Chapter 47

Effects of Strain’s Error on Residual Stresses Calculated by HDM

C. Casavola, G. Pappalettera, C. Pappalettere, and F. Tursi

Abstract The hole-drilling method is an effective and popular semi-destructive technique for residual stress (RS)
measurement. It consists in drilling a very small hole into the specimen. Consequently, RS relaxes in the hole and stresses
in the surrounding region change causing strains also to change. A strain gage rosette, specifically designed and standardized
measures these strains. Using stress–strain relationships, the RS field is calculated from the measured strains. In the case of
not uniform RS, an incremental technique is used in which relieved strains are measured during a series of small hole depth
increments. In contrast to the uniform stress case, not uniform stress calculations are extremely sensitive to errors in the
measured strain: small strain measurement errors can cause significant variations in calculated stresses, particularly for
stresses far from the surface.
This paper presents a detailed analysis of the influence of the strain measurement error on the computed stresses.
Particular emphasis is placed on the analysis of the influence of both the number of total steps and the type of step increment.
Both the integral and power series stress calculation methods are investigated, and their different responses to measurement
errors are described.

Keywords Residual stress • Hole drilling method

47.1 Methodology

The power series method was introduced by Schajer [1] as an approximate, but theoretically acceptable method of
calculating non-uniform stress fields from incremental strain data. It is based upon the assumption that the stress field can
be approximated by terms of the power series, i.e., s0 ðhÞ ¼ 1, s1 ðhÞ ¼ h, s2 ðhÞ ¼ h2 , etc. The corresponding strain
responses on the surface when an hole is drilled into a stress fields is ruled by the coefficients a0 ðhÞ, a1 ðhÞ, a2 ðhÞ and,
b0 ðhÞ, b1 ðhÞ, b2 ðhÞ that can be calculated by finite element analysis. In this way, the measured strains can be divided into
components each one of them corresponding to a term in the power series stress fields. In practical calculation only the first
two terms of the power expansion are considered because the hole drilling method is not reliable for higher order terms. The
maximum depth below the surface is limited to 0.5 rm where rm is the radius of the gage circle. The least-squares analysis
is best done by applying the “normal equations” [2] to each of the transformed strains defined as p(h) ¼ (e3(h) + e1(h))/2;
q(h) ¼ (e3(h)  e1(h))/2; t(h) ¼ (e3(h) + e1(h)  2e2(h))/2. Because the use of the transformed variables decouples
the stress/strain equations, it is possible to consider each transformed stress or strain independently of the others. The
transformed stresses P(h) ¼ (sy(h) + sx(h))/2 are calculated from strains p(h) using:
P 0 P 0    P 0 
P a1 ðhÞ
a0 ðhÞ P a1 ðhÞ
a1 ðhÞ a ðhÞpðhÞ
P E
 0 ¼ P (47.1)
a ðhÞa0 ðhÞ a ðhÞa1 ðhÞ P 1þn a1 ðhÞpðhÞ

C. Casavola (*) • G. Pappalettera • C. Pappalettere • F. Tursi

Dipartimento di Ingegneria Meccanica e Gestionale, Politecnico di Bari, Viale Japigia 182, 70126 Bari, Italy
e-mail: casavola@poliba.it

C.E. Ventura et al. (eds.), Experimental and Applied Mechanics, Volume 4: Proceedings of the 2012 Annual Conference 395
on Experimental and Applied Mechanics, Conference Proceedings of the Society for Experimental Mechanics Series 34,
DOI 10.1007/978-1-4614-4226-4_47, # The Society for Experimental Mechanics, Inc. 2013
396 C. Casavola et al.

where  P and ’P are the first two power series components of the “P” stress field, and S indicates the summation of the
products of the values corresponding to all the hole depths, h, used for the strain measurements. This calculation is repeated
for transformed stresses Q(h) ¼ (sy(h)  sx(h))/2 and T(h) ¼ txy using strains q(h) and t(h) with coefficients b0 ðhÞ and
b1 ðhÞ instead of a0 ðhÞ and a1 ðhÞ, and omitting the factor 1 + v. The Cartesian stress field can be recovered using the
transformed stresses P(h), Q(h) and T(h). An advantage of the Power series method is that the least squares procedure forms
a best fit curve through the measured strain data and this makes the algorithm quite robust. This averaging effect is
particularly effective when strain measurements are made at many hole depth increments. A limitation of the method is that
it is suitable only for smoothly varying stress fields.
The use of finite element calculations as a calibration procedure has also made application of the integral method a
practical possibility. Initial developments in this area were made by Bijak-Zochowski, Niku-Lari et al., Flaman and Manning
[2]. In the integral method, the surface strain relief measured after completing hole depth step j (ej) is supposed to be related
through Eq. 47.1 with RS originally present in the material in all the hole depth steps k with 1  k  j:

