]\gt ; [ 147 ] IV .
On the lastic Equilibrinm of Circular linders umder Certain Practical stems of Load .
L. N. G. FILON , .
A. , B. Sc. , Research Student of mbridge ; Fellow of University College , London ; Science ) ' : , F.R.S. Received May 20 , \mdash ; Read , 1901 .
OF Pages .
Object and aims of the paper . . . . . . . . . . .
148-150 S 2 .
Method of solution dopted .
Hisborical references . . . . . . . . . . .
150-151 S 3 .
General solution for a symmetrical strain . . . . . . . . . .
152-153 S 4 .
Solution of the differential equation . . . . . . . . . .
153-155 S 5 .
Solution under given conditions of surface londing ; . . . .
155-158 S 6 .
Consideration of the approximate expressions to which the results of the last secbion when the ratio of diameter to length is small . . . . . . . . . . .
159-161 S 7 .
Numerical problem .
Expressions for and stresses . . . . .
161-165 S 8 .
Calculation of the esses on the outer surface of the cylinder . . . . .
165-168 S 9 .
Calculation of the displacements on the outer surface of the . . . . .
168-170 810 .
Numerical values of the stresses and displacenients . . . . .
170-173 S 11 .
Discussion of the resuIts . . . . .
173-181 .
The problem .
Case of a cylinder under prcssnre whose ends ftre not lowed to expand .
( First method of constraint ) . . . . . . . . . . .
182-186 S 13 .
The second problem .
Constraint effected by ove the cross-sections .
Determination of the constants . . . . . . . . . .
186-189 S 14 .
Determination of the coefficients so as to satisfy the conditions the 189-193 S 15 .
Determination of the constants , E. . . . . .
192-195 S 16 .
Expressions for the stresses . . . . . . . . . . . . . . .
195-196 S 17 .
Numerical example . . . . . . . . . .
196-197 S 18 .
Tables of the constants for the special case taken . . . . . . . . . .
198-200 S 19 .
Methods of evaluation at the curved . . . . . . . . .
200-201 S 20 .
Calculation of the series in the preceding section . . . . . . . . .
202-203 S 21 .
Numerical values of the stresses . . . . . . . . . .
204-206 .
Principal stresses at each point : lines of principal stress . . . .
207-210 S 23 .
Application to rupture .
Distribution of maximum stress , strain , and stress-diffe ] en ce 210-214 S 24 .
Distorted shape of the curved surface . . . . . . . . .
214-216 825 .
Apparent YOUNG'S modulus and POISSON'S ratio . . . . .
216-217 28.2.1902 L. N. G. ILON ON THE ELASTIC OF Pages S 26 .
tion ivolving d at the of the plane ends . . . . .
217-219 S 27 .
Snmmary of resnlts . . . . .
219-221 S 28 .
problem .
of torsion .
xpressions for the displacement and stresses . . .
221-225 S 29 .
Special case of two discontinuous rings of shear . . . . .
225-226 .
Approximations on the ] ) when the cylinder is short . . . . . . . .
226-227 S 31 .
Numerical .
Values of the coefficients and of the displacement and esses 227-230 S 3 Discnssion of the nlts . . . . .
231 conclnsion . . . . .
231-933 S 1 .
Aims of the Paper .
THHj usual solution for the extension and compression of elastic bars assumes that the latter are strained under a normal nsion or pressure uniformly distributed across the plane ends .
In like manner the solution for torsion of such bars ssumes that the external forces which cause the torsion consist of a determinate system of stresses , acting across the plane ends .
In both cases the solution is such that the torsion and extension are transmitted throughout the bar toitho lchange of ) ) .
Such terminal conditions of stress , however , do not usually occur practice , and it accordingly becomes of considerable interest to find out how the results obtained for such a theoretical system of are modified , if at all , when we consider applied extel.nal stresses which give a closer represelltation of every-day mectlanical conditions .
The present paper is an attempt towards the solution of this problem in three cases , which appear of especial practical interest .
The first case is that of a bar which is subjected to a determinate system of normal radial pressures and of axial shears all over the curved surface , the radial pressures being symmetrical about the mid-section and the shears having their sigll changed .
Thus the cylinder is subjected to a total axial pull , due to the shears , and also to a transverse pressure .
The plane ends are free from stress , except for a self. .
system of radial shears , which have little or no effect at points at some distance the ends .
A. special case worked out is that where the normal pressure is zero throughout , but a determinate axial shear , which has been taken constant , is made to act over two equal rings on the surface of the cylinder .
This will give us yaluable inforlnation about a system of stress which often occurs in practice , in testing machines , for example , in which a specimen is pulled apart by means of pressures applied to the inner of projecting collars ( see fig. 1 ) .
The shaded of the figure represent the " " grips and if be the total pull applied , this is nsmitted to the test piece by means of pressure applied along CA , Now the thinncr cylinder in the middle ideally produced inside the thicker CIRCULAR CyLINDERS UNDER CERTAIN PRAGTICAL OF LOAD .
149 ends .
It is in equilibrium under the esses , radial and tangential , between the inner core and the hollow cylinder produced by the revolution of ABDC .
Fig. 1 .
But what are these radial and tangential stresses If we consider the equilibrium of the outer hollow cylinder only , we see that the resultant of the stresses across AB , must exactly balance the pull , howeyer applied .
The radial stress will probably be small , as it has no external traction to balance , and the longitudinal shears are therefore equivalent to S. Thus the thin cylinder inside is really stretched , not by normal traction over the flat ends , but by longitudinal.shears over the curved surface , and a careful investigation will show that , in ctical case , extension is obtained by the application of an axial shear to the curved surface of the cylinder , never ) of tractions to the flat ends .
The general effects of such a distribution appear , therefore , of great practical interest .
The second problem discussed is that of a cylinder of moderate length , which is compressed between two rough rigid planes in such a way that the terminal crosssections are constrained to remain plane , but are not allowed to expand , their perimeter being kept fixed .
By adding a suitable uniform distribution of pressure to a load system of this type , we can obtain the solution for a cylinder constrained in such a way that its ends expand by a definite amount .
These two problems are of importance with reference to the behaviour of a block of stone or masonry when tested between nlillboard or metal planes , which practically hinder the block from expanding , and when tested between sheets of lead , which , on the other hand , favour the expansion of the block .
The widely divergent results obtained for the strength of the same material when tested by these two methods have troubled many elasticians .
UNWIN ( ' Testing of Materials of Construction ' ) is of opinion that the results obtained when sheet lead is used are unreliable ; whereas Professor PERRY , in his ' Applied Mechanics , ' states that the true strength of the material is the one given by the lead experiments , and should be usually taken as half the published strength .
Finally , the third problem treated is that of a cylinder sub.jected to transverse shears over the of the curved surface near the ends , these shears being equivalent to a torsion couple .
This is really the analogue , { or torsion , of the first MR. L. N. G. FILON ON THE ELASTIC EQUILIBRIUM OF problem for tension , and corresponds to the case of a bar " " gripped\ldquo ; as explained above , and twisted .
This again is the method by which tol.sion is practically produced in most cases\mdash ; almost always in laboratory expel 2 .
Method of Adopted .
Historical The method adopted has been to obtain symmetrical solutions of the equations of elasticity in cylindrical co-ordinates and to express the typical term in the form , , being the usual cylindrical co-ordinates .
The expressions the strains and stresses , over any coaxial cylinder , are thel.efere series of sines and cosines of multiples of The arbitrary constants of the coefficients determined by comparison with the coefficients of the FOURIER'S series which express applied stresses at the external boundary .
This Inethod is not a new one .
It has been indicated and CLAPEYRON Memoire sur interieur des corps solides homogenes 'Crelle 's Journal , ' vol. 7 ) , but it has been for the first time worked out with any completeness by Professor .
POCHHAMMER Beitrag zur Theorie der Biegung des Kreiscylinders ' Crelle 's ' vol. 81 , 1876 ) .
Professor POCHHAMMER obtains the general solution of the elastic equations for an infinite circular cylinder subject to any system of surface loading , repeated at regular intervals .
This he applies to the case of a built-in beam .
The solution is not restricted to be symmetrical about the axis of the cylinder , but is perfectly general .
The coml)lete accurate expressions are , howeyer , unwieldy ; but , as the esult of expanding the functions involved to the first two or three terms , Professor POCHHAMMER obtains far more manageable expressions , which he is eventually able to identify with those previously given by and DE SAINT VENANT for more special cases of loading .
It is to be noted , however , that POCHHAMMER restricts himself solely to the case of bending , and that his approximations depend upon the ratio of diameter to length being a small quantity .
The same general expressions have been independently arrived at by Mr. C. The Equations of an Isotropic Elastic Solid in Polar and Cylindrical Co-ordinates , their Solution and Application 'Camb .
Phil. rans .
) .
Here , again , the solutions are not restricted to be symmetrical .
The symmetrical terms , however , agree with the solutions of the present paper , but the latter are obtained by a process slightly different from that of Mr. .
Mr. CHREE has also given a solution 01 the symmetrical case in powers of and .
Using each form of solution , it is not possible to satisfy the condition that there shall be at all on the curved surface ; this is ected in the second problem of this paper , by means of a combination of the two types of solution .
In the paper referred to , Mr. CHREE , like PUCHHAMMER , has not , so far CIRCULAR YLINDERS UNDER RTAIN PRACTICAL OF LOAD .
151 as I am aware , applied his general solution to the problems of tension and compression .
He does give one example of torsion , which he obtains by applying an arbitrary system of cross-radial shears across the flat ends .
