The Stresses in Cemented Joints

BY M . G O L A N D . 1 B U F F A L O , N. Y., AND E. R E I S S N E R , ' C A M B R I D G E , MASS.

The determination of the stresses in cemented lap joints tion of the shearing stresses, and neither the effect of the deflec-
has become of practical importance because of the de- tion of the jointed members due to plate bending nor the effect
velopment of new methods which permit a strong bond to of the tearing-stress component was taken into consideration.
be established between wood, plastic, or metal sheets, or It is shown here that consideration of these effects is essential for

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
combinations of them by the use of cement adhesives. the analysis of cemented lap joints and the mathematical formu-
In this paper, the problem is divided into two parts, lation of the problem in this manner leads to a system of equa-
(a) determination of the loads at the edges of the joint j (b) tions which, to the authors' knowledge, has not been established
determination of the stresses in the joint due to the ap- previously.
plied loads. Solutions are obtained for two limiting
FORMULATION OP T H E PROBLEM
cases, i.e., where the cement layer is so thin that its ef-
fect on the flexibility of the joint may be neglected; and Given are two rectangular sheets of equal thickness t, of unit
where the joint flexibility is mainly due to that of the ce- width and of length (I + 2c). The two sheets are lap-jointed over
ment layer. In both cases expressions are obtained for the length 2c. The bond between the two sheets is established
the shearing stresses in the cement, and for the normal by means of a cement layer whose thickness rj is small compared
stresses in the cement in a direction perpendicular to the with those of the sheets. The unjointed ends of the sheets may
plane of the joint. be assumed to be simply supported and acted upon by tensile
forces of magnitude T per unit of sheet width, Fig. 1.
HE present paper deals with the determination of the The main problem is the determination of the shearing and

T stresses in cemented lap joints. The problem has become

of practical importance as a result of the development of
new methods which permit the joining of wooden or metal sheets,
normal stresses in the cement. Due to the fact that the sheet
width is large compared with the sheet thickness, this may be
considered as a problem in plane strain. The problem is sub-
or a combination of the two, by means of cement adhesives in divided into two parts. Part one concerns the determination of
such a manner that a very strong bond is established. the loads applied to the rectangular joint section of length 2c and
The problem is subdivided into two parts. The first part is thickness (21 + n)- Assuming that it is satisfactory to approxi-
concerned with the determination of the loads at the edges of the mate the applied loads by a linear normal stress and a parabolic
joint. It is recognized that this is a problem in which the defor- shear distribution, this problem may be considered as one of
mation of the jointed sheets must be taken into account, using beam theory, or rather, of cylindrically bent-plate theory. The
the finite-deflection theory of cylindrically bent plates. The two overlapping sheets are considered as one homogeneous plate
second part is concerned with the determination of the stresses with a discontinuous thickness variation and a neutral plane dis-
in the joint due to the applied loads. This problem is formulated continuous at the ends of the joint section. Moreover, it is found
as one in plane strain. Explicit solutions are obtained for two necessary to take into account the effect which the deflection of
limiting cases, (a) the case where the cement layer is so thin that the joint section has on the moment distribution, Fig. 2. The
its effect on the flexibility of the joint may be neglected; (b) the solution of this problem is given in Part 1 of the present paper.
case in which the joint flexibility is mainly due to that of the ce- Having the distribution of loads along the edges of the joint,
ment layer. Case (6) is significant for the analysis of cemented it remains to solve a problem in plane strain in order to determine
metal sheets. Case (a) is of practical interest in the study of the stresses in the cement. The complete solution of this bound-
cemented wooden and plastic sheets. In both cases, expressions ary problem is of considerable difficulty. Its proper execution re-
are obtained for the shearing stresses in the cement and for the quires that the deformation of the sheets due to the three stress
normal stresses in the cement in a direction perpendicular to the components a x , av and t x v be taken into account, as well as the
plane of the joint. For these latter stresses, the name "tearing deformation of the cement layer due to the corresponding three
stresses" is suggested, as they are thought to explain the tearing stress components.
failures observed to take place at the edges of lap joints. However, the relative thinness of the cement layer and the
Related problems on the strength of welded or riveted joints smallness of its elastic moduli, as compared with the elastic
have been considered by several previous writers (1, 2, 3, 4, 5, 6)3 moduli of the sheet material, allow several simplifications to be
The main interest in these previous publications was the distribu- made, i.e., neglect of the normal stress in the cement parallel to
the layer and the assumption that the remaining stresses in the
' Research Engineer, Research Laboratory, Curtiss-Wright Cor- cement do not vary across the thickness of the layer.
poration, Airplane Division. On the basis of these assumptions, the strain energy of the
2 Consultant, Research Laboratory, Curtiss-Wright Corporation,
joint, expressed in terms of the stresses, is given b}'
Airplane Division, Assistant Professor of Mathematics, Massachu-
setts Institute of Technology.
3 Numbers in parentheses refer to the Bibliography at the end of
the paper.
Presented at the Annual Meeting, New York, N. Y., Nov. 29-
Dec. 3,
GINEERS.
1943, of THE AMERICAN SOCIETY OP MECHANICAL EN-
+ dxchj + h f-J\ka°2 + k^dx 111
Discussion of this paper should be addressed to the Secretary,
A.S.M.E., 29 West 39th Street, New York, N. Y „ and will be ac- Here E, G, and v represent the Young's modulus, the shear modu-
cepted until April 10, 1944, for publication at a later date. Dis- lus, and the Poisson's ratio, respectively, of the sheet material
cussion received after the closing date will be returned.
NOTE: Statements and opinions advanced in papers are to be and Ec and Gc refer to the Young's and shear moduli, respec-
understood as individual expressions of their authors and not those tively, of the cement material. The sheet normal stress a x is
of the Society. parallel to the plane of the joint and the normal stress <r„ is per-
•17

