STRESSES IN A PLATE DUE TO THE PRESENCE OF CRACKS AND

SHARP CORNERS.
By C. E. I~GLI::;, Esq., 1\f.A., Fellow of King's College, Cambridge.
[Read at the Spring Meetings of the Fifty-fourth Session of the Institution of Naval Architects,
March 14, 1!)13; Professor J. H. BILES, LL.D., D.Sc., Vice-President, i.n the Chair.]

PART I.
THE methods of investigation employed for this probl elll a.re mathematical rather than
experimental, and this mathematical treatment is given in some detail in P art II. of
this paper. Part I. is a summary of the more important results and conclusions, and
in this part mathematics are kept as far as possible in the background.
rr he main ·work of the paper lies in the determination of the stresses around a hole
I
in a plate, the hole being elliptic in form. The results are exact, and are consequently
applicable to the extreme limits of form which an ell ipse can assume. If the axes of
j the ellipse are equal, a circular hole is obtained; by making one axis very small the
stresses due to the existence of a fine straight crack can be investigated.

!
I
The destructive influence of a erack is a matter of common knowledge, and is
particularly pronounced in the case of brittle non-ductile materials. This influence is
turned to useful account in the process of glass cutting. A fine scratch made on the
!;Urface produces such a local weakness to tension that a fracture along the line of the
scratch can be brought about by applied forces which produce in the rest of the plate
<}Uite insignificant stresses.
In ductile materials some easing off of the local stresses at the end of a crack will
be effected by plastic yield of the substance, but for the case of alternating loads the
protection against breakdo,vn gained thereby must be limited. A small load tending to
open the crack will produce overstrain at its ends. On reversing the load the crack
closes again, but not before it has set up some reversed stress at the ends of the crack.
In this mrmner a small alternating load may produce in the material an alternating
stress far in excess of its elastic range, and under these circumstances, if a crack has
once fairly started, no amount of ductility will prevent it spreading through the
substance.
220 STRESSES I N A PLATE D UE TO

In a paper read before the Sheffield Society of Engineers and Metallurgists in

January, 1910, Professor B. H opkinson made some general obserYations concerning tht'
stresses around a crack In this paper Professor Hopkinson did not a.ttempt to
compute the intensities of these stresses, but the desirability of doing so he strongly
emphasised. The present paper is an endeavour to satisfy this requirement.
Fig. 1 (Plate XXVIII.) shows an elliptic hole in a plate, the major and minor axes i
I
being 2 a and 2 b respectively. The plate is subjected to a tensile stress of intensity R i
applied in the direction sho\Hl. If the hole is nowhere near the edge of the plate
and the material is nowhere strained beyond its elastic limit, a tensile stress occurs
at A of the valueR [1 + ~t} and a compression stress at B of magnitude R. Ou
exploring the plate in the direction A P, the tensile stress across A P, which has the
2
Yalue R [ 1 + ,:'] at A, rapidly decreases, and in a short distance attains approxima.tely
its normal value R. Advancing along B Q the compression stress across B Q which
starts with t he value R at B soon changes to a small tensile stress, and this gradmtlly
dies ont entirely. Stresses in the direction s A P, B Q are like,Yise brought into
existence. T hese radial stresses are zero at A and B. At considerable distances fro1u
the edge of the hole the former dies down again to Y-cro, and the latter attains the
Ynlue R.

Th e variations in the stresses along and across A B and B Q are shown by the
c·urYes of Fig. 2, which are accurat.ely dravm for the case a = 3 u. An inspection of
these curYes clearly indicates that the tension at A is by far the greatest stress. In this
case it amounts to 7 R. The rapidity with which this stress decreases with the clistan<:e
from the e.dge of the hole is also Yery noticeable. 'J1hese effects become more mrtrk(•d
. (t .
as t l1e ra t 10 b met·eases.

When ~ = 100, the tension at A is 201 times the mean tension.

(1.
,, !J = 1,00(), ,. ,, 2,001 , ,
"
The ellipse in this latter case "·ould appear as a fine straight craeJ,, and a Yery swn.ll
pull applied to the plate across the crack would set up a tension at the ends sutlkient
to start a tear in the material. The ine;rease in the length due to the t<:>ar exagg<:>rates
the stress yet fnrtbcr and the crack continues to spread in the manner charact<:>ristic
of cracks.
So far we have taken the major axis of the ellipse to be perpendicnla.r to the
direction of the pull. If the tension is appl ied in the direction of the major nxis, t~nsile
 THE PRESENCE OF CRACKS AND SHARP CORNERS. 221
Rtress of amount R [ 1 + ~{' ] is set up at B, accompanied by a compression stress of
magnitude Rat A. H ence a crack running in the direction of the pull does not produce
great local stress, a conclusion which is almost self-evident.
If the major axis of the ellipse makes an angle () with the direction of the pull,
the tensile stress at the ends of this axis is-

R [ 1- cos 2 e ( 1 + ~) J
Por such a case, however, the greatest tension does not occur exactly at these
ends, and the value given above may be considerably exceeded. An example of special
importance illustrating this point is given in Fig. 3 (Plate XXVIII.). Here the axis
of the ellipse is inclined at an angle of 45° to the direction of the pull.
At A the tensile stress is R t
.\.t P t he tensile stress is-
R[1 + ''b + " ] f/

Between A and P t he tensile stress reaches it waximnm value, and if~ is fairly
small, a good approxim n.tion to this maximum value is-
H, rt [ 1 + v't u 1+ '!.t,~]
t. b - -,i-T -
At Q there is compression stress of magnitude R.
For t he case a = :3 b-
The tensile stress at A is 3 H. ;
The tensile stress at P is 41- R ;
rrhe compression stress at Q is R.
'rhe maximum tensile stress which occnrs between A and P has the accurate
value 4·64 R.
The approximation given above makes it 4·8.5.
For smn.llcr values of £t his approximation is yet more accurate.
ExTE.:SHIO~ TO THE CASE o.F CRAUKs WHI CH AHE NOT ~ECE!:i SAHILY ELLIPTIC IN Fon:u.
Consider an elliptic hole in the plate modified by cutting a\\·ay the portion sho,vn
shaded in Fig. 4 (Plate XXVIII.).
From the point of view of stress, this is equivalent to applying along the
boundaries P Q Rand 8 V W distributions of stress, 1Yhicl1 in each case is an eq uilibrium
 STRESSES IN A PLATE DUE TO

systelll, and \rhich dies out as the points of con tad ,,·ith the ell ipse 11:re approached.
The 11:pplication of such a system does not substantially affect the stress outside tlw
aetna\ region where the distribution is applied, and for the case shown the stresses
at A and A' will not be altered to any appreciable extent, provided that the change in the
boundary does not extend to these ends. In other words, the stresses at the ends of a
caYity depend almost entirely upon the length of the cavity and the form of its ends.
If the ends of the ca,·ity are approximately elliptic in form, it is legitimate, in
calcu lating the stresses at these points, to replace the cavity by a complete ellipse hnsing
the same total length and end formation . If p is the radius of curvature at the ends
of the major axis of an ellipse, b = ..; a p~ This substitution forb gives the tensile stress
at the ends of the ellipse in the form--
R[l+2V~J
This formula will accordingly apply to a. cavity of any shape, the length of the
cavity being 2 C£ and the ends having a radius of curvature p; provided that the cavity
near its ends merges smoothly into an ellipse.
Thus for the star-shaped hole represented in Fig. 6 (Plate XXVIII.), in which the
ends merge into ellipses in the manner indicated, the tension at A will be R [ 1 + 2 V ~· }
where p is the radius of curvature at the ends of the projecting arms. The compression
stress at B will be R .
CAsE OF A SQUARE HoLE WITH RouNDED ConNERs.

To get the square hole with corners rounded to a radius p illustrated in

Fig. 6, t he cutting away process has to be more extended, and this may ease the stress
at A somewhat below the valueR [ 1 + 2 V ~ ], but the error contained in this formula
is in all probability quite small.
To this same degree of accuracy the tensile stress at A for the case illustrated in
Fig. 7 is
R,y'I .
p
At P there is a tensile stress
R [ 1 +vf+ v ~ J
At Q there is a compression stress of magnitude R. The greatest tensile stress
along the rounded corner oecurs at a point between A and P, and, if p is small
compared with l, its approximate value is
R
2
V-z[ 1 +
I'
v' 2 t + 2 r
v'e - Jp
J
·
 THE PRESENCE OF CRACKS AND SHARP CORKERS.

