(1981) Schajer

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Application of Finite Element

Calculations to Residual Stress


G. S. Schajer
Research Assistant,
Department of Mechanical Engineering,
Measurements
University of California,
Berkeley, Calif. 94720 The use of finite element calculations is assessed as a means of analysis of the strain
Assoc. Mem. ASME relaxation data from a measurement of residual stress by a material removal
method. This application is important because it allows a greater flexibility of
choice for specimen shape, materials, and experimental procedure than would be
possible if only analytic or experimental calibrations are used. Three possible ap-
plications are described using the hole-drilling method as an example, and com-
parisons of calculated results and experimental measurements are presented.

Introduction
A common method for measuring residual stresses involves those around a through hole in an uniformly-stressed infinite
the removal of stressed material and measurement of the plate. The theoretical solution [5] for the radial strain
strain relaxation in the adjacent material, typically using relaxations is
strain gauges. The relationship between the "removed
stresses" and the measured strain relaxation in the remaining er = (A +B cos2a) crmax + (A—B cos2a)a„ (1)
material can be found analytically for simple-shaped
specimens with material removed in a simple geometric way, where
for example, uniform thickness reduction of cylinders or flat \-v / a
A=- (2)
plates. However, for more complex specimens or material 1E\ ( Tr ) '
removal procedures, empirical solutions are usually em-
ployed. l+v
In this study, the finite element method [1] is examined as a
B-
IE (^(TYATY) (3)
means of relating the "removed stresses" to the measured
strains in components of general shape. Finite element ^max
maximum principal stress
calculations have been used for the prediction of residual ^min
minimum principal stress
stresses, notably in welded structures, but only rarely has the a hole radius
technique been applied to measurements of residual stress. r general radius, r > a
The application of finite element calculations to residual E Youngs modulus
stress measurements is described here with reference to the v Poisson's ratio
hole-drilling method [2,3,4]. This example is chosen because a angular coordinate measured anticlockwise from
the method is widely used, it causes a minimum of damage to the maximum principal stress direction.
the specimen, and also because the usual residual stress This relationship can be inverted to give the magnitude and
analysis involves experimentally-determined calibration direction of the two principal stresses in terms of the
constants. measured radial strains. For the 45 deg rectangular rosette in
Fig. 1, the result is
Summary of the Hole Drilling Method
61 +63 ((2e2-61-e3)2+(61-e3)2)1/2
Figure 1 shows a specially-designed rosette gauge for (4)
residual stress measurement by the hole-drilling method. The 4,4 AB
rosette is fixed to the surface of the specimen and readings are / 262-6, - 6 3 \
taken both before and after a hole of specified diameter and 0=-tan- (5)
depth is drilled at the geometric center. Alternatively, a series e, - e 3
of readings can be taken as the hole depth is increased by e2, and e3 are the radial strains measured by the
small increments. When calculating the residual stresses that gauges as numbered in Fig. 1, and /3 is the angle measured
existed at the hole, which are assumed to be uniform with clockwise from gauge 3 to the maximum principal stress
depth from the surface, the assumption is made that the stress direction.
and strain distributions around the hole are similar in kind to It is assumed that equation (1) also applies in the case of a
blind hole rather than a through hole, and for measurement
Contributed by the Materials Division for publication in the JOURNAL OF
ENGINEERING MATERIALS AND TECHNOLOGY. Manuscript received by the
of strain over a finite gauge area and not at a point. However,
Materials Division, December 9. 1980. the constants A and B for a blind hole differ from the values

Journal of Engineering Materials and Technology APRIL1981, Vol. 103/157


Copyright © 1981 by ASME
Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
R2