 
ð1 þ nÞ X 1 X  sx  sy  1 X 
j j j
sx þ sy
ej ¼ ajk þ bjk cos 2y þ bjk txy k sin 2y (47.2)
E k¼1
2 k E k¼1 2 k E k¼1

where y is the angle of strain gage from the x-axis. The calibration constants ajk and bjk indicate the relieved strains in a hole j
steps deep, due to unit stresses within hole step k. Numerical values of the calibration constants have been determined by finite
element calculations for standard rosette patterns, and are tabulated in Ref. [3] by Schajer and in ASTM E837-08. They are
dimensionless, almost material-independent and hole geometry (depth and diameter) dependent constants. Using the integral
method, stress calculation are effective when few hole depth steps are used [3]. For large number of drilling steps, the
calibration matrices a and b become numerically ill-conditioned: small errors in experimental measurements can cause much
larger errors in calculated residual stresses. Moreover a fundamental physical limitation of the hole-drilling method is due to
the fact that strain are measured on the surface, so the sensitivity of the measurements decreases by increasing the hole depth.
The various mathematical methods use different approaches to minimize the adverse effects of experimental errors. However,
they all have to face the same trade-off between spatial resolution and stress uncertainty. If good spatial resolution of the
variation of residual stresses with depth is required, there will be substantial noise and uncertainty in the individual stress
values. On the other hand, if low sensitivity to experimental errors is required, spatial resolution has to be sacrificed. As with
any other mathematical calculation, the quality of the calculated residual stresses depends directly on the quality of the input
data. The extreme sensitivity to the effects of small experimental errors in strain measurements makes hole-drilling residual
stress calculations particularly dependent on having high quality measured data. Thus, meticulous experimental procedure is
essential, with careful attention given to getting accurate measurements that are as free as possible from noise and other sources
of errors [4]. To reduce this effect, ASTM E837-08 [5] and the H-Drill software [6] adopt the Tikhonov regularization [7, 8].

47.2 Experimental Plan

Several works dealing with the non uniform residual stress measurement by the hole drilling method [9] have emphasized its
high sensitivity to errors in the measured data. The main error sources are related to the measurement of relaxed strains, hole
depth, hole diameter, eccentricity and material properties estimation. In particular, the experimental practice shows that, as is
well known in the literature [10], larger errors in the computed stresses are due to the strain measurement error, since other
errors are generally lower. For this reason, the present study concentrates on the strain measurement error, which is the
dominant error source. In particular we have considered a set of linear deformation (e1L, e2L, e3L), the same set both with
random strain error included between 3 me (e1L3, e2L3, e3L3) and with random strain error included between 15 me
(e1L15, e2L15, e3L15). This last sequence was introduced in order to simulate an effectively different signal from the original
one beyond the measurement error. The problem of optimizing the step distribution was considered by Schajer and Altus [10]
and Stefanescu et al. [11]. In order to provide indications for the optimal step selection, both constant calculation step (20, 10, 4
in a total depth of 1 mm) and incremental calculation step (12, 10, 8, 6, 4 in a total depth of 1 mm) are considered. The errors on
the calculated stresses normalized to the corresponding error on the measured strain are obtained from the equations:

smin  s3 e1

E3 ¼

min
 3
 100

(47.3)
s min e 1L  e1L
47 Effects of Strain’s Error on Residual Stresses Calculated by HDM 397

smin  s15 e1

E15 ¼

min
 15
 100

(47.4)
s min e 1L  e1L

where s is the stresses calculated considering the linear deformation, s3 and s15 are the stresses calculated considering
the linear deformation with random strain error included between 3 and 15 me respectively.

47.3 Results and Discussion

Figures 47.1 and 47.2 show the maximum value of error E3 for minimum residual stress calculated by means of integral
method and power series method of Restan [12] and H-Drill software [6], both for the case of constant and incremental steps.
Experimental data show that the error obtained by elaborating data by the integral Restan software are higher than those
obtained by the integral H-Drill, both in case of constant and incremental step (Fig. 47.1a and Fig. 47.2). This result can be
partially explained in view of the fact that Tikhonov regularization is implemented in the H-Drill software and not by Restan
software. In fact if the regularization procedure is disabled in the H-Drill software errors grow up even if they still remain
smaller at higher number of drilling steps. It also worth noting that the influence the implementation of the Tikhonov

Fig. 47.1 Plot of the max

percentage error E3 versus
the number of steps used for
the calculation of the residual
stress profile in the case of
constant calculation steps.
Results are reported for the
case of the H-Drill calculation
procedure both integral (with
and without Tikhonov
regularization) and power
series and for the case of the
Restan calculation procedure
both integral and power series:
(a) constant steps, (b) constant
steps – detail
398 C. Casavola et al.