Such a system , we have seen , would not usually correspond to what occurs in practice .
Mr. CHREHj has written several other papers On some Compound Vibrating Systems , 'Camb .
Phil. Trans. , ' vol. , Part II .
; " " On Longitudinal Vibrations ' Quarterly Journal of Mathematics , ' 1889 ; " " Longitudinal Vibrations in Solid and Hollow Cylinders 'Phil .
Mag. , ' 1899 ; " " On Long Rotating Circular Cylinders ' Camb .
Phil. Soc. Proc vol. , Part , \amp ; c which deal with the solutions of the equations of elasticity in cylindrical co-ordinates , with special application to vibrations and rotating shafts ; but I cannot find that he has anywhere returned to the statical problem and its solution by means of sine and cosine expansions .
.\mdash ; Professor SCHIFF Journal Liouville , has attempted the solution of the problem of the cylinder compressed between parallel planes , which is one of those treated of in the present paper .
His solution is expressed in a series , not of circular functions , but of hyperbolic sines and cosines , the successive values of being obtained as roots of a certain transcendental equation .
This enables him to satisfy the conditions at the curved surface , but the arbitrary coefficients finally determined by the conditions over the plane ends .
He assumes both the radial shear and the molecular rotation in a diametral plane to be given by known functions , and , ovel the plane ends , and from these he succeeds in obtaining the coefficients .
As he has only a single set of the latter left to out the identification , his functions and are not really independent .
Theoretically only the shear should be required , and in a practical problem even this is unknown , the total pressure being all that is given .
The actual distribution of this pressure does not appear to enter into Professor solution .
Also the fact that the values of are roots of a transcendental equation singularly complicates the solution from a numerical point of view , and Professor SCHIFF appears to have made no attempt to translate his results into numbers.]* It has therefore appeared worth while to apply the solutions involving circular functions of to problems such as those sketched above .
Of each of these I have given a concrete numerical example .
Indeed , the greater part of the work has been spent on these numerical examples .
The labour of calculation has in most cases been considerable , owing to the slow convergence of many of the series involved , which has necessitated special methods of approximation .
* Since writing the ctbove , I find that the problem of the circular cylinder tder a symmetrioal strain has been considered by J. THOMAE in two papers ( " " Uber eine einfache Aufgabe der Theorie der Elasticitat ' Leipzig Berichte , ' vols .
37-38 ) .
The author has used expansions in sines and cosines of but , as far I out , the only problem he considers is that of the vertic.al pillar under its own weight .
S 3 .
eneral Solution for Let , be the usual cylindrical co-ordinates ; also , the notation of TODHUNTER and ' History of Elasticity , ' denote the stress , parallel to across an element of lrface perpendicular to nding for any two of the letters Let ?
denote the radial , cross-radial , and longitudinal displacements respectively , then we have , ' cons sir sticit if of : ( 1 ) .
( 3 ) .
( 4 ) , nd being the elastic constants of We see from the above that and depend only on , the other stresses only on and .
Also the equation ( 2 ) contains only , ( 1 ) and ( 3 ) contain and ) only .
The solution for transverse displacements is therefore absolutely independent of the solution for radial and longitudinal displacements .
Let us now denote the operators .
by , and by D. Differentiate ( 1 ) with regard to with regard to , and remember that the order of the symbols I ) and is indifferent , then we find . . . . .
( 5 ) .
( 6 ) . .
CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
153 Eliminate either or ) between ( 6 ) and ( 7 ) : it is found that either of these quantities satisfies the partial differential equation . . . .
( 8 ) .
The whole problem of the determination of the elastical equilibrium of the circular cylinder under any symmetrical system of stress depends therefore on the tion of this differential equation .
S4 .
Solniion of the Equation .
The differential equation is really identical with equation in cylindrical co-ordinates , namely , Suppose V independent of and differentiate with regard to , we have If therefore V be such a solution of LAPLACE'S equation , will be a solution of the given differential equation .
For our purpose , however , it will be simpler to proceed from the equation itself Assume a typical solution where is a function of only , a function of only .
We find easily ( 9 ) , ( 10 ) .
The solutions of ( 9 ) are of the form ( kr ) and ( kr ) , where ( See GRAY and MATHEWS , ' Bessel 's Functions , ' pp. 66-7 .
) V0L .
Consider now the equation Let be any solution of the equation Then if be a solution of it is also a solution of Now if what is the condition that we can obtain a second product solution ?
We have , substituting ( 11 ) , or function of 2 only function of onlyproduct function in and If ( 11 ) is to be identically satisfied , this product function must be a function of only or of only .
Case .
where is a constant .
We find ( 12 ) , or , since is a solution of ( 13 ) , which is the same , is a solution of which is not at the same time a solution of ( 13 ) .
Now the solutions of this equation are , , , ( ) .
But ( ) , and similarly ( ) .
The four independent integrals are therefore ( kr ) , ( kr ) , , , and therefore the required values of are CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
155 Case .
; we fmd , using ( 13 ) , and therefore is a solution of , which is not at the same time a solution of The possible values of are Hence the possible sets of product functions satisfying the equation are as follows:\mdash ; ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( kr ) ( 14 ) .
( kr ) S5 .
Solution conditions of ; the firstproblem .
Let us now consider first the case of a circular cylinder under the following system of stress : .
a given even function of over the curved surface given odd function of over the curved surface over the plane ends Since are both solutions of ( 8 ) we may have them composed of a series of terms as follows : ( 15 ) .
( kr ) ( 16 ) .
( kr ) No -functions have been introduced in this case , as they lead to infinite terms at the axis .
1 .
56 MR. L. N. G. FILON ON THE ELASTIC EQUILIBRIUM OF Also the conditions of the problem require that ?
shall be an even function of and ?
an odd function of Hence Integrating ( 15 ) and ( 16 ) we have To find the relations between the constants we must substitute in equations ( 1 ) and ( 3 ) .
We then find the following relations : . . . .
( 17 ) , . . . .
( 18 ) , ( 19 ) , ( 20 ) , , say . . .
( 21 ) , , say . . .
( 22 ) .
In virtue of equations ( 21 ) and ( 22 ) , ( 19 ) and ( 20 ) reduce to the single equation . . .
( 23 ) .
Also from ( 17 ) and ( 18 ) remembering that is odd in and that is not to be infinite when For the stresses we find from ( 4 ) , after some obvious reductions , .
( 24 ) .
( 25 ) .
( 26 ) .
CIRCULAR UNDER CERTAIN PRACTICAL OF LOAD .
157 But clearly if is to be zero all over the plane ends we must have ( 27 ) , .
( 28 ) , and . . . . .
( 29 ) .
The expressions for the displacements and stresses then reduce to the following , writing for shortness : ( SO ) , , \mdash ; kz .
( 32 ) , .
( 38 ) , ( 34 ) , where ( 35 ) .
Over the surface of the cylinder , we find ( 36 ) , ( 37 ) , where written for and for Now being an even function , we can expand it in a FOURIER'S series between the limits in the form ( 38 ) , where and being an odd function can be expanded in the form .
where Now since , , we have , comparing ( 36 ) and ( S7 ) with ( 38 ) and ( 39 ) , whence ( 40 ) , ( 41 ) , ( 42 ) , and therefore from ( 35 ) and ( 27 ) .
( 43 ) , ( 44 ) .
CIRCLTLAR CYLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
159 S6 .
of the Expressions to which the Results of the last Section tead , when the Ratio of Diameter to Length is If we can treat the diameter of the cylinder as small compared with its length , we can obtain a first approximation by following the method of Professor POCHHAMMER ( ' Crelle , ' vol. 81 ) , and expanding , A in powers of , which is then a small quantity , proyided the index is not too large.$ If we do this we find and , expanding and in powers of , and dropping all the terms except the first ( which is really equiyalent to a second approximation , since the indices go up two at a time ) , we find using the Fouriel expansions ( 38 ) and ( 39 ) Now is equal to the total longitudinal pull exerted on the bar by all the forces on one side of the cross-section considered .
It represents , in other words , the total tension at that cross-section .
Denoting it by , where is the mean tension at that cross-section , ( 45 ) , which shows that the radial displacement is exactly the same as if the only forces on a thin lamina between two cross-sections were an external radial tension and a uniform tension across the plane faces .
* For the restrictions necessary in suoh a cttse , see S28 .
In like manner it can be shown that ( 46 ) , being the stretch parallel to the axis in a cylinder which is under a tension across its plane faces and a radial tension Thus the longitudinal and radial displacements are , to a first approximation , the same as if the cylinder were supposed made up of any numbel of thin circular laminae , piled up on top of each other , the longitudinal tension in ally lamina being uniform and giving a total tension equal to the total pull of all the external forces acting on the cylinder on one side of the section considered .
Further , the shearin stress at a point inside is found to the same approximation to be given by so that , in the parts of the cylinder to which external shearing stress is applied , and in these only , there is shearin stress inside the cylinder , which shearing stress is proportional to the distance from the axis .
The other stresses , , are found to the same approximation to be all constants for any given value of .
It follows from the above that the action of any radial pressure will be purely local , and also that , whatever the manner in which the cylinder is " " gripped\ldquo ; and the pull is applied , the stress in the portions of the bar between the points of application of the pull reduces practically to a uniform tension .
The above results are somewhat remarkable as tending to show how very restricted is the effect of local stresses , provided they leave no total resultant , and how , when CIRCULAR CYLINDFRS ) CERTAIN PRACTICAL OF LOAD .
1 they do leave a total resultant , the effect of this resultant is practically independent of the manner in which it is applied .
