Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 10 (2013) 54 – 60

12th CIRP Conference on Computer Aided Tolerancing

Inspection strategies and multiple geometric tolerances

Giovanni Moronia, Stefano Petròa*
a
Department of Mechanical Engineering, Politecnico di Milano, Via La Masa 1, 20156, Milan, Italy

Abstract

Recent years have seen an increase in the adoption of geometric tolerances. It is often possible to find several geometric tolerances
defined on a single part. However, this poses inspection issues: the values of the geometric error may be interrelated; therefore, the
presence of multiple tolerances should be considered in inspection design. In this work, a methodology is proposed for planning
CMM sampling strategies for multiple tolerances based on the minimization of inspection costs. A model for inspection costs is
proposed, which takes into account the influence of the inspection strategy on measurement and inspection errors costs, both
directly and through its impact on measurement uncertainty. The cost is then minimized by means of a suitable optimization
algorithm, thus defining an optimal sampling strategy. The approach can be adopted both to optimize generic, uniform, sampling
strategies, and to generate manufacturing specific strategies, which consider the manufacturing signature, i.e., the part shape
deviation from design nominal inherent to a specific manufacturing process. The latter kind of strategies is shown to be the most
effective to minimize costs. A case study which illustrates the methodology is presented.

©
© 2013
2012 The
The Authors.
Authors. Published
Publishedby
byElsevier
ElsevierB.V.
B.V.Open accessand/or
Selection under CC BY-NC-NDunder
peer-review license.
responsibility of Professor Xiangqian Jiang.
Selection and peer-review under responsibility of Professor Xiangqian (Jane) Jiang
Keywords: inspection; uncertainty; geometric tolerances; coordinate measurement; sampling strategy

measurements tends to be expensive, so “measurement

1. Introductiona
Measurement strategies for product inspection and
manufacturing process monitoring are strictly
interrelated. In fact, any quality control system is based
on measurements performed on pre-selected key quality
characteristics. Perhaps the most apparent aspect of this
link is conformance testing, which should guarantee
products functionality. However, any measurement
comes with a cost. Moreover, tests may generate
additional costs due to two kinds of errors [1, 2]: type A
errors or “False Failures” (i.e. stating as non-conforming
a conforming part) and type B errors or “Missed Faults”
(i.e. stating as conforming a non-conforming part). The
appearance of these errors is usually proportional to
uncertainty, so “error cost” increases as uncertainty
increases. Unfortunately, conducting accurate
a
* Corresponding author. Tel.: +39 02 2399 8530; fax:+39 02 2399 8585.
E-mail address: Stefano.petro@polimi.it.
cost” is usually inversely proportional to uncertainty. In
order to optimize the overall “inspection cost”, which
sums both costs, uncertainty should be carefully chosen.
Geometric tolerances must often be checked to
guarantee functionality of mechanical parts. A geometric
tolerance states how much a real part can differ from its
ideal geometry defined at design level. The process of
verifying geometric tolerances involves products
measurements, so measurement uncertainty has to be
evaluated.
An often adopted measurement system for geometric
error evaluation is a “Coordinate Measuring Machine”
(CMM), which is a very flexible instrument, being able
to sample discrete points on a large class of surfaces.
However, this flexibility makes evaluation of
measurement uncertainty difficult: different
measurement tasks are characterized by different
uncertainties; hence, a “task specific uncertainty” should
be considered. Wilhelm et al. [3] have identified several
sources of CMMs measurement uncertainty such as:
hardware, workpiece geometry, sampling strategy,