ExTE~l;ION TO THE UM;g oF A CRACK STAHTING FRO)I THE EDGE OF A PLATE.

Consider once again the plate illustrated in Fig. 1 (Plate XXYIII.). If one half 1:::;
isoln,ted from the other, the notched phtte shown in Fig. 8 is thE> result. This plate, in
addition to a, hori;~,ontal pull of intensity R, is subjected to normal actions distributed
along B Q, B ' Q', according to one of the curves shown in Fig. 2. If this latter
distribution can be neutralised, the stresses in a plate with an elliptic notch in its top
edge, subjected to longitudinal tension only, can be deduced.

The tensile stress at A is accordingly R [ 1 + 2 V ~ J, less the tension due to the

distribution of normal stress along the top edge. \Vhen the notch is narrow and deep,
these actions along the top edge produce little or no effect at A. If the pressure at B
were maintained all along B Q, the tension produced at B would only amount to R.
Actually this pressure dies out rapidly, and since t.he total action on B Q amounts
to a zero force, the tensile stress set up at A will be much less than R; it may very
likely vanish altogether. At any rate, we may conclude that these actions along the top
edge are unimportant, so far as the stress at A is concerned, and for the case of a narro>v,
deep, elliptic notch, the tensile stress at the bottom of the notch lies between the
limits-
R[ 1+2 V~ ]and R[2.vf].
where a is the depth of the notch and p is the radius of curvature at its end. Th e
former hmit -
R[ 1+2 '\/~]
is probably the closer approximation, but for the case of a deep sharp-ended notch, the
difference bet>veen these limits is relatively unimportant.

EXTENSION TO THE CASE OF A NoTCH WHICH IS NOT NECESSARILY ELLIPTIU IN F omf.

Having deduced the tensile stress at the bottom of a narrow notch of elliptic form,
the formula cn.n be seen to have a wider application.
Thus, in Fig. 9 (Plate XXVIII.), the shaded portions can be remoYed without
appreciably modifying the tension at A, which will continue to have the value-

R[l+2V ~ ].
I'

where a 1s the depti1 of the notch and r> the rn.dius of curvature at its end. The
 STRESSES IN A PLATE DUE TO

argument which leads to this conclusion is identical with that given in connection with
Fig. :3, and calls for no repetition.
By similar reasoning the tensile stress at the point A for the cases illustrated m
Fig. 10 (Plate XXVIII.) will be-

For the rectangular notch represented in Fig. 11-

The tensi le stress at A is
RVlp .
The tensile stress at P is

l{ [ 1 +.y ~ +v ~ }
'rhe compression stress of Q is R .
The greatest tensile stress along the rounded corner occurs at some point between
A and P, and, if r is small compared with 7, a good approximation to its va.lue is

R,Yf [l + ~-~~ + ~PJ·

2 p Vl-Vp

'l'he la::;t ce:1.se to be considered is a small crack or notch springing from the side of a
hole in the manner shown in Fig. 12 (Plate XXVIII.).
In this case there is a double ma-gnification of stress. 'l'h e m ean stress R is
concentrated to the value R [ 1 + 2 V~] in the neighbourhootl of the cr::tck, and this,
n.gain, is magnified to the value

at the end of the crack, where ct and fJ refer to the hole and a' ,l refer to the c·.rack.
'!'his exampl e offers an explanation of the weakening of a plate which bas been
punched with holes. .-\.ronnel the edge of the holes fine cracks may be started. These
cracks will have every inducement to extend, becc.\use the meta,! round the hole has been
rendered hard and brittle by the punch. By tbe time the crack has extended through
this hardened region, it has got such a hold on the plate that no amount of ductility will
prevent the crack from advancing. The advisabil ity of removing this bard and probably
fn\ctured rim ronnel the edge of a punched hole is very apparent.
 THE PRE. ' EXCE OF CRACK::> AXD , 'HARP CORXERS.

P.\HT li.
l)J.:TE IL\11;\.\TW:\ oF THE 8THESSES IX _\ PL.\TE WHICH 11.\S :\;\ Eu.I PTW Hor,E.
'l'hronghont thr work <·nrYilinear co·ordina,tf's n,n· <.'lllployrd.
J ~ct u= <·onsln.nt n,nd j-3 = eonstn,nt, defint' two systc iiiS of <:\li'Yt~s intersecting at
ri ght n.uglrs .
At n,ny point lrt ''··-' 11 ,1 denote the sh ifts in the direct ion of Llw normalR to a and {3.
L<'t ''.... , '',u. ''.. ,1 denote t he two stretches and the slide <·OrrC'sponding to these
direttions.
'flwn -

f',J,J = h: ~_,.,
'ltp + 1
11 1
c ( 1)
ft : 11 .. • -
G/) cu h,

t'., 1; = 1h, , ~- ('

ll: U ,1) c ('t/
+ 1h.:. ;--; 1 " •
)
(/~(.(1 fl,<p
where -
h/ = (~ (I)~ + (~ (I) ~
G ,(' C !/

li/ = (~.!2)2 + (~ f3)2

( _,. ( !/
The diln,tion f:l. is given by-
.l = 71 , "~ [ l- (I(") +-)~,/> ("·')]
I') (( ft2 Jt (' I

'l'he rotation 'v is giYen by-

For th e particuln.r problem under consideration the curvilinear co-ordinates a, {3, are
such thn,t-
x = c cosh << cos p
!I = c sinh n :,;in (3.
,.~ y2
11 =constant is accordingly the ellipse -.. - -
c• cosh~ u
+ c:-.,--:--
h.
sm • a
= l.

_,.2 y2 -
il = constant is the hyperbola .. cos---~
~- /> - c:-.• ,.;m-
. ., />
-, - 1.

In this case

Q
226 STRESSES IN A PLATE DUE TO

GE:\ERAL STRESS STRA[N EQ UATIO:\S .

These takr the forru-
(l - a) ~
0a
- (1 - 2 IT) ~ "' = 0}
() /i

oJ>~ + (1 -
0
(1 - u) 2 u) ?w = 0
ca
'Where u is the value of Poisson's Ratio.
These two equations state that-
(1 - a) 6. + 1 (1- 2a) w is a function of n + 1 J3
e - " (ro + 1 /1)
Take (1 - a)~ + 1 (1 - 2 a) w = constant x - . - -
smh (a + 1 {3)
( L- a) O.=const.·- e -Cn -1) .. cos (n+_ll t3_+ c - C••+~~co:~0 -1)!3
So that
cosh:.! n - cos :l /3
) _ t e - (n - 1lo sin (n+l) ,1 -e-< 11 +1l usin (ll- l )i3
(l - -9 6' w - cons . .
cosh2a - cos :l O
Ha U
tt = - a and v = -
h h
0

6.
li.2
= ..!!.!!.._
1-a
[e - (II - llo cos (n + 1),0 - e- (n + ll«cos (11- - 1) (3 ]
0_2'· _ ou = -2~= 2~n [ e - (•1- ll «sin (ll + 1)/3- p- (n+ ll<•sin (u -1) {3]
o {3 oa ~ 1-~a

From these two partial differential equations u n.nd v can be determined--

H = A,. ( (11 + 11) p - <11 - ll a cos (n + 1) {3 + (n - 11 ) t: - (n+ ll " cos (n - 1) {3 J + q•
1' = A, ((11 - Ji )C - tn-1) .. sin (n + 1){3 + ( n + Ji )c - (n+ l ) .. f!in ( 11 - 1)/3 J + ;1

