ANALYTIC SOLUTION FOR AXIAL POINT LOAD STRENGTH TEST

ON SOLID CIRCULAR CYLINDERS

By X. X. Wei,1 K. T. Chau,2 and R. H. C. Wong3

ABSTRACT: This paper presents a new analytic solution for the stresses within an elastic solid finite cylinder
subjected to the axial point load strength test (PLST). The ‘‘displacement potential approach’’ is used to uncouple
the equations of equilibrium; then the contact stresses on the end surfaces induced by the point load indentors
are expanded in terms of a Fourier-Bessel expansion to yield the unknown constants of the appropriate form of
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

the displacement potential. Our solution shows that a zone of higher tensile stress is developed in the vicinity
of the applied point loads, compared with the roughly uniform tensile stress in the central portion of the line
between the two point loads. This peak tensile stress within the cylinder decreases with increasing Poisson’s
ratio and the size of the loading area, and it increases with increasing Young’s modulus. The tensile stress
distributions along the axis of symmetry in a cylinder under the axial PLST are remarkably similar to that
observed in a sphere under the diametral PLST. Our solution also demonstrates both size and shape effects on
the point load strength index (PLSI) that we observed in our experiments. In particular, for a fixed length-to-
diameter ratio, the larger the specimen, the smaller the PLSI, whereas for a fixed diameter, the longer the
specimen the smaller the PLSI.

INTRODUCTION between the indentors and the surfaces of the specimens and
idealized the applied tractions as either point forces [e.g., Wijk
The direct tensile strength test on rock is seldom used due (1978)] or uniform traction (Hiramatsu and Oka 1966), al-
to the fact that tensile force is difficult to apply to rock samples though, in reality, the so-called point loads are actually applied
without inducing eccentric moments and the fact that the ten- through a pair of steel cones with spherical heads (ISRM
sile stress is normally induced indirectly by compressions in 1985). The finite-element method has been applied to the axial
rock masses in real situations. The point load strength test PLST [e.g., Peng (1976)], but the contact problem between
(PLST), which is a popular index test for strength estimation the steel cones and the end surfaces was not considered.
of rock, has been extensively used both in the field and in the Stress analysis for finite cylinders under various boundary
laboratory because of its simple testing procedures, its cheap conditions has been one of the most fundamental problems in
operating cost, and its short testing time [e.g., Chau and Wong the theory of elasticity. The most notable early development
(1996)]. In estimating the strength of rocks, cylindrical cores was the analysis by Filon (1902) for finite solid cylinders un-
obtained by drilling are the commonly used samples in the der uniaxial compression with end friction. Dougall (1914)
PLST. Specimens are normally split apart along the line join- proposed a general analysis for solid circular cylinders under
ing the two applied point loads. various types of loading. The application of Dougall’s ap-
Extensive experimental studies have been done on the PLST proach (1914) to the axial PLST is, however, not straightfor-
[see the review by Chau and Wong (1996)], but there are rel- ward, as the physical meanings of the so-called permanent free
atively few theoretical studies for the PLST. The analytic stud- modes and transitory free modes are not readily understanda-
ies for the PLST include the stress analyses of the diametral ble.
PLST on isotropic spheres (Hiramatsu and Oka 1966), spher- Using Love’s stress function for axisymmetric problems
ically isotropic spheres (Wei and Chau 1998; Chau and Wei [e.g., Love (1944, Section 188)], Saito (1952, 1954) proposed
1999), and the finite cylinders (Wijk 1980; Chau 1998a) and a Fourier-Bessel expansion of the stress function for satisfying
the stress analysis of the axial PLST on finite cylinders (Wijk any axisymmetric boundary loading conditions. By adopting
1978). However, it should be emphasized that the solution for a two stress function approach, Ogaki and Nakajima (1983)
the axial PLST obtained by Wijk (1978) is only an approxi- also proposed a general procedure and derived an analytic so-
mation of the tensile stress at the center of the cylinder. Nev- lution for finite cylinders with parabolic distributed loads act-
ertheless, the solution by Wijk (1978) was found useful in ing on the central part of the two end surfaces. Applying
estimating analytically the index-strength conversion factor for Saito’s approach (1952, 1954), Watanabe (1996) obtained the
the axial PLST (Chau and Wong 1996, 1998). analytic solution for cylinders under compression with per-
The tensile stress field within a cylinder under the axial fectly or imperfectly constrained radial displacement on two
PLST has not been solved analytically; the only available ap- loading end surfaces. This approach can also be used to find
proximation, as mentioned previously, is the stress at the cen- Young’s modulus from compression tests with end friction
ter of the cylinder (Wijk 1978). In addition, all previous anal- (Watanabe 1998) and provides a check for other approximate
yses [except Chau (1998a)] neglected the contact problem methods (Chau 1997, 1998b).
1
One important feature of the PLST is that all specimens are
PhD Student, Dept. of Civ. and Struct. Engrg., Hong Kong Polytech- split apart into two or three pieces in tensile mode, which is
nic Univ., Hung Hom, Kowloon, Hong Kong, China, on leave from Dept. believed to be caused by the tensile stress between the two
of Mech., Lanzhou Univ., Lanzhou, China.
2
Assoc. Prof., Dept. of Civ. and Struct. Engrg., Hong Kong Polytechnic point loads. Thus, the stress distribution along the line joining
Univ., Hung Hom, Kowloon, Hong Kong, China. the two point loads is of great importance to the PLST. As
3
Asst. Prof., Dept. of Civ. and Struct. Engrg., Hong Kong Polytechnic remarked earlier, there is no solution for the tensile stress dis-
Univ., Hung Hom, Kowloon, Hong Kong, China. tribution along the two applied point loads within cylindrical
Note. Associate Editor: Gilles Pijaudier-Cabot. Discussion open until specimens under the axial PLST.
May 1, 2000. To extend the closing date one month, a written request Therefore, this paper considers the stress distribution in a
must be filed with the ASCE Manager of Journals. The manuscript for
this paper was submitted for review and possible publication on March
finite circular cylinder subjected to the axial PLST. The
29, 1999. This paper is part of the Journal of Engineering Mechanics, method of solution, in contrast to the stress approach by Saito
Vol. 125, No. 12, December, 1999. qASCE, ISSN 0733-9399/99/0012- (1952, 1954) and Ogaki and Nakajima (1983), uses the dis-
1349–1357/$8.00 1 $.50 per page. Paper No. 20575. placement potential approach, and the boundary contact forces,
JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999 / 1349