Copyright © 1944 by ASME

 A-18 JOURNAL OF APPLIED MECHANICS MARCH, 1944

pendicular to this plane. The sheet shear stress rxv acts on planes consist of a tension T, a moment M„ and a transverse shearing
parallel to the directions of a x and <r„. The cement stresses a0 and force V0, each per unit of joint width, Fig. 1. The moment M„
T0 represent, respectively, the normal "tearing stress" in the is directed so as to introduce tensile stresses in the sheet fibers
cement and the shear stress acting on planes parallel to that of the nearest to the cement layer. It is found, in section 1, that the
cement layer. magnitudes of the moment M0 and the shearing force V„ are
The correct stress distribution is distinguished from all stress dependent upon the value of the tensile stress p in the unjointed
distributions satisfying the equilibrium equations and the bound- sheet, the joint dimensions, and the physical properties of the

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
ary conditions of the joint problem by the fact that it minimizes sheet material. The precise nature of this dependence is indi-
the strain energy W. cated by Table 1 and Fig. 3 of this paper. Fig. 3 is a plot of the
On the basis of this energy consideration, an estimate can be
made of the range of validity of the two approximate treatments
which are given in what follows.
The first approximate treatment neglects the flexibility of the
nondimensional factor k = — - versus the parameter -
pt2 t 4
where the factor k is the ratio of the magnitude of the edge moment
cement layer; i.e., the work of the stresses a„ and T„ in the ce- M„ and the value of this moment in a nondeformable system,
ment compared with the work of the stresses av and Tt„ in the and p represents the mean stress T/t in the sheets away from
sheet. Since, approximately the joint. Values of —° for various - 4/ L are given in Table 1.
J pt2 tJE h

f. 0
7V*
"
E
J
di/
J
«
l
—
2E
<

with a corresponding estimate for T,„, it follows that the flexi-

.[2]
In representative joints composed of cemented aluminum-
alloy sheets, the value of k is reduced to as little as 35 per cent of
its initial value, as the sheet stress p is raised from zero to 40,000
psi. The edge-shearing force in these sheets is never in excess of
bility of the cement layer may be neglected when the following 0.027\
order of magnitude relation is satisfied:
For joints with relatively inflexible cement layers, i.e., for
, 1 1 1
[3] joints where — < — — and 77 •, very high concentra-
Ec E G. 0' E ~ 10 E Gc 10 G'
tions of stre s are found at the joint edges. Fig. 7 shows the maxi-
and acceptable results should be obtained when mum values of these stresses, plotted as a function of the moment
factor k. The normal stress <r„ in the cement, acting in a direction
t I
" < 1 ' <1 • [4] perpendicular to the plane of the joint, has the surprisingly large
Ec ~ 10 E' Gc 10 G value 4.3 p for k = 1.0. As the value of k is reduced, due to an
This inequality means, for example, when the elastic moduli of increase in the applied sheet stress p, the a0 concentration also de-
the cement are 1/100, (1/1000) of the elastic moduli of the sheet creases, reaching the value 2.1 when k = 0.4. The higher sheet
material, the thickness of the cement layer must be not more stresses are thu; accompanied by lower <t0 concentrations. It is
than 1/1000, (1/10,000) of the sheet thickness, in order that this to be noted, however, that the numerical magnitude of cr0 in-
part of the theory give quantitative information. creases steadily with the sheet stress p.
The second approximate treatment neglects, in effect, the work The magnitude of the maximum shear stress r„ in the cement
of the stresses <r„ and TXU in the sheet as compared with the work of plane is likewise large. At the start of loading (k = 1.0), the
the stresses a0 and r„ in the cement. As before, Equation [2] peak shear stress value is 0.79 p; the shear-concentration factor
may be assumed to hold, and consequently the order of magni- decreases to 0.45 for k = 0.4. The numerical magnitude of the
tude relations describing the range of validity of this approximate maximum shear stress is again found to increase steadily with
theory become the sheet stress p.
It is of extreme interest to note the existence of such large ce-
L « JL, i « 1 ment stresses, particularly with regard to the <r„ stresses. It is
[5]
E Ec G Gc observed, during actual joint tests, that the start of failure is
characterized by a splitting apart of the two sheets at the joint
which conditions may be replaced by edges. This action is explained by the presence of the high tr„
stresses. For easy reference, a descriptive title for the <r0 stresses
L < 1 . 1 L < 1 . A .[6] is desirable; the name "tearing stresses" is suggested as being
E ~ 10 ' E' G ~ 10 ' Gc
appropriate for this purpose.
Thus if the elastic moduli of the cement are 1/100, (1/1000) of The third stress, shown in Fig. 7, is the longitudinal tensile
the elastic moduli of the sheet material, the thickness of the stress ay in the most highly loaded sheet fiber adjacent to the
cement layer must be not less than 1/10, (1/100) of the thickness cement layer. As do those of <r„ and r0) the concentration factor
of the sheet, in order that this part of the theory be applicable. for oy markedly decreases as the joint load increases, i.e., as the
It is recognized from these order-of-magnitude considerations value of k decreases.
that the second approximate treatment is well adapted to the Fig. 6 shows the stress distribution along the shear plane of the
analysis of metal joints of the dimensions occurring in airplane joint in the vicinity of the edges, calculated for a value k = 1.0.
structures, while the first treatment is of value in the study of The a„ stress is highest at the joint edge and rapidly damps out
the cementing of relatively thick wood and plastic sheets. to a small value (0.1 p) within a distance along the shear plane
of about two sheet thicknesses; the reversal of the a„ stress from a
R E S U L T S A N D D I S C U S S I O N OF R E S U L T S
tensile to a compressive stress within this distance is of interest.
The results of major interest are those pertaining to the values The T0 stress has the value zero at the joint edges; it rapidly
of the stresses in the cement layer of a loaded joint. Prior to rises to a peak value and once more decays to almost zero within
investigating the cement stresses, however, it is necessary to a distance of two sheet thicknesses. Finally, the sheet fiber stress
evaluate the forces and moments transmitted by the unjointed- <rf rapidly decreases to its nominal value 0.5 p within this same
sheet portions to the lapped-joint portion. distance along the shear plane. In joints with relatively inflexible
The loads transmitted by the sheets to the edges of the joint cement layers, therefore, the tearing and shear stresses in the
 GOLAND, REISSNER—THE STRESSES IN CEMENTED JOINTS A-19

cement are concentrated in small regions near the ends of the D E R I V A T I O N OF RESULTS

joint. The middle-joint portions are under practically pure ten-

1 D E T E R M I N A T I O N OF THE J O I N T E D G E LOADS
sion of amount 0.5 p in the sheets.
For joints with relatively flexible cement layers, i.e., for joints It is proposed to determine the loads acting upon the edges of
71 the joint by considering the joint and the neighboring sheet to
where ~ < — ~ and -- < — , results of entirely different act as a cylindrieally bent plate of variable cross section and
E ~ 10 Ec G ~ 10 G
nature are obtained. To begin with, whereas the stress distribu-