This is the celebrated " " principle of the equivalence of statically equipollent loads which was first enunciated by DE SAINTVECNANT on general physical principles , and has been considerably confirmed by BOUSSINESQ'S researches on the effect of small local surface actions .
It is to be borne in mind , of course , that the solution obtained in S5 , although making over the flat ends , does not at the same time ensure .
In other words , we have a determinate system of radial shears over the flat ends , but from symmetry this system must be self-equilibrating .
The disturbances due to it will therefore , by the above principle , be purely local , and , provided we remove the ends sufficiently far from the parts of the beam which we desire to study , no trouble need arise on account of all the conditions not being strictly satisfied .
S7 .
Nnmerical Problem .
Expressions for Stress .s .
Let us now return to the exact expressions and apply them to the case of a comparatively short cylinder .
Suppose that all over the curved surface and that in some way , as described in S1 , a shear , which we shall take uniform and equal to , is made to act along two rings upon the curved surface , so that when when when ) We have then The expressions for the constants , stresses , and displacements then reduce to VOL. CXCYIII.\mdash ; A. 1 ( i2 N. ON FLASTIC OF kz .
( 48 ) .
2 is the argument of the -functions , unless the argument is written .
To simplify the expressions we shall take , so that the length is about three the radius .
This makes .
Further , suppose , so that the cylinder is divided into 5 ones , as shown in fig. 2 .
Fig. 2 .
The middle one from to , unstressed ; two ings from to to oyer which a uniform shear is acting ; finally , the outer rings 2 to and to , which are unstressed .
Also , in order to simplify still more , shall suppose POISSON'S ratio to have the value 1/ 4 , or It may be objected , it is true , that in many actual materials POISSON'S ratio is not 1/ 4 .
Bnt this is not really an objection , because the object of this investigation is not so much to find out the .solute values of strains and stresses in any given material , as to calculate the alterations in these values as deduced from the hypothesis uniform stress , and this we can best do by taking a value for POISSON'S ratio which is , on the whole , well within the limits indicated by practical results , and which makes the arithmetic somewhat easier .
If we do this and calculate the values oi the constants , we find that for the first 10 terms they come to the following values : CIRCULAR LOAL ) .
of Constants .
0323.529 From these I have calculated the coefficients of the series .
the stresses and strains for , and .
For highel yalues of the convergence becomes slower and the expressions more difficult to handle .
In the case of the stresses and strains at the boundary , special metbods of ) tion have to be resorted to .
The expressions for the strains and stresses are : , , , in like manner for n ) and the stresses .
To save space , the coetlicients the ) exhibit .
in ) ' CIRCULAR CyLINDERS UNDLR TAIN PRACTICAL SYSTEMS OF LOAD .
165 Using the above values I have calculated the stresses and ains for the points , and , \amp ; c These are tabulated on pp. 171-173 .
S8 .
Calculation of he Stresses on the Outer of the Cylinder .
Along the outer surface , and we following expressions for the stresses and , and the displacenlentst and 10 : of course known .
Consider example the stresses and They are , and ( 50 ) .
Now , when is fairly large say , and may be ) laced by their semiconvergent ulsions : .
GRAY HEWS , 's Functions , .
which we find that the coefficients of in the expallsions of and approximate to the values ( remembering ) : and ISIR .
L. N. G. FILON ON THE LLASTIC EQUILIBI[IUM OF Now let us write sin ' sin ( 52 ) so that are comparable with the terms of the series which convel.ge ' fairly rapidly .
We see therefore and are made up of two kinds of tel .
Ins , of the form and , which are absolutely uniformly convergent series , and ( b ) series , in which the coefficients the proximate expressions found above .
Of the series ( b ) , those which have or are absolutely and uniformly convergent .
This , howevel , is not the case with the series formed by taking the leading terms in the .
:\mdash ; and For the series may be broken up into the sum of four other series , thus : .
Now it is easy to show that where is the ical value of CULAR CyLINDERS TAIN PRACTICAL OF LOAD .
167 The series on the left is divergent and if .
We see , therefore , that , at the points , i. e. , wherever the shear changes discontinuously , the stresses and become infinite .
The meaning of this in practice would be that , as the transition fxom the stressed to the unstressed surface becomes more abrupt , the tractions in the neighbourhood become dangerously large .
And if the shear is applied by means of a projecting rim or collar of material , on which the pull is brought to bear , as in fig. 1 , then this rim or collar must not project out of the material at a sharp angle , or in any way which tends to introduce a discontinuous tangential stress over the surface of the cylinder .
This is already recognised in practice ; test pieces , which are thicker at the ends than in the middle , being made in such a way that the tion from the smaller to the larger diameter is gradual .
The series containing can also be in finite terms : from to from to from to from to from to Thus we have only to calculate and to sum the corresponding , the rest of the expressions for the stresses being reducible to finite terms .
For , I find the values of to be given by : from which the values of can be found when .
They are tabulated , with the other stresses , upon p. 171 .
S9 .
of the xcements on the Outer of the In a precisely similar manner we find for the displacements 1 \mdash ; ( 53 ) where and is of the order ( 54 ) , CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
where and is of order It so happens that in the term of order is evaluable in finite terms , and I have included it .
It is easy to see that \mdash ; \mdash ; from to from to from to \mdash ; from to from to The leading series in cannot , however , be evaluated so easily .
It is seen to depend upon the eyaluation of the series As series of this kind are frequently turning up in investigations like the present , I have tabulated below the values of and also of for values of ranging from to at intervals of .
Intermediate values are then obtained by interpolation when required .
TABLE of .
VOL. CXCVIII .
\mdash ; A. MR. N. G. FILON ON THE ELAST'iC IBRIUM TABLE of We have thus the means of evaluating all those parts of the expressions which give rise to the most slowly convergent of the series employed .
Taking , the found for are tabulated below:\mdash ; Using these and the expressions given above for the finite terms , we can find the values of the displacements on the outer surface of the cylinder .
S10 .
Numerical of the Stresses isplaceme , The numerical values obtained in this way are tabulated below ; I have given the stresses in the form of ratio ( stress)/ Q , where is the uniform tension which would produce a pull equal to that due to the shear S. CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
171 TABLE of Stresses .
In the above tables it is to be remembered that are all even functions of is an odd function of , the points ) and are both , while at the points ' and are both The displacements and have been compared with the corresponding total elongation and lateral contraction and ?
of the same cylinder under a uniform nsion Q over its plane ends .
TABLE of Displacements .
CIRCULAR CyLINDERS UNDER CERTATN PRACTICAL SySTEMS OF LOAD .
S 11 .
Discussion of the Results .
The numerical results tabulated above are illustrated by the curves contained in Diagrams 1-6 .
Diagram 1 shows the radial shift , of course enormously exaggerated , on the diagram being taken as numerically equal to of the radius of the cylinder .
For convenience in plotting , the horizontal and vertical scales are not the same , thus and are represented by the same length on the diagram , although their actual ratio is .
The same arrangement has been adhered to in Diagram 2 .
Diagram 1.\mdash ; Distortion of a CyliIIder extended by Shearing Stress applied to the Curved Surface ( Radial Shifts ) .
( on the scale of the diagram .
) From Diagram 1 we see at once that a discontinuous change in the slope of the outer surface of the cylinder occurs at the points between which the uniform shearin stress is applied .
Referring to equation ( 58 ) we see that at those points changes abruptly by the value \mdash ; , where is the abrupt increase in the shear .
This result is exhibited in the curves referred to , and we notice that the effect of shear , applied to the outer surface of a cylinder , Diagram 2.\mdash ; Distortion of the Cross-sections of a Cylinder under Shearing Stress applied to the Curved Surface .
( on the scale of the diagram .
) 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 undistorted cross-sections , 1 cross-sections .
is to depress that part of the surface towards which the shear is acting .
In fact the greatest contraction throughout the cylinder occurs near the points and appears due to this effect .
Near the ends the cylinder broadens out again , as we should expect , though it is to be noted that the distorted generators meet the plane ends obliquely , which should not be the case if the-condition of no stress over the plane ends were accurately fulfilled .
This we know is not so : there is a system of finite shear over the plane ends , as is easily seen on referring to the table of on p. 172 .
This CIRCDLAR CYLIND RS RTAIN PRACTICAL SYSTEMS OF LOAD .
175 system of shears is , however , self-equilibrating .
The shear is zero at the centre and at the circumference , and its greatest value does not exceed about 1/ 4 of the laterally applied shear .
Its effects , at some distance inside the cylinder , will therefore be small compared with the effects of the large and unbalanced lateral distribution of shear .
We notice that , for so short a bar , the lateral contraction is very much less than the contraction we should expect according to the " " uniform tension\ldquo ; theory .
In fact it never amounts to 60 per cent. of that contraction .
For points deeper in the material , the contraction is much smaller than this .
Thus , for , the lateral contraction is 22 per cent. and for it is 9 per cent. of what it should be on the " " uniform tension\ldquo ; hypothesis .
This seems due to the fact , in itself extremely remarkable , that there are considerable radial and cross-radial tensions inside the material .
Indeed , referring to Diagram 3 , we see that the radial tension amounts to Diagram 3.\mdash ; Showing Stress for the Oylinder under a Shearing Pull .
about of the mean tension which would give the same total pull , and which has been consistently taken as the unit of comparison .
These tensions are changed to pressures after passing the ring of shear , which is in accordance with the general compressive effect mentioned above .
It may be noticed that the shape of the successive curves on Diagram 3 suggests that , as we approach the outer skin , the two bumps on either side of would MR. .
