]\gt ; V. On the General Theory of Elastic Stability .
By R. y. SOUTHWELL , B.A. , Fellow of Trinity College , Cmbridge .
by Prof. A. .
H. LovE , F.R.S. Received January 4 , \mdash ; Read January 30 , 1913 .
CONTENTS .
Page Introduction and summary of paper . . . . .
187 EQUATIONS NEUTRAL EQuILIBRIuM IN RECTANGULAR CO-ORD1NATES .
Method of derivation . . . . .
190 Example in rectangular co-ordinates .
Stability of thin plating under edge thrust . . . .
198 EQUATIONS OF NEUTRAL EQUILIBRIUM IN CYLINDRICAL CO-ORD1NATES .
Derivation of the equations . . . . . .
202 Examples in cylindrical co-ordinates .
Stability of boiler flues and tubular struts 208 Solution for boiler flue without end thrust . . . . .
217 Comparison with experimental results . . . . .
:222 Validity of investigation by the theory of thin shells .
224 Comparison with existing formulae . . . . .
225 The " " critical length\ldquo ; . . . . .
226 Solution for tubular strut .
Special case . . . . .
227 Validity of investigation by the theo1y of thin shells 230 Solution by the theory of thin shells .
General case 231 Comparison with existing formulae . . . . . .
236 Stability of tubes under combined end and surface pressure 236 GENERAL THEORY OF INSTABILITY 1N MATERIALS OF FINITE STRENGTH .
The practical value of a theory of instability .
237 Stability of short struts . . . . .
242 Need for further research .
Conclusion 244 Introduction and Summary of Paper .
PROBLEMS which deal with the stability of bodies in equilibrium under stress are so distinct from the ordinary applications of the theory of elasticity that it is legitimate to regard them as formin a special branch of the subject .
In every other case we ( 501 .
) separately , August 6 , 1913 .
188 R. .
SOUTff ON THF THEORY OF LASTIC STABILITY .
are concerned with the integration of certain differential equations , fundamentally the all problems , and of certain dary conditions ; and due to we are entitled to assume that any solution which is ique .
these problelns we are confi .
OIlted with the possibility of two or of equilibrium , and Y have to determine conditions be tisfied i that the equilibrium of any given configuration may be stable .
The development .
both ches has proceeded upon similar lines .
That is to sity , the earliest discussions vere concerned with the solution of isolated examples the tion of leral ideas .
In case of elastic stability , prehel lsive 1 not propounded until the problem of the straight strut had been that of the circular under radial pressure by G- .
H. , S and A. G. had discussed the stability of in under its own under twisting couples , and when rotatiug .
fr yhicl become the foundation of the theory in its existing form , G. H. brought these isolated problems for the first time within the range of a single lization .
the conditions which KIRCHHOFF'S theorem of fail , he led to the conclusion that instability is only in casG of such bodies as thin rods , plates , or shells , and in these only when types of tion call occur which do not involve extension of the central line or middle surface , so that it is legitimate to discuss any problem in elastic stability by methods which h been devised for the approximate treatment of such bodies .
He showed .
eover , that the stability of the equilibrium of any given configuration depends upon the condition that the potential shall be a minimum in that collfigul.ation .
A closer examination of BRYAN'S theory suggests that some of the conclusions which have been drawn from it are scarcely warranted .
The contention that no closed shell fail by instability , because any distortion would involve extension of the middle , will be discussed later .
For our present purpose it is sufficient to remark that the whole theory is based upon the assumption that the strains occurring previously to collapse must be kept to the extremely narrow limits within which , in the case of ordinary materials , HOOKE'S Law is satisfied .
This assumption , of course , xpresses a restriction necessarily imposed upon the ran ge of practical A. E. H. LOVEi , 'AIathematical Theory of sticity ' ( second editiolt ) , S118 .
'Hist .
Acad. Berlin , ' XIII .
( 1757 ) , p. 2.52 .
'LIOUVILLE'S X. ( 1884 ) , p. 5 .
S 'Comptes endus , ' XCVIII .
( 1884 ) , p. 422 .
'Proc . .
Phil. Soc , p. 65 .
'Proc .
Inst. Mech. ' 1883 , p. 182 .
'Proc .
Camb .
Phil. Soc VI .
( 1888 ) , p. 199 .
Cf .
pp. 222 , 236 .
MR. R. V. SOUTIIWELL ON THE GENERAL THEORY OF LASTIC STABILITY .
189 problems which can be treated by the ordinary theory of elasticity ; ) it is not to conclude that instability is only possible , evell if its conditions Ollly calculable , the of which obey , and there warrant for the employment of " " , formulae\ldquo ; in the desigll of short struts itnd thick boiler A more serious in existing theory of elastic stability , regarded from the lmathelnatical spoint , is the thab the it ellployt3 are admittedly only approximate .
tho elastic of the under consideration , the less adequate these lethods to deal with the whole allge of problems come its scope .
ct , we with the anomaly that , while its ordinary ( pplications the theory of elasticity is ) collcerned with the ception of all stic limit , in questions of stability th existence of finite limits is essential condition of its results .
1 an ideal material , possessing elasticity with unlimited strength , types of instability could occur with which existing luethods would be quite insufficient to deal .
The theol.y of elastic stability is thus in much position ils that of the ordinary theory of elasticity before the of ) , nd oue aim of the present paper is to remedy its defects by the ation of general equations , which may be ter1ned " " quatiol l of Neutral quilibrium and which express the condition that a given ) may be one of limiting ] uilibrium .
These equations are universally applicable only to ideal material of indefinite and the possibility of elastic break-down must receive separate ation ; but they are also applicable , even with of finite strength , to alty which comes within the restrictions imposed by BIIYAN'S discussion , and thel.efore ellable us to test the accuracy of his tl.eatment of probleIns , such as of boilel flue , for which the ordinary Theory of Thin Shells has been In every problem of this paper it is found that of Thin Shells gives a solution which is correct as a first approximation , the practical ) of the new method of investigation are , , not immediately apparent .
But it must be remembered that the approximate eory of thin plates and yet been rigorously established , and that work has recently been undertaken with the object of testing it by mparisoll with accurate solutions of isolated problems .
S Now in finding conditions for the neutrality of the equilibrium of any given configuration we are at the same time obtainin the solution of a statical problem ; for configuration of distortion from the equilibrium position will also be one * W. C. UNWIN , 'Elements of chine I ' ( 1909 ) , I. , ; S. E. , ' The Coll , of Tubes under External Prcssurc ' ' January 8 , 1909 .
By " " elastic limit\ldquo ; is intellded , hcre and throughout this paper , the limit of linear sticity .
Cf .
pp. 210 , 224 .
S LOVE , op .
, Introduction , p. 29 , and Chapter XXII .
190 .
R. .
SOUTHWELL ON GENFRAL THEORy OF ELASTIC STABILITY .
of equilibrium .
Hence every which we obtain will add to the number of these " " test \ldquo ; which has not hitherto included solutions for any but plane plates .
A more important , of the new method , from the practical point of view , is the accuracy with which it follows the actual ' stress history\ldquo ; in a body which fails by instability under a gradually increasing stress .
In cases where precedes elastic break-down this difference of method is not important ; but the discussion of instability in overstrained material , where the stress-strain relations tely ddent upon the previous stress history , its introduction is absolutely The extellsio of theory to struts of practical dimensions and materials , which forms the conclusion of this paper , gests a and new field for .
The ] umber of similar cases which call be treated , in the existing state of our of plastic strain , is very small , and indications are given below of the questions which still require an answer ; there is reason to believe that the requisite expel.imentalreseal.ch would not present insuperable difficulties , and that we hope in the future to obtain an adequate theory of experimental results which are at present very little understood .
EQUATIONS OF NEUTRAL IN RECTANGULAR CO-ORDINATES .
VIethod of Derivation .
The question of stability arises in regard to any system in which there is a possibility of slight displacement the collfiguration of equilibrium .
This possibility may be afforded either by a more or less limited degree of mechanical freedom\mdash ; in which case the problem is one of statical stability , and practically unaffected by Fig. 1 .
the tende1lcy , which any actual body displays , to distort under the influence of applied forces ; or it may be due , more or less entirely , to this tendency .
In the latter case the problem is one of elastic stability , and must be treated by distinct methods .
here is , however , no essential difference between the two types of MR. R. .
SOUTHWELL ON THE THEORY OF STIC STABILITY .
191 instability , and a general discussion of the elastic type may be very conveniently illustrated by eference to a mechanical example .
In ] this connection we may consider the system illustrated by fig. 1 , in which a uniform heavy sphere rests in ] ibrium within a hemispberical , under the action of its own weight and of the pressure exerted by a pointed , which is free to move in a vertical line through the centre of the .
This syst chosen for the illustration which it affords of collapse under a definite " " critical loading In this it bears an unusual resemblance to examples of elastic instability\mdash ; the stability of most mechanical systems beinob dependent solely upon the relative dimensions of their melnbers .
In the absence of friction , we find that the equilibrium will become unstable as the load on the plunger is increased through a Cl.itical value by , . . .