Fig. 47.2 Plot of the max percentage error E3 versus the number of steps used for the calculation of the residual stress profile in the case of
incremental calculation steps. Results are reported for the case of the H-Drill calculation procedure both integral (with and without Tikhonov
regularization) and power series and for the case of the Restan calculation procedure both integral and power series

Fig. 47.3 Plot of the percentage error E3 versus the depth of step used for the calculation of the residual stress profile in the case of 20 constant
calculation steps. Results are reported for the case of the H-Drill calculation procedure both integral (with and without Tikhonov regularization)
and power series and for the Restan calculation procedure both integral and power series: (a) 20 constant steps, (b) 20 constant steps – detail

regularization in H-drill method appears to be very strong in the constant step calculation while in the incremental step case a
reduction of few percent can be achieved. In other word it can be assessed that a proper choice of the calculating steps acts
equivalently to a regularization routine. Incremental stepping technique generally reduces error also in the integral Restan
calculation procedure. Figures 47.1 and 47.2 also show that power series approach appears less sensitive to errors on the
strain measurement.
Moreover, it seems that when strain measurements are made at many hole depth increments, errors obtained by the
H-Drill with the power series method and the integral are comparable, while error in the integral method in Restan is
considerably higher than those obtained in the power series approach in Restan. Using Restan software it seems that errors of
integral method are always larger than errors of power series method.
Figures 47.3, 47.4, and 47.5 show the value of E3 for minimum residual stress calculated by means of integral method
and power series method of Restan [12] and H-Drill software [6] for the case of constant steps.
Experimental data show that, independently of method used, the errors obtained by elaborating data grow up if stress
calculation depth increases. This result can be partially explained in view of the fact that the hole drilling method measures
47 Effects of Strain’s Error on Residual Stresses Calculated by HDM 399

Fig. 47.4 Plot of the percentage error E3 versus the depth of step used for the calculation of the residual stress profile in the case of 10 constant
calculation steps. Results are reported for the case of the H-Drill calculation procedure both integral (with and without Tikhonov regularization)
and power series and for the Restan calculation procedure both integral and power series: (a) 10 constant steps, (b) 10 constant steps – detail

Fig. 47.5 Plot of the percentage error E3 versus the depth of step used for the calculation of the residual stress profile in the case of four constant
calculation steps. Results are reported for the case of the H-Drill calculation procedure both integral (with and without Tikhonov regularization)
and power series and for the Restan calculation procedure both integral and power series

the strain on the surface. It also worth noting that by reducing the number of steps, at the same stress calculation depth, the
level of error on the stress calculation reduces.
Figures 47.6, 47.7, and 47.8 show the values of E3 for minimum residual stress calculated by means of integral method
and power series method of Restan [12] and H-Drill software [6] for the case of incremental steps.
Experimental data of integral RESTAN show that the incremental steps reduce the error in the last steps. Since the
maximum error occurs generally in the last steps, the incremental steps reduce the amount of the maximum error.
Experimental data of Integral Hdrill show that if Tikhonov regularization is not performed, the incremental steps reduce
errors (compare integral Hdrill no regol, 10 steps constant and incremental); if Tikhonov regularization is performed, the
incremental step does not change much the magnitude of the errors (compare integral Hdrill, 10 steps constant and
incremental).
Figures 47.9 and 47.10 show the maximum value of E15 for minimum residual stress calculated by means of integral
method and power series method of Restan [12] and H-Drill software [6] both for the case of constant and incremental steps.
400 C. Casavola et al.

Fig. 47.6 Plot of the

percentage error E3 versus
the depth of step used for the
calculation of the residual
stress profile in the case of 12
incremental calculation steps.
Results are reported for the
case of the H-Drill calculation
procedure both integral (with
and without Tikhonov
regularization) and power
series and for the Restan
calculation procedure both
integral and power series

Fig. 47.7 Plot of the

percentage error E3 versus
the depth of step used for the
calculation of the residual
stress profile in the case of 10
incremental calculation steps.
Results are reported for the
case of the H-Drill calculation
procedure both integral (with
and without Tikhonov
regularization) and power
series and for the Restan
calculation procedure both
integral and power series

By reminding that, in this way, the difference between two effective strain inputs is analyzed it is possible to observe that in
the case of constant steps, the integral H-drill approach appears to be more sensitive. If the incremental steps case is analyzed
instead it appears that sensitivity of the integral H-drill is reduced and becomes comparable with sensitivity of the other
methods the other methods. In other words incremental step algorithm seems to act like a filter on the output signal.