N. G. FILON ON THE ELASTIC EQUILIBRIUM OF lead to infinities , or at all events , to discontinuities in the stress .
In other words , , though we have chosen our constants so as to make formally zero , yet the limit of as given by the series is not zero when approaches .
This would suggest that the series for , considered as a series of -functions , behaves at in much the same way as a discontinuous Fourier series whose general term is behaves at .
In fact , if we rentiate r in the usual way with regard to and then put , we get a divergent series .
It is easily seen , however , that no discontinuity really occurs except at the points where the shear is applied discontinuously .
The general term in is of the form ( .dropping irrelevant factors ) : where .
Now , at the semi-convergent expansions for and we find , putting and \mdash ; S where S and is small , terms of order S/ and higher terms in and \mdash ; higher t where base of Napierian system oflogarithms .
The coefficient of then reduces to the form ermsf oinite t CIRCULAR CyLINDERS ACTICAL S LOAD .
177 Now , if we remember that is of order , we see that the successive terms in the coefficient of are of the orders ' ' respectively .
Also in considering discontinuities , we need only consider the terms towards infinity , for the terms at the beginning can introduce no discontinuity .
But clearly the series are of the order multiplied by a series , which is finite and continuous up to and including the value .
They tend therefore to the limit with , and call introduce no discontinuity in the stress .
The same will be seen to hold of the series 2 , provided ?
If however , we have to deal with the series The series under the sign of summation is divergent if .
If , however , is small , but still finite , the series can be summed , and we have the expression equal to .
This tends to zero when is small , proyided tend to zero , which is known to be the case .
Hence this series again call never introduce a discontinuity in the stress .
Now consider the series This is not of the same form .
The series under the 2 is sometimes oscillatory , and sometimes divergent , but is never convergent , if ; put equal to zero , V0L .
CXCV111 .
But if be small but still finite , the series As approaches the limit , this series also approaches the limit .
Hence , fortiori , this series multiplied by approaches the limit , and the stress is continuous .
This holds provided .
But if the series in question .
The limit of when is \mdash ; .
But when absolute zero in the series ' the series identically .
We have therefore in these cases a finite discontinuity in the stress .
This takes place at the points , i. e. , , where the shear varies discontinuously .
At all other points .
approaches the value zero continuously as we move up to the outer surface of the cylinder .
Coming now to the distortion of the cross-sections , is exhibited in Diagram 2 .
The displacements are exaggerated , as in Diagram 1 , being taken .
The cross-sections become hollowed out in the middle , the greatest longitudinal extension taking place at the sides .
Another noticeable feature is that the cross-sections are slightly curled round the rim , except over the part of the cylinder which is subjected to shear , where they slope up sharply .
This follows from the fact that Thus , where and from Diagram 1 , it follows that or , since nearer the centre , a maximum value of occurs at a comparatively small distance inside the " " outer skin\ldquo ; of the cylinder .
When , however , increases by , we have seen that increases by \mdash ; , hence increases by and is always positive at the outer surface .
At the further end , where ceases to act , the reverse takes place .
It is now easy to understand why the tension is infinite at the inner end of the shear ring and the pressure infinite at the outer .
For if we take two parallel near cross-sections , the one just inside the shear ring and the other just outside , the distorted cross-sections remain sensibly parallel until we approach the outer surface , when they diverge sharply , if near the inside boundary , and converge sharply if at the outside boundary .
In the one case we get an infinite extension , in the other an infinite compression .
Hence we should expect the stresses and to become infinite at these points , and the stress .
to vary infinitely rapidly\mdash ; and this , we have seen , is what does actually occur .
Further , we see that if we measure the elongation of the outer skin as is done with an extensometer , we shall always get too high a value for the extension .
to the table on p. 172 , we have the following table of the displacements CIRCULAR TAIN PACTlCAL S LOAD .
I 7 Diagram 4.\mdash ; Showing Stress for the Cylinder umder Shearing Pull .
80 L. N. .
FILON ON THE ELASTIC OF measured by the extensometer , as compared with the displacements calculated from theory , over the free of the bar:\mdash ; cements .
We see , therefore , that in such a case very corrections have to be applied to extensometer readings .
Diagram 5.\mdash ; Showing Stress for the Cylinder under Shearing Pull .
CIRCULAR UNDER CERTAIN ACTICAL S OF LOAD .
181 Diagrams 3-6 giye curves showing the variations of the stresses , with , for the values of equal to , I have omitted the intermediate vadue , because the series used converge in this case inconveniently slowly , and no methods of approximation , such as were employed in the case , are here avail .
Observation of the curves for the smaller values of will , however , in most cases suggest the process by which they are deformed continuously into the curve Diagram 6.\mdash ; Showing Stress for the Cylinder under Shearing Pull .
Axis of cylider corresponding to .
In Diagram , of course , this is not obvious , but here , as has been shown , discontinuous changes occur .
In Diagram 6 it is also not quite clear how the curve for becomes transformed into the rectangle corresponding to .
The curve for ] , however , already developed a double hump , and its righthandmost .
ordinate 's rate of increase is fast diminishing .
This suggests that the two humps will rise and approach each other , ultimately covering the rectangle , whilst the two " " tails\ldquo ; will dwindle down to zero .
Remarks of a similar character apply to Diagram 5 .
Mr L. N. .
FILON ON THE OF Second Problem : of Cylinde under Pressure whose Ends are not allowed to expand .
JIethod of Consider a cylinder ( fig. 3 ) subjected to the following system of load:\mdash ; ( 1 .
) There is shear along the curved surface .
Ovel two rings of the ends a radial pressure made to act , this pressure being so adjusted is no radial shift at the points ; the breadth being in the limit to be made indefinitely small .
Fig. 3 ( 2 .
) The plane ends AB , , are constrained to relnain plane , and are subject to total essure The above would fit the case of cylinder compressed between two rigid planes , into which shallow circulal depressions had been cut , to fit the ends of the compressed cylinder and prevent them from expanding .
If we return to the expressions for the stresses in the general case , ( 24 ) , ( 25 ) , ( 26 ) , anl also to those for the displacements , we find that if is to be constant for . . . . . .
( 55 ) , . . . . .
( 56 ) .
Also this gives ovel ' the ends , so that we may suppose our rigid conplane tu be also sllooth .
CYLINI ) CERTAIN CTICAL SOF ) .
183 The condition that is to be no shear over the curved surface now gives ( 57 ) , writing before ) .
From ( 56 ) and ( 57 ) ( 5 ) .
In what follows , the argument of the -functions , when not written , will always be assumed to be We find ( 59 ) .
Putting in this Now expand the given pressure in the form .
Where , . . . . . .
are determined , remains a free constant .
We have at once , comparing coefficients , , ( 62 ) Next we have when . . . . . . . .
( 64 ) , where .
This condition gives in terms of , and hence by ( 63 ) ( 66 ) .
We have now to find such a value for that the mean pressure on the plane ends is Q. whence , applying the well-known theorem , we have Using the relation ( 67 ) , and putting in for their values in terms of , we find that the terms under th vanish identically .
Hence ( 68 ) .
Now suppose the distribution of stress is such that ?
from to and from to , and from to , we find whence ( 69 ) .
When is at all large , the terms of are comparable with those of the series CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
185 which is equal to ( 70 ) .
( 70 ) will giye the approximate value of wheneyer is at all large .
If this method of approximation will already be quite fair .
We see therefore that if tends to zero , also tends to zero , but tends to become logarithmically infinite .
Now from ( 68 ) Hence , since - tends to when tends to zero , tends to zero when tends to zero .
And similarly , for any finite value of tends to zero when tends to zero .
But if we write down the expressions for the stresses , they are : . . .
( 71 ) .
Now the above series.are absolutely convergent for all of , , except where indeed they are discontinuous .
Leaving the neighbourhood of out of account , we see that for points inside the material , when the space over which the constraining pressure acts is indefinitely reduced , and ; therefore outside the rim , where plastic deformation may be expected to occur , the stresses are exactly the same as on the ordinary hypothesis .
We come then to the conclusion that this method of preventing the ends from VOL. CXCVIII.\mdash ; A. 2 expanding is not adequate , and that to obtain any real effect , we require to make the constraining rim of a certain definite thickness .
In so doing , we are really introducing an additional condition , besides the nonexpansion of the ends , the cylinder being now , as it were , built-in .
The probJem as it stands did not appear of sufficient interest to warrant the expenditure of arithmetical labour upon it , so I have contented myself with stating the algebraical results .
S13 .
The Second Problem : Constraint effected by over the -sections .
of Constants .
Suppose now that we consider our cylinder subject to the following conditions : A total pressure over the plane ends , the distribution of this pressure being unknown .
( ii .
) The ends constrained to remain plane , so that const .
when ( iii .
) The ends not to expand along the perimeter when This condition is satisfied by allowing a shear over the plane ends , its distribution being , however , unknown .
No stress across the curved surface , i. e. , when when These conditions will represent the state of things which we may expect to hold it the cylinder be compressed between two rigid planes which are sufficiently rough to prevent the expansion of the ends .
Now , in such a case as this , it is obvious that the expressions for the stresses and strains as purely periodic series in break down , for if we take the expressions ( 24 ) and ( 26 ) for and the condition that const .
when will give us , as before , , and the vanishing of the stresses at the curved surface will give two homogeneous equations of condition between , and C. These , taken in coniunction with equation ( 35 ) , give three linear homogeneous equations in , and , which are general inconsistent unless , which would destroy the periodic solution altogether .
We have therefore to assume that and are made up of two parts .