J. Eng. Mech. 1999.125:1349-1357.

 similar to the approach by Saito (1952, 1954) and Ogaki and ing respectively to those stresses; and u and w = displacements
Nakajima (1983), are expanded into Fourier-Bessel series. In in the r- and z-directions, respectively.
contrast to all previous studies [except Chau (1998a)], the two The specimen is assumed to be linear elastic homogeneous
point forces are modeled more realistically as the contact stress and isotropic; thus the following Hooke’s law applies:
between the flat end surfaces and the spherical heads of the
indentors. In verifying the applicability of the present solution 1
ε rr = [srr 2 n(suu 1 szz)] (4)
to real PLST, a series of the axial PLST have been carried out E
on rock-like plaster specimens. Plaster is used in our PLST
1
because man-made plaster is generally more homogeneous ε uu = [suu 2 n(szz 1 srr)] (5)
than natural rock, in which inherent defects inevitably exist. E
Except in the vicinity of the contact zone between the in-
1
ε zz =
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

dentors and the end surfaces, where crushing is the dominant [szz 2 n(suu 1 srr)] (6)
mode of deformation, the present linear elastic solution should E
provide meaningful results for interpreting the failure mecha- 1
nism and strength of the rock specimens under the axial PLST. ε zr = szr (7)
2G
GOVERNING EQUATIONS where E, n, and G = Young’s modulus, Poisson’s ratio, and
Consider a cylindrical specimen of diameter D (or 2R) and the shear modulus, respectively.
length H (or 2h). For mathematical simplicity, we take the As shown in Fig. 1, all the surfaces of the specimen are
center of the cylinder as the origin of our cylindrical coordi- traction free, except for the two small circular areas where the
nates (r, u, z) with the z-axis coinciding with the axis of sym- spherical heads of the steel cones come into contact with the
metry of the specimen, as shown in Fig. 1. Because the axial end surfaces of the specimen (one on the top and the other at
point forces are applied by pressing two collinear indentors the bottom). The standard size for the radius of the spherical
along the axis of the symmetry of the cylinder, all variables heads in 5 mm, as suggested by ISRM (1985), whereas our
of the problem are independent of u. In the absence of body experimental results show that the size of the contact area is
force, the equations of equilibrium are as follows (Love 1944): in the range of 1 to 5 mm. Therefore, the contact is expected
to develop only between the spherical heads of the steel cones
­srr ­srz srr 2 suu and the end surfaces of the specimen (probably except for very
1 1 =0 (1) soft rocks). In addition, the size of the contact area is much
­r ­z r
smaller than the standard diameter of the specimen (i.e., D =
­szz ­srz srz 50 mm), as suggested by Broch and Franklin (1972) and IRSM
1 1 =0 (2)
­z ­r r (1985). Hence, the contact stress acting between the steel
cones and the end surfaces of the specimen can be approxi-
where the usual notation for Cauchy stress tensor is adopted. mated by considering the Hertz contact problem between a
For axisymmetric problems, strains and displacements are spherical surface of radius R2 and a semi-infinite flat surface.
related by In particular, the formulas for contact stress in Section 140 of

ε rr =
­u
­r
;
u
ε uu = ; ε zz =
­w
­z
; ε rz =
1
S
­u
­z
1 D
­w
­r
(3a–d )
Timoshenko and Goodier (1982) can be specialized to the
present case as

H
r 2
2p(r) r # R0
where szz, srr, and suu = normal stresses in the axial, radial, szz = (8)
0 r > R0
and circumferential directions, respectively; srz = shearing
stress on the rz-plane; ε zz, ε rr, ε uu, and ε rz = strains correspond- where
p0
p(r) = ÏR 02 2 r 2 (9)
R0

3P
p0 = (10)
2pR 02

F G
1/3
3pP(d1 1 d2)R2
R0 = (11)
4

1 2 n2 1 2 n 22
d1 = ; d2 = (12a,b)
pE pE2
where P = magnitude of the applied point force; R0 = radius
of the circular contact area; and R2, E2, and n2 = radius of the
spherical heads, Young’s modulus, and Poisson’s ratio of the
spherical heads, respectively. As suggested by ISRM (1985),
the spherical head is of radius R2 = 5 mm and is made of
tungsten carbide. Thus, Young’s modulus E2 can be assumed
large enough such that d2 = 0 will be used in (11)–(12).
Therefore, the boundary conditions for a solid cylinder sub-
jected to the axial PLST are as follows:
srr = 0, srz = 0 on r=R (13)
FIG. 1. Sketch for Axial PLST for Cylindrical Rock Core of Ra-
dius R and Length 2h. Origin and z -Axis of Coordinates Is Set at
Center of Cylinder and Axis of Symmetry
srz = 0, szz = H 2p(r) r # R0
0 r > R0
on z = 6h (14)

1350 / JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999

J. Eng. Mech. 1999.125:1349-1357.

 where p(r) = contact pressure given by (9). and J1(x) = Bessel functions; I0(x), and I1(x) = modified Bessel
The next task is to find the solution for the stresses governed function of the first kind of zero and first order, respectively;
by the equations of equilibrium [(1)–(2)] together with the A 0, C0, A n, Bn, Cs, and Ds are unknown coefficients to be de-
boundary conditions [(13) and (14)]. termined. This solution form is similar to those proposed by
Saito (1952), if the minor typographical errors on the functions
METHOD OF SOLUTION sinh and cosh are corrected. Note that the product of the mod-
ified Bessel functions and trigonometric functions in the first
In contrast to the stress potential approach used by Saito summation and the product of hyperbolic functions and Bessel
(1952, 1954) and by Ogaki and Nakajima (1983), the displace- functions in the second summation are general solutions of the
ment potential approach is adopted. In particular, the following biharmonic equation. The first two terms in (22) account for
displacement potential F is introduced: the cases of constant normal stresses being applied on the sur-