( 1 ) where is the weight of spbere , .
is radius of ) and is the radius of bowl .
The above solution rests upon the assumption that th , bowl and plunger are absolutely smooth and rigid , and the possibility of slight displacument is afforded by the freedom of the sphere to take up ally position of contact with the bowl .
To discuss the equilibrium of the sphere in the position illustrated we must consider the forces which act upon it in a position of slight displacement .
These include two systems , one tending to restore the initial conditions , the other tending to increase the distortion , and stability depends upon the relative magnitude of the two effects .
We may investigate the problem by three methods , fundamentally equivalent , which are described below:\mdash ; ( 1 ) The Energy Method.\mdash ; We may derive expressions for the potential energy of the system in a position of slight displacement from the position .
The condition of stability requires that the expression for the potential energy shall have a minimum value in the equilibrium position .
( 2 ) The Method of Vibrations.\mdash ; We assume that the displacement ] been effected by any cause , the types of possible to system when this cause is removed .
The condition of stability ] es all such types shall ( 3 ) Statical confine our attention to the in which the stability the position is neutral .
, there exist some type of mGnt for which and effects , discussed above , are exactly need , it intained by the system applied forces .
, have , , find conditions the equilibrium of of small given system ) applied rees .
192 .
R. .
SOUTHWELL ON { GENERAL OF ELASTIC STABILITy .
Any of these methods is valid for the investigation of elastic stability , and all have in fact been employed , the displacel1lent considered being that of the central-line or lmiddle-surface of the rod or shell , and the resultant actions over cross-sections being derived in terlns of this displacement , by the approximate theory first suggested by KIRCHHOFF .
The third method is generally to be preferable , and is the basis of to be described below , but the actual procedure will be found to possess one or two novel features .
the first place , an endeavour will be made to dispense with the assumption that elastic break-down occurs at very small values of the strains ; instead , we shall deal with an ideal material possessing perfect elasticity combined with unlimited strength .
Such a material not by instability , and our )lems will no longer be ined to rods , , or shells .
It follows that we can only obtain sufficient accuracy in our conditions for neutral stability by deriving them with reference to a volur -cler aent of , material .
Further , since instability will in some cases not occnr until the strains in the material reached , values , we shall have to introduce an unusual precision into our ideas of stress and ) .
discussi o of is merely a problem kinenlatics , has out with some completenes ; but the correrelations in our ideal material are necessarily less certain , since they must be based upon } iments in which only small strains are permissible .
For example , if we assume that HOOKE'S Law is satisfied at all stresses , we must decide whether our ition of stress is to be .
or .
For the ordinary purposes of elastic the two definitions may be regarded as equivalent , and the distinction is too fine to be settled experimentally .
In the absence of any generally-accepted lmolecular theory which migbt indicate the correct result , it seems ate to the simplest possible assumptions which do not involve self-c dictions , and yield the usual results when the strains are very snnall .
It may be shown that in a distortion of any nitude three orthogonal linear elements issue from any after distortion , which were also orthogonal in the unstrained configuration , and that these linear elements stationary or minimnm or ) extension .
, an elelnentary parallelopiped constructed at the point , with sides parallel to these linear elements , undergoes change of angle in the distortion .
It is clear that only lormal stresses will act upon its faces * For a discussion of the theory , with references , see LOVE , .
cit. , Appendix to Chapter I. LOVE , , SS26 , 27 .
MR. R. V. SOUTHWELL ON THE GENERAL OF ELASTIC STABILITy .
193 after , and that if these stresses be expressed terlns of the extensions of the sides we have complete relations between stress and strain .
We shall therefore assume that these stresses and ncipal whatever thei , .
magnitude , are connected by the ordinary equations of HOOKE'S ; that is to say , if the extensions in the principal directions are , and the corresponding stresses are then \amp ; c. , where is YOUNG'S Modulus , and is POISSON'S ratio for the material under consideration .
These relations may be written the form \amp ; .
, ( 2 ) where is the Modulus of Rigidity .
In these relations the measure of sion is assumed to be and of stress* We have then the usual expressionf for the energy of strain , per unit volume of the unstrained material , in terms of the principal extensions , vi :\mdash ; .
( 8 ) The above assumptions yield sufficient data for the calculation of the stress system in any configuration of equilibrium , even when the strain are not small .
Assuming that the calculation has been effected , we have to show how conditions for the stability of the system may be obtained .
We must distinguish three configurations : the unstrained configuration , in the co-ordinates of any point are given by ; the configuration of equilibriuln under the stress-system , the bility of which we are investigating ; and a This assumption is opell to the that it re1lder possible the compression of material to zero volume by means of a finite stress .
It will not , howevcr , introduce aIly serious error , and has the advantage , which more probable assumptions do not possess , of leading to a definite energy-function .
The definitions of stress and given above are generally employed in the construction of ' ' strain diagrams\ldquo ; from a tension test , the extensions of the specimen being takeIl abscissae , and the total loads as ordinates of the plotted curve , LOVE , , S68 .
VOL. CCXIII . .
2 I 94 .
SOUTHWELL ON THE GENERAL OF ELASTIC STABILITY .
1 tion of slight distortion from the equilibrium position , which can be maintained without the introduction of additional stress at the boundaries , if the equilibrium of the second configuration is neutral .
We shall consider first a stress-system which is such that the principal stresses in the second configuration have the same magnitudes and directions throughout the body ; and we shall take these directions as axes of and .
We Inay then define the second and third configurations by saying that in thelll the co-ordinates of the point become , respectively .
We shall not limit the values of , although in practical cases must be small : ) , are infinitesimal .
In the second configuration the axes Ox , , Oz are directions of principal stress , and the stresses are \amp ; c. ( 2 ) bis In the third figuration we shall find that lines which in the first configuration were slightly inclined to Ox , Oy , Oz become directions of principal stress and strain .
The final extension of a line which originally had direction-cosines is .
( 4 ) It may be shown that has a value when ( 5 ) and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and to terms of the first order in ?
*In some cases , problem of the ility of a heavy vertical rod ( p. 188 , footnote ) , it is necessary to allow for variation in one or more of the stresses ; the necessary alterations are easily made , and as they are nob required for the examples of this paper their consideration would involve unnecessary ) plexity .
[ A May 1.\mdash ; The approximation of these expressions is insufficient if any two of the principal strains in the second configuration are equal ; in this oase additional terms must be retained in the denominators .
The equilibrium under ydrostatic stress is necessarily and obviously MR. R. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
195 Thus the line initially given by the direction-cosines becomes a direction of principal stress in the final configuration .
Its direction-cosines ( referred to Ox , Oy , and Oz ) are then or , ( 6 ) which we shall write as 1 , ; and its final extension , to terms of the first order in , ?
, is ( 7 ) In the same way we find that the other directions of principal strain in the final configuration are given by the direction-cosines and ( 8 ) 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' and 1 , ' which write , and that the final in these directions are and . . . . . . . . .
( 9 ) The stresses in these directions , which we shall call the directions of , and referred to the original areas of the faces on which they act , therefore ' , \amp ; c. ( 10 ) *Cf .
the assumption of p. 193 .
Referred to the new areas of the faces on which they act , they are and to the required degree of approximation we may write , \amp ; .
( 11 ) Then if denote the co-ordinates in the final configuration , referred to the original axes , of the point which was originally at , so that , , \amp ; .
, we may find the stress components in the third configuration , referred to the original axes , and to the strained areas of the faces upon which they act , by the scheme of transformation The following expressions are thus obtained ( to the required order of approxima , ( 12 ) Now the stress-components ( 12 ) must satisfy the ordinary equations of equilibrium , which are three of the type , ( 13 ) MR V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
197 and since the co-ordinates of the )oint which ultimately goes to were originally ' we have , , \amp ; c. It follows that ( 13 ) may be written ( to our approximation ) as follows:\mdash ; .
( 14 ) Substituting for , we have finally .
( 15 ) and two similar equations .
In any ordinary problem we may neglect in comparison with The equations thus obtained may also be written ( with LAME'S notation for the elastic constants ) as follows:\mdash ; \amp ; c. , ( 16 ) where and in this form they may be conveniently compared with the ordinary equations of elasticity.ae The three equations of the type ( 15 ) we shall term Equations of Neutral The equilibrium of the stress-system will be neutral , provided that solutions for ?
exist which satisfy certain boundary conditions .
These boundary conditions are peculiar to each problem , but usually express the condition that the additional stresses involved by shall vanish on certain boundary surfaces .
never the of ?
, so that our solution gives the form only of the tion which tends to occur in the body under consideration when its equilibriunl becomes unstable .
It gives a defi11ite veen the stress-system and the dimellsions of the body , which must be satisfied in order that any distortion may be permallent ; but if this relation be satisfied , no limits are imposed by the equations upon the magnitude of the distortion which may occur .
xample iRectangular Co-ordinates .
Stability of Thin Plating unde ?
Edge Thrust .
It seems advisable , before we employ a new method on problems which have not as yet received satisfactory treatment , in some degree to test its validity by the result to which it leads in a more familiar example .