R1

^ ^
a\

ho le y<h strain gauge

Fig. 2 Strain gauge geometry

are the strains in the direction of the


principal stress <jmax and crmin, respectively. By comparison
with (1), it follows that
k,=E(B+A) (10)
vk2=E(B-A) (11)
3 Strain Gauge
Elements The form of the equations (8) and (9) is motivated by
Gauge
analogy with the generalized Hooke's law. However it is not
Length correct to assume this dependence on Poisson's ratio, because
the two strain components emax and emin are not measured at
Fig. 1 Rosette gauge for residual stress measurements,
Micromeasurements type 062-RE. After Beaney, 1976, [3]
the same point. Also, the theoretical solution for the idealized
case, equation (1) with a — 0, is not of the form of (8) and (9).
Beaney [3], groups together vk2/kl = (B-A)/(B+A) as
given by (2) and (3), and these are usually determined through his second constant in place of k2 alone, and finds that in
measurements with known stress fields. The strain relaxation practice this group has an almost fixed value of 0.3 for a wide
depends on hole depth, and typically A and B are determined class of materials. Equations (10) and (11) with (6) and (7)
from the measured strains at a hole depth equal to the hole indicate that both k] and vk2/k, are functions of c, but
diameter. fortuitously this functional dependence is a weak one. Finite
Some investigators [4] have used the idealized constants A element calculations, to be described in the next section,
and B defined in (2) and (3), or other related constants [6], show that for a hole whose depth is equal to its diameter, k{
and ignore the influence of the limited hole size and finite varies within a 2 percent range and vk2/ki from 0.32 to 0.27
gauge area. If the effect of finite gauge area is included, the for changes in Poisson's ratio in the range 0.25 to 0.35. This
solutions for A and B become weak dependence on Poisson's ratio explains the success of
this procedure in practice.
a 2 (l + c)
A= - 1Y2-Y1] (6) Alternative definitions of material independent constants
fi-area which overcome the theoretical shortcomings of (8) and (9)
a2(\ + v) p d[2(1-p) can be made through consideration of equations (6) and (7).
B= - y ] (Y2-Yi) + sin2
Y2-sin2Y| The following dimensionless constants are suggested
£"«area L l + v
IE
C,=2EB (12)
1 +v
( — J (2sin272+sin472) The definition of C, follows directly form (6), but there is no
similar simple definition for C2. The dimensionless group
(7) given in (12) is the best of several simple expressions which
\R~J ( 2 s i n 2 7 1 + s m 4 7i)J were examined. Finite element calculations show that both C,
where a is the hole radius, "area" is the area of each strain and C2 vary by less than 2 percent for changes in Poisson's
gauge element, and Ru R2, y\ and y2 a r e t n e radii and angles ratio in the range 0.25 to 0.35. The factors of two are included
shown in Fig. 2. Equations (6) and (7) ignore the effect of in (12) so that C, and C2 are associated with the mean biaxial
finite hole depth, but however, the actual constants should and shear stresses, ("max + " m m ) / 2 a n d ((7 max - <7 m i n )/2,
approach these values at large hole depths. For the standard respectively.
strain gauge rosette shown in Fig. 1, the values of A from (6)
are almost exactly the same as from (2) for a wide range of Finite Element Approach
hole radii, but the values of B from (7) are about 20 percent The idealized solution for the constants A and B, equations
less than from (3). (6) and (7), is based on the assumption that the strain
The constants A and B are material dependent, and if distribution around a blind hole is essentially the same as
determined experimentally, a calibration measurement must around a through hole in a thin plate. This idealization can be
be made for each material to be studied. This has led some removed using finite element calculations, so that the in-
authors [2,3,9] to introduce "general" or "material in- fluence of hole depth on the relaxed strain can be considered
dependent" constants kx and k2 such that in detail.
Consider the material in the gauge area before the hole is
h E
(8) drilled. A cross-section view is shown in Fig. 3, where the
E boundary curve is assumed to be sufficiently distant from the
kx k2 hole that the drilling does not significantly influence the
>— < (9)
E stresses at the boundary. The initial stresses at the hole

158/Vol. 103, APRIL 1981 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Hole location. Strain gauge d

Plating i rt=d/4
Substrate

Fig. 3 Cross-section of strain gauge area before the hole is drilled

Fig. 5 Cross-section of the gauge area in Example 1

-IOO

(a) (b) (c)