The first part , which I shall denote by , consists of the periodic solution hitherto obtained .
The second part is a finite power series in and .
The resulting expression is a combination of the two types of solution , which are discussed separately by Mr. CHREE ( ' Camb .
Phil. Soc. Trans. , ' vol. 14 ) .
Either of these two types , taken by CIRCULAR OYLINDERS UNDER CERTAIN ACTICAL SySTEMS OF LOAD .
187 itself , is of comparatively restricted application , but by combining the two we are enabled to deal with far more general problems .
Assume therefore ( 72 ) , ( 73 ) .
The above power series are the most general expressions of the fifth degree consistent with the conditions that must be odd in .
and even in , and must be odd in and even in In the above we have , as before , ( 74 ) , ( 75 ) .
Consider first of all the condition that is to be constant when .
This fixes : ( 76 ) .
Further we have . . . . .
( 77 ) , . . . .
( 78 ) .
Now remember that and have to satisfy the differential equations The parts and we have seen already will satisfy these equations , provided . . . . .
( 79 ) .
Consider therefore only the algebraic terms .
Of these and always satisfy the above equations .
The third order terms require . . . . .
( 80 ) , MR. L. N. G. FILON ON THE STIC QUILIBRIUM OF The fifth order terms give which imply the four relations ( 82 ) , . . . . .
( 84 ) , . . . . .
( 85 ) .
There is , however , a further relation to be satisfied among these constants , and that is obtained as follows .
If we proceed to write down the expressions for and and to put in them , we shall obtain expressions of the form algebraic polynomial in series of cosines of algebraic polynomial in series of sines of where the coefficients of , contain the two undetermined constants and We may now proceed to expand the two polynomials in series of cosines or sines of .
Equating then the coefficient of each cosine and sine to zero , we can make and zero oyer the whole of the curved surface , and at the same time we obtain two equations for and But it is clear that , if the Fourier expressions in the second case are to be continuous , then the algebraic polynomial part of must reduce to zero when otherwise its expansion in sines of is discontinuous , and at the perimeter of the flat ends the shear is discontinuous .
This introduces infinite stresses at this point which render the solution inconvenient .
Now we have at our disposal nine constants ; these have already been made to satisfy the seven homogeneous equations ( 78 ) , , and therefore we are free to make them satisfy an eighth homogeneous equation .
Choose then the constants so as to make ( polynomial part of when zero , and we have . . . . .
( 86 ) .
If now we express all the other constants in terms of the constant , we find ; CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
189 ( 87 ) , and these satisfy the equations and the condition ( 86 ) .
It is noticeable that a solution can be obtained , in the form which can be made to satisfy all the conditions except ( 86 ) .
If , however , one works out this solution , it is found , as we should expect , to give infinite values for the stresses , all round the perimeter of the plane ends .
Thus , though simpler in form , this solution is not really simpler to work with .
I have given on pp. 217-219 the expressions for the stresses and displacements obtained from such a solution .
S 14 .
Determination of the Coefficients so to }Conditions at the Surface .
If we write down the expressions for the stresses , we find We have therefore to make Now we find easily Hence if we expand in a series of the we find The first term is zero in virtue of equation ( 86 ) , and using the relations ( 87 ) we find , whence , comparing coefficients ( 88 ) .
This gives us consistently zero right up to the plane end .
Next we have Hence we have to make -- ( 89 ) .
Now we have Hence if the right-hand side of equation ( 89 ) be expanded in the form CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
191 we find - .
whence using equations ( 87 ) we find after long but otherwise straightforward reductions Hence , equating coefficients on both sides of equation ( 89 ) we obtain the relations ( 91 ) , and it in ( 91 ) we substitute for their values in terms of deduced from ( 79 ) and ( 88 ) , we have , or .
( 92 ) .
and being known in 1 rms of , all the constants are now determined , except , and E. S15 .
of the Constants , E. Let us now apply the condition that : rzzdr If we write down the expression for , using the expressions for and in terms of , we find : CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
193 Hence But by(67 ) , and , so that we have to make .
Now it is easy to show that whence finally 2A ?
( 93 ) .
( 90 ) and ( 93 ) thus give us already two equations for , and .
We require a third equation .
This is obtained from the condition that VOL. CXCVIII .
\mdash ; A. 2 MR. L. N. G. FILON THE ELASTIC EQUILIBRIUM OF This gives Now where ( 94 ) , so that is a known constant .
We then find , putting in for its value ( 95 ) .
If now we write ( 96 ) , so that are known constants , then equations ( 90 ) , ( 93 ) , ( 95 ) may be re-written as follows:\mdash ; ( 97 ) .
Solving , we have : ( 98 ) , , CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
195 ( 100 ) .
All the constants are therefore absolutely determinate and the solution is complete .
S16 .
Expressions for the Stresses .
The reduced expressions for the four stresses are given below : ( 102 ) . .
( 103 ) .
( 104 ) , where , putting in for in ( 98 ) their values , \mdash ; ( 105 ) .
S17 .
Numerical Example .
Let us now consider a concrete example .
Take a cylinder whose diameter is nearly equal to its length .
This corresponds about to the dimensions used in practice for test .
pieces under pressure .
We will take in order to simplify the calculation of the -functions .
Further , we shall assume uniconstant isotropy , so that We then find : , , , ' ( 106 ) , , , , CIRCULAR UNDER PRACTICAL SYSTEMS OF LOAD .
and the stresses become : \mdash ; sm where MR. L. N. G. FILON ON THE ELASTIC EQUILIBRIUM OF S 18 .
Tables of the Constants the special taken .
The values of these constants I have calculated for the values 1 to 6 of and the values , i.e. , remembering that in our case , for .
These values are given in the following tables:\mdash ; TABLE of Constants .
CIRCULAR INDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
1 159100 . . ?
2 00 } MR. L. N. FILON ON THE ELASTIC EQUIh BRIUM OF The above , when substituted in the formulae , give quite fairly rapid yence when , th conyergency ratio being in this case less than unity by a finite amount .
But when , the series becomes mparable with the series 2 , where is a positiye integer , and , as in the first a special procedure to be adopted .
S 19 .
Methods the Curved Boundary .
When and are of course zero ; but the stresses and require separate evalution .
Now , if we use the series for and in descending powers of the argument , the first few terms of these give a very good representation of the function when is at all large , and will be quite sufficient for , at which point the tables of the last paragraph stop .
Replacing by these series , we find . . . . . . . . . . . . . .
( 107 ) , and CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SYSTEAIS OF LOAD .
201 Now write , , .
, where denotes the first four terms of the above series for , with a similar meaning for and .
Then if we substitute in the equations we find : Now remembering that we find , using \mdash ; ( 108 ) .
And , in a similar manner , .
VOL. CXCVIII .
\mdash ; A. 2 S20 .
Calculation of the Series in the Section .
If we work out the values we find they are as tabulated below : Hence the parts of the which depend upon converge quite rapidly enough to allow us to stop after the sixth term .
It therefore merely remains to evaluate the series and These cannot be expressed in finite terms , and although we may apply the EULERMACLAURIN sum-formula to these series directly , though in a slightly modified form , this sum-formula is not really of very great advantage , as its rapidity of convergence depends on , and is such , for certain values of this variable , as to render the formula useless as an approximation to the remainder .
As a matter of fact , however , the series were to be calculated only for values of being any integer from to 6 .
But for such values of , the cosine terms repeat themselves after . . . . . .
A precisely similar formula holds for .
CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
203 Thus we see we need only work out the series and where has integral values rangin from 1 to 12 .
These series are easily calculated , and , to them , the sum-formula is quite applicable .
By this means it was found that and , and the calculation of the stresses on the boundary then became a simple matter .
S21 .
Numerical Values of the Stresses .
The values of the stresses , referred to the mean pressure as unit , are tabulated below : of Stresses . .
13436 .
00896 94843 CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
205 In the above the values of for an additionalvalue of have been computed in order to exhibit more clearly the variation in the pressure along the radius .
The numerical results here given are graphically in Diagrams 7-10 .
We see at once that , save near the ends , the stresses , and do not differ very much from zero , which is the value they should have on the uniform pressure hypothesis .
On the other hand , the axial pressure deviates throughout from uniformity over the cross-section , the total variation in any section remaining tolerably constant over nearly two-thirds of the length of the cylinder , and equal to about 25 per cent. of the mean pressure .
Diagram 7.\mdash ; Showing Stress for Cylinder compressed between Rough Planes ( second example ) .
of cylinder Diagram 8.\mdash ; Showing Stress for Cylinder compressed between Rough Planes ( second example ) .
of We notice also , that , near the centre of cylinder , the greatest pressures occur at the centre of the cross-section ; whereas the reverse takes place at the ends , the pressure at the perimeter of the ends amounting to .
about times the mean pressure .
If we bear in mind the suggestion of the first problem of the present paper , that surface shear depresses those parts of the material towards which it acts , it is easy to see , physically , why such a distribution of pressure should be expected in practice .
The system of fi.ictional shears required to prevent the ends expanding will be towards the centre : the parts of the material round the centre will therefore be depressed , and the compressing planes ( supposed rigid ) ] have to compress the outer portions more than the inner , if the cross-section is to retain its original lane form , i.e. , remain in contact with the compressing planes throughout .
It is thus not surprising that the greatest pressure should be at the perimeter , being in fact nearly times the pressure at the centre .