F G
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

­2F ­2F face of the cylinders.

u= ; w = 2 2(1 2 n)=1F 1 (1 2 2n) (15a,b) It is straightforward to see that (22) satisfies the biharmonic
­r­z ­z 2
equation [(17)]; thus (22) is a probable solution for the prob-
where lem. The next step is to see whether the boundary conditions
can be satisfied exactly by the series expansion [(22)] by as-
­2 1 ­ signing appropriate values to the unknown coefficients.
=1 = 2 1 (16)
­r r ­r
Substitution of (15) into (1)–(2) and by the virtue of (3)– DETERMINATION OF UNKNOWN COEFFICIENTS
(7), the two equations of equilibrium are reduced to the fol-
lowing governing equation for the displacement function F: To see whether the series representation given in (22) is
=4F = =2=2F = 0 (17) appropriate for our problem of the axial PLST, the shear stress
is first expressed in terms of the unknown coefficients by sub-
where =2 = Laplacian operator, or =2 = =1 1 ­2/­z 2. That is, stituting (22) into (21)
F satisfies the biharmonic equation.

O
`
Substitution of (15) into (3) then results in (4)–(7) that leads
to the following stresses in terms of the displacement function srz = sin(nph){A n I1(bnr)
F n=1

­F ­3F 1 Bn[2(1 2 n)I1(bnr) 1 bnrI0(bnr)]}

srr = 22nG =2 1 2G

O
(18)
­z ­z­r 2 `

1 J1(lsr)[(Cs 1 2nDs)sinh(gsh) 1 Dsgsh cosh(gsh)]

­F 1 ­2F (23)
suu = 22nG =2 1 2G
s=1
(19)
­z r ­z­r

F G
By noting that the shear stress srz is identically zero on the
­ 2 ­3 curved surface r = R (or r = 1), we obtain the following ex-
szz = 22G (2 2 n) = 2 3 F (20)
­z ­z pression between A n and Bn:

srz = 2G F
2(1 2 n)
­
­r
=1 1 n
­3
­r­z 2
G F (21) An
=
Bn
=
En
2(1 2 n)I1(bn) 1 bn I0(bn) 2I1(bn) Dn
(24)
Except for differing by a constant, these expressions are, in
essence, the same as those obtained by Love (1944, Section where En and Dn = constants introduced to simplify the sub-
188), although a different approach is adopted here. sequent analysis.
Similarly, the shear stress is also identically zero on the two
APPROPRIATE FORM OF DISPLACEMENT end surfaces z = 6h (or h = 61), and this condition yields
POTENTIAL the following relation between Cs and Ds:
The most difficult task for solving our problem is to find Cs Ds 2Fs
the appropriate form for the displacement potential such that = = (25)
2n sinh gs 1 gs cosh gs 2sinh gs Vs
both the biharmonic equation [(17)] and the boundary condi-
tions [(13)–(14)] can be satisfied. As shown in Fig. 1, the axial
where Fs and Vs = constants introduced to simplify the ex-
PLST is symmetric with respect to the center of the cylinder.
pression involved in the subsequent analysis.
Due to this symmetry, srr, suu, and szz must be even functions
with respect to z, whereas srz is an odd function with respect On the curved surface r = R, the radial stress can be found
to z. Thus, the following series representation for the displace- by substituting (22) into (18) and setting r = 1 in the resulting
ment function F is adopted: equation

F=2
R3
2G
HA0
k3h3
6
1 C0
khr2
2 srr = A 0n 1 (2n 2 1)C0 2 O
`

cos(nph){A n[I0(bn) 2 I1(bn)/bn]

O
n=1
`
sin(nph) 1 Bn[(1 2 2n)I0(bn) 1 bn I1(bn)]}
1 [A n I0(bnr) 1 BnbnrI1(bnr)]
b3n

O
n=1
`

O J
`
J0(lsr) 1 J0(ls){[Cs 1 (2n 1 1)Ds]cosh(gsh) 1 Dsgs sinh(gsh)}
1 [Cs sinh(gsh) 1 Dsgsh cosh(gsh)]
l3s
s=1
s=1 (22)
(26)
where the two normalized coordinates r and h are defined as
r = r/R and h = z/h; k = geometrical shape ratio defined as k In terms of the newly introduced constants in (24)–(25), (26)
= h/R; ls = sth root of J1(x) = 0; gs = lsk; bn = np/k; J0(x) can be rewritten as follows:
JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999 / 1351

J. Eng. Mech. 1999.125:1349-1357.

 O
`
[bn I0(bn) 2 2I1(bn)]I0(bnr) 2 bnrI1(bn)I1(bnr)
srr = A 0n 1 (2n 2 1)C0 1
O
En cos(nph) `
n=1 ls2

O
= 24b I (bn) 2
J0(lsr)
[l 1 (bn)2]2J0(ls)
` n 1 2
Fs J0(ls) s=1 s (37)
1 [(sinh gs 2 gs cosh gs)cosh(gsh)
s=1 Vs Thus, szz can be expressed in terms of a series of J0(lsr) only.
In particular, substitution of (37) into (35) leads to
1 gsh sinh gs sinh(gsh)] (27)

OF O G
` `
In obtaining (27), we have set szz = A 0(1 2 n) 1 2(2 2 n)C0 1 Fs 1 En Rsn J0(lsr)
s=1 n=1
Dn = {[2(1 2 n) 1 b ]I (bn) 2 b I (bn)}/bn
2
n
2
1
2 2
n 0 (28) (38)
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

To apply the boundary condition for srr on r = R (or r = where

1), we first consider the following Fourier series expansion for
the terms in brackets in the second summation of (27) 4(21)(n11)l2s bn I 12(bn)
Rsn = (39)
Dn[ls2 1 (bn)2]2J0(ls)
(sinh gs 2 gs cosh gs)cosh(gsh) 1 gsh sinh gs sinh(gsh)