Fig. 4 Superposition of loadings to find the strain relaxation due to
hole drilling (a) original stresses, (b) stress change due to hole drilling,
(c) final stresses

location are marked and are assumed to be uniform over the


hole cross-section so that there are no shear stresses at the hole
surfaces. Now imagine drilling the hole, and externally
loading the hole surfaces with the same stresses as were
previously existing. Under these conditions, the stress
distribution in the remaining material is unchanged.
Consider Fig. 4, where the material with the loading applied
to the hole surface is shown in part (a). Since the system is
linear elastic, the loading shown in (&) may be superposed,
where equal and opposite stresses to those in (a) are applied at
the hole surface, but with zero stress at the distant boundary.
The sum of these two, shown in (c), is the stress distribution 0.2 0.3 0.4 0.5
which exists after the hole has been drilled. Thus, loading (b) Hole d e p t h / r m
corresponds to the redistribution of stresses due to the drilling
of the hole and the associated strains are those strain Fig. 6 Calculated relaxed strain versus hole depth for a thin elec-
relaxations which are measured by the strain gauges. It also troplating. Hole radius, a = 0.3rm = 0.77mm. For plating:
follows that the measured strain relaxations are dependent thickness t = 0.15 rm = 0.39 mm. E = 193000N/mm2, v = 0.33, stress
= 160N/mm 2 For substrate: E = 69000N/mm 2 , v = 0.33, unstressed.
only on the original residual stresses at the hole boundary. An
analysis of loading case (b) using finite element techniques
gives the relaxed surface displacements, from which the
average strain over the gauge area is found by numerical stress field in the plating is uniform biaxial and axisymmetric
integration. elements with axisymmetric loading were used to compute the
surface radial displacements. Since the substrate material is
relatively massive compared with the plating, it is assumed
Example 1 that the substrate is essentially unstressed, as in Fig. 5. The
result of the calculations is shown in Fig. 6 as a curve of
In a recent paper, Dini, et al. [7] used the hole drilling relaxed strain versus hole depth, normalized with respect to
method to measure the residual stresses in a nickel-cobalt the mean gauge radius /•„,. The strain at large hole depths is
electroplating on an aluminum substrate. They used ex- about -9(Ve, a readily measurable value. In an actual
perimentally-determined values of the constants A and B and residual stress measurement, one would calculate the relaxed
so had to meet the associated specimen size conditions, strain for the given plating thickness and hole size with a unit
notably the requirement that, for a hole depth equal to the stress in the plating and then use this figure as a calibration
hole diameter, the specimen thickness should be at least four factor for the measured strain relaxation.
times the whole diameter. In order to do this, they used a
plating thickness of up to 7.6mm. In this example, the ap-
plication of finite element calculations is examined to see
whether it would be practicable to use a much thinner plating, Example 2
say a thickness of 0.39mm, a quarter of the hole diameter, hole In their paper, Rendler and Vigness [2] did an extensive
diameter. series of calibration measurements for strain gauge rosettes of
Using the given material data and measured stresses, a the configuration shown in Fig. 1 and they summarized their
series of finite element calculations was done for the system results in a table of values of the constants A and B for
shown in Fig. 5, in order to compute the strains which would various hole depths. In this example, the finite element
have been measured had the thinner plating been used. The method is used to calculate the same constants, and the results

Journal of Engineering Materials and Technology APRIL 1981, Vol. 103/159

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


o Rendler and Vigness, 1966
-0.6 — - Finite element

-0.4-

-o.i

-0.2

0.5
Hole d e p t h / r m
Fig. 8 Calculated constant C 1 for a uniform stress field with the strain
gauge rosette for Fig. 1 and a square-ended hole
0.2 0.4 0.6
Hole d e p t h / r m

Fig. 7 Comparison of the finite element calculated constants A and B


with the experimental results of Rendler and Vigness, 1966 [2]

are compared with the experimentally-determined values.