Where JJ stands for 3- -1 u , and 4> and !/;are conjugate functions of a and j3 satisfying
L aplace's equa.tion.
Suitable values for 4> and ·!/; are, const. X e - nu cos n (3 n-nd co nst. X t -~~ .. sin n {3.
From t hese general Yalues of u and v values of IJ.," e,1, 1 e.. J1 can be deduced by means
of the relations-

c. ,1 = • Z (h 2 v ) + ~i, (112 11 )
'' a ( Jl
 THE PRESENCE OF CRACKS AND SHARP CORKERS. 227
Thr stress components can then be determined by the equations-

R.... = -- E -
l + a
[e. . + -1-tu
"-.- ~ J

These two sets of calcnlations are too lengthy to be given in detai l ; the results are
as follows :-
(n+ 1) e - <n- 1)« cos (n+:1) 1J+(n- l) I' -(n+llu cos (n - ~ )13]
- { -1 e - l•+llu + (n + e -<n-~)u}
:n cos (11 + 1)/3 A ,,
[
., ( h2 , _ +{ -1e - <n - l)u _ (n-3)e -<1~'·3lu } cos(n-1)f3
I·' uu COS Cl - COS 2 p) 2 -
11 e llu cos (n + :1) t3 + (n + 2) e-<-.+llu
- (IH cos (n- 1)/3] B
+ [ - { (n + 2) e - (n- l)u + 11 e- ut3'u} cos (n + 1)/3 "

- (n- 3)e-<11 - h · Col:l(ll +:~)P -(11+ :3) e-<n+lluco:> (n - :~)p]

+{(n - 1)e-<n-3lu - .J e-<u+t>u} cos (n + 1)/3 A.
[
Hp, (cosh 2 a - cos 2 J3l
1
= + Hn +1) e-<,+Sla +
! e-<n- l)u} COS (n - 1)/3
_ [n e- <n+l)a cos (11 + 3) /~+(II+ 2) e- (>I+IJ.• cos (n - 1)!3]
- { (n + 2) e-rn - IJu + n e- (ll+~l"} cos (n + l )f3 B,.

. [ <n - 1) e - <u- 1).. sin (n+ :I) ,3 + (n+ 1) e-<n+l)u sin ( 11 - :J)13]
- (n+ 1)e - ln-3Ju sin (n+ 1)/3-(n - l) e-<•t+S)« sin (n - l )t3 A ,.
S.,,l (cosh 2rc - cos 2f3p = r ne-<t~+lJusin(n + :~)(3 + (n + 2) e - <n+ll" sin ( n -1 )!3 ]
+ - {(n+2)e- <11 -ll·•+ne -("J:I'" } sin (n +l)t3 B,.

In these fonuulie for H.... , R ,1;1l su;.ll Jl. can be any integer positive or negatiYe,
and the general expression for these three stresses takes the forJU of an infinite series,
i1wolviug a number of arbitrary constants, which luwe to be determined by the
conditions oxistillg n.t the inner and outer boundaries of the plate.

UMm Ol-' A Pr. ATE ~t: u.JE<.:TED TO A TENSILE DTREss R IN ALL DIRECTIONS, THE P LATE
HA \'I:-;G AN ELLIPTIC H oLE DEFINED BY cc = a 0.

rl'he boundary <·onditions for this case arc-

H..... = S..p = 0, when u = u 0 ,
ant! H..u = R, R.sil = H, S,..p = (l, when rc is lal"g·t>.
22 I STRESSES L.\ A PL.-\TE DUB TO

Thr:;c conditions <:<'tn be satisfied by making-

A_, =- Rl'i•'
({
.\+I = - ''
1:5

13 _ , = ~ cosh :! 11,..
Adding together the three terms COlTt'Rponding to A_,, A+, and B _1, the exact
values of the stress for this case are-
H _ Rsinh 2 u [cosb 2n- cosb2 11 0]
"" - [cosh :! n - cos ~ 1)]2
R _ R-~~ h 2 n [_cosh 2 u + cosh 2 <~o - 2 cos~ J3]
,l,l - (cosh;! u - CO!:li J3r
~ _ H sin 2,3 [co!'h 2 a - c~s 2 ~3]
' "; - [cosh 2 - cos 2 J3Y

..~.l ong the <'dge of the elliptic hole-

2 Hsiob 2"o
H,J,J = cosh:! u 0 - cos 2J3

l'A~E. oF A PLATE su3JECTED

(.1 7T
TO A T ENSHJE STn1;:ss
IN TRE DIRECTION f..l- R ~) THE
PLATE HA.Yl::\G A)l ELLIPTIC .HOIJE DEFI::\ED llY (I = n 0 •

Th e uotmdary conditions for this east' are-

R"" = S,.,J = 0, 'vhen a = «.,,
nntl Han=~ (1- cos 2!3), R11 ,1 = ~ (l +CO& 2J3), S0 ,J = -~sin 2 /~,
when u is ln.rge.
The conditions can be satisfied by making-

A- 1 = - ~: B _1 = ~ [ 1 + cosh 2 ~~..]
R H .. .
, .~
.\ ;. ] = - ] ()
1"\

R
B_, = - .·
;)

Adding together the five tenus corresponding to A _ 1, B_ 11 A+ 11 B _ 11 B _ 3 , selected

from the general formulre stated abon?, the exact cxprcs$iOn for H .... , R p.1, S .. ,; can be
 THE PRESEXCE O.F CRACKS A~D SHM~P COR~ERS .

d etennincd. Th ese expressions are complicated in appearance', n.nd too lengthy to be

wri tten ont here in fnll. Th t• 11tost important and interesting information they contain
is the

ExPHEssro :-> FOH THE Tt-:;.;srLE STHE:->s AT THg EDGE OF THE ELLIPTIC H oLE .

'f his latter tn.kes thr COll lparatin•ly simple form -

sinh 2 " " + e~"• cos 2 ;3 - 1
H ~~~ = I{ .. · - - - -
, cosh 2 a - cos~ 1~

If the sellli-major and minor axPs of tho ellipse are a and D respect ively:-

At tbe end of the major axis (/3 = 0) the t ensile stress is-
2
H [ 1+ ~!]
At t l1 e end of Lhe winor ax is (,~ = ;) there is a colllprcssion str ess of magnitude R.

CASE OF A PL:\TE Sl;ll.TECTEO TO A SHEAHI:\G STHESS S, THE PL.-\TJ·: 11.-\Yl:-\G AX li~LLll:'TW

HoLE DEFI:\ ED BY u = uo·
'l'he boundary tond itions for this case are-
R "" = s .. ,J = 0, 'vheu II = a,•.

If t he shea.rin g stress is app li ed to plctnes at rig ht angles to t he area of the el lipse,

Lhe uont1iLion s whith hold wbeu c« is large are-

T o dcn.l " ·ith this c:tse, the Y<l.ltws of H .... , Hp,1, 8 ,,11, consecJtlC'IIt on taking--·
e - utu + ,,1.
( I - ,.) .l + 1 ( I - 2 u) •<~ = 1 x constant x
Sill
. h(
n + 1 11)
httYe to be t'lllp loycd.
H,. .., H,1,1 wi ll I)(' founll to hn.ve exn.ctly the same forw as before, except that s ines
rq>ln.ce cosi1ws. 'l'he IW\Y form of S .. ,1 is obtainPd hy rcYersing t lw s ig n of the original
form n.ncl replacing sinl'!:i e\·ery\rhen• \\·ith tosin ('s.
·~:30 STRESSES IN A .PLATE DUE TO CRACKS AND SH ARP CORNERS.

The condi tions stated above can be satisfied by making-

AI = 4S e .,- ''u
B+ l = -
s C ~"
4
B -:: = - :rS

Adeli ng together the three corresponding tenns, the va,l ne of H,1 ,1 a,t the edge of
the hoiP takes the form-

c.-.sr-: OF .\ PL.-\TE HA \"1::-<G AX ELLif'TI(' HoLE DEFINED BY a = «o SUBJF.('T TO A Tr·:XsJLE

STREss R APPLIED IN A DIRECTIOX i\IAKll\G AN ANGLE </' WITH THE .:\Luon Ax1s
OF THE ELLll'SF..