For this purpose we may consider the stability of an infinite strip of flat plating under edge thrusts in its plane .
The accepted formula for the thrust necessary to produce instability , per unit length of edge , is , . . . . .
( 17 ) where thickness of plate , breadth of plate , and the opposite edges are simply supported .
If the edges are built in , the thrust required has four times this value .
To investigate this problem by the new method we take axes Ox and Oz in the middle surface of the plate , in the direction of its breadth and length respectively , and Oy perpendicular to the middle surface .
The initial stress-system is then given by const .
( 18 ) ; LoyE , op .
cit. , S91 , equation ( 19 ) .
The equations are , however , rigorous only in the case of displacements ; cf. footnote , p. 240 .
LOVE , , S337 , whence the above be MR. R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
199 and we may assume that the system of strain which is introduced at collapse will be two-dimensional , so that const . . . . . . . .
( 19 ) The third equation of neutral stability ( for the direction Oz ) is then satisfied identically , and the other two equations beco1ne we neglect terms of order and ( 20 ) Let us assume a solution of the form , ( 21 ) where and are functions of only .
It is easy to show that this assumption as to the phase-relation of and is justified .
We have then and ( 22 ) The solution of these equations is of the form ( 23 ) where are constants .
The boundary conditions now demand attention .
It is clear that the stresses introduced by must vanish at the surfaces of the plate .
Hence these surfaces will still be planes of principal stress , and , moreover , the normal stress upon them must vanish .
But , as we have already seen , the line which becomes a direction Oy ' of principal stress has initially the direction-cosines : it follows that at the surfaces of the plate the expression for and must vanish identically ; moreover , at these surfaces , must vanish .
These conditions may be written in the form identically , when . . .
( 24 ) The first condition is already satisfied .
The other two give we neglect terms of order and , when , ( 25 ) or a ay sin , . . . . . . . . .
( 26 ) and ay ay sin . . . .
( 27 ) MR. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
201 Thus we obtain ( 28 ) -at There are two solutions of the equations ( 28 ) .
Either at , and , ( 29 ) or and . . . . . .
( 30 ) The criterion for neutral stability is in the first case , and in the second case , VOL. CCXIII .
so that the values of , for which collapse by instability may be expected to occur , are given by ( 31 ) and ( 32 ) respectively the total thrust , per unit length of edge , being . . . .
( 33 ) The first nationS to a solution , in terms of , are ( 34 ) and ( 35 ) respectively .
Since the complete wave-length of the corrugations into which the plate distorts is , . . . . . . . . . .
( 36 ) we see that ( 34 ) is equivalent to ( 17 ) , and that the latter formula is therefore supported by our investigation as a first cition .
The second solution ( 35 ) is without practical interest , owing to the magnitude of the thrust required to produce collapse .
It refers to a type of distortion , theoretically possible for an ideal material without limits of elasticity , which oximately realized in actual specimens of ductile material , when tested to failure under compressive stress .
Since we see from ( 23 ) that in this type the middle surface .
In the first type of failure , where , we find that when , so that the iddle ' of extension in the distortion given by EQUATIONS OF rRAL EQUILIBRIUM IN YLINDRICAL Derie atio of the Equations .
The equations ( 15 ) of neutral equilibrium are expressed in a form which is unsuitable for the investigation of problems concerned with the stability of thin tubes , and we have next to obtain the corresponding equations in cylindrical * Besides the harmonic solutions to we may have , and vanish in virtue of the boundary conditions .
MR. R. .
SOUTHWELL ON THE GENERAL THEORY OF LASTIC STABILITY .
203 co-ordinates .
We shall limit our discussion to stress-systems which produce a displacement symmetrical about an axis , up to the instant at which the equilibrium becomes unstable and distortion occurs : in PEARSON'S notation , the principal stresses in the equilibrium configuration are , and and quantities are functions of only .
The new equations are derived by a method very similar to that which has already been explained .
The co-ordinates of a point in the unstrained configuration are ? .
, ; in the second configuration ( of equilibriuln ) they are and in the third configuration ( of slight distortion from the position of equilibrium ) they are ( the radial , tangential , and axial displacements bein ultimately taken as infinitesimal ) .
The extension of a line-element joining the point to the point is , ( 37 ) where , and .
; and this has a stationary value line very slightly inclined to the radius , by . . . . .
where are written , and respectively .
The extension of this line-element in the third configuration is , ( 39 ) and its inclination to lines issuing from the point ( in its final position ) in the radial , tangential , and axial directions is given by the direction cosines where and ( 40 ) where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and where and We find also that the other directions of principal stress in the final configuration are initially inclined to radial , tangential , and axial lines through , z ) at angles whose direction cosines are and , and that in the third configuration they have corresponding inclinations to radial , tangential , and axial lines through \mdash ; in its final position\mdash ; which are given by and , where and ' ( 41 ) The extensions of the corresponding line-elements in the final configuration are respectively and . . . . .
( 42 ) MR. R. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
205 so that the principal stresses in the third configuration , referred to the final areas of the faces on which they act , are ( 43 ) Then by the scheme of transformation we find for the stress components in the third configuration , at the poin which was initially at \mdash ; referred to axes in the radial , tangential , and axial directions through the final position of the point , and to the areas of the faces upon which they act\mdash ; the expressions ( to the required approximation ) , \amp ; c. . . . . . . . . .
( 44 ) 206 MR. .
SOUTHWELL ON THE GENERAL THEORy OF ELASTIC STABILITY .
The equations of equilibrium , to be satisfied by the stress components ( 44 ) , are* where and Moreover , since and the equations ( 45 ) may be expressed in differentials with respect to .
The terms which do not inyolve vanish in virtue of the equilibrium conditions for the second configuration , and only terms of the first order in these quantities need be retained .
In general , , 2 : } may all be functions of , but in this paper shall only consider problems in which , and have constant values .
Moreover , in all problems of practical importance we may terms of order in comparison with , S59( ) .
MR. R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
207 terms of order , We then obtain , as the equations of neutral equilibrium in cylindrical co-ordinates , , . . .
( 46 ) ( 47 ) and .
( 48 ) Equations ( 46-48 ) represent the conditions neutral stability in the of a body subjected to a stress-system , where is constant , and and are metions of only , which satisfy the condition of equilibrium .
( 49 ) For comparison with the ordinary equations of equilibrium in cylindrical co-ordinates they may be written ( with LAME'S notation for the elastic constants ) in the forms* , ( 50 ) and where *Cf .
Lovll , .
cit. , S 199 .
ON LASTIC TABILITY .
-o 1r tion s of Equilibrium in Cylindrical Co-ordinates enable us to deal fnlly some difficult connected with the stability of cylindrical tubes .
) of considel.able importance will ) discussed in this paper\mdash ; the collapse of ) oiler flues the strength of tubular struts .
It sbould be noticed that neither of been quite tisfactorily treated by the ordinary theory ofells , the ) that the middle surface of the shell is u1lextended , and outer from applied ; hence their is a , even apart practical considerations , has cted a deal 01 .
It will be convenient at this point to vork been question of stability of struts is to the frequency .
their elmployment 11 practice .
economy of matel.ial the cylindrical tube all over truts of solid cross-section , and both the theory of thS .
, CLXXXI .
, AYLEIGH , 'London Math. I , 1 .
MR. .
SOUTHWFLL ON THE GENERAL THEORy OF ELASTIC STABILITY .
209 EULER* and LAGRANGE and the more practical formula of RANKINE suggest that this advantage increases without limit as the thickness of the tube is reduced .
Such a conclusion is , however , inaccurate , for types of distortion are possible in the case of a tube which do not involve flexure of the axis , and when the tube is thin these types , of which some practical examples are shown in fig. 2 , may be maintained by a smaller thrust than would be required to produce failure of the kin discussed by EULER .
Moreover , the natural wave-length , for these symmetrical types of distortiol } , is in general small , so that distortion can occur without hindrance quite short tubes .
Hence , for a considerable rallge of length the strength of a tube to resist end-thrust is practically constant , and is not given by any of the usual formulae for struts .
The determination of the strength of tubes to resist these sylnmelrical types of distortion is obviously a problem of the highest practical importance , and attracted a great deal of attention in recent years .
Illustrations of collapsed tubes , showing symmetrical types of distortion , have been published by A. MALLOCK and R. LORENZ , S and a great deal of experimental work has been carried out by W. E. LILLY Theoretical discussions , by approximate methods , have been proposed by A. GROS , W. E. LILLy , .
TIMOSCHENKO and R. LORENZ .
The problem of the boiler flue seems first to have been suggested by the ments commenced by FAIRBAIRN in 1858.SS These showed that the collapse of tubes under external pressure was in some degree analogous to thab of straight columlls under end-thrust , and a discussion of the pheno1nenon , based on EULER'S theory of struts , was given by W. C. UNWIN who assisted FAIRBAIRN in his research .
The similar problem of a circular wire rin subjected to radial pressure has been discussed by M. and M. and rational theories of the boilerflue problem have been given by G. H. BRYAN , A. FoPPL , P. " " Sir la force des colonnes ' Hist. Acad. Berlin , ' XIII .