O
` To match the load-induced contact stress [(8)] on the end
(21)n(np)2
= 4gs sinh2gs cos(nph) surfaces with the stress field given in (38), (8) is again ex-
n=1 {g2s 1 (np)2}2 (29) panded into Fourier-Bessel expansions

O S D
`
Consequently, (27) can be expressed as a function of cos(nph) 2p0 R ls R0 ls R0 ls R0
only szz = cos 2 sin J0(lsr) (40)
s=1 l3s R0 J 20(ls) R R R

OF O G
` `
The details of this expansion are described in Appendix I for
srr = A 0n 1 (2n 2 1)C0 1 En 1 FsQ sn cos(nph) (30)
n=1 s=1
the sake of completeness.
Finally, by comparing the coefficients of (38) and (40), we
where have
4(21)ngs(np)2J0(ls)sinh2gs A 0(1 2 n) 1 2(2 2 n)C0 = 0 (41)
Q sn = (31)

O S D
Vs{gs2 1 (np)2}2 `
2p0 R ls R0 ls R0 ls R0
Fs 1 En Rsn = cos 2 sin (42)
Thus, substitution of (30) into the boundary condition of (13) n=1 l3s R0 J 20(ls) R R R
on the curved surface r = 1 leads to the following relations
between A 0 and C0 and between En and Fs: It is straightforward to see from (32) and (41) that the only
solution for A 0 and C0 is
A 0n 1 (2n 2 1)C0 = 0 (32)
A 0 = C0 = 0 (43)

O
`

En 1 FsQ sn = 0 (33) For the unknown coefficients En and Fs, the two coupled sys-
n=s tem of equations [(33) and (42)] have to be solved simulta-
neously. For numerical implementation, we can truncate the
Finally, by combining (20), (22), (24), and (25), we obtain the infinite series in (33) and (42) and only retain the first n and
following expression for szz: s terms. Then, there will be (s 1 n) linear equations for the

O (s 1 n) unknown coefficients of Fs and En. Once these coef-

`
En cos(nph)
szz = A 0(1 2 n) 1 2(2 2 n)C0 1 ficients are solved, the following expressions can be used to
n=1 Dn evaluate the stress field inside the cylinder:
? {[bn I0(bn) 2 2I1(bn)]I0(bnr) 2 bnrI1(bn)I1(bnr)}
O
`
En cos(nph)

O
` szz = {[bn I0(bn) 2 2I1(bn)]I0(bnr)
Fs J0(lsr) n=1 Dn
1 [(sinh gs 1 gs cosh gs)cosh(gsh)
Vs
O
`
s=1 Fs J0(lsr)
2 bnrI1(bn)I1(bnr)} 1
2 gsh sinh gs sinh(gsh)] (34) s=1 Vs

On the end surface z = 6h, we can set h = 61 in (34) and it ? [(gs cosh gs 1 sinh gs)cosh(gsh) 2 gsh sinh gs sinh(gsh)]
is reduced to (44)

O O H
` `
En(21)n En cos(nph)
szz = A 0(1 2 n) 1 2(2 2 n)C0 1 srr = 2[I1(bn) 1 bn I0(bn)]I0(bnr)
n=1 Dn n=1 Dn

?{[bn I0(bn) 2 2I1(bn)]I0(bnr) 2 bnrI1(bn)I1(bnr)}

1 bnrI1(bn)I1(bnr) 1 [2(1 2 n)I1(bn) 1 bn I0(bn)]
I1(bnr)
J
O
`
bnr
1 Fs J0(lsr)
O H
`
Fs
s=1 (35) 1 J0(lsr)[(sinh gs 2 gs cosh gs)cosh(gsh)
s=1 Vs
Note that in obtaining (35), we have set
J1(lsr)
Vs = sinh gs cosh gs 1 gs (36) 1 gsh sinh gs sinh(gsh)] 2 [gsh sinh gs sinh(gsh)
lsr

J
To apply the end boundary condition for szz, we first expand
the modified Bessel functions in the first summation of (35) 1 [(1 2 2n)sinh gs 2 gs cosh gs]cosh(gsh)]
to a Fourier-Bessel expansion (Watson 1944) (45)

1352 / JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999

J. Eng. Mech. 1999.125:1349-1357.

 O
`
En cos(nph)
suu = {(1 2 2n)I1(bn)I0(bnr) 1 [2(n 2 1)I1(bn)
n=1 Dn

O
`
Fs
2 bn I0(bn)]I1(bnr)/(bnr)} 1
s=1 Vs

H
? 2n sinh gs cosh(gsh) J0(lsr) 1
J1(lsr)
lsr

? [[(1 2 2n)sinh gs 2 gs cosh gs]cosh(gsh)

J
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

1 gsh sinh gs sinh(gsh)]

(46)

O FIG. 2. Normalized Stresses srr /s0 and szz /s0 versus Normal-
`
En sin(nph)
srz = {bn I0(bn)I1(bnr) 2 bnrI1(bn)I0(bnr)} ized Distance z /h along z-Axis for Various Values of Poisson’s
n=1 Dn Ratio n for H/D 5 1.1, D 5 2R 5 50 mm, and R 0 = 0.05R