This r::tse can be arriYed at by a, combination of cases previously consi<11·n·cl . Tht'

expression fo r t he tension along the inside ~:'dgP of the taYity is-
u _ si nh 2 rt, . +co:; 21/' - f'~"" cos 2 (rp- 1{)
" tl Ji - .
cosh 2 n., - cos 2 J{

H _ ttinh 2 u., - e~... sin 2 {3

,1,1- I ·> •>
COS 1 .., n., - COB ,., 1i

::\IAxnru~r V.u.u-: 01~ R ,1p WHEN ¢ = ~.

Let '" be the slope of the tangent ::tt t he point where 11 11 11 has its lllnximmu va in!:' .
m is gi>en by the expression c~"·· + v 1Tc ·· !-;;.:.

TIH~ \'a luo of {3 for this point of maximum st.ress is giwn by-
tan 1i = -
1 tanh a,,,
m
and t he \·alnc of the stress is-
u sin 2!3 f ?•llfl•
-f'l. , -
.,
cos~ ,~

At the point "·here the slope of the tangent is + 1

H,11, = H [1 + tanh u ,, + _!_ .

tan b ....
l
.\t the point where the slope of the tn.ngent is - 1
 DISCUSSION ON THE TWO PRECEDING PAPERS.

The SECRETARY read the following translation of a communication from Professor Mesnager :-
Professor A. MESNAGER (Ingcnieur-en-chef des Ponts et Chauss£es) : Professor Coker's paper is an
extremely interesting one. The complete study of the stresses in riveted structures or in the different
parts of a plate perforated by rivet holes has never before been worked out so completely or by such
interesting methods. Nevertheless the present investiga.tion is limited to the sections perpendicular to
the direction of stress, and parallel to the same, both passing through the centre of the rivet hole.
These sections are those for which one usually calculates the stress in order to settle the dimensions
of the plate. Unfortunately, these a.re not the sections in which the first permanent deformation
usually occurs. I n a test piece perforated by rivet holes the first permanent deformation occurs in
those sections which are inclined at an angle of 45° to the direction of the stress, although rather closer
to the direction perpendicular to the stress. Professor Coker's work forms, therefore, in my opinion,
only a beginning of the investigation of this question of the fatigue of riveted plates, and it would be
very desirable if this could be extended to other sections, notably to those which are most likely to
be subject to permanent deformation. Some remarkable work has, within recent years, been carried
out in E ngland on this question of permanent deformation. I t has been shown that the importance
of the shearing stresses is very great, and we should therefore seek to find the locality in which they
reach their n1aximum values (see in Enginee·ring, December 24, 1909, a description of t he work of
Messrs. Guest, Mason, and Smith, and the discussion thereon). For such work polarised light offers
considerable advantages, as the greatest shearing stress is exactly equal to half the difference between
the extreme principal stresses. Now the stress perpendicular to the ends of the plate can only be very
small; therefore, when the principal stresses which arc parallel to the plane of stress are of contrary
sign, a simple reading of the calibrator enables us to obtain the required difference. When they are
of the same sign, the difference is equal to the greatest of these two stresses.
Professor W. E. DALBY, M.A., B.Sc., F.R.S. (Associate Member of Council): Mr. Chairman and
Gentlemen, I thought that possibly the discussion on the two papers would be taken together. If the
Chairman will allo\\· a reference to Mr. Inglis' paper, I should like to say that seldom have two more
remarkable papers been presented to the Institution in one evening. On the one hand, Mr. Inglis gives
a rigid mathematical solution of the problem of finding the stress distribution caused by an elliptical
hole in a plate under tension, including the particular case of the circular hole. On the other hand,
Professor Coker shows how the stress distribution caused by applying tension through a rivet in a plate (or
strictly speaking a pin) can be determined experimentally by his optical method. Comparing t he results
obtained by the mathematical method of Mr. Inglis with the results found optically by Professor Coker,
so far as can be seen from the diagrams and from the details in the papers, the stress distributions as
determined agree with oue another. The one paper is a complement to the other, and the agreement
in comparison enables us to use Professor Coker's method with increasing confidence to solve problems
in stress distribution which elude the mathematical machinery at present available for their solution.
I think I am right in saying that one of the main features of Professor Coker's paper~and I hope
he will correct me if I am wrong- is the description of an extension of his optical method in order to
find the magnitude of one of the two vectors which is implicitly used by Rankine in his investigations of
the stress distribution in a plane. Rankine's method consists essentially in reducing the principal stresses
to two vectors ; one equal in magnitude to half the sum of the principal stresses, and the other equal
in magnitude to half the difference. The sum of these two vectors gives the direction of stress in a
 DL'KT SSWX 0:\ THE TWO PRECEDI:\:U PAPERS.

plane to which the vector which is halEthe sum of the principal stresses is at right angles. In Professor
Coker's former paper to the Institution he showed ho'v to determine optically t he vector which is the
difference of the principal stresses. To- night he shows bow the vector whic h is the sum of the principal
stress can be found. H ence the principal stresses can be found, and also the direction of the normal
and shearing stresses at any point in a plane. I think you will all agree that a great advance has been
made, au advauce which will greatly increase the utility of Professor Coker's met.hod. Professor Coker
has shown on a slide the distribution of stress caused by a rivet or pin ncar cHough to t be margin of
the plate to cause the indications of bending stress to appear. H as he determined how far the margin
\YaS from the centre of the hole when bending began t o appear, and, if so, does t he distance bear any
relation to the usual rule used in designin::r ri,·eted joints, namely, that the edge of t he plate must be
at least at a dista nce equ al to the diameter of the rivet from the edge of the hole 1
I should like to conclude my remarks b.'· expressing my admirat ion for the Yer~· beautiful results
P rofessor Coker has obtained b_,. his optica l method.

Professor B. HOPKI~sox, F .R.S. (Associate) : Mr. Chairma n and Gentlemen, I have not much
more to say than that I echo Professor Dalb)·'s admiration of t he work done here by Professor Coker.
This is t he first time I have had the opportunity of seeing t he beaut iful coloured diagrams t hat he
gets when he strains his pieces and urings t o the e.ve, in a wa.y mathematics cannot do, the manner in
which the stress is distributed. I have only one remark in the way of criticism to make about the
paper, and it merely concems the title. I think that what Professor Coker has determiHeu is not t he
stress in the neighbourhood of a rivet, but rather the stress in t he neighbourhood o[ a pin joint. I
believe t hat when t wo plates are rivet ed together the strength of the joint against pull is not t he shear
strength of the rivet itself, but t hey are pinched together by t cusion in the rivet which may not fill
the hole at all, and, practically speaking, the joint holds by friction . I cannot cite any evidence on that
at the moment, but 1 think that is the general impression of people who have had to do with riveting
joints of the ordinar:· kintl such as are found in structural steel work, and I am pretty s ure t here is a
good deal of experimental evidence iu faYour of it. Of course, that does not in the least detract from
the value of t he paper. but it seems to me that the title is perhaps not quite correct. Mr. Inglis just
now suggested to me that the real boundary condit ion, looking at it as a problem in mathematical
elasticity, in the neighbomhood of a rivet ed joint is that the metal just round the rivet is held so
that it cannot move radially . It is t he same as though the rivet were completely inelast ic, and filled
the hole, adhering to the metal of the plate all round, so as to prevent all radial displacement. That
seems to me to be the correct boundary condition, expressing, as fa r as mathematics ca n, what happens
in the neighbourhood of a riveted joint.
Turning now t o ~lr. Inglis' paper. it is. I think. one of great. importa nce. Most. failures in engineering
structures originate, I s uppose, in a crack of some sort, whether they be started, us he suggests, by the
presence of some hard material or possibly by a caYity containing some impurity such as slag. They
originate in a centre of high stress such as exists at the end of a crack. Mathematicians, of course,
have told us before now that t he stress in t he neighbourhoou of a circular hole is as st ated in Professor
Coker's paper, a nd we ha,·e all learned that if t he edge of the crack is an absolutcl;.' sharp <:Orner, then
you get an i1tfinite stress, but the intermed iate condit ion of a crack that has a very high curvature at the
edge but is not. infinitely sharp, has. so far as I am a ware, been tackled by no one unti l no,v, and
t hat, of course, is the importa nt case. Mr. Inglis has shown us exactly ho\\· the stress at t he end of the
crack varies with its curvature and size. H e has s hown that it is proportional t o the square root of
the length of the crack, and inversely proportional to its radius of cun·a ture. H e has pointed out,
 DI. 'Cl'SSIOX OX T HE TWO PRECEDIXG PAPERS.