, ) " " Sir la figure des colonnes ' Miscellanea Taurinensia , ' .
( 1773 ) .
'Roy .
Soc. Proc , LXXXI .
( 1908 ) , p. 389 .
S 'Physikalische Zeitschrift , ' XII .
( 1911 ) , p. 241 .
'Proc .
Inst. Mech. Eng 1905 ; 'Inst .
Civ .
Eng. Irelalld , ' 1906 ; gineering , , 1908 .
' Comptes Rendus , ' CXXXIV .
( 1902 ) , p. 1041 .
'Inst .
Civ .
Eng. Ireland , ' 1906 .
'Zeitschrift .
Mathematik .
Physik , ' LVIII .
( 1910 ) , p. 337 .
'Zeitschrift des Vereines Deutscher Ingenieurc , October 24 , 1908 ; ische Z ( 1911 ) , p. 241 .
SS 'Phil .
Trans. Roy .
Soc CXLVIII .
, p. 389 .
'Proc .
Inst. C.E. , ' XLVI .
( 1875 ) , p. 225 .
' Course de Me'canique ppliquee , ' 1 .
Partic , Paris , 's Jourllal , ' X. ( 1884 ) , p. 5 .
'Proc .
Camb .
Phil. Soc .
( 1888 ) , p. 287 .
'Resistance des ' 1901 ) , p. 286 .
SSS 'Zeitschrift des Oesterreichischen Ingenieur- u1ld Architekten-Vercincs , ' 1904 .
VOL. CCXIII.\mdash ; A. 2 and R. LORENZ .
* W. E. LILLY has ndicated the correct form of the result for an infinitely long flue , and A. E. H. has discussed the strengthening effects of constraints which keep the tube circular at its ends .
A. B. has given a very clear exposition of the difficulties which are encountered an attempt to construct a theory of flue collapse by usual methods .
To sufficient equations we must assume that the middle surface undergoes no extension ; and the existence of pressure on one or both surfaces of the tube not only makes this assumption very improbable , but violates an essential condition upon which theory of shells is based .
When one surface only is subjected to pressure , there is reason to believe that BRYAN'S solution is substantially correct ; but no treatment can be looked upon as rigorous which neglects the cross-stresses in the material .
The experimental researches of A. P. and R. T. have revived interest in this problem , since they offer the first information which has been obtained as to behaviour under practical conditions of tubes which in circularity , uniformity of thickness and eneity are fair approximations to the ideal tube of theoretical analysis .
We commence our discussion by considering the stability of a thin cylindrical tube , subjected to the combined action of end and surface pressures .
We shall thus be able to derive the required solutions for the thin tubular strut , and for a boiler flue without end-thrust , as particular cases , and from the general solution we may obtain indications of the way in which end-thrust tends to promote the collapse of a boiler flue .
In the most general form of the boiler-flue problem , as enunciated by BASSET , ff pressures are on surfaces of the tube , and we shall therefore investigate collditions for neutral stability in a ) subjected to the following system of stresses :\mdash ; An end-thrust of total amount @ , distributed ; ( ii .
) An external hydrostatic pressure , of intensity ; and ( iii .
) An internal hydrostatic pressure , of intensity ' Physikalische Zoitschrift , ' XII .
( 1911 ) , p. 241 .
'Inst .
Civ .
Eng. ' 1910 .
'Proc .
Lond. Math. Soc XXIV .
( 1893 ) , p. S 'Phil .
hTag XXXIV .
( 1892 ) , p. 221 .
" " Resistance of Tubes to Collapse ' Bulletin of the Univ. of Illinois , ' No. 17 , 1906 .
' ' Collapsing Prcssures of emcr Steel -Welded .
American Soc. Mech. Lng 1906 , p. 730 .
The experiments of FAIRBAIRN were restricted to tubes hich were constructed from sheet metal , with brazed and riveted sennls .
, p. 223 .
MR. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
211 We shall consider a tube of indefinite length , of which the inner and outer radii are ( so that the thickness is ) , and we shall write for the ratio The corresponding stress-system , for the position of equilibrium , is easily obtained .
* We have . . .
( 51 ) and It can also be shown that is constant , and equations ( 46-48 ) may therefore be taken to express the conditions of neutral equilibrium .
The degree of approximation to which these equations have been obtained ( p. 206 ) will be maintained for the rest of this paper , i. e. , terms of order will be neglected .
They may also be written as follows:\mdash ; , ( 52 ) , ( 53 ) and , ( 54 ) LOVE , ' Mathematical Theory of Elasticity ' ( 2nd editio1l ) , S100 .
212 R. .
SOUTHWELL ON THE GENERAL THEOR OF ELASTIC STABILITy .
where ( 55 ) and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and We may assume a solution for equations ( 52-54 ) of the form* , ( 56 ) where must be integral , and are functions of only , which satisfy the differential equations , ( 57 ) , . . . . .
( 58 ) and . .
( 59 ) It is easy to show that the phase-relations assumed in equations ( 56 ) are necessary .
MR. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
] The boundary conditions now require investigation .
From the consideration that the cylindrical boundary surfaces of the tnbe must continue to be tangent to principal planes of stress , in any possible type of distortion , we deduce the conditions , identically , when .
( 60 ) The other boundary conditions are more complex .
Since the pressures acting on the surfaces of the tube are hydrostatic , it is clear that the stress , as defined on p. 193 , is increased at points on the surfaces of the tube where the distortion involves positive extension .
In the notation employed above , we have , when , when and from ( 43 ) we deduce the following equations , which must be satisfied , when .
( 61 ) Substituting from ( 56 ) in the identities ( 60 ) and ( 61 ) , we finally obtain , as the required boundary conditions in ' and ( 62 ) and and and and and and and and and and and and and and and and when * In obtaining these equations it should be noticed that before distortion occurs and - are the values at the boundary of ' and not of , if we retain the significance for which was assumed on p. 193 .
Ths distinction is not really needed for the approximation of the following work , but it may lead to confusion if neglected .
The differential relations ) , with the boundary conditions ( 62 ) , are theoretically sufficient for an exact solution of our problem : we shall , however , content ourselves with approximate solutions for and @ , correct to terms in .
To obtain these , we assume solutions for , in series of ascending powers of the quantity .
Thus we write , ( 63 ) where We may now derive , from equations ( 57-59 ) , any required number of relations between the undetermined coefficients , and the boundary conditions ( 62 ) take the form of equations in series of ascending powers of the small quantity , in which the sums of the odd and of the even powers must vanish separately .
If we neglect in these equations terms of order higher than some definite power , we may obtain corresponding approximations to the values of A and , by the elimination of the ulldetermined coefficients .
The approximate boundary conditions , correct to terms in , .
( 64 ) , ( 65 ) , .
( 66 ) , ( 67 ) , .
( 68 ) * In deriving these boundary conditions it is to be noticed that is to a first approximation equa to , so that to our approximation may be written for MR. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
215 and and these equations involve the fifteen coefficients From equations ( 57-59 ) we may obtain nine other relations between these coefficients , as follows:\mdash ; .
( 70 ) , ( 71 ) , ( 72 ) , . . . .
( 73 ) , ( 74 ) , . . .
( 75 ) , ( 76 ) , . . . . .
( 77 ) MR. R. .
SOUTHWELL ON THE GENERAL THEORy OF ELASTIC STABILITY .
217 .
( 78 ) We may now eliminate the coefficients from equations ( 64-78 ) , and obtain a determinantal equation , of fifteen rows , which gives a relation between and the dimensions of the tube .
This relation is the condition for neutral equilibrium of the initial stress-system , and is clearly correct to terms in ; but by further consideration of the terms involved we may show that the labour which would be required for its complete evaluation is unnecessary , and as the fifteen-row determinant may be written down directly from the above equations it will not be given here .
Solution for Boiler Flue wiihout End Thrust .
We shall begin by deriving a sufficiently approximate expression for the difference of pressure required to produce collapse of a tube , when there is no resultant end thrust or tension ; and in the first case we shall deal with a form of collapse possible only in the case of tubes of infinite length .
That is to say , we make and equal to zero in the fifteen-row determinant , which may then be reduced to one of ten rows .
In the latter determinant we may treat A as a quantity of order ; for if A be put equal to zero , and the determinant be expanded , the terms which are independent of vanish identically .
Hence -A may be written for , and may be neglected .
The ten-row determinant , simplified by these and other obvious modifications , is given on , 218 and 219 .
Expanding it from the top row , with the neglect of terms of order higher than , we obtain whence .
( 79 ) VOL. CCXIII.\mdash ; A. 2 \ldquo ; , \ldquo ; ' , , , MR. R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
219 , , , , , , , 2 , ' , , , , , , , 220 MR. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
But to a first approximation , so that we have , ( 80 ) which agrees with BRYAN'S result .
* To complete our discussion of this problem we must consider types of distortion in which the axial wave-length is finite , and thus obtain a theoretical estimate of the strength of short flues with fixed ends .