O
`
Fs J1(lsr)
2 [gs cosh gs sinh(gsh) 2 gs sinh gs cosh(gsh)]
s=1 Vs ulate that the point where the crack initiates should be the
(47) same as the point where the maximum tensile stress is induced.
Indeed, we will show a later discussion that our experimental
Therefore, the problem has been solved analytically, and thus observation basically agrees with this speculation. If the tensile
the assumed form for the displacement potential [(22)] has strength or PLSI is proportional to the maximum tensile stress
been shown to be appropriate and complete. at the instant of failure, Fig. 2 shows that PLSI is extremely
sensitive to the values of Poisson’s ratio n of the rock. How-
NUMERICAL RESULTS AND DISCUSSIONS ever, the tensile stress at the central part of the axis of sym-
The fracture surface of a failed rock specimen under the metry and the axial compressive stress are insensitive to the
axial PLST must pass through both loading points, if the result change in Poisson’s ratio. This conclusion is similar to that for
is to be accepted as valid data [ISRM (1985)]. Indeed, our an isotropic sphere or a spherically isotropic sphere subjected
experiments show that almost all specimens are split into two to the diametral PLST (Wei and Chau 1998; Chau and Wei
or three prismatic fragments along the line joining the two 1999).
loading points. Thus, the tensile stress along the line of the Fig. 3 illustrates the effect of the radius R0 of the contact
two point loads (i.e., the axis of symmetry as shown in Fig. zone on the magnitude of the local peak of tensile stress for
1) is of crucial importance to interpret the data from the axial H/D = 1.1, D = 2R = 50 mm, and n = 0.25. Except for the
PLST. Therefore, subsequent discussions will focus on the varying R0, the other parameters used in the plot are the same
stress distribution along the z-axis. It should be noted that the as those used in Fig. 2. It is clear that the maximum tensile
radial stress equals the tangential stress along the z-axis, and stress increases drastically with the decrease of R0, particularly
thus only the radial and vertical stresses are plotted in the for R0 < 0.05R. For example, the maximum tensile stress rises
following numerical examples. to about 250% as R0 decreases from 0.05R to 0.01R, and the
Numerical results show that, if h = R (i.e., a roughly equi- tensile stress at the central part of the z-axis and the axial
dimensional specimen), about 25 terms in both n and s are compressive stress are relatively insensitive to the change of
needed to achieve a result in the tensile stress field such that R0. Physically, an increasing R0 implies a decreasing Young’s
the contribution from the next term in the summation will be modulus E of the specimen [see (11) and (12)]. Therefore, the
<3%. Note that n and s are the summation indices of the first stiffer the specimen, the smaller the radius R0, and thus a larger
and second sums of (44)–(47). In general, more terms in n maximum tensile stress will be induced. It appears that the
and less terms in s are needed for h > R, whereas more terms stiffer the rock, the smaller the PLSI. This conclusion appears
in s and less terms in n are needed for h < R; however, a total to be contrary to most experimental observations. In obtaining
of 50 terms for n and s are normally needed to limit the error this result, we have actually assumed that the local tensile
to within 3%. strength for fracture is independent of the stiffness of the ma-
Fig. 2 plots the variations of the normalized stresses srr /s0 terial. However, in reality for most natural rock, strength and
and szz /s0 (where s0 = P/HD) versus the normalized distance stiffness increase simultaneously. Nevertheless, if there were a
z/h along the z-axis for various values of Poisson’s ratio n.
The results are obtained for H/D = 1.1, D = 2R = 50 mm, and
R0 = 0.05R. Note that H/D = 1.1 is the recommended shape
ratio for the axial PLST and that D = 50 mm is the standard
size of rock core for determining the point load strength index
(PLSI) suggested by ISRM (1985). To reflect only the effect
of Poisson’s ratio of the specimen on the stress concentrations,
we have fixed the size of the contact area in Fig. 2. That is,
the effect of contact is neglected in Fig. 2. Following the usual
sign convention of continuum mechanics, tension is plotted as
positive. In contrast to the elastic stress field in the Brazilian
test, Fig. 2 shows that the tensile stress distribution along the
z-axis is not uniform, and a local tensile stress concentration
is developed near z/h = 0.9. As shown in Fig. 2, this maximum
tensile stress increases with the decrease of Poisson’s ratio n. FIG. 3. Normalized Stresses srr /s0 and szz /s0 versus Normal-
Because the failure of the specimen under the axial PLST ized Distance z /h along z-Axis for Various Diameter R 0 of Con-
caused by brittle fracture is believed to be dominant, we spec- tact Area for H/D 5 1.1, D 5 2R 5 50 mm, and n 5 0.25

JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999 / 1353

J. Eng. Mech. 1999.125:1349-1357.

 number of solids having the same local tensile strength, the
softer the rock, the higher the PLSI.
To verify the applicability of our analytic solution to actual
axial PLST, the next section will compare our analytic solution
with our experimental data for the axial PLST on cylinders
made of a type of rock-like plaster.

EXPERIMENTS AND COMPARISON WITH

THEORETICAL PREDICTIONS
Experiments have also been done on rock-like plaster to
verify the applicability of our theoretical solution to the axial
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