Diagram 9.\mdash ; Showing Strcss Cylinder compressed between ough Planes ( second example ) .
engchs along Axis of cylinder Diagram 10.\mdash ; Showing Stress for Cylinder compressed between Rough Planes ( second example ) .
along Axis of cylinder We see also that , near the mid-sections , the -radial traction changes from a pressure to a tension as we go towards the circumference , so that the outer parts of the material are subject to two pressures , parallel to the axis and the radius respectively and to a tension , at right angles to these .
It should be noticed here that , in the diagrams , the ordinates representing the stresses increase negatively upwards .
This been found convenient in this case , where , owing to the general predominance of pressures , the greater part of the stresses have the negative sign .
CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
207 S22 .
Principal Stresses Point .
Lines of Stress .
Now , when we have a distribution of stress which is symmetrical about an axis , then the principal stresses are ; where is the same as before , and are two tractions in the meridian plane , making an angle with , where ( 110 ) , and the values of and are given by . .
Whence , using the tables in S21 , we find the values for and compared with the mean pressure over the ends .
TABLE of Principal Stresses .
The values of for the same points are given in the following table:\mdash ; of Slope of Lines of Principal Stress .
: 8806 By the aid of the above , we may draw the lines of principal stress , which is done in Diagram 11 , the slope being exaggerated in the ratio 10 : 1 .
In order to do so , we suppose the lin of principal stress to remain always in the neighbourhood of the same generator , so that , in the aboye table , the values in any row apply to the same gram 1.\mdash ; Lincs of Stress for Cylinder compressed between Rough Planes ( second example line of principal stress .
This of course is not correct near the ends , but it is sufficient for our purpose , the diagram bein merely intended to show the general course of the lines of stress .
These are sensibly parallel to the generators throughout the middle part of the cylinder , but slope outwards near the ends .
Diagram shows the distribution of the lines of equal principal stress In Diagram 12 , as in several others , is represen ted as having the same value as , so that the horizontal and vertical scales aro ot strictly the same , but differ in the ratio .
This has been done for convenience in plotting .
CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
209 inside the cylinder .
I have chosen the stress in preference to the others , as , except in the neighbourhood of the centres of the plane ends , it is everywhere the Diagram 12.\mdash ; Distribution of Principal Stress inside the Cylinder ( case of compression between Rough Rigid Planes ) .
The lines are drawn atintervals of of the stress .
The critical line corresponds to nearly greatest of the three principal stresses , and therefore will be the quantity when we come to discuss the maximum stress .
The diagram has been constructed by careful interpolation from the values tabulated above , and from it we see that the surfaces of equal principal stl.ess in the V0L .
CXCVII .
cylinder are in general made up of three sheets , and they fall into two classes : ( n ) those for which has a value less than a certain critical , which , as nearly as I find out from graphical methods , is about , and ( b ) those for which has a value greater than Q. The surfaces ( a ) consist of two solid caps or buttons , round the centres of the end sections , together with a hollow cylindrical shell surrounding the middle of the cylinder .
For values sensibly the latter sheet disappears , and only the caps remain , their volume gradually dwindling down to zero as falls to 686 Q. The surfaces consist of an elongated core , resembling a cylinder close by curved ends , surrounding the centre of the compressed block , together with two annuli at the ends , as shown in the figure eferred to .
The critical surface consists of two nearly plane sheets , roughly coinciding with the cross-sections , and one cylindrical sheet , which bends inwards towards the end , though without completely closing in , and which roughly coincides with the cylinder over the greater part of its surface .
S23 .
to Rupture .
Distribulion of Strcss , rence .
In considering what when material breaks , we have to ask , first of all , whether it be brittle or ductile .
In the first case , the law of stress to strain will be oximately linear ) to the where rupture takes place ; in the second case , the stress-strain relation remains approximately linear until a point is reached ( called the yield-point ) at which a large and sudden change occurs in the stress-strain curve , aftel which the material becomes sensibly plastic , so that rupture finally takes place after a large permanent deformation .
In applying an elastic theory to practiee , we can , in strictness , treat of rupture only in the case of a brittle solid .
Even then it has to be borne in mind that the mathematical theory of upon which the equations of elasticity depend\mdash ; equires the stl.ains to be so small that their squares are negligible .
It is possible that , even in the case of the most brittle solids known , this condition may cease to hold before upture occurs , although the stress-strain relation may continue to be linear .
Nevertheless , the calculated values of even the breaking straius in a material like cast iron , for instance , are so small as to render this unlikely .
For a ductile metal , such as mild steel , the elastic results only tell you where the lnaterial will begin to take permanent set .
In the case of stone or cement , however , to which the present results would be applied , there seems to no definite yield-point or elastic limit , the material bein in fact , only imperfectly elastic throughout .
Still , we may consider that- the result , \mdash ; CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
211 of elastic theory give in such a case an indication of the state of stress when the specimen breaks .
There are three distinct theories , both of rupture and of failure of elasticity .
According to and NAVIER , failure occurs when the greatest stress at any exceeds a certain limiting value .
This is also often taken as the criterion of absolute rupture .
According to SAINT-VENANT , the maximum strain , and not the maximum stress , is that which determines failure .
Finally , a theory has lately been put forward by various elasticians to the effect that failure occurs when the gTeatest shear at any point , that is , the greatest principal stress-difference , exceeds a definite amount .
Professor PERRY has proposed another criterion , suggested by the angle at which compressed cylinders shear ( see ' Applied Mechanics , ' pp. 344-345 ) , namely , that rupture takes place when exceeds a certain value , where is the shear across any element of area at a point , is the normal pressure on this element of area , and is a constant depending on the material .
This theory , however , need not concern us so much , as it appears more specially applicable to the final breakdown of ductile materials after they have become plastic .
On the other hand , it has been shown by Mr. J. J. GITEST ( ' Phil. Mag July , 1900 ) that the beginning of plasticity was very probably determined by the maximum stress-difference .
Let us now proceed to apply these three criteria , namely , those of the maximum stress , maximum strain , and maximum stress-difference to the cylinder in the present example , and see what results they give us , on the hypothesis that for materials like stone and cement , plastic yielding and rupture are simultaneous .
Consider first the greatest stress theory .
This would make failure of elasticity first begin to occur round the perimeter of the plane ends , and that as soon as a certain limiting value .
If the pressure be uniforlnly applied , and the ends expand , we get failure of elasticity when Hence the apparent strength of a cylinder tested in this way would be about of the strength of a cylinder tested under a distribution such as is usually assumed .
Further , if we consider the where the stress is greater than a given value we find that they consist of separate spaces , which join on to each other as diminishes , the critical value for which this occurs being given by Q. The regions of greatest stress consist therefore of a central core , which spreads out into a sort of hollow cone near the ends .
If then we suppose fracture to occur over regions of greatest stress , we see why it is that the material breaks off in conical pieces at the ends .
Consider now the greatest strain theory .
Let be ths three principal stresses , and the corresponding stretches .
Then , 212 .
N. G. FiLO ON THE so that the greatest will correspond to the greatest , if have the same sign .
This is our case everywhere , except in cases where , and then is so small that it still lsaves the strain corresponding to numerically the greatest .
We have then , remembering we have assumed , to investigate the values of .
This will be proportional to the greatest strain , except near , where should be taken .
It is found , however , that at this point the strain is comparatively small , and the maximum strain there is a matter of indifference .
TABLE of , where maximum stretch under the same uniform pressure .
It we take therefore the " " greatest stretch " " theory , failure of elasticity still occurs at the perimeter of the ends , but this tinle only when the stress is ( limiting stress in the case of uniformly compressed cylinder ) , so that although the apparent strength is less than in the uniform case , it is greater than if we adopt the " " greatest stress\ldquo ; theory .
The lines of equal principal stretch are shown in Diagram 13 .
They are drawn for only one quarter of the meridian plane , the rest being symmetrical .
They present the same general characteristics as the curves of equal stress , with this difference , that the critical line corresponds to .
Again , the caps or buttons at the ends are far larger ; so that if pieces are cut out , they will be considerably larger than on the " " greatest stress\ldquo ; theory .
Also , looking at the inclination of the linesjoining the corners to the critical points , i.e. , the points where the two branches of the critical line intersect , we see that the fragments , if approximately conical near their base , will probably be cut off at a much higher angle than in the previous case .
Let us now proceed to consider what happens if we adopt the third or " " greatest stress-difference " " theory of rupture .
It is easy to see from the tables of and that the greatest stress-difference is either or .
In the sixteen cases tabulated , for which , the first of these is the greatest stress- CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
213 difference , and in the twelve remaining cases the second is the greatest , although , as a matter of fact , the two stress-differences , for these twelve cases , do not diverge very much .
Diagram ) istribution of Principal Stretch , , inside the Cylinder ( case of Compression between Rough Rigid Planes ) .
The number corresponding to each lins value of for that line .
critical line .
The actual greatest stress-difference is given in the following table:\mdash ; TABLE of Maximum difference ) Here again plastic deformation will first occur round the perimeter of the ends when of the value it should have , on the same theory of rupture , in order to produce failure of elasticity in a uniformly compressed cylinder .
So far , then , this theory leads to the same results as the maximum stress theory .
Diagram 14 shows the distribution of maximum stress-difference .
The lines of equal maximum stress-difference present very similar characteristics to those of equal maximum stretch .
The critical line corresponds to a maximum stress-difference Q. Remarks similar to those used in the last case apply in Diagram 14.\mdash ; Distribution of Principal Stress-difference inside the Cylinder ( case of Compression between Rough Rigid Planes ) .
The corresponding to each line denotes the the ratio ( maximum stress-difference : for that line .
----------- , critical line .
Stress-difference Q. S24 .