A solution giving A correctly to terms in may be derived the complete fifteen-row determinant ; but we may show that for practical purposes the labour which this evaluation would entail is quite unnecessary .
We find first of all that those terms in the expression for A which are independent of contain as a factor .
Now being approximately equal to the mean hoop stress in the tube before collapse , it is clear that A must in all cases of practical importance be a very small quantity .
It follows that in the expanded equation the terms in A are of primary importance , and and higher powers may be neglected ; further , since must also be small , that terms in and higher powers of may be neglected in comparison with terms in , and that of the terms in those which involve are negligible in comparison with the terms already found .
In accordance with these principles we may derive the terms which are required to complete our solution from a nine-row determinant , obtained by omitting terms in from the general determinant .
This simplified determinant is given on pp. 218 and 219 .
Further , we may neglect in the expansion , and in coefficient of retain only those terms which do not involve ; we thus obtain the equation .
( 81 ) But , by equations ( 55 ) , and therefore , to the approximation of equation ( 81 ) , .
( 82 ) *Cf .
footnote , p. 209 .
MR. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
221 Combining this result with ( 80 ) we have , as our final expression for the pressuredifference which can produce collapse of the flue , .
( 83 ) In this equation is the ratio of the thickness to the diameter of the tube , and is the number of lobes in the distorted form of its cross-section .
The quantity is connected with the axial wave-length of the distortion by the relation . . . . . .
( 84 ) We may imagine a flue subjected at its ends to constraints which merely keep the ends circular , without imposing any other restrictions upon the type of distortion.ae In this case the end conditions may be written in the form wheu , ( 85 ) and from ( 56 ) it is clear that , the length of the flue , is equal to Fig. 3 .
Added June 8.\mdash ; Thin circular discs , inserted into the tube at its ends , but not fixed to it , would approximately realize these conditions .
] In practice , the end constraints will also tend to maintain the cylindrical form at the ends of the flue , and this effect will strengthen the tube , by an amount which is not easy to determine exactly .
In any case we may say that ' and we may strate the way in which the end effects die out by plotting the pressure differences against the quantity .
To do this we must take some definite value of the ratio , and plot different curves for the values 2 , 3 , , \amp ; c. , of .
The result is shown by fig. 3 , in which the following values have been assulned for the constants:\mdash ; pounds per sq .
inch , dynes per sq .
cm .
From an inspection of the different curves we see that long tubes will always tend to collapse into the two-lobed form , since the curve for then gives the least value for the collapsing pressure , but that at a length corresponding to the point A the -lobed distortion becomes natural to the tube , and for shorter lengths still , of which the point gives the upper limit , the four-lobed form requires least pressure for its maintenance .
Thus the true curve connecting pressure and length is the discontinuous curve , shown in the diagram by a thickened line .
Whatever be the relation between and the length of the flue , it is clear that instability is theoretically possible in where the distortion involved is not even approximately " " inextensional For is sufficiently small , the collapsing pressure , as given , need not elastic break-down in the position of equilibrium , even though the first ( or " " extensional term in ( 83 ) be equal to , or even greater than , the second .
Of course , elastic break-down will occur by reason of the extension very soon after the commencement of the distortion .
Nevertheless , failure in such a case must be regarded as due entirely to instability ; for if this source of weakness were removed , effective resistance could be offered for an indefinite period to pressures which actually result in collapse .
Comparison with Results .
Although , as we have just remarked , it is theoretically possible for failure to occur by true elastic instability in comparatively short tubes , yet the relative dimensions of the tubes must be such as it would be quite impossible to test experimentally .
In any practical case , instability will not occur until the properties of the material have been altered by overstrain , and the value of the pressure at collapse is therefore very much less than the foregoing theory would suggest .
MR. R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
223 It is , however , of interest to compare the general shape of the theoretical curve CBAE ( fig. 3 ) with the results of experiment , and fig. 4 has been constructed for this purpose .
It represents a number of tests conducted by the author seamless steel tube ( inches thick and 1 inch in external diameter ) , and shows the } amounts of resistance to external pressure offered by different lengths of tube .
In these experiments ( selected for fig. 4 from a more comprehensive series which is still Fig. 4 .
in progress ) the ends of the tube were gripped by means of slightly conical plugs and sockets , the interior being kept in free connection with the atmosphere , and no attempt was made to balance the axial thrust due to hydrostatic pressure on the plugged end of the tube .
Other experiments have shown hat the existence of this thrust is not seriously important .
It will be seen that the general shape of the theoretical curve is well reproduced , as well as the changes in the number of lobes which characterize the distorted crosssection .
Similar results to those of fig. 4 have been obtained by CARMAN , his experiments were not sufficiently numerous for a satisfactory comparison with the theoretical curve of fig. 3 , his object in conducting them being merely to discover what is the limit of length beyond which the strength of a tube may be taken as *Cf .
footnote , p. 210 .
sensibly the same for all lengths .
The main interest both of CARMAN and of STEWART* was confined to tubes in excess of this limit , experiments on which may fairly be compared with the theoretical formuIa ( 80 ) ; their results showed that this formula gives a satisfactory estimate of the strength of very thin brass and steel tubes , but must not be taken as a basis for design throughout the whole range of dimensions employed in practice .
The experiments of FAIRBAIRN , on the other hand , were restricted to tubes of such relatively small length that he failed to realize the existence of a definite minimum below which the strength of a tube , however long , will not fall .
He also neglected the possibility of discontinuities in the curve of collapsing pressure at points where there is a change in the form of the distorted cross-section .
In the light of these facts , figs. 3 and 4 help to explain his well-known formula , by which the collapsing pressure is given as inversely proportional to the length of the flue ; for a curve of hyperbolic form will represent as well as any other single curve the scattered points of fig. 4 , and trial shows that the hyperbola ( 86 ) is very closely an envelope of the discontinuous curve CBAE in fig. 3 , in each case donvn to the point of least collapsing pressure .
Validity of Investigation by the Theory of Thin Shells .
One important result of our investigation , which is apparently new , is shown by equation ( 83 ) .
It may be seen that collapse is practically dependent upon the pressuredifference alone , and that the absolute values of the pressures are immaterial .
In view of this result , the objections raised by BASSET against BRYAN'S treatment of the problem require further consideration .
These objections are : first , that the ordinary expressions for the stress-couples in a plate or shell , in terms of the curvature of its middle surface , are not valid when the surfaces are subject to pressure ; and secondly , that it is not legitimate to assume , as we must if sufficient equations are to be obtained , that the middle surface is unextended in a configuration of slight distortion .
Hence the theory of thin shells is not applicable to this problem .
The above difficulties may be almost entirely overcome by a change in the method of investigation which is employed .
It is customary to derive equations for the equilibrium of the distorted shell directly , and without reference to the position of equilibrium .
Such procedure renders it necessary to make BRYAN'S assumptions , that the middle surface is unextended , and that the usual expressions for the stress-couples *Cf .
footnoto , p. 210 .
Cf .
footnote , p. 209 .
See footnote , p. 210 .
MR. R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
225 are valid .
But we may also proceed , as in the foregoing discussion , by first determining the stress-system for the equilibrium position , and then equations for an infinitesimal displacement .
The stress-couples which appear in these equations will be due to the adra tional stresses introduced by the distortion , and , to a first , vanish the of the tube , they ooith sufficient by tloe usual expressions .
Moreover , when the distortion is two-dimensional ( as in BASSET'S problem ) , the change in the " " hoop\ldquo ; stress-resultant will be of an order which is negligible , so that the middle surface may be regarded as undergoing no extension to equilibrium position , even though its area may be sensibly changed in comparison with the unstrained configuration .
* The method of investigation just described , which follows the actual sequence of occurrences in the material , is suggested as in every way preferable to methods , for the investigation of any problem in elastic stability .
For the present example , in particular , it leads to the same results as the more rigorous methods this paper .
Comparison with xistiw F Previous discussion of the boiler-flue problem by analytical methods have , exception , dealt with a tube subjected to pressure on one surface only , almost ffi of them have been restricted to the case of an indefinitely long flue .
Their results have , therefore , to be compared with our equation ( 80 ) , when is zero .
It will be found that this equation agrees with the formula obtained by BRYAN and FoPPL'S formula omits the factor , which measures the increased resistance to flexure of a long tube as compared with a circular ring .
The more general formula may be compared with that of LORENZ , if be put equal to zero .
It will be found that there is a serious want of agreement in regard to both terms in the expression ( 83 ) .
In support of the latter result , it may be urged that LORENZ ' solution gives for the indefinitely long flue a result which does not agree with equation ( 80 ) ( and , as we have just noticed , this is supported by previous investigations ) , and which vanishes , not when , but when .
Now the value 1 , in the case of an itely long flue , corresponds to translation of the tube as a whole , without distortion , and the value to a change in the diameter of the tube , without any departure from circularity .
It is clear that the applied pressures can have no tendency to maintain such a form of distortion , so that LORENZ ' formula can hardly be correct .
Added June 8.\mdash ; The arguments of this section are more fully developed in paper by the tuthor " " On the Collapse of Tubes by External published in the 'Philosophical Magazine ' for 1913 ( pp. 687-698 ) .