however. that his results being based 011 the mathematical theory of elasticit-y appl_,. on!_,. "·ithin the
limits of elasticity of the material. I hope t hat he will carry the inYestigation a stage further in
the nea r future, and tell us what happens in a. ductile mat erial when t he elast ic limit is locally
exceeded. Taking an elliptical cavity in mi ld steel, the material will begin to yield and stretch at t he
e nd of t he cavity· as soon as the stress there exceeds the yield point, and, as Mr. Inglis points out,
that \\'ill lead to a re-distribution of st ress. the result of which can b e represented by taking off the top
of t he s harp-pointed curve shown in :Fig. 2 (Plate XXVIIT.) and raising t he stress at points from there to
the right so as to keep the total area the same. Unless t he curvature of the crack is more than a certain
amount this will, no doub t, save the material and the crack will not go any further. At t he same time, it
is obYious that that process of stretching puts an excessive loca l tax on the ductility of t.he material in
the immediate neighbourhood of the end of the cavity . Whether the crack spreads or not depends on
the relation of t he a mount of the local stretch to t he ductility of the material. It would be a ,-aluable
settnel to Mr. Inglis' work to find out t he dimensions of a cavity which would cause the material to
stretch, say, :30 p er cent. at t he ends when the metal as a whole is stressed to near its elastic limit..
:-3uch a cavity would presumably spread as a crack even in mild steel. I t is known that good mild steel
camwt be tom by any ordinary n1echanica l means. If a nry t hin slot be cut with a hacksaw at the
edge of a mild steel plate, at rig ht a ng les to the length of the p late, and t he plate be broken in a
testing machine, a lthough, of course, the stretch is Yery much local ised. t he plate does not break b~' a
t ear starting from the slit, but t he meta l in and near the p lane of t he slot pulls out just as a longer
p iece would pull out with a good reduction of area. On the other hand, it is known that d uctile material
does n uder some circumstances break by t earing, and )h. Inglis in his paper cites the commonest instance
of that, viz. , where a crack starts fro m a punched r i,·et b ole, or from the edge of a sheared plate.
Owi ng to the g reater elastic stress which exists in t he neighbourhood of t he hole. a crack is started
there in the hardeneJ materia l, and. as he says, that materia l has sufficient hold on the ductile material
beneath it to eause the erad: to spread. Once started in that way a crack may spread to any extent
e,·en in ductile material, as has sometimes happened in t he plates 0f boilers.
The effect of case-hardening in making m ild steel brittle is a nother instance of the same thing
on which I ha,·e made a few simple experiment-s. The Wolsd ey Tool a nd Motor Company were good
enoug h to suppl,, · me with some bars case-hardened according to their ordinary· process for motor car
part::;. 'l'be bars were of good mild steel (about ·12 per cent. of carbon) i of a n inch in diameter, and
tht'_,. \\'ere case-hardened, some to a depth of 1-:.i2nd of an inch, a nd some to l-l6th of a n inch. They
were broken in a 0-ton machine by bending . A bar case- hardened like that, although the central part. is
entirely ductile, snaps like a carrot under bending stress, and does not y ield at a ll. That, I suppose,
is a common exp erience. I then took a bar that had been case-hardened to a depth of 1-lGth of an
i neh and had all the hard metal ground off. The piect- so treated was first bent in the testing machine.
and t lten fiua ll:-· bent double in a vice wit hout s howing any defect. T hat shows that the interior still
possesses a ll the ductility of mild steel. This was one half of the same bar which, before the ca:;e-hardening
was removed , had broken q uite short b_,. bending. l t broke wi tho ut a trace of permanent set. Then I tried
t he experiment of grinding off most of the case-hardening, but not all of it. I left a litt le strip of hard
naaterial on the tension side of the bar when it was bent. When a bar so treated is loaded b~- bending
the process of t earing in ductile material is wr:· well shown . At a certain load the hardened tension
part at the bottom cracks with an a udible snap, and when t he load is a little increa:>ed t he ductile stuH
tears anoss like a bit of paper. T hat is, as fa r us I kno\\', t lte only \\'ay in which mild steel can be made
t o tear. I han! not tried the same experiment with iron instead of steel. but it is genera l!.\- belie,·ed
tlt,tt iron is not utfectt'd b_,. casc- h anl c nin ~ in th e sa me wa\·. If so, it is perhaps due to its greater
 DISCUSSION OK THE TWO PRECEDING PAPERS.

ductility, in virtue of which it can stretch sufficiently to relieve the stress even at the base of a
fine crack starting from the hard part. A full experimental and mathematical examination of this problem
of the ductility necessary to prevent cracks from spreading would, I think, be of great scientHic and
practical interest. Mr. Inglis has taken the first step, and a very important one, towards its solution,
and I hope that he will be able to carry it further.

M_r. C. E. STROMEYER (Member of Council) : Mr. Chairman and Gentlemen, like the previous
speakers, I very much admire the work that Dr. Coker is doing in regard to the examination of stresses.
I attempted to deal with a similar subject (Proceedings Royal Society, vol. lv. p. 374) by utilising the
interference of monochromatic light for measuring the cross contraction of test pieces. I should imagine
that this method might be more safely applied to Professor Coker's problems, using metal, of course,
instead of celluloid for determining the value of p,. + p,. The instrument was an interferometer, of
which there are now several designs, which allowed changes of dimensions to be measured by counting
the number of interference bands which passed a cross wu·e in the microscope. This instrument had
to be securely clamped to the test piece, and other precautions had to be taken to guard against outside
disturbances ; and I cannot understand how an instrument which differs from mine in having levers
should be not only ten times more sensitive than a simple interferometer, but also more reliable in its
readings.
Professor Dalby has already mentioned Guest's Law, which seems in general to be reasonably correct-,
for mild steel seems to break down, not when a single stress or the sum of two principal stresses
exceeds a certain limit, but when their difference exceeds a certain limit, and, therefore, from an engineer·
ing point of view, the weakness of a structure is measured by the differences of two stresses, which are
measured by the polariscope. The measuring of p, + p,. is of far less importance. Experilnents,
which to a certain extent confirm Professor Coker's results, have been carried out on curved and 011
holed plates by E. Preuss (Zeitsch. d. Vereine.s Deut.sche:r l ngenieure, 1912, vol. xliv. pp. 1349 and 1780).
The radii of the curvatures of the inner sides o£ the curved beams or hooks varied from 0 to 35 mm.
on a width of 40 mm., and the ratios of holes to widths of the perforated plates were 1 to 9, to 4,
to 2·5 and to 1·71 respectively; the ratios of maximum to mean strains being 2·35, 2·34, 2·13, and 2·25.
I should like to suggest to Professor Coker, when he continues the experiments, as, of course, I assume
he will, if he would perhaps map out or plot on a chart the differences and the sums of the stresses, in order
that, instead of having simply one ljne across the rivet hole, we should have a regular chart of the whole
neighbourhood. It will help us very considerabl:v, I think, in applying to other forms the results
which Professor Coker is getting for one form.
As regards Mr. Inglis' paper and the formuloo in Professor Coker's paper, I regret to say
that, although I have studied them with Professor Love's book on elasticity at my elbow, I
find that the notations, which are doubtless very convenient for the discussion of vibrations
and other complicated subjects, are so elaborate that I have not been able to do more
than apply some simple tests to the results, and' do not now feel satisfied that the mathematical
deductions are fair representations of practical rases; in other words, the holes, corners, and cracks
with whjch Mr. Inglis' paper deals are mathematical, and not real ones. For instance, fatigue cracks
were mentioned, but far too little is as yet known about this subject for anybody to say whether these
mathematical solutions apply in those cases or not. I have personally had to deal with about sixty
cracked crank shafts, and I do not know with how many cracked tail shafts ; and I remember Yery
well the first which I disroverecl and condemned, for I became aware that marine engineers do run
their crank shafts when t.hey are cracked, without very much danger. If the assumed cracks in the
 DISC:t.TSSION ON THE TWO PRECEDING PAPERS.