Two sets of calculations are required, one for each of the
constants A and B. The finite elements used here are members
of a class of elements for axi-symmetric solids [8] whose
nodal displacements are of the form
ur(r,z,6) =u„,(r,z) cosnd
ue(r,z,6) = u0„{r,z) sinnd (13)
u, (r,z,6) =uz„ (r,z) cos nd
where ur and ur„, ue and ue„, uz, and uzn are the radial,
circumferential, and axial displacement distributions,
respectively. The corresponding load distribution for each
value of n has the same harmonic form and would typically be
the nth component of a Fourier expansion of a given non-
symmetric loading. 0.5 I.O
In the present example, there are two Fourier components Hole depth/r m
to consider, n = 0 and n = 2. The constant A is found from Fig. 9 Calculated constant C 2 for a uniform stress field with the strain
calculations using n = 0, corresponding to a uniform biaxial gauge rosette of Fig. 1 and a square-ended hole
stress field, by integrating the surface radial displacement
profiles over the strain gauge area. This type of calculation
was also done in Example 1. The constant B is similarly found
from calculations using n = 2, corresponding to a uniform way as for Fig. 7, using the same finite element calculations;
shear stress field, by integrating both the surface radial and changes in the hole diameter are simulated by appropriate
circumferential displacements over the strain gauge area and scaling of the gauge layout dimensions and hole depth. The
taking into account their harmonic angular variation. "theoretical" values from equations (6), (7), and (12), shown
Figure 7 shows a comparison between the calculated and by the dashed lines, are approached assymptotically at large
measured values of A and B for various hole depths. The hole depths. Here, and also in Figs. 6 and 7, the hole depth
agreement for the constant A is within 1 percent and for B has been normalized with respect to the mean gauge radius,
there is a difference of about 5 percent. All calculations were /•„,, and not with the hole diameter as is the usual practice.
done using quadrilateral elements with linear nodal in- When normalized in this way, the curves in each set have a
terpolation and tests using both fine and coarse meshes similar shape and reach their maximum values at the same
showed that the calculated strains are accurate to within 2 nondimensional hole depth. Increasing the hole diameter
percent. The calculated average strain for a gauge aligned increases the magnitude of the curves, but has little effect on
circumferentially is slightly smaller than for a gauge aligned their shape. This shows that the character of relaxed strain
radially, and so the effect of strain gauge transverse sensitivity measurements is dependent on the gauge layout, while only
should typically be less than 0.5 percent. There is a further the sensitivity is dependent on the hole diameter.
uncertainty of up to 1 percent because the effective area of the Any disturbance to the stresses and strains in the strain
strain gauge is not precisely known. gauge area due to the proximity of the specimen boundaries is
Figures 8 and 9 show the variation with hole depth of the characterized by distances in terms of gauge dimensions, such
dimensionless constants C, and C 2 , defined in (12), for as mean gauge radius. Dimensions in terms of hole diameter
different hole diameters. These curves are derived in the same are only appropriate when considering the influence of the

160/Vol. 103, APRIL 1981 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


CONSTANT CI

HOLF HOLE RADIUS h


DEPTH 0,25 0 30 0.35 0.40 0,45 0,50 'dh
0.1 -0,018 -0 026 -0.035 -0.049 -0.059 -0.081
0.2 -0,041 -0 057 -0.077 -0.104 -0.132 -0.172
0.3 -0.059 -0 0B3 -0.111 -0.148 -0.191 -0.233
0.4 -0.070 -0 099 -0.134 -0.174 -0.221 -0.265
0,5 -0.077 -0 108 -0.144 -0.188 -0.234 -0.283
0,4 -0.079 -0 111 -0.148 -0,192 -0,239 -0.288
(a)
0.7 -0.079 -0 111 -0.148 -0.191 -0.238 -0.287
' O.S -0.078 -0 110 -0.146 -0.189 -0.234 -0.283
0.9 -0.076 -0 107 -0.143 -0.185 -0.230 -0.278
1.0 -0.075 -0 105 -0.139 -0.181 -0.226 -0.272