47.4 Conclusions

In this paper the influence of the strain measurement error on the computed stresses has been analyzed. Integral and power
series approaches were compared both for RESTAN and H-drill commercial systems. The effect of regularization was also
investigated. Furthermore the effect of using incremental optimized steps of calculation was explored.
47 Effects of Strain’s Error on Residual Stresses Calculated by HDM 401

Fig. 47.8 Plot of the

percentage error E3 versus
the depth of step used for the
calculation of the residual
stress profile in the case of
four incremental calculation
steps. Results are reported for
the case of the H-Drill
calculation procedure both
integral (with and without
Tikhonov regularization) and
power series and for the
Restan calculation procedure
both integral and power series

Fig. 47.9 Plot of the max

percentage error E15 versus
the number of steps used for
the calculation of the residual
stress profile in the case of
constant calculation steps.
Results are reported for the
case of the H-Drill calculation
procedure both integral (with
Tikhonov regularization) and
power series and for the
Restan calculation procedure
with the power series
approach

Fig. 47.10 Plot of the max

percentage error E15 versus
the number of steps used for
the calculation of the residual
stress profile in the case of
incremental calculation steps.
Results are reported for the
case of the H-Drill calculation
procedure both integral (with
and without Tikhonov
regularization) and power
series and for the case of the
Restan calculation procedure
both integral and power series
402 C. Casavola et al.

It was found that integral RESTAN method and integral H-drill without regularization are strongly sensitive to strain
errors. This sensitivity increases by increasing the number of calculation steps. This situation can be improved in the H-drill
system by properly implementing the Tikhonov regularization.
Also the adoption of incremental steps of calculation appears to be effective in reducing the effects of strain errors. In this
case final results appear to be also not influenced by the number of steps.
Furtherly it was observed that errors on strain measurements introduce larger errors at higher drilling depth. This can be
attributed to the fact that strain gage rosette measures the strain at the surface of the drilled hole so that it becomes less
efficient in detecting strain induced by stress relaxation at higher depth. Moreover with regard to the integral method errors
are also introduced by the mathematical instability of solution especially at higher depth.
Finally, it should be observed that care should be paid to the choice of the elaboration parameters in order to guarantee a
good stability with respect to measurement errors and at the same time to avoid losing of sensitivity which could mask
important features in the stress distribution profile.

References

1. Schajer GS (1981) Application of finite element calculations to residual stress measurements. J Eng Mater Technol 103:157–163
2. Schajer GS (1988) Measurement of non-uniform residual stresses using the hole drilling method. Part I – Stress calculation procedures. J Eng
Mater Technol 110:344–349
3. Schajer GS (1988) Measurement of non-uniform residual stresses using the hole drilling method. Part II – Practical application of the integral
method. J Eng Mater Technol 110:344–349
4. Vishay Micro Measurements (2007) Measurement of residual stresses by the hole drilling strain gage method. Tech Note TN-503-6, 2007
5. ASTM (2008) Determining residual stresses by the hole-drilling strain-gage method. Standard Test Method E837-08. American Society for
Testing and Materials, West Conshohocken, PA
6. Schajer GS (2006) Hole-drilling residual stress calculation program Version 3.01, Vishay, User guide, 2505 West Sixth Avenue Vancouver
7. Schajer GS (2007) Hole-drilling residual stress profiling with automated smoothing. J Eng Mater Technol 129:440–445
8. Schajer GS, Prime MB (2006) Use of inverse solutions for residual stress measurements. J Eng Mater Technol 128:375–382
9. Zuccarello B (1999) Optimal calculation steps for the evaluation of residual stress by the incremental hole-drilling method. Exp Mech
39:117–124
10. Schajer GS, Altus E (1996) Stress calculation error analysis for incremental hole-drilling residual stress measurements. J Eng Mater Technol
118:120–126
11. Stefanescu D, Truman CE, Smith DJ, Whitehead PS (2006) Improvements in residual stress measurement by the incremental centre hole
drilling technique. Exp Mech 46:417–427
12. SINT Technology srl (1995) RESTAN – Sistema per la misura dello stress residuo mediante il metodo del foro. Manuale d’uso e manutenzione,
Calenzano, Firenze