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
tion in the cement for the inflexible-layer case is independent of l — -- ? c - I, —
both the joint dimensions and the phj'sical properties of the I
cement material, the cement stresses in flexible layers are depend-
— o
ent upon both of these factors. Furthermore, the cement stresses 1
are no longer restricted to small end zones in flexible layers, but
are of appreciable magnitude over the whole of the shear plane. 11

Fig. 9 shows typical <r„ and r„ distributions in joints with flexi- Fio. 1 P R O F I I . F . V I E W OK O E M E N - T F . D - J O I N T S Y S T E M
ble cement layers, calculated for several k values. The joints
are composed of representative materials and dimensions (say,
>M
aluminum-alloy sheets cemented with a typical resin adhesive Ti 1 • 1. X I 11
and with a joint overlap to joint thickness ratio of 5), as specified _1 2 1 111)
by the values of the parameters given on the sketch. It is evi- (b)
dent that the increased flexibility of the cement layer causes the (a)
shear stress t„ in the cement to be more uniformly distributed FIG. 2 L O A D A P P L I E D TO C E M E N T E D - J O I N T SYSTEM IN FIG. 1
over the joint shear plane. Nevertheless, the distribution still ( a , N e u t r a l p l a n e o f j o i n t b e f o r e a n d a f t e r a p p l i c a t i o n o f l o a d , b, Positive
indicates a sizable concentration of shear in the end regions of the c o n v e n t i o n s f o r M, V, a n d T, a s u s e d i n S e c t i o n 1.)

joint. The <r0 stresses again reverse their sign near the joint
edges, but are not damped away as quickly as they were in in- variable neutral plane. The required results indicate that account
flexible cement layers. The peak t0 and a0 values still occur at must be taken of the effects of the transverse deflections of the
the joint edges. sheet and joint on the values of the edge loads.
It is at once clear that the magnitudes of the peak cement A graphical representation of a section of an unloaded joint is
stresses are very much lower in joints with flexible-cement layers shown in Fig. 1. The joint (overlap) length is notated 2c, the
than in joints with relatively inflexible layers. Their orders of two joined sheets are considered to be of equal thickness t and
magnitudes for representative joints are indicated in Fig. 9. are presumed to extend a distance I to either side of the joint.
c The joint width, in a direction normal to the plane of the section
For - = 5, the maximum a„ value is about 0.12 p and that of r„ shown, is assumed to be large compared with the sheet thickness.
If the system is now loaded by tensile forces T per unit of sheet
is approximately the same magnitude. It is to be noted that
width at the supports a and b, the line of action ol the forces will
even more pronounced reductions in the a0 values than in the r„
be aob. The order of magnitudes of the joint edge loads can at
values are brought about by increased cement-layer flexibilities.
once be recognized. For reasonably small deformations of the
Figs. 10 and 11 are plots of the maximum values of the tearing
system, the tensile force in the sheet per unit of width remains
stress <r0 and the shear stress T„ in the cement. The plots are ar-
very nearly equal to T, and the sheet moment at the joint edges
ranged to include the effects of variations in the semi-lap length
is of the order l/2Tt. Since the order of magnitude of the trans-
c
to sheet thickness ratio as Well as the effects of variations in the verse deflections observed during actual joint tests is t, the need
t for including the effects of the sheet and joint deflections in de-
sheet and cement materials. The definitions of /3 and y, which termining the edge loads is apparent.
appear on the figures are The solid line of Fig. 2 represents the elastic axis of a longitud-
inal cross section of the lap joint. Since the form of the axis will

V6
Gc Ec always remain antisymmetrical about the mid-point o of the
= "4/8 7 = V 6
E E joint, only the left-hand sheet and the left portion of the joint
are shown in Fig. 2. In the cemented region, the axis is along the
The flexible-cement-layer analysis is applicable to joints com- mid-plane of the joint and hence, presuming always a cement
posed of thin metal sheets cemented together. The adhesive layer of small thickness compared with that of the sheet, the elas-
materials used in such joints are particularly weak in tension. tic axis has a discontinuity of amount t/2 at the joint edges.
Hence the a0 stresses, although smaller than those encountered (The discontinuity is drawn to a magnified scale on the figure for
in inflexible-layer joints, are still of sufficient magnitude to cause purposes of clarity.) Under load, the elastic axis deflects to some
a splitting failure at the joint edges. form such as that of the clotted line in Fig. 2. In order to study
It may be noted that several interesting extensions of the work the system deformations, it is convenient to introduce two co-
presented in this paper are possible. For example, the joint edge ordinate systems (xh wi) and (x2, w2). The first system is used
loads could be determined for sheets which are not initially flat, but to analyze the behavior of the left-hand sheet under load; the
which are instead crimped near the overlap region. By proper xi co-ordinate has its origin at the support a, extends along the
crimping, it is possible to greatly reduce the values of the edge gravity axis of the sheet and is positive to the right. The w, co-
bending moments and thereby the values of the tearing stress ordinate represents transverse deflections of the sheet from the
concentration. The effects of lateral loads, normal to the plane unloaded condition and is positive downward. A similar defini-
of the sheets, could also be studied. Furthermore, the methods tion is taken for (x2, w2) with reference to the joint, as shown in
of this paper can be extended so as to cover such cases as the join- Fig. 2.
ing of two sheets of unequal thickness and the problem of one Setting Mi equal to the bending moment in the sheet at sta-
sheet, to which are cemented two side sheets (the double shear tion xi and Mi equal to the moment in the joint at station x2,
joint). each per unit of width, then
A-20 JOURNAL OF APPLIED MECHANICS MARCH, 1944

