (RAD) l~g.005 0.010 eI 0. intercepts of the lines in each region.000 0. 5.025 O.020 0.015 0. Plot of angular stiffness of bellows corrugation. The gradient of the line in region I is easily calculated since for a linear elastic bellows 3 M~ 0.

The validity of this assumption will be commented upon in Section 3.) For simplicity. The experimental results plotted are applicable to 0. (12) and (16) are indistinguishable at the scale indicated in Fig. In region III adjacent convolutions of the bellows come into point contact. (A similar effect occurs when an ordinary helical spring is fully compressed. together with the solution of eqn (11). 3) is valid for small lateral deflections. W.) It is worth pointing out that the individual solutions of eqns (11). Details of the relevant bellows geometry and materials data are given in Appendix 2. it may be assumed that the slope of the line in region III is infinity. The value of bellows equivalent bending stiffness used in the derivation of the theoretical line shown was determined from the formula by Haringx (see Appendix 1). are presented in Fig. 6. 6 there is sufficiently good correlation between theory and experiment to establish that the simple bellows beam model (Fig. .b. As can be seen from Fig. viz. 3 RESULTS Experimental data for the elastic lateral stiffness of a bellows taken from Ref. Snedden the maximum permissible equivalent axial displacement ~ of the convolution. no further deformation can take place once the convolutions touch. 15. It should be evident that the convolutions come into contact when (23) where h = bellows wall thickness. the limiting angle 01 can be estimated from O1 ~ (22) The intercept of the lines in regions II and III can be determined from the assumption of undistorted cross-sections. 6.156 N. As a result the angular stiffness of the convolution will be significantly increased compared to the elastic bending stiffness. (n.040 in lateral deflection.

~*. Typical experimental curves for the lateral stiffness of a bellows subject to m u c h larger displacements are plotted in Fig. . ... ..... . . .. . .:::::::: :::: -::: ========================== ~... ......~ ~ : :q-..i ?:: t i Fig..::: -'--:::::t::-:~ !~ ~ .. . .. B e y o n d this deflection the discrepancy between theory and experiment can be mainly attributed to the optimistic assumption for the bellows bending stiffness once adjacent convolutions touch. . ... Despite the relative crudeness of the analytical m e t h o d employed the a g r e e m e n t up to about 1 in lateral deflection is very good between the two sets of curves for p = 0 and 70 psig..~:~:: ~r:.. l .25 in lateral deflection. t .... . Lateral stiffnessof bellows expansion joints 157 I!!!: • i :i i. :. ... Plot of bellows elastic lateral stiffness. .. ~ i ~ !L.. It can be inferred from the test data s h o w n in Fig.. . .-::-.. .... . .~:. . . .. .. 6..7 in the bellows is shown to be unstable.-~ ~.... . in the case where internal pressure is applied the onset of instability is correctly predicted at 0. . Nevertheless.. .. --.. . . . .. This figure also shows the theoretical predictions obtained using eqns (17) to (23) and the tri-linear function for the angular stiffness of one convolution prescribed earlier. . respectively.. .. ~ . 7 that the convolutions continue to deform after initial contact is made... . . .. ... .. .-..... . Between this displacement and approximately 0. . . ..... 7.. . .. .. .. . ~ .:~: :::: -::: :: ~:÷:: :~q~:.: I~o0~l~ili?ilIiiiii:ii~i:iiii~ili~li~i~]ii!ilii~i[i!ili~i~ i i!]i!iiiTiliiiiii~::i::i:::::~i~!!i ~iii i ~il ~t : ::: :q: : : : : : ~. . : ~. :h~o ~ . .

6 show that for the particular bellows tested the elastic lateral stiffness decreases linearly with increasing internal pressure.9 1. This result is also evident from the form of eqn (12).0 1. Plot of bellows lateral stiffness curves. (12) and (13). 4 DISCUSSION The data plotted in Fig. It is possible to determine the bellows elastic buckling pressure either by linear extrapolation of the experimental data or.? 0. W. 4 t.0 0.8 0.1 2 Ii ~ !.~ 0.5 0.5 LATERAL DEFLECTION [IN) Fill.. Snedden 5000 ii IF 4500 ii .6 0. from eqns (6). as previously indicated. 30DO -[ M Experiment ~2500 2000 1500 // / / I000 / / 500 I I i I I I I I I 0.4 0.158 N.2 0. When subject to combined internal pressure and lateral movement the .i ~ Theory 4000 i' 3500 .3 0. 7.

7For the bellows to remain stable the lateral stiffness Ke must exceed zero.if Po = p. . 7 . 5 DESIGN RECOMMENDATIONS It is recommended that. can be a temporary event impeded by the increased bending stiffness due to contact between convolutions.])8 p=0 p~p. or indirectly from eqn (2) using either the measured value of bellows axial stiffness or the result for Ka calculated from the EJMA standard. can lead to catastrophic failure of the bellows. (12) and (14) as follows F. (24) On the other hand. if the bellows design stresses are within the elastic limit. instability.2 E I / g 2. The test results shown in Fig. for high internal pressure. However. .e. it is recommended that the lateral load/deflection characteristic of the bellows should be verified by experiment or calculated using eqns (17) through (23). In either instance. In the event that the bellows design stresses exceed yield. i. 7 corroborate this assertion. gross buckling due to an excessively large destabilising force. if the bellows convolutions undergo localised plastic deformation the destabilising force due to pressure is simply F~ = ( r ] . If displacements are small and the material remains elastic this force is easily calculated from eqns (6). or plastic collapse. ] . the bellows lateral stiffness should be determined using eqn (12) if the bellows axial compressive load P < . If the bellows operating pressure is relatively low. = ( K . or with eqn (11) if P >. Lateral stiffne~ of beUows expansion joints 159 expansion and compression of the convolutions as the bellows bends produces a destabilising force. However. An alternative function from that used by the author may be employed to define eqn (17). uncontrolled movement.6"rr'r2p°85e ifpo<p' or -12EI8.2 E I / g 2. the quantity E1 should be derived from Haringx's formulae given in Appendix 1.K.4 ) (25) p = 0 P=Po where Fis computed from eqn (18) for a given lateral deflection 8.