not
CONSTANT C2

HOLE HOLE RATI I US


DEPTH 0.25 0 30 0.35 0.40 0.45 0.50
(d) (e)
• -0.035 -0 052 -0.069 -0.092 -0.120 -0.150
0.2
0.1 -0.080 -0 110 -0.144 -0.187 -0.236 -0.292 Fig. 11 Superposition of loadings due to stresses at different depths
0.3 -0,124 -0 171 -0.221 -0,286 -0.350 -0.428
0.4 -0.158 -0 218 -0,282 -0,357 -0.436 -0.520
0.5 -0.181 -0 250 -0.323 -0.407 -0.490 -0.576

TO
0.6 -0.196 • -0 270 -0.350 -0.437 -0.524 -0.610
0.7 - 0.205 0 282 -0.365 -0.455 -0.544 -0.632
0.8 0.209 -0 2BB -0.373 -0,466 -0,554 -0.642
0.9 -0.211 -0 290 -0,377 -0.467 -0,559 -0,646
1 .0 -0.211 -0 291 -0.378 -0.460 -0,560 -0.648

Fig. 10 Calculated constants C 1 and C 2 for a uniform stress field with


the strain gauge rosette of Fig. 1 and a square-ended hole. Hole (Q) (b) (c) (d)
dimensions are normalized with respect to rm.

boundaries on the stresses and strains at the edge of the hole S^~~ e € €
itself. Thus, specifications regarding recommended hole
depth and minimum width and depth of the specimen should + + +
be given in terms of gauge dimensions rather than hole / •
diameter. The difference is significant if holes of different
diameters are used with rosettes of a given size. In terms of Depth Depth Depth Depth
mean gauge radius, the recommendations of Rendler and (e) (f) (g) (h)
Vigness [2] regarding hole and specimen sizes are Fig. 12 (a), (b), (c), (d) Decomposition of an arbitrary stress profile into
a power series of stress profiles, (e), (f), (g), (/)) Corresponding relaxed
Hole depth = 0.6 r,„ strain versus depth responses.
Minimum boundary distance from hole = 5.0 rm
Minimum specimen thickness = 2.5 rm
curve of the relaxed strain versus depth is measured for a hole
where r,„ = mean gauge radius drilled into a known uniform stress field. A similar curve is
= 2.57mm for MM 062-RE gauge then measured for a hole drilled into the unknown stress field
= 5.15mm for MM 125-RE gauge and the variation of the second stress field is found by taking
The effect of violating these boundary specifications could the ratio of the gradients of the two strain curves point by
be assessed by doing finite element calculations with different point.
mesh areas. The mesh used here was designed to approximate Although this procedure gives satisfactory results for the
"infinte" boundaries. stress distribution close to the surface, the method is unsound
A tabulation of the calculated constants C, and C2 for a because the initial assumption for the strain change with a
range of hole depths and diameters is given in Fig. 10, from small hole depth increment is incorrect. In Fig. 11, case (d)
which the constants A and B for any material can be found by shows the stresses corresponding to the measured strain
using (12). The figures are for a square-ended hole and do not relaxation at hole depth h and case (a) to the strain relaxation
make any allowance for machining stresses. However, the use at depth h + dh, but the difference between the two is not
of a specially-prepared end mill and a drilling guide together case (e), corresponding to the strain relaxation due only to the
with the manufacturers' recommended drilling procedures [4] stresses within the hole depth increment dh. An example of a
should keep these machining stresses to a minimum. Air- false prediction which would be made with the above residual
abrasive machining, such as is described by Beaney [3], does strain assumption is shown in Fig. 6, where the relaxed strain
not induce significant machining stresses, but the results in continues to increase after the plating has been penetrated,
Fig. 10 cannot then be used because the hole produced by this even though the substrate material is assumed to be un-
process is slightly tapered and has a rounded bottom. Beaney stressed. The assumption that the relaxed strain change is
made some finite element calculations using axisymmetric proportional to the stress within the hole depth increment,
elements only, in order to examine the influence of hole shape would predict no increase.
on his calibration constants, but his final results are based on The calculation of the stress distribution which gives rise to
experimental measurements. a given strain response may not proceed directly, as with the
inverse problem, Fig. 4. In this example, a power series
Example 3 method is described whereby an approximate solution can be
The hole-drilling method has been adapted by previous found using measured relaxed strain data.
authors [9,10] so that an assessment of the stress variation An arbitrary stress field can be decomposed into an equal
with depth can be made. It is assumed that the measured biaxial and a shear stress component. Consider first the
change in strain for a small hole depth increment is biaxial component, where the variation with depth from the
proportional to the stress at that depth. Firstly, a calibration surface, h, is ab(h). This unknown stress distribution is