Mi = T[anxi — w, ] for 0 < Xi < V ratio, respectively, of the sheet material, then since the joint is
t_ [7] very nearly twice the thickness of the sheet
Ms = T a„(l + s2) - w2 - for 0 < 22 < c
2 D2 Et3
Di
8 12(1 —
Here a„ represents the angle between the Xi (or x2) co-ordinate
and
and the line of action aob of the applied forces and is very nearly
equal to

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
I = 2 V3(l - | ^
_ 1 t [13]
"" ~ 2 (l + c) J 3 ( l — ?») c I
s
j
-
- •; X.E
The positive conventions chosen for the moments Mi and Ms
and for the transverse shearing forces per unit of width Vi in In all practical cases, the value of ud is sufficiently large to permit
the sheet and V? in the joint, to be introduced in the later work, taking
are shown in Fig. 2.
According to the theory for the small bending of thin, cylin- sinh ud « cosh ud « ^ eul1 [14]
drically bent plates, the differential equations for the transverse
deflections of the sheet and joint become Equations [11] and [12] then reduce to the simplified forms

dho 1
dx? ~
d*w3
Mi
•Di "
M2 _
T
Di
T
M0 = k
Tt
V. = kT

where the value of the factor k is dependent upon the sheet

V 3(1 - x2) ~E- . . . [15]

a„(J + x2) — w2 — , 0 < s2 <c

dx7* ~ D2 ~ d2 stress p, the joint dimensions and the physical properties of the
• [8] sheet material in the following manner:

where Di and D> are the flexural rigidities of the sheet and joint cosh w2c
k = [16]
respectively. cosh u2c sinh u2c
The two preceding Equations [8] have solutions of the form
Values of k = and of for various values of - -4/— are
u>i = Ai cosh MiSi + Bi sinh UiXi + anxi, 0 <<_ I Tt Tt t IE
t
Wt = A2 cosh UiX2 + B2 sinh «2s2 + a„ ^ Z + E 2 — - I, 0 £ s 2 ^ c| TABLE 1 V A L U E S OF h = A N D k' = ~ FOR VARIOUS
Tt Tt
P
• [9] i V "
' E
where 2 Mo VoC
i V 5 k = Tt
( v E Tt

=VI 0 1.00 0
Ml and Ui = 0.1 0.75 0.039
D2 0.2 0.61 0.064
0.3 0.51 0.080
The four constants of Equations [9] are readily evaluated from 0.4 0.45 0.094
0.5 0.40 0.104
the four conditions 0.6 0.37 0.116
0.65 0.36 0.122
at si = 0 Wi = 0
given in Table 1; Fig. 3 is a graphical plot of the factor k versus
at si = I, X2 — 0 Wi = Wi
0 Ip
dw i •[10] For average values of overlap and for plate stresses as high
at Si = I, X2 = 0 ^™E
tZsi dx2
at s2 = c w2 = 0 as 40,000 psi in joints composed of cemented aluminum-alloy

The moment and shearing force at the transition section then 1.00
become
0.90

0.80

0.70
cosh M2C sinh uiZ • [11]
«1 ^0.60
2 jsinh MiZ cosh u2c -\—1 cosh Uil sinh u2c{ 0.50
V "2 J
0.40
and
0.30,
0.10 0.20 0.30 0.40 0.50 0.60
= (F,) m =i = (dMl\
\ dxi / x ^ i £/ZL
^ i cosh M2C cosh Uil j [12] f VE

FIG. 3 P L O T O P M O M E N T F A C T O R FC V E R S U S C --\ / 2
2 Jsinh MiZ cosh u2c + —1 cosh wiZ sinh w2c( t\E<E
v Ms ;
sheets, a practical lower limit of about 0.35 is established for k.
Letting p represent the applied tensile stress in the plate (so Hence it is evident that the joint-edge moments are considerably
that T — pt) and E and v the Young's modulus and the Poisson's reduced under load from the initial value V2 Tt, the amount of
 GOLAND, REISSNER—THE STRESSES IN CEMENTED JOINTS A-21

A
+ y
-- 1
+4p
u
1

c y
tx

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
(a)

-?t ?t +X
-2t

FIG. 4 DIAGRAM OF JOINT W I T H RELATIVELY INFLEXIBLE CEMENT LAYER

(a, Profile of joint, showing edge stress distribution; b, periodic function for stress on joint edge y -c; c, periodic function
for stress on joint edge y = c. For k = 1.0.)