Distorted the If we work out the values of when , we find , some reductions , .
( 112 ) where ( 113 ) .
Now in the particular case we are dealing with , if be large , approximates to Write then ( 114 ) , and substitute in ( 112 ) .
We find , putting in for 2 and 2 their known values , CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
215 , where , putting in now , we have the following values for :\mdash ; . . . . . .
Whence , using the methods of S 20 , we find the following values for , where lateral expan sion of a cylinder of the same dimensions under the same total pressure uniformly distributed : \mdash ; . . . . . . . .
30863 These results exhibited in Diagram 15 .
We see that the cylinder has a single bulge in the centre , and indeed it is easy to verify that by differentiating equation ( 115 ) and putting in the numerical values .
Diagram 15.\mdash ; Showing Distortiou of Curved Surface ( second example ) .
of cybinder This agrees with the figures shown by BACH in his ' Elasticitat und Festigkeit ' figs. of the distorted sbapes of such cylinders .
There are , however , in the possession of Professor KARL PEARSON , at University College , London , of iron which have been strongly compressed , so that the strain has been large and permanent and the meridian section of the distorted curved surface is an cbviously wavy curve , with two maxima , one on either side of the midsection .
With regard to the apparent disagreement here between practice , I would obserye that these specimens have been subjected to enormous stresses , for which the equations of elasticity certainly do not apply , and probably are not even an approximation ; in the second place , the specimens I have seen are longer , compared with their diameter , than the cylinder of the present numerical example , so that it is easy to see why the conclusions above need not apply to these specimens .
S25 .
Apparent Young 's and Poisson 's Ratio .
We find that the total shortening of the bar ( 116 ) , where .
Hence the apparent YOUNG'S modulus Ey ' ( 117 ) Now if be small , i. e. , if the bar be very long . . .
and , retainin only terms of 4th order , where is the greatest integral value for which makes the above approximation sufficient .
CULAR C UNDER CERTAIN PRACTICAL OF LOAD .
217 Hence if is very small , may be large , and thus although the first term in cancels , yet the terms in curled brackets will become indefinitely compared with the other terms .
Thus , for a very long cylinder , Ey ' where Ey is the true YOUNG'S modulus .
On the other hand , for a very short cylindel is very small , and being of the , the two leading terms in the numerator and of the right-hand side of ( 117 ) are negligible and This is identical with the modulus of compression for a cylindel is prevented from expanding laterally by a constant pressure applied to the sides .
that we see that for a very flat disc the effect on the modulus of compression is the same , whether the lateral expansion be prevented by means of shearing-stress over the flat ends , by hydrostatic pressure over the curved surface .
The apparent YOUNG'S modulus for intermediate cases will be between those two values these , uniconstant isotropy , being .
Thus , in the given example , where Ey ' PolSSON'S ratio comes out to be apparently instead of 2500 .
Thus the errors in the values of ) modulus and POISSON'S ratio , as deduced an experiment with cylindel .
S under the given conditions , will be 5 ) cent. per cent. respectively .
S26 .
Sohption involving Discontinuities at the Perimeter of iloe Ends .
In S 13 it was stated that a solution , obtained by methods strictly to those used in that section , but which neglected the condition that the shear should be continuous at the perimeter of the plane ends , could be found .
It seems of interest to give , for purposes of comparison , the expressions for the displacements and stresses , deduced from this solution .
They are : VOL. CXCVI11 .
Mlt L. N. G. FILON THE tSTIO EQUILIBRIUM where ( 119 ) , ( 120 ) , where Also ( 121 ) , an where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) where ) / ( 122 ) .
The method by which the constants are obtained is precisely the same as that used in SS14 , 15 .
The expressions for the stresses are given below : ( 123 ) , ( 124 ) , , ( 126 ) .
CIRCULAR CYLINDERS PRACTICAL SySTEMS OF LOAD .
219 It is now easy to see , if we bear in mind that when and therefore is large , remains finite , as appears on examination of ( 121 ) , how it is that the stresses at the boundary become infinite .
For in both and we have , when , terms of order 1 , when is laxge .
These are of alternate sign , containing if they become ] of the same sign , and the series become logarithmically infinite .
S27 .
of Resnlts .
Looking back upon the results obtain ed , we notice : That the three solutions we have been considering are only the simplest of an infinite series of solutions , which are continually growing more complicated ; for we need not necessarily stop , as has been done , at terms of the fifth degree , but might go on to terms of any degree in and and thus construct , as it were , solutions of successive orders .
We should then have an infinite numl ) of free constants , which might be determined by introducing further limitations at the plane ends , such as , for instance , restricting , to be zero at every point and not merely along the perimeter .
The analytical complexity of such a complete solution would , however , be very great , and would render it quite beyond the reach of arithmetical expression , and consequently valueless for the purposes of the ngineer and the physicist .
No attempt has therefore been made to develop this solution , although , as an analytical possibility , it appears interesting .
That the different solutions all agree in giving the perimeter of the plane ends as the locus of the points where the elastic limit will first be passed , one of these solutions actually making the stress infinite at this perinneter .
In the more important solution , however , where continuity and finiteness are preserved , the conclusion still holds , and , further , is independent of whatever theory of tendency to rupture we adopt , whether we suppose it due to maximum stress , to maximum stretch or squeeze , or to maximum shear or stress-difference .
That in the numerical example considered , plastic deformation begins to occur round the perimeter for a stress between 2/ 3 and 1/ 2 of that which is required to cause a cylinder under uniform pressure to pass the elastic limit .
This is apparently in contradiction with the results of engineering experience , both UNWIN and PERRY stating that blocks of stone or cement , pressed between millboard , which hinders the expansion of the ends , show greater strength than the blocks when the ends are allowed to expand .
The key to this appears to be found in a emark of UNWIN , which Professor EWING confirms , that the lead sheets do not merely allow the expansion of the block , they force it , i.e. , lead in its plastic state will expand more than the stone or cement would do laterally under a uniform axial pressure .
N. .
FILON ON TH ELASTIC OF Bnt the solution , when the ends are compelled to expand by a given quantity is easily deducible from that given for non-expanding ends .
Thus , let be the of and in the case worked out , when , and let us write this will satisfy all the conditions , provided nnd Therefore ?
' the solution under a given mean pressure , which produces a flow of lead plate , and thereby constrains the ends of the test piece to expand by that amount .
The principal stress at the perimeter of the section is now and if , be made large enough , this can be made much smaller than Q. It begins to be smaller than as soon as the expansion induced by the flow of the lead is greater than the natural expansion of the stone under uniform pressure .
On the other hand , the principal stress at the centre of the plane ends is and this again may be made great by making large .
The principal stress-differences are : at the perimeter at the centre P. Hence we see that whatever theory of failure we adopt , if the ends are forced to expand , so that , the material first becomes plastic ( or else breaks ) at the centre of the cross-section , the strength of the test-piece diminishing as increases , but having no definite value .
That some such thing as this does really occur in practice is very well shown by the results published by UNWIN ( ' The Testing of Materials of CIRCULAR RTAIN PRACTICAL OF LOAD .
Construction where blocks show less strength when three sheets of lead are introduced between the compressing planes and the test piece than when one sheet only is introduced , the lateral flow being greater in the first case owing to the amount of lead .
It would seem , therefore , as if the true strength of a cylinder were really greater than its strength as tested either between millboards or between lead sheets , and not , as Professor PERRY states in his 'Applied Mechanics , ' equal to the strength shown in the lead test\mdash ; this test , as we see , leading to results that are not definite , but vary with the expansion of the lead .
The millboard test , however , which is advocated by UNWIN , a constant value , although it is not the value which would hold for a cylinder umder uniform pressure .
Diagrams 12-14 suggest an explanation of the fact that , when short cylinders are strongly compressed between very hard surfaces , pieces are sometimes cut ont the ends of an approximately conical shape .
The same occurs when spherical pieces of metal , such as ball-bearings , are compressed between parallel plates .
This is usually explained by saying that the material breaks along the planes of principal shear .
On the other hand , it may be argued simply that rupture should take place over the regions of greatest stress .
These are near the perimeter at the ends , and gradually close in upon the centre , forming hollow caps .
Further , in the case of the lead tests , where , this state of things is reversed , and the material should give way from the inside , so that we should expect it to split axially , and possibly along meridian planes as well .
That this is what really occurs can be ified by referring to the figures in the chapter on testing of stone in UNWIN'S ' Testing of Materials of Construction .
' The results both of this and of the first problem show us how unreliable any experiments on short cylinders must be , which have in view the determination , by tensile strain , of either YOUNG'S modulus or POISSON'S ratio .
Thus any results obtained in such a case without the dimensions and the mode of application of the stress being exactly specified , would not justify us in general in drawing any conclusions as to whether a given material possesses or not uniconstant isotropy . .
The Third Problem .
of Torsion .
for the Displacement Stresses .
I now proceed to consider a case where ?
and are zero , that is , where we have to deal with the solution in , which we have seen is independent of the others .
We have in the notation of Hence , excluding -functions , since the solution mnst be finite nnd continuous at the origin , we have , .
G. FfLON ON TH F , LASTIC EQUILTBRIUM OF Now , if the torsion be symmetrical on either side of the mid-section , we have .
, when , therefore But also at the ends .
erefore or .
the well-known property of the -functions Now suppose that the cylinder is subjected to a certain system of transverse surface sbears , so can be expanded in a FOURIER'S series in the form or Hence .