PLST. Plaster, instead of natural rocks, was used because the

specimens will be more homogeneous and the results are more
reproducible. Our plaster specimens were made by using the FIG. 4. Stresses srr and szz versus Normalized Distance z /h
along z-Axis for Four Specimens of H/D 5 1.1, D 5 56 mm, and n
following procedures. A mixture of cool distilled water and 5 0.25 under Critical Load P
gypsum with a mass ratio of 3:10 was poured into a cylindrical
mold made of a plastic hollow tube. The mold was filled with
plaster paste in three consecutive phases, and each phase local tensile stress field of 22.78 MPa is needed to initiate the
covers one-third of the length of the mold. When each third unstable crack growth from a defect of typical size in the plas-
of the mold is filled, a tiny wooden hammer was used to tap ter material, in the presence of a high compressive stress in
the mixture 50–70 times to yield a homogeneous sample and the axial direction.
to minimize the air content of the specimen. The procedure Many previous experiments showed that the PLSI (i.e., Is =
has to be completed swiftly because the mixture starts to co- Pp/4HD) depends on both the length and diameter of the spec-
agulate within 8–10 min. If warmer water were used, the set- imens (Broch and Franklin 1972; ISRM 1985). By applying
ting time of the plaster paste will be even shorter. The mold the concept of the local tensile strength (as described above)
was removed after 10–15 min. The curing of specimens was to our analytic solution, prediction on both shape and size
done at room temperature and humidity (under air-condition- effects on the PLSI can be made. In particular, the predicted
ing). The weight of each specimen was recorded daily until a PLSI (or Is) is plotted against the diameter in Fig. 5 for a fixed
steady weight was maintained. Normally, the daily change in length-to-diameter ratio H/D = 1.1 together with our experi-
mass will be <0.1% after about 45 days. To ensure that a mental results. Because of the limited vertical clearance (10–
steady strength had been developed when the specimens were 89 mm) of the loading frame of the PLST apparatus, only
tested, all tests were conducted 60 days after mixing. The den- specimens of diameter from 41 to 67 mm were tested. Fig. 6
sity of the specimens is about 1.75–1.78 g/cm3. For investi- plots the PLSI versus the length-to-diameter ratio for a fixed
gation of the shape and size effects, a total of 18 specimens diameter D = 56 mm. As shown in Figs. 5 and 6, the predic-
were initially made for the axial PLST. Two samples were tions on both shape and size effects on the PLSI agree ex-
made for each of the following combinations of size and tremely well with our experimental results. Because the shape
shape: (1) Specimens of the same shape (H/D = 1.1) with effect was examined experimentally only for specimens of
various diameters of 41, 51, 56, and 67 mm; and (2) specimens D = 56 mm (Fig. 6), Fig. 7 shows our theoretical predications
of the same diameter (56 mm) with various lengths of 34, 44, for the PLSI versus length-to-diameter ratio for different di-
58, 63, and 69 mm. All end surfaces of the specimens were ameters (D = 40, 56, and 67 mm). These results illustrate the
polished to a tolerance of 0.02 mm of flatness, and the max- well-known observations that for a fixed H/D, the larger the
imum departure of the perpendicularity of the end surfaces specimen, the lower the PLSI, and for a fixed diameter D, the
from the axis of the specimens was controlled to within 0.001 more slender the specimen, the weaker the rock (Broch and
rad or 0.05 mm in 50 mm. By following the testing procedure Franklin 1972; Brook 1977, 1980; Forster 1983).
suggested by ASTM for the uniaxial compression test, the av- Note that the size effect demonstrated by Fig. 5 is primarily
erage Young’s modulus, Poisson’s ratio, indirect tensile due to the size of the contact, which is obviously a function
strength by Brazilian test, and compressive strength of the of the diameter of the specimen. Thus, a length scale (i.e., the
specimens were obtained as 17.5 GPa, 0.25, 5.79 MPa, and contact size) has been introduced into the analysis. This is
57.54 MPa, respectively. In addition, the uniaxial stress-strain
curve of our samples show clearly the sign of dilatancy. [For
more discussion on the requirements of rock-like solids, see
Wong and Chau (1998).]
To calibrate the maximum allowable local tensile field of
the plaster under the PLST, two more specimens with H/D =
1.1 and D = 56 mm were cast. The point load P under which
the specimen was broken was recorded for each specimen.
Then, these data were input into our solutions, and the cal-
culated stresses srr and szz were plotted versus the normalized
distance z/h from the center in Fig. 4. The calculated maxi-
mum tensile stresses for these four specimens are 23.06, 23.35,
21.50, and 23.21 MPa with a mean of 22.78 MPa. To predict
the failure load for each sample, we make a major assumption
that if the maximum tensile stress reaches 22.78 MPa, which
will be called local tensile strength hereafter, along the z-axis
in the specimen, the specimen fails. It should be noted that
this local tensile strength is about 4 times larger than the ten- FIG. 5. Comparisons of Theoretical Predictions and Experi-
sile strength obtained from the Brazilian test. From a micro- mental Results for PLSI Is versus Diameter for Fixed Length-to-
scopic point of view, this assumption implies that a critical Diameter Ratio H/D 5 1.1, and n = 0.25

1354 / JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999

J. Eng. Mech. 1999.125:1349-1357.

 similar to that observed in the double-punch test used in con-
crete testing [see Figs. 2 and 3 of Chen and Yuan (1980)].
These fall-off cones are in the shape of a single ridge (only
one line of a sharp ridge) if the specimen breaks into two
pieces along roughly a flat fracture surface, but in the shape
of a triangular pyramid with a circular base if the specimen
breaks into three pieces (of roughly the same size). There is
also evidence of crushing at the inclined surfaces of the fall-
off cones, as the plaster turns to pale yellow comparing it with
the yellow color in the intact regions. The heights of these
fall-off cones (i.e., L b in Fig. 8 and 9) are about 3–5, 5–6,
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

and 6–7 mm for specimens of diameter of 40, 56, and 67 mm,

whereas the diameters of the circular base of the fall-off cones
(i.e., L a in Figs. 8 and 9) are about 3–4 mm for specimens of
diameters 40 and 56 mm and 4–5 mm for specimens of di-
FIG. 6. Comparisons of Theoretical Predictions and Experi- ameter 67 mm. The size of the circular base of the fall-off
mental Results for PLSI Is versus Length-to-Diameter Ratio for cones indicates the size of the contact zone.
Fixed Diameter D 5 56 mm, and n 5 0.25
For the height of the fall-off cones, we speculate that it
indicates the depth of the crushing zone under the contact
points. To illustrate this idea, we propose a very simple model
to estimate the size of the crushing zone. In particular, we
estimate the crushing pressure for our plaster using the grain
crushing model proposed by Zhang et al. (1990). Zhang et al.
(1990) concluded that the porosity and the grain radius are the
two major parameters strongly correlated with the critical pres-
sure for the onset of grain crushing, which is illustrated in Fig.
10 of Zhang et al. (1990). For our plaster, the porosity is es-
timated as 37.4%, which is obtained from the weight differ-
ence between the dried and saturated samples, as suggested by
Zhang et al. (1990). The average radius of the grain of the
plaster, however, is very difficult to be determined because the
grain size cannot be unambiguously defined after the harden-
ing of the plaster paste. However, the grain size is very fine
(e.g., <63 mm, which is the borderline between fine and coarse
grains in soil mechanics). An average radius of the grain of 5
mm is observed by using a Nikon UFX-II polarizing micro-
FIG. 7. Theoretical Predictions for PLSI Is versus Length-to- scope with magnification power of 4003. From Fig. 10 of
Diameter Ratio for Various Diameters for Poisson’s Ratio n 5 Zhang et al. (1990), we found that the crushing pressure is
0.25 about 19 MPa. Theoretically, the average hydrostatic pressure
at any point within the specimen can be estimated by taking
different from the size effect due to the length scale from the the average of the normal stresses (the axial, radial, and tan-
defect within the specimen [e.g., Bazant et al. (1991)]. gential stresses) of our elastic solution. Fig. 10 plots this pre-
To further justify our solution, the failure patterns are ex- dicted hydrostatic stress (negative for pressure) versus the nor-
amined in more detail here. It was observed that all specimens malized distance z/h along the z-axis for various diameters (D
are broken into either two or three prismatic fragments with = 40, 56, and 67 mm). Other parameters used in the calculation
two additional small pieces of fall-off cones under the point are H/D = 1.1, n = 0.25, and E = 17.5 GPa. If the hydrostatic
loads, as shown in Figs. 8 and 9. This failure pattern is very pressure is larger than 19 MPa, grain crushing is expected to