Journal of Engineering Materials and Technology APRIL 1981, Vol. 103/161

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


assumed to be expandable into a power series with un- CONSTANT CI

determined coefficients £>,; i = 0,1,2, . . . HOLE HOLE RADIUS


OEF'TH 0.25 0.30 0.35 0.40 0.45 0.50
ab(h) = b0+blh + b2h2 + . . . (14)
0. 1 -0.0007 -0.0011 -0.0015 -0.0020 -0,0024 -0.0031
as shown schematically in Fig. 12(a), (b), (c), and (d). The 0.2 -0.0036 -0.0050 -0.0065 -0.0087 -0.0108 -0.0140
0.3 -0.0071 -0.0098 -0.0130 -0.0170 -0.0205 -0.0245
measured strain response eb(h), Fig. 12(e), corresponding to 0. 4 -0.0102 -0.0140 -0.0183 -0.0233 -0.0279 -0.0321
ah(h) is 0.5 -0.0123 -0.0169 -0.0216 -0.0270 -0.0318 -0.0361
0.6 -0.0134 -0.0178 -0.0229 -0.0279 -0.0325 -0.0367
e„(h) =b0\0(h) +&,X,(A) +h\ W + (15) 0.7 -0.0131 -0.0173 -0.0221 • 0,0268 -0,0311 -0.0346
0.8 -0.0120 -0.0158 -0.0201 -0.0243 -0,0279 -0.0309
where A, (h) are the strain responses to unit power series stress 0.9 -0.0106 -0.0140 -0.0175 -0.0209 -0.0241 -0.0266
1.0 -0.0089 -0.0117 -0.0145 -0.0173 -0.0197 -0.0220
fields h'. These responses are shown schematically in Fig.
12(/), (g), and (h) and can be calculated by the finite element
procedure previously described. CONSTANT C2
The coefficients b, in (15) can be determined by a least HOLE HOLE RADIUS
squares minimisation of error [11] between the measured HEF'TH 0.25 0.30 0.35 0.40 0.45 0.50
ti,(h) and (15). The original stress distribution can then be O.l -0.0016 -0.0025 -0.0035 -0.0045 -0,0056 -0.0066
estimated from (14), assuming that the calculated b, also 0.2 -0.0070 -0.0097 -0.0128 -0.0166 -0.0202 -0.0250
approximately minimize the error between the actual ah(h) 0,3 -0.0159 •0.0218 -0.0279 -0.0355 -0.0415 -0.0479
0.4 -0.0256 -0.0348 -0.0446 -0.0550 -0.0638 -0.0718
and (14). 0.5 -0.0344 -0.0466 -0.0595 -0.07.17 -0.0850 -0.0950
0.6 -0.0417 -0.0558 -0.0713 -0.0860 -0.1000 -0,1106
The same calculation can be done for the shear stress 0.7 -0.0465 •0.0625 -0.0795 -0.0956 -0.1093 -0.1207
component as(h), given that the unit response functions O.B -0.0190 -0.0666 -0.0847 -0.1010 -0.1153 -0.1269
0.9 0.0504 -0.0683 -0.0873 -0.1037 -0.1187 -0.1305
\,{h) are replaced by the equivalent functions K,(/?) 1.0 -0.0506 -0.0686 -0.0B81 -0.1045 -0.1197 -0.1317
calculated for a shear stress field.
Fig. 13 Calculated constants C 1 and C 2 for a linear stress field with
In practice, an assessment of a nonuniform residual stress the strain gauge rosette of Fig. 1 and a square-ended hole. Hole
field would proceed as follows. Firstly, a set of strain readings dimensions are normalized with respect to rm.
is taken from each of the gauges of a rosette such as Fig. 1, for
a series of hole depth increments. These sets of readings are
denoted e, (h), e2 (h), and e3 (h), where the subscripts refer The responses \x (h) and K, (h) are similarly found from the
to the gauge numbers shown in Fig. 1. For this 45 deg rec- constants C, and C2 for a linear stress field given in Fig. 13.
tangular rosette, the biaxial and shear components of strain Practical tests of the power series procedure using
are published strain data were not done here because the depths of
the specimens used in the experimental work were ap-
e„(h)=1-[e](h) + ei(h)] (16) proximately equal to the gauge mean radius, rm. The results in
Figs. 10 and 13 are for specimens which are deep compared to
gauge dimensions, depth > 2.5 /•„,.
The limited precision achieved in the trial calculations is to