] Cf .
footnote , p. 209 .
Cf .
footnote , p. 210 .
VOL. .
R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
The " " Critical Length A. E. H. has investigated the rate at which the strengthening effect of circular ends falls off when the length of a boiler flue is increased .
His result suggests that at a distance which is great compared with the quantity the influence of the ends becomes negligible , and the flue collapses under sensibly the same pressure as a tube of infinite length ; hence , in order that " " collapse rings\ldquo ; may any appreciable effect , their distance apart must not exceed some experimentallydeferlnined multiple of this quantity .
The greatest length of tube over which the ends exert any appreciable strengthening influence , or the least length for which collapse is possible under a pressure sensibly equal to the critical pressure , has been called by Prof LOVE the " " critical length It is a conception of great importance in experimental work ; for , as we have seen , tests any length of tube in excess of this limit may be taken to give the strength of an length of the same tube , and their results compared with the theoretical formula ( 80)S : but as a basis for the spacing of " " collapse rings\ldquo ; it is superseded by the theory of this paper , which yields an expression for the greatest length of tube consistent with stability , when the thickness and diameter of the flue , and also the collapsin pressure , are given ; and Prof. LOVE has suggested to the author that it would be better now to employ the term " " critical length\ldquo ; in this more general significance .
As we have seen ( p. 222 ) , the length of the tube is some multiple of the quantity , and we may therefore obtain from ( 83 ) the following formula:\mdash ; length , ( 87 ) where is a constant , depending upon the type of the collapse rin , and has that integral value which gives the least value for the right-hand expression of equation ( 87 ) .
Before this subject is dismissed , it should be noticed that the theory of this paper does not support Prof. estimate , mentioned above , of the rate of decay of end effects .
The term in equation ( 83 ) which depends upon the length of the tube may be regarded as negligible , compared with the constant term , when the ratio * Proc. Lond. Math. Soc XXIV .
( 1893 ) , p. 208 .
'Theory of Elasticity ' ( 2nd edition ) , S337 Page 224 .
S In this sense the term " " critical length\ldquo ; has also been employed by CARMAN , who began his research investigating the strcngthening effects of the GIld plugs with which he sealed his tubes for test .
MR. R. .
SOUTHWELL ON THE GNERAL THEORY OF ELASTIC STABILITY .
227 has some sufficiently small value ; and being inversely proportional to the length of the tube , we deduce for the critical length in the original sense of the term , an equation of the form where is constant .
Prof. LovE , as has been said , has obtained an equation of the form which is very different ; but he has informed the author that in.the light of the above investigation ( pp. 210-222 ) he does not regard his method as adequate .
* Solution for .
Strut : Case .
We may obtain another simplification of the general determinant to ten rows by taking a zero value for .
This corresponds to a type of distortion , possible in the case of a tubular strut , in which the axis remains straight and the cross-sections circular , the diameter varying in a sinusoidal manner .
The ten-row determinant for this case is given on pp. 228 and 229 ; the factor has been cancelled from the sixth column , and terms in A have been omitted , so as to yield a result for tubes collapsed by end pressure alone .
The expansion is only correct to terms of order , and for a first approximation we may also neglect the square and higher powers of , which must be small in any case of practical importance .
Investigating first the terms which are independent of , we obtain .
( 88 ) We may now employ the substitution in the determinant , and expand it from the top row , neglecting terms of higher order than A considerable amount of unnecessary labour may be avoided by a preliminary examination of the relative importance of the various terms involved .
It will be argument in favour of the new formula be from considerations .
The offered by a ) to any given form of distortion is due partly to the Sion and partly to the flexure which such distortion entails ; and it is clear that the rslative importanco of the extensional part increases as the thickness is reduced .
Hence , other things being equal , the effects of the ends , which necessitate extension of the middle surface , are more important in a thin than in a thick tube ; that is to say , bhey are sensible over a greater length .
] MR. R. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
229 \ldquo ; , .
R. .
SOUTHWELL ON THE GENERAL OF ELASTIC STABILITY .
found that contains terms in ) and , as well as terms independent of .
Thus the complete expression for is of the form , and it is clear that has a minimum value when the axial wave-length has a finite value , given by This minimum value , which alone is of practical importance , is given , to a first pproximation in terms , by the equation , ( 89 ) so that the determination of and is not required .
By expansion of the determinant we find and from ( 55 ) we deduce , for the minimum thrust required to produce collapse , .
( 90 ) This expression is correct to terms in Validity of by the Theory of Thin Shells .
A complete investigation of the tubular strut problem must deal with lobed formoe of deformation , since it is possible that one of these may require a smaller end-pressure for its maintenance than the circular form treated above .
We have , therefore , to obtain a general expression for ( when A is zero ) in terms both of and The derivation of this expression , if we employ the rigorous methods of the present paper , will entail nothing less than the evaluation of the complete fifteen-row determinant ; for the existence of a " " favourite type of distortion of finite axial wave-length , which we have noticed in the particular case , is found by practical experiment to be equally a feature of the lobed forms of distortion , and shows that the terms in are important .
Now it will be shown that the value of , when , may be obtained , correctly to terms in , by the ordinary theory of thin shells ; and as there is no reason to believe that the latter theory will lead to MR. R. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
231 less accurate results when has a finite value , it does seem necessary to employ our more rigorous method , with the very laborious calculations which it entails .
We shall therefore rely upon the approximate theory for the treatment of the tubular strut problem in its general form .
Slight modifications in method will be introduced , as suggested above ( pp. 224-225 ) , and only the more important steps will be given here .
Solution by the Theory of Thin Shells : Case .
We consider the stability of an element of the tube , originally bounded by the planes , and as shown below\mdash ; Fig. 5 .
The other dimension of the element is the full thickness of the tube , denoted in this paper by .
The radius of the middle surface is The initial stress system is const .
( say ) .
In the distorted position this system produces a radial force on the element , of amount where is the radius of curvature of a section of the distorted elemen by an axial plane ( see fig. 5 ) .
It also produces a tangential force , in the direction of increasing , of amount where ( see fig. ) The above system of distorting forces must be exactly balanced by the restoring system shown in the upper part of the figure .
Tfence we obtain the following equations of neutral stability:\mdash ; ( 91 ) Now and may be expressed in terms of the displacements of the middle surface , as follows:\mdash ; ; . .
, . .
( 92 ) and the restoring system of stress resultants may also be expressed in terms of this systeln , as ( 93 ) where is the quantity .
( 94 ) *Cf .
LovE , 'Mathematical Theory of Elasticity ' ( second edition ) , Chap. XXIV .
have the same significance as in the part of this paper , except that they now refer to the middle surface , and functions only of and MR. R. V. SOUTHWELL ON THE GENERAL THEORy OF ELASTIC STABILITY .
233 Eliminating and from equations ( 91 ) , and substituting from ( 92 ) and ( 93 ) , we have ( 95 ) where .
( 96 ) Assuming a solution of the type ( 56 ) , we find , as the criterion for neutral stability , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . . . . . . . . .
( 97 ) This equation , in its expanded form , is .
( 98 ) Taking first the terms which are independent of , and neglecting the square of ( which must be small ) , we find , as the first term in our solution , .
( 99 ) VOL. CCXIII.\mdash ; A. 234 MR. R. V. SOUTHWELL ON THE GENERAL THEORy OF ELASTIC STABILITY .
When , this becomes , ( 100 ) which shows that must be great , if has a value possible in practice .
Similarly , when , we see that must be small , and the approximate expression in this case is . . . . . .
( 101 ) We may now determine sufficiently approximate expressions for the terms in by treating as great when , and as small when .
That is to say , we retain only the highest and the lowest powers of in the two cases .
Thus , when , the important terms are and we have or .
( 102 ) When , the important terms are , ( 103 ) whence , to terms in with sufficient accuracy , when is small .
This leads to the result .
( 104 ) For practical purposes only the stationary values of are important .
It is readily seen that the mimmum value obtained from ( 102 ) agrees with ( 90 ) , and is therefore In every case it is legitimate for practical purposes to neglect the term in MR. R. .
SOUTHWELL ON THE THEORY OF ELASTIC STABILITY .
235 accurate as far as terms in ; we shall assume that ( 104 ) gives the same approximation , which for practical purposes is quite sufficient .
We then find , for values of other than and 1 , the expression . . . . .
( 105 ) When , the axis does not remain straight after distortion of the tube has occurred .
This is the type of distortion ( sometimes called " " primary flexure which was discussed by EULER , and it is easy to see that his result is identical with that of equation ( 104 ) , which becomes in this case . . . . . .
( 106 ) The exact expression for the length of the tube , in terms of , is not a matter of great importance in the present problem , because the wave-length corresponding to a minimum value of the collapsing pressure is in all cases small , and the strength of any strut of ordinary dimensions will therefore be given by equations ( 90 ) or ( 105 ) , into which the length does not enter .
As in the case of the boiler-flue problem , we Fig. 6 .
Strength of Tubular Struts .