FIG. 8. Sketch Showing Failure Pattern of Solid Cylinder of Plaster Material under Axial PLST if Specimen Breaks into Two Pieces
along Flat Surface

JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999 / 1355

J. Eng. Mech. 1999.125:1349-1357.

Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

FIG. 9. Sketch Showing Failure Pattern of Solid Cylinder of Plaster Material under Axial PLST if Specimen Breaks into Three Pieces

the point load strength index (PLSI), and for a fixed diameter,
the more slender the specimen, the weaker the PLSI. A series
of the axial PLST were also carried out on cylinders of a rock-
like plaster material, and our predictions on both size and
shape effects agree well with our experimental observations.

APPENDIX I. FOURIER-BESSEL EXPANSION OF

CONTACT STRESS
Eq. (8) for szz can be expanded into series of J0(lsr) by
using the following formula [e.g., Watson (1944)]:

O O S D
` `
r
szz = Es J0(lsr) = Es J0 ls (48)
s=1 s=1 R
FIG. 10. Theoretical Predictions for Hydrostatic Stress versus where

E S D
Normalized Distance z /h along z-Axis for Different Diameters R
for Poisson’s Ratio n 5 0.25, Together with Crushing Pressure 2 r
Interpreted from Zhang et al. (1990) Es = 2 2 r szz J0 ls dr (49)
R J 0(ls) 0 R

occur. Interpreting from Fig. 10, the predicted height of the Because szz is nonzero only within the circular contact area,
fall-off cones for specimens of diameter 40, 56, and 67 mm (49) can be evaluated by substituting (14) into (49) as
are 5.5, 5.9, and 7.0 mm, respectively (corresponding to the
E S D
R0
2 r r
size of the region with a hydrostatic pressure larger than the Es = 2 2 2 p0 ÏR 20 2 r 2J0 ls dr (50)
crushing pressure). These values agree qualitatively with our R J 0(ls) 0 R0 R
experimental observations. Although our analysis is more il- Applying the change of variable r = R0 sin u, (50) becomes

E S D
lustrative than definitive, the present analysis does indicate that p/2
the solution is capable of yielding the size of the crushing zone 2p0 R 20 R0
comparable to those observed in experiments. In addition, the Es = 2 2 2 sin u cos2u J0 ls sin u du (51)
R J 0(ls) 0 R
location of the maximum tensile stress field also coincides
with the tip of the fall-off cones. This lends further credence This integration can be done exactly by setting m = 0 and d
that the point of maximum tensile stress beneath the contact = 1/2 into the following formula [formula 6.683 of Gradshteyn
zone is indeed the site of crack initiation. and Ryzhik (1980)]:

E
p/2

CONCLUSIONS Jm(a sin u)(sin u)m11(cos u)2d11 du = 2dG(d 1 1)a2d21Jd1m11(a)

0
In this paper, we have presented a new analytical solution
for the stress concentrations in an elastic, isotropic solid cir- for Re(d) > 21, Re(m) > 21 (52)
cular cylinder of finite length subjected to the axial PLST. Our
The result of this integration gives
analytic solution predicts that a local tensile zone will develop
near z/h = 0.9 along the line joining the two point loads (or
the axis of symmetry). Our numerical results show that the
maximum tensile stress increases with the decrease of Pois-
Es = 2
p0(2pR0)1/2
l3/2 1/2 2
s R J 0(ls)
J3/2 S D
R0
R
ls (53)

son’s ratio n and the size of loading area but increases with Because J3/2(ls R0 /R) can be written in terms of sin(ls R0 /R)
Young’s modulus. The nonuniform distributions of the stresses and cos(ls R0 /R) as follows [e.g., formula 8.464 of Gradshteyn
along the line joining the two point loads bear a close resem- and Ryzhik (1980)]:

S DÎ F S D S DG
blance with those predictions for an isotropic sphere under the
diametral PLST. Numerical results show that, for a fixed R0 2R R R0 R0
J3/2 ls = sin ls 2 cos ls (54)
length-to-diameter ratio, the larger the specimen, the weaker R pR0ls R0ls R R

1356 / JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999

J. Eng. Mech. 1999.125:1349-1357.

 (53) can further be simplified to the following function of sine certain practical systems of load.’’ Philosophical Trans. Royal Soc.,
and cosine London, A198, 147–233.

S D
Forster, I. R. (1983). ‘‘The influence of core sample geometry on the
2p0 R ls R0 ls R0 ls R0 axial point load test.’’ Int. J. Rock Mech. Min. Sci. and Geomechanics
Es = cos 2 sin (55) Abstract, 20, 291–295.
l3s R0 J 20(ls) R R R Gradshteyn, I. S., and Ryzhik, I. M. (1980). Table of integrals, series,
and products, Corrected and Enlarged Ed., Academic, New York.
Thus, finally we arrive at
Hiramatsu, Y., and Oka, Y. (1966). ‘‘Determination of the tensile strength

O S D
`
2p0 R ls R0 ls R0 ls R0 of rock by a compression test of an irregular test piece.’’ Int. J. Rock
szz = cos 2 sin J0(lsr) (56) Mech. Min. Sci., 3, 89–99.
s=1 l3s R0 J 20(ls) R R R International Society for Rock Mechanics (ISRM). (1985). ‘‘Suggested
method for determining point load strength.’’ Int. J. Rock Mech. Min.
which is the result given in (40) of the text. Sci. and Geomechanics Abstract, 22, 53–60.
Downloaded from ascelibrary.org by UNIV OF CONNECTICUT LIBRARIES on 05/26/14. Copyright ASCE. For personal use only; all rights reserved.