be expected from consideration of St. Venant's principle. The
+ (e,(/0+ e, (h))2V (17) magnitude of the relaxed strains depend mainly on the stresses
within the first 0.3 /•„, depth of material. However, the
where the negative root is taken in (17) because the constants associated strain relaxations are measured at a distance 0.2 /•„,
B and C2 < 0. The directions of the principal stress axes can to 0.8 /•„, from the hole boundary. Thus, variations of the
be determined from (5) and it is assumed that these directions stress field in the subsurface layer about a given mean value
do not change significantly with depth. Given eh(h) and will only slightly affect the measured strains. For example,
ts(h), the two stress profiles ah(h) and a,(h) are estimated consider two cases of an equal biaxial stress field in aluminum
by the power series procedure, from which the two principal with a hole of radius 0.4 r,„ and depth up to 0.6 r,„. In one
stress profiles <7max (h) and <jmin (h) are found using case the stress varies linearly from 100 N/mm 2 at the surface
to zero at depth 0.6 /•„,, and in the second the stress is constant
°™Ah),°m\n{h)=ah {h)±as(h) (18) at 75 N/mm 2 . The measured relaxed strain vs. depth curves
Numerical trials of this procedure were made for a test case would have the same maximum value of -280^e and would
of MM 125-RE gauge on aluminum with a hole of radius a = differ by less than 20/ze at any depth.
0.4 r,„ = 2.06mm. The exact strain responses eh(h) and e, (h) Thus, the modest precision achieved in the trial calculations
were calculated for a sequence of six equal hole depth in- is not a function of the analysis method. On the contrary, the
crements up to A = 0.6 r ,„ for several specified stress profiles procedure works fairly well given that the strain gauge layout
ff
maxC0 and ami„(h). These strain responses were then is not well adapted to the assessment of nonuniform stress
corrupted by random amounts to simulate measurement fields. An improved rosette gauge could be made by reducing
errors, assumed to be within ±3fie for the strain readings and the gauge area radially and enlarging it circumferentially, so
±0.02mm for the depth measurements. The associated stress that a greater proportion of the active area is close to the hole
profiles were then calculated by the power series procedure boundary. Circumferential gauge orientation would be
and were compared with the original profiles. It was found preferable under these conditions.
that the functions \(h) and K,(A) are not sufficiently well-
conditioned to allow more than two terms to be used reliably Conclusion
in the least squares analysis. For this case, the errors in the Finite element calculations can successfully be applied in
calculated stress field are within ±3N/mm 2 for the uniform association with measurements of residual stress. This ap-
stress component and ±25N/mm 2 per /•„, depth for the linear plication is important because it allows a much greater
stress component. flexibility of choice of specimen shape, materials and ex-
perimental procedure than would be possible if only analytic
The unit responses XQ(/!) and K0(h) are calculated from
solutions are used. Experimental calibration for complex
the constants Cx and C2 given in Fig. 10, using the relations
geometry specimens has previously been employed, but this is
d + ")C, C, time-consuming and may have to be repeated if different
\o(h) = KQ(h) = (19)
materials are to be examined.