1tlay illustrate the effects of length upon the collapsing thrust by plotting the lntensity of stress , or , against .
For this purpose we must take some de , finite value of the ratio , and draw separate curves for different integral values of .
The result is shown by fig. 6 , in which the following values are assumed:\mdash ; pounds per sq .
inch .
dynes per sq .
From an inspection of these curves it is easily seen that as the axial wave-length increases the type of distortion which involves the least value for the collapsing thrust ( and which the tube therefore tends naturally to assume ) changes .
For very short lengths we shall expect the circular type ; then , as the length increases , lobed forms of distortion , in which the value of becomes less as the length increases .
The limit is reached when ; hence , the tendency of very long tubes is always to collapse in the manner discussed by EULER .
It is also to be noticed that those parts of the different curves which lie to the right of their lowest points have no practical significance .
The actual curve , which shows the effect of length upon the value of the collapsing thrust , will approximate to the form shown in thick lines , since the wave-length ( which varies as ) will naturally not increase beyond that value which involves the least collapsing thrust .
Comparison with Existing Formuloe .
The formulae of equations ( 90 ) and ( 104 ) may be compared with the results of other discussions of this problem .
Equation ( 90 ) has been obtained by LORENZ , has given the same result , except that the factor is omitted .
The only existing solution for lobed forms of distortion is due to LORENZ , this is not in agreement with equation ( 104 ) .
In support of the latter formula it may be urged that LORENZ ' formula does not agree with EULER'S result when It may also be remarked that the foregoing results for the tubular strut problem contradict BRYAN'S theorem , that a closed shell cannot fail by instability , because distortion would involve extension of the middle surface ; for although the first terms in equations ( 102 ) and ( 104 ) are due solely to extension of the middle surface , yet the compressive stress at collapse , as given by ( 90 ) or ( 105 ) , may be insufficient to produce elastic breakdown in the position of equilibrium , if the ratio has a sufficiently low value .
Stability of Tubes under Combined End and Pressure .
We shall not treat this case in any detail , but it requires notice in connection with the " " localization of collapse\ldquo ; which is observed in experiments conducted upon long tubes tested under hydrostatic pressure , the permanent distortion being generally confined to a portion only of the length of the tube .
This result is not predicted by the theoretical formula ( 83 ) , which suggests a steady fall in the value of the collapsing pressure as the wave-length increases ; and a partial explanation may possibly be found in the fact that the method of test has generally left a wholly or partially *Cf .
footnote , p. 209 .
For a similar omission in a solution of the boiler-flue problem cf. p. 226 .
MR. R. .
SOUTHWELL ON THE GENERAL THEORy OF ELASTIC STABILITY .
237 unbalanced end-thrust , due to the water pressure acting upon the closed ends of the tube .
It is clear that the expansion of the general fifteen-row determinant will give an equation of the form where depend upon the dimensions of the tube and the type of the distortion .
But in any practical case , as we have already observed , A and must be very small quantities .
It follows that an approximate solution may be obtained from the terms . . . . . .
( 107 ) Let and be the values of the external pressure and of the end-thrust , each of which , acting alone , could produce collapse into the assumed type of distortion .
Then equation ( 107 ) may clearly be written as follows:\mdash ; , ( 108 ) where and @ are the of the external pressure and end-thrust can produce collapse when acting in con .
It may be seen from this equation that can have a minimum value for some finite value of the axial wave-length when , and only when , @ exists .
If the end-thrust be entirely unbalanced , we have , . . . . .
( 109 ) and the collapsing pressure may , in this case , be determined from equation ( 108 ) .
GENERAL THEORY INSTABILITY 1N 0F FINITE STRENGTH .
The Practical Value of Theory of Instability .
In the concluding section of this paper an attempt will be made to estimate the practical value of a theory of elastic instability ; to suggest ways in which we may hope to increase this value ; and to indicate the questions to which answers must be found in order that further may be possible .
The first point which must be noticed is the non-realization in practice of our conception of a " " critical loading owing to imperfections which always exist , and which violate our ideal assumptions .
In any actual example the displacement of the system increases continuously with the load , and the system collapses at a smaller value of the load than our theory would dictate .
It is necessary to inquire whether serious discrepancies are to be expected .
In some mechanical problems the effects of imperfections may be calculated .
We may take , as an example , the system illustrated in fig. 1 , and consider any one of the 238 MR. .
SOUTHWELL ON THE GENERAL THEORy OF ELASTIC TABILITY .
many imperfections which occur in practice .
For simplicity , let us assume that the sphere , bowl , and plunger are still smooth , rigid , and accurately formed , but that the line of thrust of the plunger is eccentric by an amount .
It is easy to see from fig. 7 that the displacement of the sphere from the line of thrust of the plunger , when the system is in equilibrium under a load , is where ' ( 110 ) and these equations enable us to trace the steady increase of the sphere 's displacement as the load on the plunger is increased from a zero value .
Fig. 7 .
Fig. 8 .
Thus in fig. 8 curves are drawn to connect and , for a value 3 of the ratio when the initial displacement has the values , and respectively .
At the points on these curves for which has a maximum value , " " collapse\ldquo ; will occur , since the equilibrium then becomes unstable .
The locus of these points is shown in the figure by a broken lin , and a dot-and-dash line shows the connection between and the maximum value of P. From the latter curve it is evident that a small initial inaccuracy may cause a material reduction in the " " collapsing load\ldquo ; ; nevertheless the " " critical load\ldquo ; ives a limit which will be more and more nearly attained as our experimental accuracy is improved , and its investigation is by no means useless for practical purposes .
When the problem is one of elastic stability , the discussion of imperfections by analytical methods will , in general , be beyond our power ; but it is clear that similar remarks will apply .
An " " exchange of stabilities\ldquo ; at some " " point of bifurcation * must be regarded as a purely ideal conception , and in practice there will always be a steady increase of distortion as the load is increased , owing principally to practical imperfections of form .
A strut , for example , may be very accurately loaded , if suitable methods are employed , but its centre-line will never be quite straight ; the initial deflection which characterizes it may be regarded as composed of a series of * H. POINCARE , ' Acta Mathematica , ' 7 .
( 1885 ) , p. 259 . .
R. .
SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
239 harmonic terms , and when the load is applied one of these harmonics will be developed very much mors than the others , just as one constituent harmonic may be developed by " " resonance\ldquo ; in an alternating current wave of irregular shape .
In the ordinary strut problem this magnified harmonic is such that one-half wave occupies the length of the strut , but in other problems , such as that of the tubular strut , though there is always a " " favourite\ldquo ; or " " natural harmonic\ldquo ; which is especially magnified , its relation to the dimensions may be more complicated .
* In any case the effects of practical imperfections of form might be studied , if the analytical difficulties could be surmounted , by ating the rate at which the amplitude of this " " natural harmonic\ldquo ; increases with the load , when its alue in the initial configuration is given ; and the results of the investigation might be shown graphically by curves of distortion , similar in character to the curves of fig. 8 , in which the abscissae represented the amplitude of the natural harmonic , and the ordinates represented the magnitude of the applied stress-system , or " " load These " " curves of distortion\ldquo ; are of considerable utility for the study problems in elastic , even though their true form can only be guessed .
They help us , for example , to explain , and in some degree to remedy , the serious discrepancy existing between EULER'S theory and the results of experiments on short struts .
The discrepancy has often been attributed to practical imperfections of form ; but it should hardly be necessary to point out that practical imperfections are likely to diminish rather than to increase in importance , as the dimensions of an elastic solid become more nearly comparable , so that they will never be more effective as causes of weakness than in struts of great length , which , as a matter of fact , give results in close agreement with EULER'S formula .
A more satisfactory explanation of this , and of similar discrepancies in other problems , may be found in the fact that the ordinary theory of elastic stability neglects the possibility of elastic break-down .
If we attempt to draw " " curves of distortion\ldquo ; for any single problem , we shall find that , apart from the other data of the problem , three possible cases exist , depending upon the elastic limit of the material under consideration:\mdash ; ( 1 ) The material may be of infinite strength ; ( 2 ) Its elastic limit may be so hi gh that the critical load , as determined by the theory of instability , is not sufficient to cause elastic break-down in the configuration of equilibrium ; ( 3 ) Elastic break-down may occur , even in the position of equilibrium , at a load less than the critical value .
In the first case ( which is , of course , purely ideal ) , the distortion due to loading will vanish when the loading is removed , and in this sense we may say that the *In the problem of the strut , the " " favourite harmonic\ldquo ; is , of course , defined by that of which corresponds to a minimum value of 5 in equationls ( or ( 104 ) .
240 MR. V. SOUTHWELL ON THE GENERAL THEORy OF ELASTIC STABILITY .
material will never fail .
The " " curves of distortion if we could determine their true shape , would probably be approximately of the form shown in fig. 9 .
The theoretical methods of this paper enable us to fix the position of , the " " point of bifurcation but give no information as to the form of AB , beyond the fact that it cuts OA at right angles .
* The other curves of the diagram will approach more and more closely the limiting form OAB as the initial value of the amplitude is decreased .