Love, A. E. H. (1944). A treatise on the mathematical theory of elasticity,

ACKNOWLEDGMENTS 4th Ed., Dover, New York.
Ogaki, Y., and Nakajima, N. (1983). ‘‘Stress analysis of a circular cylinder
This research was supported by Research Grants Council of the Hong of finite length subjected to loads symmetrical to middle plane and to
Kong Special Administrative Region Government under the Competitive axis of revolution.’’ Sci. Engrg. Rev. of Doshisha Univ., 23(4), 195–
Earmarked Research Grant No. PolyU 70/96E. 205.
Peng, S. S. (1976). ‘‘Stress analysis of cylindrical rock discs subjected to
APPENDIX II. REFERENCES axial double point load.’’ Int. J. Rock Mech. Min. Sci. and Geome-
chanics Abstract, 13, 97–101.
Bazant, Z. P., Kazemi, M. T., Hasegawa, T., and Mazars, J. (1991). ‘‘Size Saito, H. (1952). ‘‘The axially symmetrical deformation of a short cyl-
effect in Brazilian split-cylinder tests: Measurement and fracture anal- inder.’’ Trans. JSME, Tokyo, 18(68), 21–28 (in Japanese with English
ysis.’’ ACI Mat. J., 88(3), 325–332. abstract).
Broch, E., and Franklin, J. A. (1972). ‘‘The point load strength test.’’ Int. Saito, H. (1954). ‘‘The axially symmetrical deformation of a short cyl-
J. Rock Mech. Min. Sci., 9, 669–697. inder.’’ Trans. JSME, Tokyo, 20(91), 185–190 (in Japanese with En-
Brook, N. (1977). ‘‘The use of irregular specimens for rock strength glish abstract).
tests.’’ Int. J. Rock Mech. Min. Sci. and Geomechanics Abstract, 14,
Timoshenko, S. P., and Goodier, J. N. (1982). Theory of elasticity, 3rd
193–202.
Ed., McGraw-Hill, New York.
Brook, N. (1980). ‘‘Size correction for point load testing.’’ Int. J. Rock
Watanabe, S. (1996). ‘‘Elastic analysis of axi-symmetric finite cylinder
Mech. Min. Sci. and Geomechanics Abstract, 17, 231–235.
constrained radial displacement on the loading end.’’ Struct. Engrg./
Chau, K. T. (1997). ‘‘Young’s modulus interpreted from compression tests
Earthquake Engrg., Tokyo, 13(2), 175s–185s.
with end friction.’’ J. Engrg. Mech., ASCE, 123(1), 1–7.
Chau, K. T. (1998a). ‘‘Analytic solutions for diametral point load strength Watanabe, S. (1998). ‘‘Discussion on ‘Young’s modulus interpreted from
tests.’’ J. Engrg. Mech., ASCE, 124(8), 875–883. compression tests with end friction,’ by K. T. Chau.’’ J. Engrg. Mech.,
Chau, K. T. (1998b). ‘‘Closure to ‘Young’s modulus interpreted from ASCE, 124(10), 1170–1171.
compression tests with end friction.’ ’’ J. Engrg. Mech., ASCE, Watson, G. N. (1944). A treatise on the theory of Bessel functions, 2nd
124(10), 1171–1174. Ed., Cambridge University Press, Cambridge, England.
Chau, K. T., and Wei, X. X. (1999). ‘‘A spherically isotropic sphere sub- Wei, X. X., and Chau, K. T. (1998). ‘‘Spherically isotropic spheres subject
ject to diametral point loads.’’ Int. J. Solids Struct., 36(29), 4473–4496. to diametral point load test: Analytic solutions.’’ Int. J. Rock Mech.
Chau, K. T., and Wong, R. H. C. (1996). ‘‘Uniaxial compression strength Min. Sci., 35(4–5), Paper No. 006.
and point load strength of rocks.’’ Int. J. Rock Mech. Min. Sci. and Wijk, G. (1978). ‘‘Some new theoretical aspects of indirect measurements
Geomechanics Abstract, 33, 183–188. of the tensile strength of rocks.’’ Int. J. Rock Mech. Min. Sci. and
Chau, K. T., and Wong, R. H. C. (1998). ‘‘Author’s reply to discussion Geomechanics Abstract, 15, 149–160.
by T. Y. Irfan and L. S. Cheung on ‘Uniaxial compressive strength and Wijk, G. (1980). ‘‘The point load test for the tensile strength of rock.’’
point load strength of rocks,’ by K. T. Chau and R. H. C. Wong.’’ Int. Geotech. Testing J., 3(2), 49–54.
J. Rock Mech. Min. Sci., 35(1), 103–106. Wong, R. H. C., and Chau, K. T. (1998). ‘‘Patterns of coalescence in a
Chen, W. F., and Yuan, R. L. (1980). ‘‘Tensile strength of concrete: Dou- rock-like material containing two parallel inclined frictional cracks un-
ble-punch test.’’ J. Struct. Div., ASCE, 106(8), 1673–1693. der uniaxial compression.’’ Int. J. Rock Mech. Min. Sci., 35(2), 147–
Dougall, J. (1914). ‘‘An analytical theory of the equilibrium of an isotro- 164.
pic elastic rod of circular section.’’ Trans. Royal Soc., Edinburgh, Scot- Zhang, J., Wong, T., and Davis, D. (1990). ‘‘Micromechanics of pressure-
land, 59(Part 4), 895–978. induced grain crushing in porous rocks.’’ J. Geophys. Res., 95(B1),
Filon, L. N. G. (1902). ‘‘On the equilibrium of circular cylinders under 341–352.

JOURNAL OF ENGINEERING MECHANICS / DECEMBER 1999 / 1357

J. Eng. Mech. 1999.125:1349-1357.