162/Vol. 103, APRIL 1981 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


As an example, the application of finite element References
calculations to the hole drilling method was studied, and the
1 Zienkiewicz, O.C., The Finite Element Method, 3rd, ed., McGraw-Hill,
calculated calibration curves for uniform stress fields agreed New York, 1977.
to within 5 percent of the experimentally-determined values. 2 Rendler, N.J., and Vigness, I., "Hole-Drilling Strain-Gage Method of
A frequently-used analytical procedure for nonuniform stress Measuring Residual Stresses," Experimental Mechanics, Vol. 6, No. 12, Dec.
fields was found not to be satisfactory and an alternative 1966, pp.577-586.
3 Beaney, E.M., "Accurate Measurement of Residual Stress on any Steel
approach was proposed. However, the standard rosette using the Centre Hole Method," Strain, Vol. 12, No. 3, July 1976, pp. 99-106.
geometry is not well adapted to such assessments and only 4 Anon, "Measurement of Residual Stresses by the Blind-Hole-Drilling
modest precision was found possible. Method," Technical Data Bulletin T-403. Photoelastic Inc., Malvern, Pa.
5 Muskhelishvili, N.I., Some Basic Problems in the Mathematical Theory
of Elasticity, Noordhoff, Groningen, Holland, 1953, pp. 202-204.
Acknowledgments 6 Nawwar, A.M., McLachlan, K., and Shewchuk, J., "A Modified Hole-
drilling Technique for Determining Residual Stresses in Thin Plates," Ex-
The author sincerely thanks Prof. R.L. Taylor of the perimental Mechanics, Vol. 16, June 1976, pp. 226-232.
University of California, Berkeley, for his advice and 7 Dini, J.W., Benedetti, G.A., and Johnson, H.R., "Residual Stresses in
assistance with the finite element calculations and also Prof. Thick Electrodeposits of a Nickel-Cobalt Alloy," Experimental Mechanics,
CD. Mote, Jr. and Prof. I. Finnie for their support and Vol. 16, No. 2, Feb. 1976, pp. 56-60.
8 Wilson, E.L., "Structural Analysis of Axisymmetric Solids," AIAA
interest in this study. He is also grateful to the University of Journal, Vol. 3, No. 12, Dec. 1965, pp. 2269-2274.
California Forest Products Laboratory, the California Cedar 9 Kelsey, R.A., "Measuring Non-uniform Residual Stresses by the Hole
Products Company, California Saw, Knife and Grinding, Drilling Method," Proc. S.E.S.A., Vol. 14, No. 1, 1956, pp. 181-194.
Inc., the Hudson Lumber Co., the Potlatch Corp., the 10 Bathgate, R.G., "Measurement of Non-Uniform Bi-Axial Residual
Stresses by the Hole Drilling Method," Strain, Vol. 4, No. 2, 1968, pp. 20-29.
Simpson Timber Co., Sun Studs, Inc. and the Weyerhaeuser 11 Dahlquist, G., Bjork, A., and Anderson, N., Numerical Methods,
Company for their financial support. Prentice-Hall, Englewood Cliffs, N. J. 1974, pp. 92-95.

Change of Address Form for Journal of Engineering Materials and Technology

Present Address—Affix Label or Copy Information from Label


If you are planning
To Move, Please
Notify The
ASME-Order Dep't
345 East 47th St.
N.Y..N.Y. 10017
Don't Wait! IDrint New Address Below
Don't Miss An Issue!
Allow Ample Time to Name
Effect Change. Atten tinn
Addre:ss
City State or Country Zip

Journal of Engineering Materials and Technology APRIL 1981, Vol. 103/163

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like