( 127 ) , ( 128 ) , ( 129 ) , where have the same meanings as before , In the case where is very small , or the cylinder is very long relatively .
to its diameter , we may obtain a first approximation by retaining only the first terms in the expression for the , Proceeding as in S6 , we find CIRCULAR CYLINDERS CERTAIN ACTICAL SYSTEIIS OF .
223 sin .
Now if , we see that Now if be the total torsion moment to cross-section , angle turned through by a radius say .
efore torsion at the point herefore Therefore But these are the actual formulae connecting the torsion with the applied couple with the shear across the section for a circular cylinder .
We see , , that the usual continue to hold , to the first ) tion , when the forces applied to the surface of the cylinder with ) rovided we define our torsion-couple at any section ( much as the bending moment at any section of beam is defined ) , as the couple of all the external applied forces to one side of that cross-section .
It is interesting to note also that , to this , there is no distortion of the -sections , being constant for the section .
Straight radii therefore emain straight radii .
Further , ( its value at the boundary ) .
In other words , the transverse shearing-stress across cylinders coaxial with the give one is zero sections where there is no such external applied stress , and other sections diminishes rapidly along the radius as we go inwards , so that near the centre it is always small compared with There is one very important point to be noted with regard to this method of approximation : and increase with , and therefore , however small may be , so long as it remains finite , we still reach a value of , for which it is notjustifiable to take for and the first terms of their expansions in positive integral powers of the argument .
If , however , we stop at the term , where is finite , then if is the remainder after terms of the series , for example , and if , on the other hand , the numerical value of the difference \mdash ; ) for all values of not greater than , where is a quantity which depends upon , and which can be made as small as we please by making small enough , then the rence must be numel.ically less than where denotes the numerical value or modulus of and R is the remainder after terms of the sel.ies Now if both the original series and the approximate series are unifo1mly convergent , then by giving a certain value , and can both be made less than a certain small quantity which tends to zero when tends to infinity , and that for all values of Now so small that , which we can always do so long as is finite .
Then the difference between the two series is numerically , and the approximation holds .
If , however , for any value of , it becomes impossible to assign an upper limit to or , i.e. , if eithel series cease to be uniformly convergent , then we should have to increase indefinitely in order to make , and thel.efol.e to modify , so CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
225 that no limiting value of ( which should , of course , be independent of z ) could be found , and the approximation need not necessarily hold .
As a matter of fact , it is hown in S29 to fail for particular cases .
This is true a fortiori , if either series cease to be convergent at all .
The same remarks apply in their entirety to the process of approximation given in S6 , and further , to the approximate expressions given by Professor POCHHAMMER in his investigation on the bending of beams ( ' Crelle , ' vol. 81 ) .
S29 .
Special Case of Two Rings of Suppose that we have the following system of values for if il if so that we have a cylinder twisted by two equal and opposite rings of transverse shear extending over lengths of the cylinder , near the ends .
we find easily with the following values of the displacements and stresses : Now it is easy to see that in this case the conditions for uniform convergency are satisfied , except at the boundary , and except with regard to the stress , whose approximate expression is not uniformly convergent , being in fact discontinuous for .
At the boundary , tends to unity , its approximate expression , when is large , being ( 132 ) .
Hence is always uniformly convergent and its approximate expression likewise , so for it the approximation , for sufficiently small values of , holds throughout .
VOL. CXCVIII . .
2 For the series is non-uniformly convergent for in the neighbourhood of the owing to the series having a finite discontinuity .
For the approximation certainly fails , for , in the neighbourhood of for part of the expression for is the series , which is divergent .
Hence , it we are in such a case to use our approximations for the stresses , we must exolude the sections where the applied stress is discontinuous and their immediate neighbourhood from consideration .
It will be found that similar remarks apply to the example of pull given in S 7 , and also to the example given by Professor POCHHAMMER in his paper on bending ( loc. cit in which he also deals with discontinuous systems of stress , so that his approximate expressions leave us in the dark as to what does really happen at points support , the cross-sections in the neighbourhood of such points being , for reasons analogous to those developed above , excluded from the regions where his approxima- tions hold .
Before proceeding to an actual numerical concrete case , we may notice that becomes infinite at the points for the causes stated above .
Hence any discontinuity in a system of transverse shears applied to the surface of a cylinder , or any such shear transmitted by a grip applied to a portion of the material at a sharp angle , will produce in the neighbourhood a very great stress across the section .
It would seem , therefore , that a cylinder treated in this way would be likely to experience plastic flow , or to rupture , not in the middle , but near the points where it is seized .
S30 .
Approximations on the Bound when the ylinder is short .
When the cylinder is , so that becomes rapidly large , we may use the method employed in SS 8 , 9 , and 19 , availing ourselves of the approximation ( 132 ) .
We then iind : CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF LOAD .
227 , ( 133 ) , andc -from z in ' , where in the last 2 in both equations only a comparatively small number of terms need be taken .
S31 .
Nurnerical Example .
Values of the Coefficients of the Displacement and Stresses .
Take a cylinder such that , and let , so that the distribution of stress is as shown in fig. 4 .
Then , and we find : . . .
the law of the signs being obvious , and the , and 's being given : MR. L. N. G. FILON ON THE ELASTIC EQUILIBRIUM OF TABLE of and TABLE of .
oooooo .
oooooo From these , using the formulae of approximation given in the last section , when , we obtain the values of and the two stresses .
I have tabulated them in the form and stress ) where are the greatest values of the displacement and of the shear respectively in a cylinder of the same length , subject to a uniform CIRCULAR CyLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
torsion over its whole length , the total couple applied being the same as in the present case .
We find and for the given example .
Tables are given on page 229 .
Looking at these tables we see that , oyer the length free from external applied shear , the strains and esses inside the cylinder are sensibly the same as what they would be on the hypothesis of a uniform torsion .
Outside the torsion couple diminishes , and the stresses diminish in consequence .
It is interesting to compare these results with those that we should have obtained if we had supposed the approximate results given on p. 223 to hold good in this case .
Denoting by the values of t , he displacement and stresses calculated on this hypothesis , we have\mdash ; TABLE of TABLE of TABLE of 000402060 CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SySTEMS OF LOAD .
231 S32 .
Discussion the Results .
From the above tables we see that the radii in each cross-section do not remain straight lines , but assume distorted shapes , which are shown , on a very exaggerated scale , on Diagram 16 , where , for each of the ten cross-sections , the curve of , which indicates the deviation from the straight lin in the distorted form of the radii , has been plotted .
The variation from the straight lin increases rapidly as we approach the region where the stress is applied , as can be seen from curves ( 1 ) on Diagram16 .
On the other hand , towards the ends , the distortion remains fairly constant .
The distorted radii meet the bounding circles at right angles when at the sul.face , but they meet it a finite angle where From the values of see that as soon as we get at all away from the ends the conditions that , which hold for uniform torsion , are very closely satisfied , that , more generally , except where the abrupt change takes place in the shearing stress at the surface , the approximate expressions given in S 28 do not differ widely from the true expressions , the law that varies as the square of the radius being , near the ends , tolerably well verified .
It is to be noted also that , where the approximations would give a discontinuity in the material ( viz. at ) , the true values are almost exactly the mean of the two discontinuous values obtained from the approximate formulae assumed correct .
In like manner is nearly the same as , except near , where , as we have seen , an infinite stress really occurs , of which the approximations give no hint .
We note , however , that does llot strictly vary as all over the section , being smaller than should be expected inside and larger at the boundary .
The theoretical result , that the stress is infinite where the transverse applied shear is discontinuous , throws much light on the case of a cylinder whose cross-section abruptly changes , as in fig. 1 , with the difference that now the stress applied to the collar is transverse .
We see that in such a case we should expect the material to give way at the points of sudden change .
This conclusion is in accordance with practical experience , the tail ends of propeller shafts , for instance , breaking almost invariably in this manner .
S33 .
General Conclusion .
This example concludes the series of three which it was proposed to treat of .
The object has been to obtain a clear idea of the effects of certain surface distributions of stress which come much nearer to the cases arising in practice than does the uniform distribution ordinarily taken .
Diagram 16.\mdash ; Showing Distortion of a Radius originally Straight in the case of Torsion produced by applying Shearing Stress to the Curved Surface .
The curve corresponding to the section .
is numbered .
The first four curves have had the ordinates exaggerated in the ratio of 10 : 1 .
They are ) by the dotted lines , and to them refer the numbers in brackets .
CIRCULAR CYLINDERS UNDER CERTAIN PRACTICAL SYSTEMS OF No doubt the cases treated involve somewhat arbitrary conditions , not strictly obtained in practice , but it appeared useful to ascertain how far they gave results diverging from those which would be found on the ordinary hypothesis of uniform extension or torsion .
This furnishes us with a test of how far we may accept DE principle of " " equipollent\ldquo ; systems of load for a bar whose length is gradually made smaller compared with its diameter .
The results we have here obtained indicate that , as we go away from the points of application of the stress , " " uniform\ldquo ; solution is reached much sooner in the case of torsion than in that of either tension or pressure .
With regard to the arithmetic of the paper , the results have been as far as possible checked .
It is believed that they are correct to the number of figures given , but owing to the slow convergence of certain of the series , accumulated errors may in some cases affect the last and even the second last figure .
Even this , however , would not sensibly disturb the conclusions .
For the -functions the tables in GRAY and MATHEW'S " " Bessel 's Functions\ldquo ; were used , but the range of the tables is so limited that a large number of these functions had to be independently calculated .
The semi-convergent expansions were employed , the argument being large in each case .
My very best thanks are due to Professor EWING for his kindness in coming to my aid with suggestions and advice .
cxcvIII .