reduction being as much as 65 per cent in practical cases. It is , 12M0

also clear, from the values of Table 1, that only small transverse v H
r
z
t3
shearing forces are introduced at the joint edges of loaded systems
with reasonable overlap. The shearing forces are never in excess where z refers to the distance from the elastic axis of the sheet to
of 0.02 T in actual joints. a sheet fiber stressed in tension by the bending moment M,. The
2 STRESS D I S T R I B U T I O N IN JOINT FOR R E L A T I V E L Y I N F L E X I B L E
normal stress distribution at the two edges is shown schematically
CEMENT LAYERS
in Fig. 4(a).
Let a co-ordinate x be defined in the direction of the joint
If the thickness of the cement layer in a joint is exceedingly
thickness, with its origin at the upper surface of the joint, Fig.
small compared with that of the sheet, the presence of the cement
4(a). The co-ordinate y is taken with its origin at the mid-point
can be ignored in the calculation of the stress distribution in the
of the joint length and is measured positively to the right. The
joint. For, in the case of very thin layers, the variations in stress
normal stress distribution of Fig. 4(a) then becomes
and the magnitudes of the deformations in the cement are in-
significantly small and, consequently, do not appreciably affect
the stress distribution in the two adjoining sheet portions. For = p + pk [6 ~ — 3 J, 0 < x <t
at y = —c, (<ry)„=-c
purposes of analysis, therefore, the joint can be assumed to con- = 0, I < x < 2t
sist of a homogeneous slab of thickness 2 t, length 2 c, large
width and with the same physical properties throughout as those = 0, 0 < x < t
of the sheet material. It is to be noted that this treatment is at y = c, (c7„)„ = c
valid even for cement layers of appreciable thickness, providing = p + pk 6 - J, t < x < 2t
the cement material is relatively stiff, i.e., providing its physical
[17]
properties are at least of the same order of magnitude as those of
the sheet material. The evaluation of the stress distribution in the joint is thus
The loads transmitted by the sheet to the edges of the joint transformed to the solution of the plane-strain problem involving
have been determined in Section 1 of the paper. They are found the boundary stresses of expressions [17], acting on the edges of
Tt a wide, rectangular slab of length 2 c and thickness 2 t. The
to consist of a tension T, a moment M0 = k — (each per unit of boundary stresses, expressions [17], will first be written as Four-
z ier series, and a stress distribution, satisfying the conditions at
joint width), and a small transverse shearing force. In this first the edges y — — c and y = c, will be determined with their aid.
analysis, the transverse shearing forces are neglected, the tension This solution will, however, introduce undesirable normal and
T is assumed to be uniformly distributed over the sheet cross shear stresses on the joint surfaces x = 0 and x = 21. An ap-
section and the moment M„ is considered to arise from the usual proximate method is used to remove these stresses, employing the
elementary linear stress distribution. It is evident, from past known results for a series of concentrated forces distributed over
experience with similar problems, that the first two assumptions the plane boundary of a semi-infinite elastic solid.
do not introduce appreciable error; the last assumption implies The edge stress conditions, expressions [17], are made into
that sheet cross sections remain plane near the edges of the periodic functions as shown in Figs. 4(6) and (c).
joint and has, for verification, the results of a series of photo- Writing
elastic tests. The tests, reported by Tylecote (7), were made on
homogeneous, short joints. The departure from a linear normal nir
stress distribution at the joint edges should presumably be a
maximum in short joints; the test results indicate, nevertheless, a
(av)y= -C = a„ +
E-
71=1
cos — x
21
• [18]
remarkable degree of linearity. OO

The normal stresses on the joint edges over the regions where nir
the sheet attaches to the joint are then given by expressions of
the type
W i / = ( = 6„ +
x> cos — x
21
A-22 JOURNAL OF APPLIED MECHANICS MARCH, 1944

the Fourier constants become cosh ay (sinh ac — ac cosh ac) + ay sinh ac. sinh ay
2/i = ; • • • [26]
1 sinh 2ac + 2ac
«. = K = g V
The fact that the length of the joint is alwaj's large compared
2p (12k( nir \ nir I with its thickness leads to a considerable simplification of ex-
a„ — — \ — cos - — 1 + {3k + 1) sin ..[19] pressions [25]. For then c is large compared with t and
nir (nir \ 2 / ~2 j

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
2p (l2fc ( nir \ ,„,,,.. nir) sinh ac ~ cosh ac
bn - —• I cos — cos nir j — ( 3 s + 1) sin — f
(nir \ 2 / 2 ) [27]
„2 ac
According to the theory for plane strain, there exists a stress sinh 2ac ± 2ac « sinh 2ac
function f>i such that

dVi ay AVI
Furthermore, confining attention to the joint regions near the
•[20] edges, where the stress distribution is of most interest, then with
£>xJ' dxdy
small error
where o^, o-„, TXV are stress notations as defined in reference 8.
It is appropriate to take sinh ay ~ cosh ay « - eay
£i
px2 \ \ nir With these approximations, the complexity of the forms describ-
u>i = — + / Y„ cos — x. [21]
v 4 / 4 21 ing ?/i, 2/2, . . . ., 1/e is reduced and the stresses can be written as
n=1

where F„ is a function of n and the variable y only. The require-

ment that V 4 v>i = 0 fixes the form of Y„ as ^ = + B„) za{y-c) [a{y — c) + 1] | cos ax
n=l
Yn = Ci cosh ay + C2 sinh ay + C3y cosh ay + Ciy sinh ay... [22] CO

nir
where a = — . Then, denoting the derivatives of F„ with respect ° v = l + ^2(A n + B„) j ^ - c > [ a ( c - 2 / ) + l] j c o s a s
21
n= l
to the variable y by primes, the three stresses become CO

7xv = + sin ax

- x: n = l
Y." cos ax n=1
.[28]

_?_ V1
CO
where
asY„ cos ax [23]
4 p I 12k•(/ nir \
An + Bn = 26„ = — S — II cos cos — — cos
cosnir
nirI
CO nir ( nir V\ 2
2 J

E aY ' sin ax — (3k + 1) sin ^ j [29]

The four constants appearing in the function Yn are determined During the calculation of the values of the stresses, Equations
from the boundary conditions, previously decided upon [28] in the vicinity of the joint edges, the series converge fairly
rapidly.
at y = —c, r,v = 0 It has already been mentioned that the stress function 45, does
"v - W P - < .[24] not maintain the necessary stress-free conditions on the joint
at y = c, Tx„ = 0 surfaces x = 0 and x = 21. In fact, it is found that large normal
"v ~ (°V)!/ = C and shearing stresses remain on both of these faces. In order to
remove these objectionable stresses, an approximate method de-
and the three stresses <rx, <rv, and TXU then assume the forms
scribed in reference (8) is used. It employs the results, first ob-
tained by Boussinesq, for the plane-strain problem of an inclined

-T.
<rx =
n=1
/ ^{A„yi + B„y2] cos ax
force applied to the plane edge of a semi-infinite elastic slab.
The left joint edge is shown in Fig. 5 with the joint material
arbitrarily extended infinitely to either side of the joint in the x
direction. The plane boundary at y = —c is maintained. A