paper truly represent practical cracks, then as soon as a crack showed itself in a fillet of a shaft it ought
rapidly to break in two, and, at any rate, the crack ought to extend; but I have noticed in many
cases that fatigue cracks in crank fillets are not single cracks. They usually start in many places about
the same time, and gradually merge into each other. It would almost seem that instead of the first crack
being the weakest part of the shaft it is merely a centre of relief, and other parts beyond the crack
give way. It is, perhaps, not fair to apply mathematics to ductile material like mild steel; but
the tests I made on nickel samples (Journal of the Iron and Steel Institute, 1907, i. p. 229; iii . p. 9:3)
may be mentioned here, as they have some bearing on the question. I nickelled my samples for the
purpose of ascertaining whether the material grew brittle-whether it aged, as it were. I certainly found
t.hat those samples which were bent at once after being nickelled were more ductile than those which
were bent after waiting ; but the interesting part, from a mathematical point of view, is that, although
some of these nicks did crack-they did not all crack, for some of them simply tore with a silky frac-
ture-yet those that did so cracked for only short distances into the metal, and the material at the end
of this crack tore, so that, instead of the crack being really a source of weakness, it seemed as if the
cracking of the end of a nick was mereiy a question of induced brittleness. Do not let it be thought
I wish to favour sharp corners or cracks. I kuow they are very dangerous, but I do not believe that
mathematics can thoroughly represent the case. They should apply, a.bove all things, to hard materials;
but even there I do not find that mathematical cracks are practical ones. For instance, anybody
who tries the experiment of continuing a crack in a glass plate by producing strains by light hammering
will find it exceedingly difficult to draw the crack in the desired direction, as it should do if the stresses
were infinitely great at the end of the crack. Then, again, if these mathematical solutions were quite
correct, the fine nick which is cut into a glass by means of a diamond should lead to its brea.king in two
quite easily, and yet, it ;., perfectly well known that a thick glass cannot be cut with a small diamo:1d;
it is, in a sense, the external shape or size of the diamond, and not the edge itself, or the nick which
it produces, which does the cuttiag. Diamonds of considerable size have to be used to cut thick plates.
Then, again, if you wish to break a gl'a.nite stone, which is also a hard material, you do not nick it with
a sharp instrument, but you hit the block along the desired line of fracture with a blunt but heavy
hammer, and the stone breaks as desired. It is the external shape of the hammer, and not the weakness
produced by a fine crack, which causes the fracture.

Professor J. B. HENDERSON, D.Sc. (Associate): Mr. Chairman and Gentlemen, the Institution is
indeed to be congratulated on having two such splendid papers contributed at one meeting, the one
experimental and the other mathematical. I shall deal, first of all, with Professor Coker's paper. I
should like to ask Professor Coker about the probable accuracy in the determination of p 1 + p,,
because, in determining two quantities by a difference method, the percentage error is notoriously
low, depending as it does altogether on the relative values of p, and p,, and one might, by making a
1 per cent. error in the measurement, have a 20 per cent. error in the resulting values of Pt and p,.
I t occurred to me that the determination of Pt + p,. might be liable to an error due to the specimen
being slightly bent, either initially or due to the weight of the extensometer, or due to the pull not being
rigorously axial. The extensometer grasps the specimen at two points, and the weight of the extenso-
meter tends to bend it. I do not know what the weight was ; it may have been quite small, but one
knows the difficttlt.ies of getting a specimen initially straight, of keeping it straight, and of applying
an axial stress. The mirror is attached to one end, so tha.t any straightening of the specimen gives a
+
deflection which corresponds to a variation of p 1 p,. I had a similar measurement to make
some time ago, and I found considerable difficulty due to the bending of the specimen. I got over it
 DISCTRF-:IOX OX THE TWO PREC'EDI};G PAPERS.

in a way which I think might impr<n-c the acc-uracy of Professor Coker's measm cmcnt. by putting two
mirrors at tight angles, one attached to the frame and the other to the moving le,·er. and then \·iewing
the image of the scale formed b~· double reflection in the t wo min·ors . An.' · bending of the specimen
then produces no deflection of the scale at all.
I should like to ha ,.e Professor Coker's opinion as to the elastic properties of x.d onite. I s it quite
a reliable elastic material ? I " ·as making experiments three yea rs a.go on the stresses round deck-
hatches. using a glass plate. and Brewster's interference band method, just like Professor Coker's. The
p late ,,·as optically worked , but unfort unately the students who were mak ing the experiment o,·er-strcssed
it during the first few exp eriments and broke it, and as the plate costs several pounds, it \\·as not possible
to renew it. X.donite is a beautiful substa nce if it can be relied on to give accurate results as to elastic
properties. As a matter of history, I think Professor Cok er is wrong in ca lling the s:·mbol 111 Poisson's
ratio; Poisson 's ratio was really the r eciproca l of 111.
Now, ttu·niug to Mr. I nglis' paper, I can on]_,. hold up my hands in admiration of this beautiful
piece of mathematics. Looking at the Appendix, I see mathematical tools whose existence I had a lmost
forgotten, and. like the true artist .. he has kep t his tools in t he background. One ca n on ly adm ire the
results; these must be ,·ery valuable to p ractical men, because t here is uo doubt as to their
r eliability. The t ensions and the stresses round cracks have been a puzzle to most engineers, and the
method by which a crack spreads has also been a puzzle. Mr. I nglis s hows ho,,· it is t hat the crack
spreads as soon as it is subj ect ed to an alternat ing stress exceeding t he elastic limit, by the opening and
closing at the edge. I should like to ask Mr. Ing lis if he has considered the ques tion of the direction in
which a crack will d evelop. If a crack is acros:; the plate. i:; t here not a t endency , when it cle\"\'lops.
to turn parallel to the p late ? D oes not the plate give wa _v a-long the plane of shear. according t o the
present theor.•·? In this case it would go at 4:) 0 to the direct stress, antl \\·ould g radua lly ,,·ork
ronnel till it \\·as practically para llel to t he plate, so that t he plate would be sa ved. The met.hod "rr.
Inglis has used for dealing wit.h holes which are other than elliptical is extremely ingenious. By the
met hod o f superimposed s:·stems of stresses he has s hown t hat we need not bot her much about the
shape of the hole, except at the com er, ''"hich is a great simplification for practica l men ; and a lso tha t
the cra<:ks near t he edge of a p late can be treated wit h eynal simplicity.
:\Ir. \\'. A. Sconu:, B .Se., Wh .Sc.: P erha ps, Sir, you will excuse me if I deal with some o f t hese points in
the order in which they ha,·e been raised. '\"i th reference to the paper by Professor Coker a nd myself.
the first point was raised by Professor Mesnager in his letter, in which he n ifen ed t o t he fact, t hat the
greatest stress occurs at an angle of .J.5° to the d irection of t he pull. You rememb er that, in the
photographs which Professor Coker showed :·ou, undoubtedly there was a widening out- of the colour
bands at those angles. I must confess that in ma king the experiments on one occasion I overstrained
t he specimen, and the point at which ;·ield took p lace was immediately below the ri,·et. at a point which
we investigated. Professor Da lby asked about the stress at the outside of the plate. and the rela t ion
between that stress and the a mount of overlap. Yo u will notice that we dealt with two cases. one
in which the amount of t he overlap was somewhat considerable compared with t he dillrneter of the hole.
The distance b elow the hole was approximately three times the ri,·et diameter. In the other case. the
space below the hole was equal to the dia mct<> r. In the first case the stress acting across the axis of
the specimen below the hole was a compression just near the hole, but it became a ,·er.'· sma ll tension
to the end of tlic plate. When t he o\·edap was n1ade smaller, t he distribution of stress ''"as more ncar!~·
similar to that in the case of a bea m.
A point was ment ioned which I thittk we expected , that o ur case d eals rather with the
pin than with the ri,·et. It is sometimes a little difficult t o get tt title that exactly fits
 DISGCS.SIO:X O:X THE TWO PRECEDl:XG PAPERS.