London. if the d e s ~ stresses are greater than yield. Philips Research Report. Space Technology Laboratories. It is also firmly r e c o m m e n d e d that any bellows expansion joint whose failure could be catastrophic should be adequately restrained to prevent excessive deformation due to instability..2-1973. J. 1981. 1983.~. Code of Practice for the Selection and Application of Bellows Expansion Joints for Use in Pressure Systems. October 1958. Rules for Construction of Nuclear Power Plant Components. . Standards of the Expansion Joint Manufacturers Association. A. Instability of thin-walled cylinders subjected to internal pressure. 7(2) (1952). The effect of pressure on the bending characteristics of an actuator system. It is recognised that these design recommendations are based on cx tremety limited data. REFERENCES 1. The Flixborough Disaster--Report of the Court of Inquiry.. if the bellows design stresses are elastic. .. Buckling and rupture of the double bellows expansion joint assembly at Flixborough. Haringx. EM 8-20. D. ANS-51. W. .. Instability of bellows subjected to internal pressure. ASME Boiler and Pressure Vessel Code. Report No. A. Newland. Part l--Metallic Bellows Expansion Joints. Section Ill--Division 1. Inc. . 11. ..~. D. Ser. P. 12. .. .. An Investigation of the Stability of Double Bellows Assemblies. Proceedings of the Royal Society of London.. J.160 N. BS6129. 6. 1974. 5. P. or from eqn (25). Buckling of double bellows expansion joints under internal pressure. ~--~'- ~u. Journal of Mechanical Engineering Science. D. New York. Snedden e x t r e m e care should be taken in establishing this function so as not to underestimate the lateral deflection at which instability initially occurs. u. 1975. Fitzgibbon. G. Department of Employment.. . .1/N18. . Experimental measurements of the stiffness of a bellows system. Webster. . E. 6(3) (1964). A. 8. .~^v"~. 9.. H o w e v e r . Journal of Applied Mechanics. 351(1667) (1976). Haringx. Newland. 7(3) (1952). Seide. 4. 3. HMSO. . 7. Report to the Flixborough Court of Inquiry. If simple lateral supports are e m p l o y e d these should be capable of sustaining at least the destabilising force due to pressure predicted from eqn (24). E.~'". American National Standard Nuclear Safety Criteria for the Design of Stationary Pressurised Water Reactor P/ants.. London. 2. 27(3) (1960). . White Plains. and-theoretical w o r k should enable the recommendations to be substantiated and developed. 10. A. Fifth Edition 1980. Philips Research Report.

F. 14. Snedden.3) 1[ (1+p2)(1-p) 3 ] E = -~ 1-p2+(l+p2)lnpJ (A1. Bellows spring rate for seven typical convolution shapes. Y. N. Bulletin of the JSME. Laterally loaded elastic shells of revolution. Wan. Bending deform- ation of U-shaped bellows.p2 + 2p In p ) (1 -. D. (A1.7) ro(1 . Matheny. K. lngenieur- Archiv.Ex(h3r 3 E1 = zr(1 .p2)(1 -. PhD thesis.p4 + 4p21np) x = 4(1 .5) ri p ...2x (A1. M. 15. 14(71) (1971). Haringx 1derived the following formulae for the bending stiffness of a bellows ~. J. Nakagawa. Lateral stiffness of bellows expansion joints 161 13.. K. 42(4) (1973). (A1.. 16.6) ro X = 1 + 2c[(1 -p2+2p21np)2ho+p(1 -p2+21np)2h~] (A1.r33 (A1.p2 _ 2p In p ) (A 1.p2) [(1 . M. University of Cambridge. 1981. and Nakade. Machine Design.4) (1 . Hamada. The strength and stability of corrugated bellows expansion joints.p2)2 _ 4p2(ln p)2] .1) where T 7' .v2)n(ro. APPENDIX 1 Based u p o n the theory of symmetrical bending of fiat circular plates and circular cylindrical shells..2) X ~2~ ~". W.. 34(1) (1962). K. Miyata.

2. S n e d d e n sinh 2/xc + sin 2/xc k= (A1. W.022 in Number of convolutions 16 Number of plies 1 TABLE A. 1 and A.C-4n (A1.2 Properties of Type 321 Stainless Steel % Reduction in thickness 0 12.3 76-4 Young's modulus (lbf/sq in) 28-0 E6 Poisson's ratio 0-3 aApproximate value of thinning at root and crown of convolution.2a 21. .9) APPENDIX 2 The bellows test specimen employed to obtain the results presented in Figs 6 and 7 was manufactured from type 321 stainless steel.8) 2/xc(cosh 2/~c + cos 2/xc) I-t = 4 . respectively.386 in Outside radius 7.680 in Wall thickness 0.136 in Convoluted length 7.162 N. TABLE A. Data under this column have been applied in the large deflection analysis.9 50-0 0-2% Proof stress (tonf/sq in) 13.1 Bellows Dimensions Inside radius 6. Nominal dimensions of the bellows and mechanical properties of the material for a range of values of work hardening up to 50% reduction in section thickness are given. in Tables A.9 29-9 42-3 73-8 Ultimate tensile strength (tonf/sq in) 37-2 43-6 51.