E [A„y3 + B n yi] cos ax • [25] series of concentrated forces may now be applied to the plane
2 boundary outside of the region from x = 0 to x = 2t without
introducing any' surface stresses over the joint edge. Either
forces parallel or perpendicular to the plane boundary can be ap-

=J2[a» ,2/6 + B„f/»] sin ax plied at will. Let m such separate forces be applied and consider
the resulting stresses at points along the two planes x = 0 and
x = 21. It is obvious that if any combination of in normal and
shear stresses a x and r xy , acting at arbitrarily specified points
where A„ = 6„ + a„, £ „ = 6„ — o„ and yu y2, , ya are six
along these planes, are chosen for reference, then the magnitudes
functions of y only. A typical form is of the m applied forces can be so adjusted as to fix these stresses
 GOLAND, REISSNER—THE STRESSES IN CEMENTED JOINTS A-23

at any desired values. It is the purpose here to set these m

values of <rx and t x v of equal magnitude and opposite sign to those
remaining on the joint surfaces due to the stress function </>, and
by choosing m sufficiently large, to obtain reasonable correspond-
ence between the surface stresses due to the two stress systems
over the entire semi-length c of the joint. By superposition of
the two systems, the surfaces x = 0 and x = 21 are then left free
of stress. It is allowable then to ignore the portions of the slab
outside of the two surfaces and to study only the joint itself.

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
If <p-i refers to the stress function describing the application of the
m forces, the stress distribution in the joint is obtained from the
total stress function
<p = <pi + <P2 [30]
The form of the stress function <pi is obtained from well-known
results (8) and will not be given here in detail for reasons of brev-
ity.
By trial and error, it is found that six concentrated forces, ap-

p
! \
Jfi Surface of Joint x=0
p ' ~ 7 l
V
r, Shear Plane of Joint %-t

Surface of Joint X=2t \ FIG. 6 STRESS DISTRIBUTION ALONG SHEAR P L A N E IN JOINTS WITH
-
r"-1/ 7 RELATIVELY INFLEXIBLE CEMENT; FOR k = 1

P3 - -i /
^—T
FIG. 5 F O R C E P A T T E R N U S E D TO C L E A R THE S U R F A C E S x — 0 AND
x = 21 O F S T R E S S

plied as shown in Fig. 5, are sufficient to remove almost com-

pletely the surface shearing stresses introduced by <pi. Their
positions are dimensioned in Fig. 5, and their magnitudes are,
4.00
each per unit of joint width

Pi = (—0.07 + 0.77/c)jrfp
P2 = —0.30k irtp
3.00
Pi = (0.14 — 0.30 k)ntp [311
P, = 0.16 wtp
Q, = (0.13 — 0.7ok)irtp
Q, = (0.25 — 0.25k)idp
0,2.00
The values of the shearing stresses yet remaining on the surfaces
x — 0, x = 21 of the joint in the neighborhood of the joint edge
are given in Table 2 and are seen to be quite small for all values 1.00
of k. The remaining normal stresses <rx are also given and are of
considerably larger magnitude. Since the joint is relatively
thin compared with its length, it is allowable to consider the mean
of the two remaining surface-pressure distributions to be trans- 0 1.0 0.9 0.8 0.T 0.& 0.5 0.4
mitted directly through the sheet without change of form. Hence, k
this component can be subtracted from the total stress distribu-
FIG. 7 MAXIMUM STRESSES IN SHEAR PLANE FOR JOINTS WITH
tion and the net remaining surface pressures are given by the dif- RELATIVELY INFLEXIBLE CEMENT LAYERS, PLOTTED VERSUS MO-
ferences between the original and the removable mean stresses, MENT FACTOR k

TABLE 2 S T R E S S E S R E M A I N I N G ON J O I N T S U R F A C E S , x = 0 A N D x = 21
Distance
from Mean normal
joint Shear stress Shear stress Normal stress Normal stress stress ax on Final normal Final normal
edge Txy on x = 0 Txy on x = 2t uioiii = 0 ax on x = 21 x= 0, x = 21 stress ax on x — 0 stress aj on x = 2/
c—y -4- p tp -j-p -i-p -T- p + p
0 0 0 —0.100 + 0.100/t 0.166 — 0.166ft 0.033 — 0.033fc — 0 . 1 3 3 + 0.1334 0.133 — 0.133ft
0.5 ( 0.01S — 0.047ft — 0 . 0 7 5 + 0.093ft —0.090 + 0.544ft 0.150 + 0.256ft 0 . 0 3 0 + 0.400ft — 0 . 1 2 0 + 0.144ft 0.120 — 0 . 1 4 4 f t
I.0( 0.021 + 0.058ft —0.017 —0.079ft 0.011 + 0.309ft 0.008 + 0.158ft 0.010 + 0.234ft 0.001 + 0.075ft — 0 . 0 0 2 — 0.076ft
1.51 0.026 + 0.070ft — 0 . 0 0 5 — 0.054ft 0.029 + 0.1564 —0.040 + 0.07oft — 0 . 0 0 5 + 0.116ft 0 . 0 3 4 + 0.040ft — 0 . 0 3 5 — 0.041ft
2.Of 0.025 + 0.048ft + 0 . 0 0 6 — 0.037ft 0.023 + 0.067ft —0.042 + 0.042ft — 0 . 0 1 0 + 0.055ft 0.033 + 0.0124 — 0 . 0 3 2 — 0.013fe
A-24 JOURNAL OF APPLIED MECHANICS MARCH, 1944

as indicated by the last column of Table 2. No attempt will be of the joint and its positive direction to the right. The bending
made to remove the small normal and shearing stresses which moments, vertical shear, and axial tension in the sheets are de-
persist on the joint surfaces. noted by M, V, and T, respectively, and the subscripts u and I
It is pertinent to study the probable errors resulting from the designate quantities pertaining to the upper or lower sheet,
small stresses remaining on the surfaces x = 0 and x = 2t. respectively. The transverse normal (tearing) stress and the
These are best investigated by noting the effect of changing the shear stress in the cement are notated A„ and T0, respectively.
magnitudes of the applied concentrated forces on the stress dis- Figs. 8(b) and (c) show elements of the upper and lower sheet,