a paper, and this point was raised in giving the title to the paper. I think that you should
remember that this is the first case in which the two methods have been combined, and it should
be looked upon rather as showing the possibility of the combination of the two methods to distinguil:>h
between the principal stresses. Our case was that of a pin rather than a rivet, because there was no
force pressing the t"·o plates together, no force in the direction of the axis of the rivet. There is
another consideration which is, I think, equally important, and which we were not able to take into
account up to the present.. There is often an intense radial pressure due to the rivet being pressed hot
into the hole, and there will be stresses around the hole in addition to t.hose which we have im·est.igated.
It appears to me that there will be stresses somewhat as if the hole was subjected to internal fluid
pressure. \Ye have not dealt with the complete case, but when we are discussing the matter, I think
the view to take is this, that when we make experiments it is impossible to exactly reproduce the practical
conditions. I n fact, frequently it is advisable that they should not be so reproduced. Therefore
we have dealt with the .first case of a rivet in a hole where the stress in the direction of the axis of
the rivet has not been considered, nor the radial stress, which can be compared to an internal fluid
pressure; those can be added later.
Mr. Stromeyer referred to the instrument for measuring the sum of the principal stresses, and he
said that he could hardly believe that such a simple instrument would measure the stresses. I attribute
the accuracy of the results to the simplicity of the instrument. I haYe read his paper, but I forget
the details of the construction of his apparatus. It struck me that the lag which he observed might
haYe been due to the instrument, and not to the lag in the material. He also said that since we
pretty much agreed that, accordiug to Guest's La,,·, the shear stress determines the fai lure of a plate,
therefore all we need to measure is the difference in the principal stresses. ,,·hich, of course, is double
the shear stress. I think we need much more in all cases where we ha,·e an unequal distribution of
stress; and I think that also has a bearing on the experiment by Professor Hopkinson, in which he
took a tension specimen, put. a saw-cut into it, and found under test that the specimen fractured Yery
much Jjke an ordinary tension specimen. I must confess that is just what I should expect it to do;
and I think that unless all these results "·hich arc giYen arc considered in relation to ~·ield-though it is
difficult to experiment after yield- we shall get quite wrong impressions of what happens.
Referring to ~lr. Inglis' diagram-because his case is very much like ours-and considering a
tension specimen \\'ith a hole, which may be either a circular or an elliptical one. we have at the hole a
wry intense stress, which diminishes towards the edge of the plate. Supposing the tension be graduall)·
increased, the material will yield a little at the hole. When it has ~·ielded to the order of. sa,v,
1- LOOOth of au inch, all this other material across the section is under the full stress, so that after
a little yield we baYe prar.t.ically a uniform distribution of stress across that section. I think Professor
Hopkinson's experiment confirms that, because he tells us that with a slot across his specimen the appear-
a nee of a fracture across the 1·emainder of the section is very much as i11 the ordinary tension test.
Professor Henderson mentioned the bending action due to the instrument for measuring the sum of
t he principal stresses. It was made so light that the bending action was ent.irel,r negligible. I suppose the
weight of the whole instrnment was not more than three ounces, or something of t hat order. The bar,
for instance, which :·ou see carrying the supports for the lever, is made of ·110 in. steel, and i.-s about
.} in. wide. That will giYe an idea of the djmensions of the whole apparatus. The clamp which
supports the measuring portion is made of 1 11, in. steel strip about J in. \\'ide. With reference to
the prope1ties of xylonitc, we ha,·e made tests on the xylonite at least to t.he stresses we ha\'e had
to deal with in this paper. Personally, I was surprised to find it behaved as well as it did. From
the nat.ure of the material I did not expect it to ba ,.c a nry extensiYe elastic range. I was afraid it
 018l'U~SIO:X OX THE TWO PREOEDI~G PAPERS.

would have a considerable amount of creep, but I was surprised to find it behaved very much like a
specimen of steel
Professor E. G. CoKER, M.A., D.Sc. (Associate) : Mr. Chairman and Gentlemen, the questions raised
in connection with the paper by Mr. Scobie and myself are somewhat numerous, and I will endeavour
to reply briefly to those ·which have not been dealt with by Mr. Scobie. With regard to Professor
Mesnager's view of the distribution, the pictm:es which we exhibit show that there must be a very
considerable amount of truth m what he says, and we hope, very soon, to investigate the problem
in much greater detail. Professor Dalby has very happi ly pointed out the connection between an
optical measurement and Rankine's ellipse of stress construction. As he points out, the methods described
here give the sum and difference of the principal stresses, and these are the vectors Rankine uses.
The angular positions of the planes of principal stress can also be determined experimentally, but not
in quite so easy and direct a manner as in the Rankine construction.
Professor H opkinson has also mentioned a matter whjch we felt sure would be raised in this discus-
sion, namely, the function of the rivet in our experiments. We qwte agree that it acts as a pin in
these cases, and not as a true rivet. I n order to get the tension of the rivet it would be necessary
to obstruct the view and hide some of the neighbouring parts we wish to _see. We have not, therefore,
attempted the rivet problem in its entirety. Mr. Stromeyer mentions a very interesting fact that
some years ago he measured the thickness of plates to get the values of the sum of the principal
stresses. That is a new fact to us, and a most interesting one. I think he made a good deal of the
difficulty with regard to the question of the measurements. There is no real difficulty, because in the
ordinary extensometers, like the Ewing instrument for compression, for example, the unit of the scale
is one division for a quarter of a millionth of an inch over a length of from one to two inches. The
required increase in delicacy is not very great in order to measure accurately the change in a thickness
of ~thor ):~ths of an inch to as nearly as is reqwred. Another point was raised by Professor Henderson
in regard to the accuracy of the measurements. It is perfectly true that it is a severe test to bring
two entirely different types of measurement together, and then add and subtract them to obtain the
quantities required. Any experimental error you make is bound to assume a large proportion of the
difference value you obtain, and in the paper we have not concealed the fact that we did not get absolutely
concordant results. I think we were really surprised to find them agree so nearly as they did. They were
brought together at the last, and from the way in which they are given it was impossible for us to
tell whether they would agree or not.
In this connection I might refer to Mr. Inglis' paper. We are delighted to see t.hat we
agree ";th him, without knowing anything beforehand of what he was going to say. For a
particula.r problem Mr. I nglis has an accurate solution, of which t here can be no doubt whatever.
It complies with all the conditions of an accurate solution, and, therefore, it must be correct. You
will observe that the readings agree fairly well, so far as we have got them for cases in which
the two papers can be compared. Professor Henderson also referred to the question of the
properties of xylonite. I am glad that my colleague, Mr. Scobie, has mentioned this point independently,
because one is always biassed in favour of one's own particular way of doing things, but Mr. Scobie was
not.. He did not at first believe that the material would give measurements as accurate as we required
them, but he is now quite converted from actual experience. If sufficient care is taken, and every·
thing is measured in t he very best way, it is possible to get an accuracy of within ± 2 per cent.
You cannot always be sure of getting as close as this, unless you are making very careful measure-
ments, and everything is in first-class order. In previous work we have, without knowing exactly
what results ought to be obtained until the very end, got within these limits of accuracy. I do not think
 DISCUSSIOK ON THE TWO PRECEDIXG P.-\PER:-:i. 239
the work in the present paper can be taken as close as this all through. There is some creep with time,
but not very much. The material is not so perfectly elastic as glass, but then it has this great
advantage over glass, that you can put a very considerable stress on it without fracture taking place,
and this gives the opportunity of a larger range of measurement than with glass. I dealt with that
matter somewhat fully in a previous paper at the Spring Meeting of 1911 ,*and it is shown there that
glass is a superior body for this purpose as regards its elastic properties, but it is inconvenient in almost
every other way. With regard to Poisson's ratio, we have used it here in the sense which most engineers
use it, fully recognising that we are not in perfect agreement with mathematicians on the point. It is a
small matter, but we thought it better to give it in the way we have done.
Mr. C. E. INGLIS, M.A. : Mr. Chairman and Gentlemen, I am fortw1ate in not having to answer
many difficult questions, so my reply can be brief. The only adverse criticism comes from :Mr. Stromeyer,
and it takes the form of a general, though somewhat vague, distrust of mathematics. Mathematical
reasoning is the soundest possible form of argument, and the conclusions arrived at are perfectly reliable,
provided that the fundamental conditions laid down are properly respected. This Mr. Stromeyer fails
to do. The paper deals with cracks in plates subjected to a two dimensional distribution of stress. Mr.
Stromeyer apparently thinks it ought to explain the behaviour of cracks in crank shafts. From the
mathematical standpoint these cases have almost nothing in common. To discredit the validity of
this work because it fails to explain cases which a.re definitely excluded by the fundamental conditions
laid down, is, to put it mildly, somewhat unreasonable. Concerning the direction in which a crack
"Will spread, theory, I think, tells us little or nothing. Professor Henderson suggests that the plate gives
way along the plane of greatest shear. I think this is even more tha.n we can at present claim to know.
Certainly, it seems fairly well established that the first indication of loss of elasti<'ity occurs in this
plane, but whether the ultimate rupture of the material is decided by tension or shear remains to be
found out. To settle this point, a knowledge of the stress in a body strained beyond its elastic limit
is required, and this takes us into a region at present almost entirely unexplored.
The CHAIRMAN (Professor J. H. Biles, LL.D., D.Sc., Vice-President): Gentlemen, I am sure you
will allow me to propose a very hearty vote of thanks to Professor Coker and Mr. Scobie for their paper,
and to Mr. Inglis for his admirable explanation of what his paper is intended to convey ; it is most
interesting from the theoretical and purely mathematical side to see this elegant analysis, and from the
experimental side to see the results which Professor Coker and his colleague have brought before us.
Those who have been away to-night, attra.cted by other engagements, cannot, I am sure, have had a
better entertainment than you have had this evening. At any rate, those who have remained faithful
have had their reward. I ask you to give a very hearty vote of thanks to the readers of both these
paperl!.
WRITTEN CONTRIBU'fiONS TO THE DISCUSSLO:\.
Mr. H. l\1. RouNTHWAlTE (Member) : Referring to Professor Hopkinson's remarks in the above
discussion, it occurs to me that the case described below may be of interest; it is one that has been a
puzzle to me ever since it occurred. I have mentioned it to many engineers and metallurgists, and have
never obtained any helpful suggestion in reference to the cause. The engines of H.M.S. Gladiaror were
being erected in Messrs. Maudslay's shops (1895-G), and the piston, piston-rod, crosshead, and guide
blocks (as shown in the sketch) were suspended from a crane, the piston being in the cylinder, and
the crosshead in its working position between the two pairs of guides attached to the four columns
0 Tran~. I.N.A., Yo!.' LUI. li.) p. 265.
240 DISGGSSIOX ON THE TWO PRECEDIXG PAPERc.