In the second case , we have the additional complication of elastic break-down under finite stress , which reduces the resistance of the material and causes the new " " curves of disbortion shown by thick lines in fig. 10 , to begin at certain points to fall away from the corresponding curves of fig. 9 ( reproduced in fine lines for comparison ) ; these points will lie on some line such as , cutting OA at a point above , and it is clear that to the right of CD curves of distortion refer to displacements which do not wholly vanish when the load is removed .
Total collapse of the system will obviously occur at the points of maximum load on the curves of distortion , and the locus of these points , which is shown on the diagram by the dot-and-dash line , may be termed the " " line of final collapse Fig. 9 .
Fig. 10 .
Fig. 11 .
A knowledge of the true form of OF would enable us , when we are given the initial value of the amplitude , to predict the load at which the system will collapse ; and these quantities could be connected by another curve AG , which would show at once whether the resistance of the system to collapse is seriously reduced by practical inaccuracies of form .
A complete theory of any problem in elastic stability must yield information on this very important point , as well as an expression for the critical load\ldquo ; ; but in most cases more powerful methods would be needed for its derivation than are at present available .
The investigation of the " " critical load\ldquo ; is therefore not without utility , for although never realized in practice , this forms a limit which should be fairly closely approached when considerable accuracy is possible .
In our third case the " " critical load as deduced by theoretical methods , is more than sufficient to cause elastic break-down .
We may proceed as before to draw hypothetical curves of distortion .
The line ( fig. 11 ) , which corresponds to the * It must not be assumed that AB is a horizontal straight line ; in general , since the distorting effect of the applied stress-system , which varies as the deflection , increases less rapidly than the resistance , which as the curvature , AB will tend to rise from A. MR. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITY .
241 line CD of fig. 10 , will intersect OA at a point below , and the other curves of distortion at correspondingly lower points .
We have seen that the effect of local elastic breakdown upon fig. 10 was to deflect the curves of distortion from the forms which they would have assumed if the material had possessed strength ; and it is clear that this deflection will begin at lower values of the loads in the present case .
We may therefore expect curves of the type shown in thick lines in fig. 11 , where the curves already obtained are reproduced in fine lines for comparison .
As before , we may draw a line " " of collapse\ldquo ; through the points of maximum load on the curves of distortion , and connect the collapsing load with the initial value of the amplitude by another curve It is clear that the curves of distortion must tend to a limit which is no longer OAH , but some other curve OA'H ' , where OA ' , the itical load under the ) conditions , is more than cicnt to produce elastic breakdown , but less than OA .
We can see further that the curve is not likely to fall away much more steeply than AG from A in fig. 10 .
The great weakness of short struts in practice , compared with theoretical nate , is now explained .
Whereas long struts come within the conditions of fig. 10 , the failure of short struts will be represented by , and occurs at comparatively low stresses , not because practical imperfections have a greater effect upon the strength , but because OA ' , the true value of the critical load , is less than OA , the value which EULER'S theory would dictate .
* It is the rule , rather than the exception , that the critical load , as found by the ordinary theory of elastic stability , is more than sufficient in practice to produce elastic break-down .
This may be readily seen in reference to any particular example .
In the case of the tubular strut , fig. 10 is only applicable when the ratio of diameter to thickness is greater than 560 ( for an average quality of mild steel ) , and for thicker tubes the critical load falls , apparently by a very considerable amount below the theoretical estimate .
The determination of the critical load , in cases where this is more than sufficient to produce elastic break-down , is thus a problem of great importance , since it forms a limit which can never , under any circumstances , be exceeded .
In the ordinary strut problem the determination can be effected without difficulty , and an apparently new field is thus indicated for research .
The distinguishing feature of its problems is the dependence of the stress-strain relations upon the past history of the material , rendering absolutely necessary a method which follows the actual cycle of events up to the occurrence of collapse .
Added Mny ll.\mdash ; Since this paper was written , the author 's attention has been drawn to a disscrtation by T. VON KARMAN ( ' Untersuchungen uber Knickfestigkcit , ' Berlin , 1909 ) , in which the forms of these " " curves of distortion for solid struts of practical dimensionls , are deduced both from theory and from experiments .
KARMAN also gives a relation equivalent to that of equation ( 112 ) .
] Experiments conducted by the author upon seamless steel tubes showed failure under loads which were in every case little more than sufficiont to produce " " manent set VOL. CCXIII.\mdash ; A. 2 I Stability of Short Struts .
This problem has been discussed elsewhere by the author , it will be noticed here only at sufficient length to indicate the directions in which further research is needed .
We have to derive an expression for the collapsing load of a straight strut , when this is more than sufficient to cause elastic break-down of the material ; and we proceed as before by considering three configurations of the strut : ( 1 ) before strain ; ( 2 ) in a position of neutral equilibrium under uniform end-thrust ; and ( 3 ) in a position of infinitesimal distortion from the second configuration .
For a first approximation we may say that cross-sections remain- plane in the third configuration , so that the diagram of longitudinal oompressive strain for any cross- section is as shown in the upper part of fig. 12 ; the horizontal line shows the Fig. 12 .
Fig. 13 .
uniform strain of the second configuration .
Then , if fig. 13 be the stress-strain diagram for a compression test of our material , and this uniform strain corresponds to a stress which is represented by the point , we see that to the right of the point in fig. 12 the longitudinal compressive stress in the third configuration must be greater , and to the left less than Now it is a well-known property of metals that if at any point on the stressstrain diagram , beyond the elastic limit , we begin to decrease the load , the diagram is not retraced , but that we obtain a line BC which is parallel to OA .
It follows that the ratio decrease of stress decrease of strain is still given by , YOUNG'S Modulus for the material .
On the other hand , the diagram shows that if we increase the load beyond by an infinitesimal amount , the ratio increase of stress increase of strain is a smaller quantity , which may be found from the slope of the diagram at B. * ' Engineering , ' August 23 , 1912 .
A. NIORLEY , ' Strength of rials , ' S42 .
MR. R. V. SOUTHWELL ON THE GENERAL THEORY OF ELASTIC STABILITy .
243 We are considering an infinitesimal distortion in the third position , so that if we represent the increase of strain in fig. 12 by , the increase of longitudinal compressive stress to the right of may be taken as , and the decrease of this stress to the left of as .
Hence we obtain , for the section under consideration , the diagram of longitudinal stress which is shown in the lower part of fig. 12 .
The uniform stress of the second configuration is shown by the horizontal line , and it is a condition for neutral stability in the second configuration that no increase of thrust shall be required to maintain the distortion .
If the cross-section of the strut is rectangular , of dimensioloe , it follows that the triangles and must be equal in area , or . . . . . .
( 111 ) This relation fixes the position of on the cross-section , and in terms of the dimensions shown in fig. 12 we may write for the moment of resistance about ( E-E ' ) .
But if is the deflection of the strut at the point , in the infinitesimal distortion , we have , as in the ordinary theory of bending , where denotes the distance of the section from one end ; and for equilibrium in the third configuration must be equal to the bending moment due to thrust , or : hence , . . . . .
( 112 ) Equation ( 112 ) shows the modification which must be introduced , to take account of elastic break-down , into EULER'S equation , ( 113 ) and it is easy to see that if is the length , calculated from ( 113 ) , of a strut which can just support the stress , and the length as calculated from ( 112 ) , then .
But , from ( 111 ) , so that , finally , This result leads to a simple method by which the collapsing loads of short struts may be obtained graphically from the compressive stress-strain diagram .
A full explanation is iven in the paper to which reference has already been made , and a comparison , satisfactory agreement , is made with the results of experiments .
conclusion of some practical importance may be noticed : the curves of collapsing stress show that great ductility of material is by no means desirable in struts , the primary requisite bein a high elastic limit .
Need for Further Research .
Conclusion .
With slight modification , the theory just given for short struts might be applied to the problem of circular rings under radial pressure ; but these appear to be the only cases in which we can at present discuss the stability of overstrained material .
In any problem dealing with plates or shells distortion from the equilibrium position must introduce new stresses , in directions perpendicular to that of the stress which has caused elastic failure .
The circular type of distortion in a tubular strut , for nple , will introduce " " hoop\ldquo ; stresses , and at present we have no knowledge of the corresponding stress-strain relations when " " set\ldquo ; has occurred .
This and many other stability probIems may be regarded as special cases of a general problem , viz. , the determination of the changes of strain which occur when an stress-system , defined by principal stresses , is impressed upon a material already overstrained by a simple stress .
The problem is not simple , and its solution would probably entail much theoretical and experimental work ; but this would be justified by the importance , both for theory and practice , of its applications .
In conclusion , the author desires to express his indebtedness to Profs .
LOVE and HOPKINSON , for valuable criticism and advice ; to Mr. L. S. PALMER , for the photographs reproduced in fig. 2 ; and to Messrs. H. J. HOWARD and D. P. SCOTT , for assistance in the prosecution of the experiments described on p. 223 .
He also takes this opportunity of thanking Messrs. Stewarts and Lloyds , Ltd. , of Glasgow , for gifts of very accurate steel tube for experimental purposes .