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
tribution in the joint. It is found that large changes in the with the sign conventions chosen for moments and forces indi-
values of these forces result in relatively small changes in the cated.
stress distribution across the shear plane x = I of the joint. Since The conditions of moment equilibrium for the elements of the
this is the region of greatest practical significance, it is concluded sheet are
that the approximate method used to remove the undesirable
dMu t
surface stresses is of sufficient accuracy. -rJi-F u + r0 - = 0
dx 2
Fig. 6 is a plot of the stress distribution along the mid-plane (32]
x = t of the joint for the case k = 1. This plane corresponds with dM, t
that of the cement layer and hence the stresses TXU and AX shown ~r ~ V, + r0 - = 0
dx 2
are those to which the cement material is subjected in a loaded
joint. The stresses are highest near the joint edges and only this The conditions of horizontal force equilibrium are
region is shown in Fig. 6. The stresses in the other regions of the
joint are of lesser interest, since joint failure is usually precipi- dT,.
~ rn = 0
tated by fracture of the relatively weak cement material. dx
[33]
It is recognized that the peak values of <RX and TXY at the shear dT.
plane will change with the factor k. Fig. 7 shows this variation -rl + T» = 0
dx
by plotting the maximum values of the cement stresses <sx =
and TXU = T0 as a function of k. The change in the longitudinal The conditions of vertical force equilibrium are
stress af in the sheet fibers adjoining the cement, brought about
by a change in the value of k, is also indicated on this plot. The (IK
significance of Figs. 6 and 7 in explaining joint behavior has al- • t„ = 0
dx
ready been discussed. [34]
dV,
— + <r0 = 0
3 STRESS D I S T R I B U T I O N IN JOINT FOR R E L A T I V E L Y FLEXIBLE dx
CEMENT LAYERS
Denoting by vu and vt the transverse deflections of the upper
In this section, joints are considered for which the following
assumption is permissible: The transverse normal strain and and lower sheets, respectively, both measured positively upward,
shear strain in the jointed sheets are negligibly small compared then from thin plate theory
with the corresponding strains in the cement layer. Assuming,
Mu
then, that the deformation of the sheets is due solely to the longi-
tudinal normal stress <rx, it is apparent that the sheets may be dx2 D
• [35]
treated as cylindrically bent plates, whereas the role played by Mi
the cement layer is analogous to that of a system of infinitesimal dx2 D
coil springs positioned between the two plates.
Fig. 8 represents the cemented joint with the loads, as deter- Et°
where D = refers to the flexural rigidity of the sheets.
mined in Section 1, applied to it. The co-ordinate x is defined in 12(1 — v2)
the direction of the joint length, with its origin at the mid-point Let iiu and ut represent the longitudinal displacements of the

+4p
A

I- 4

(a)
crdx
rf.T
•To
Vu
Mu
Tu

Tgdx- clx
(c)
(h) dx

FIG. S DIAGRAM OF JOINT WITH RELATIVELY FLEXIBLE CEMENT LAYER

(a, Profile of joint, showing longitudinal stress distribution on edges; b, element of upper sheet; c, element of lower sheet. For k = 1.0.)
 GOLAND, REISSNER—THE STRESSES IN CEMENTED JOINTS A-25

sheet at the sheet boundaries adjacent to the cement. Then, The boundary conditions for t„ and a„ are obtained by com-
the stress-strain relations give bining Equations [38], [39], [42], and [43], The conditions for
t0 are
duu
dx at
• 136]
dui _ 1 (TI + q^I)
to which is added the equilibrium condition
dx E\t t*J

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
where the terms on the right side of Equations [36] are the r0dx + T = 0. .[47]
known combinations of bending and direct stress in the outer- —c
most fibers of a loaded beam. The conditions for a, are
The required system of equations is completed by the relations
between the stresses and strains in the cement. For an elastic Ec
cement they are dx2 TJZ) "
at x = =•= c • [48]
Uu UI
^ = ± * V
dx3 VJ) °
[37]
It may be noted that Equations [45] and [48] for a0 are analo-
E. gous to the equations for the deflection of a beam on elastic
It might be remarked that the linear stress-strain relations of foundation, acted upon by edge forces and moments.
Equations [37] could be replaced by nonlinear expressions with- On the basis of the differential Equation [41] and the boundary
out making the solution of the problem of prohibitive difficulty. conditions, Equations [46] and [47], the distribution of shear
The edge conditions for the two sheets are stress in the cement is obtained in the form
Pc x
Mu = Tu = cosh — • -
at x = c,
7 „ = 0
/3c t c
Ml - M „ 7 , = 7 „ T, = T v ( 1 + 3A') — +3(1 [49]
. [ 38] V t 8 \ t ,,f3c
MI = T, = 7, = 0 sinh —
at x = — c, t
M U = — M„, 7 „ = 7 0 , T„ = 7'
where
The system of Equations [32] to [38], inclusive, is now to be
reduced to two differential equations for <r0 and t0 and to bound-
ary conditions in terms of these two quantities. Combining
Equations [36] and the first of Equations [37], there follows

I Tu - T, _ il/„ +2 Jf,
E < < [39]
dx 0.14

Differentiating once and observing Equations [32] and [33]

0.12
— . _ 1 / o If 4. f Vu + 7,N
[40]
Gc' dx* E \ t t
-0.10
Differentiating once more and noting Equations [34], an equa-
tion in t„ only results,

8GC dr. -0.08

= 0. [41]
dx3 Eh, dx
An equation for a0 is found by first differentiating the second f°k
0.06
of Equations [37] twice and introducing Equations [35], Then

M, — Mu
[42] 0.04
Er_ dx2 D
From Equations [32], it follows that
-0.02

1
(V, - 7 ). •[43]
E, dx3 vD
U

and from Equations [34], finally

I.^V? =
• [44]
Ec ' dx1 vD'