which carried the cylinder. The piston was at some point above the half-stroke when the sling chain
snapped, and the parts shown in the sketch fell until the piston came to Yiolent contact with the
bottom of the cylinder. The cylinder was not in any way injured, but one arm of the cmsshead broke
short off at " A" in the sketch. The fracture resembles a fine tool steel, that is to say, it showed
a very fine crystalline structure all o,·er and quite uniform,
excepting round the circumferenc<', where the brighter line of the
case-hardening showed for perhaps l-20th inch in. Apparent ly
the only force acting to Lrea k the cross head was the momentum
of the portion bc.voll(\ the break. The forging was specified and
had been ordered as Swedish iron, and the two connecting-rod
journals bad been case-hardened. After the accident a piece
of the metal was sent to Professor Amold, who at once
demonstrated by means of micrographs that the material was
not Swedish iron at all. but ordinary mild st eel. In my
apprentice days I was often told by old workmen that one
should never case-harden steel, as the process would make it
quite rotten, but I could ne,·er obtain scientific confirmati on of
this view, and in the last twPnty years I ha\'e uoticed that
many ma nufacturers mah a practice of ease-hardening steel
pius, &c.

Professor B. HOPKL'-: SO X. F.R.S. (Assoc-iate) : I am ,·et·y

much interested in Mr. Rountll\\'aite's account of t,l)e accident to
the crosshead of H.M.S. Gladiator. I do not think that the mass
of the part to the right of the fracture woulrl be sufficient to
break the case-hardening, if it were perfectly sound, especially as
there wou ld be a certain amount o! spring on the piston. Of
course, a minute crack in the case on tLe top side would be quite sufficient to account for the breakage.
The character of the fracture described is very interesting, and is precisely s imilar to that of the
smaller case-hardened bars which I described at the Meeting. It is evidently characteristic of case-
hardening of this kind that it causes the material within to break short wh<'n the latter is mild
steel. The statement that it is a common opinion among workmen that mild steel breaks in this
way, whereas iron does not, is in accordance with my own experience, but I ha\'C neYer actually t.ried
the experiment with iron.

Mr. L L OYD WooLLARD, R.. C.N.C. (Member): I t hink the lnstitution is to be congr:ttulat.ed on t his
paper by Mr. Inglis, as it is one of the comparatively few attempts that have been made to adapt the
mathematical theory of elasticity to the practical problems met with in Naval Architecture. As one
of the speakers remarked, there is a tendency in text-books dealing with this subject to limit its appli-
cation to such problems as are more readily susceptible to mathematical treatment; e.g. , t he vibration
of elastic spheres and kindred investigations having little application to actual structures are very fully
treated. In this case, however, the author has selected , on its practical merits, a problem which by no
means readily lends itself to theoretical investigation, and has applied the results in a way that is
directly useful and interesting to us. It is to be hoped t hat he may be able to continue his researches
to the many problems affecting the strength of ships and other structures that are still unsolved ; e.g.,
the stresses in plate web girders, the stiffening and riveting of bu lkheads, &c. I have only one point
 DISCUSSION ON THE TWO PRECEDING PAPERS. 241
to raise, which has already been dealt with in part by some of the speakers in the discussion. In a·.
ductile material, like mild steel, how far can a slight overstrain, due to the limit of elasticity being
exceeded over a limited region of material, be regarded as really dangerous ? There are two instances,
well known in naval architecture, where such overstrain is constantly produced and accepted. The
first is found in the fiat plating near the bottom of some ships, under the action of the water pressure
and longitudinal bending moment. In many battleships the thickness of such plating is i in. ; with
a 4-ft. frame spacing and a momentary head of 35 ft. of water, it can be shown (see paper read
in 1902,* by Mons. I. G. Boobno:ff) that a stress is caused between 7 and 18 tons per square
inch, probably nearel' the upper limit. On combining this with the stress due to longitudinal
bending, it is evident that the elastic limit of mild steel must be exceeded ; yet, as far as is known,
no sign of weakness has been observed in such cases. A second instance is furnished by an iron chain
ring, whose internal diameter is considerable, say, 8 times the diameter of the iron. The ring can be
proof tested, and the stress caused thereby estimated fairly readily by a simple mathematical process.
I t is found that the maximum intensity of stress on the assumptions becomes 30 tons per square inch.
I should like to ask Mr. Inglis whether he has considered this point; and if he can indicate the·
extent to which the factor of safety ought to be increased in practice to allow for possible small defects
in the material.
Mr. C. E. INGLIS, M.A. : Concerning the two instances of high resi:,;tance to stress mentioned by
Mr. Woollard, a possible explanation may be found in the remarkable fortification sometimes produced
in a material which has been partially overstrained. For instance, if a thick cylinder is subjected to a.
large "interna l pressUTe, the inner layers will be overstrained. When the presstue is released the strains.
do not entu:ely disappear. The elasticity of the outer material produces compression in the overstrained
under layers, and this initial state of stress fortifies the material to withstand the next application of
pressure. The initial stresses produced by the overstrain are, in fact, veiy similar to those set up by
wire winding on guns. Something of this sort may very likely occur in the two cases mentioned by Mr.
Woollard, and the material may be able to withstand large added stresses owing to the existence of
beneficial initial stresses brought about in the first instance by a partial onrstrain of the material.

- - ·0 - - - - - - - - - ·- -
Tran~. T.N.A., Vol. XLI V. page !fl.
R