-0.04
d*aO 24(1 — v*)Ec
• [45] FIG. 9 STBESS DISTRIBUTION IN CEMENT IN JOINTS WITH RELA-
dx* + J3E =
TIVELY FLEXIBLE C E M E N T LAYERS
A-26 JOURNAL OF APPLIED MECHANICS MARCH, 1944

Ri = cosh X sin X + sinh X cos X

Rz = s i n h X c o s X — c o s h X s i n X

A = -(sinh 2X + sin 2X)

and, from Section 1

2 M. V0c
= k. = k'

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
pt2 pt2

The maximum value of <r0 acts at the joint edges and is given by

(<r„)max /c\2 . k sinh 2X •- sin 2X . cosh 2X + cos 2X

+ Xfc'
V \tj 2 sinh
sir 2X + sin 2X sinh 2X + sin 2X
[54]
For long joints, defined by X greater than, say, 5/2 this becomes

M-
v # t
v
,, t
- + k' -
c wJ E V
. . .[55]

It is noteworthy that, for long joints, the transverse normal

stress and shear stress in the cement layer depend in essentially
the same way on the elastic-moduli ratio and the thickness ratio
of the sheet and cement. While, however, the shear stress T„
is due in equal measure to the direct sheet stress and the sheet
bending stress, the normal stress a0 is almost entirely due to the
FIG. 10 MAXIMUM SHEAB STBESS IN CEMENT IN JOINTS WITH
existence of the sheet bendingL g stress.
BUlt
RELATIVELY FLEXIBLE CEMENT LAYERS
... ., . , , - (Omi
Fig. 11 is a plot of versus X for various values
This expression may be compared with the average shear P
of the factor k. A typical <r„ distribution in a representative
• [50] joint is shown in Fig. 9. The end concentration of a„ is of inter-
(T ° )avg ~ 2c est, showing the existence of tearing stresses of appreciable mag-
nitude (see section on "Results and Discussion of Results").
Fig. 9 shows a typical distribution of T„ in a joint of representa- It is of interest also to note that increasing overlap ratios c/t
tive dimensions and materials and for values of the factor k of result in decreasing values of the maximum tearing stress (<T<,)mai
1 and 0.60. It should be noted that, as expected, Equation [49] at the joint edges.
indicates a substantially uniform shear distribution in short
ACKNOWLEDGMENT

0c
joints. For longer joints, say for — > 0.5, this is no longer the The authors are indebted to Mr. E. S. Jenkins, Chief of Ma-
case.
The maximum value of T0 is found at the edge of the joint and 70 A-Loi

/
(r.)
't 8Li
(1 +3k) coth J + 3 (1- k) .[51] 60

/
/ 0.3^

/
( o)max: C . , ,
T

Fig. 10 is a plot of — • - versus — for various values of k.

r P
50 •0 H
p t t
An inspection of Equation [51] indicates that increasing the
/ /k-O.i/

//
0c
40
length of joint beyond the limit — » 25 has no effect on the
t /
magnitude of the maximum shear, which remains
/
/
/
/ / /
30
k-0
G,.
/ / /
8
(On (1 + k)
3*)
E
• [52]
/
/ V/
70
The <t0 distribution is found to be, on the basis of Equations
[45] and [48] / /
10 y
—[- | = i (R-iY1 + \k' cosh \ cos X) cosh X X cos X ~
p\{/ A 2 c c

+ (RiX1 - + Xfc' sinh X sin X) sinh X - sin X - • [53] <o 10

2 c c
r f
Ec t c
where = 6 — ' X = y - FIG. 11 MAXIMUM TEARING STRESS IN CEMENT IN JOINTS WITH
Ei, t RELATIVELY FLEXIBLE CEMENT LAYERS
 GOLAND, R E I S S N E R — T H E STRESSES IN C E M E N T E D JOINTS A-27

terials and Structures Research, Research Laboratory, Cur- tions," by W. Hovgaard, Proceedings National Academy of Sci-
tiss-Wright Corporation, Airplane Division, for the suggestion ences, vol. 17, 1931, pp. 351-359.
3 "An Investigation of the Stresses in Longitudinal Welds,"
to undertake this investigation and tor permission to publish its
by W. Hovgaard, Proceedings National Academy of Sciences, vol.
results. During the course of the work, helpful advice was re- 20, 1934, pp. 17-23.
ceived from Mr. Jenkins and from Mr. A. F. Donovan, Head of 4 "The Stress Distribution in Longitudinal Welds and Adjoining
Structures Group, Research Laboratory. Curtiss-Wright Cor- Structures," by W. Hovgaard, Journal of Mathematics and Physics,
poration. vol. 13, 1934, pp. 195-248.
5 "Distribution of Shear in Welded Connections," by H. W .

Downloaded from http://asmedigitalcollection.asme.org/appliedmechanics/article-pdf/11/1/A17/6744732/a17_1.pdf by Virginia Polytechnic Institute and State University user on 06 October 2021
Troelseh, Trans. A.S.C.E., vol. 99, 1934, pp. 409-436.
BIBLIOGRAPHY 6 "Stress Distribution in Side-Welded Joints," by W. H. Weis-
1 "The Stress Distribution in Welds," by W. Hovgaard, Pro- kopf and M. Male, Journal of the American Welding Society, vol. 9.
ceedings National Academy of Sciences, vol. 16, 1930, pp. 667-673; Sept., 1930, pp. 23-48, and Dec., 1930, pp. 68-82.
also "The Stress Distribution in Welded Overlapped Joints." by W. 7 "Spot Welding," by R. F. Tylecote, Welding Journal, Welding
Hovgaard, Proceedings National Academy of Sciences, vol. 16, 1930, Research Supplement, vol. 20, Aug., 1941, pp. 359-368.
pp. 673-678. 8 "Theory of Elasticity," by S. Timoshenko, McGraw-Hill
2 "The Distribution of Stresses in Welded and Riveted Connec- Book Co., Inc., New York, N. Y., 1934, pp. 82-93.