2212-8271 © 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of Professor Xiangqian (Jane) Jiang
doi:10.1016/j.procir.2013.08.012
 Giovanni Moroni and Stefano Petrò / Procedia CIRP 10 (2013) 54 – 60
fitting algorithms, and extrinsic uncertainty sources. In
particular several authors have pointed out that the
sampling strategy can significantly affect uncertainty [4,
5], in particular when sample size is small, which may
be a typical situation if inspection cost has to be
optimized. Because sampling strategy is usually
determined by the operator, it is the main leverage to
control uncertainty.
Moreover, measurement uncertainty tends to be
closely related to the actual part geometric error [3, 5, 6].
If geometric tolerance definitions given by the ISO 1101
[7] standard are followed, only those “anomalous” zones
of the geometric feature which deviate the most from the
nominal geometry define the geometric error. There is a
strong interrelation between the sampling strategy and
the ability to detect anomalous zones caused by
geometric error patterns of the part, and this directly
affects measurement uncertainty. However, it has also
been observed that anomalous zones of the part
profile/surface tend to be the same throughout the
production [8]. It may therefore be stated that the part
presents a process inherent “manufacturing signature”.
In recent years several studies have suggested that the
interaction between sampling strategy and
manufacturing signature could be analyzed in order to
generate very effective sampling strategies (e.g.
Summerhayes et al. [9], or Colosimo et al.[10-12],
Moroni and Petrò [13]). However, these discussions
consider only the case of form tolerances, which involve
a single geometric feature. More recently, the authors of
the present work have considered the case of tolerances
with datum [14, 15]. However, the most general
situation still lacks, i.e. the situation in which more than
one tolerance is defined on a single part.
1.1. Paper aims
This work aims to propose preliminary results to fill
the gap in the field of inspection planning when more
than single geometric tolerance is involved. The
methodology implicitly considers the interrelation
between the geometric error estimates due to the fact
that every geometric feature is measured only once,
regardless of the number of tolerances in which it is
involved. The method is based on a cost function
depending on the sampling strategy. It has to be pointed
out that several manufacturing cost models have been
proposed in the field of tolerancing [16-20], but usually
these models lack in accuracy when evaluating
inspection cost, which are supposed independent of any
inspection planning. The proposed cost function
calculates inspection cost as sum of measurement cost
and errors cost. The application of the methodology will
require an uncertainty evaluation, which will be
performed by means of an experimental procedure based
55
on the ISO/TS 15530-3 international standard [21]. The
effectiveness of manufacturing signature based sampling
strategies will also be shown. An industrial case study
will be proposed to prove methodology effectiveness.
2. Proposed methodology
The proposed methodology is based on an inspection
cost model whose applicability has already been
discussed in the case of early cost estimation [22].
Mathematic detail is now proposed.
In the proposed approach for inspection cost
evaluation, it is supposed that the manufacturing process
has been defined and that at least the process output
characteristics (capability) and costs are known.
Any function for the inspection cost CI evaluation
must consider several cost sources. In general,
contributors to CI can be grouped into measurement cost
CM (cost arising directly from performing a
measurement) and inspection error cost CE (cost derived
from inspection errors of type A and B). These cost
sources are not independent – usually an increase of the
measurement cost corresponds to a decrease in the
inspection error cost. A tradeoff between them is
required.
In the following, the case in which N identical parts
must be inspected will be considered; for every part, T
geometric tolerances are defined, and S geometric
features are involved.
2.1. Measurement cost CM
The measurement cost CM is the sum of two costs: a
cost CMM that depends on the chosen measurement
strategy and a cost CMF that is independent of it.
CMF depends essentially on the measuring instrument.
It originates from the cost of acquiring the fixtures
required for the measurement, planning the measurement
strategy and placing every part on the measuring stage
for inspection. In general, these costs can be considered
fixed, given the measuring system.
Particular care should be taken with strategy
planning. The cost of planning the measurement, which
does not depend on how many parts have to be
measured, depends on how the measurement is planned.
There are several ways to plan a sampling strategy [4]
that influence the link between the sample size and
measurement uncertainty. In particular, classical
strategies (which are not based on any knowledge of the
manufacturing process) generally are not expensive but
are also not so accurate. Adaptive strategies are more
accurate but are difficult to apply with current industrial
measuring systems. Finally, manufacturing-based
strategies, which are based on an a priori knowledge of
56 Giovanni Moroni and Stefano Petrò / Procedia CIRP 10 (2013) 54 – 60
the process, can yield lower uncertainty but require a
manufacturing analysis, which comes with a cost.
The set-up cost can be simply evaluated as hourly
cost for set up cs multiplied by N and the time ts required
for a set-up.
In general, CMM is proportional to the time required to
perform the measurement. To evaluate this time, two
different categories of instruments should be considered,
i.e., discrete measuring instruments and continuous
measuring instruments.
Discrete measuring instruments are instruments
whose probing sensors are activated to measure the
feature a discrete number of times, and each activation
requires a fixed time to sample one or more points. This
includes, for example, CMMs equipped with touch
trigger probes, structured light scanners, image probing
systems and white light interferometers. In this case, the
measurement cost depends essentially on the number of
activations of the sensor, that is,
S S
CMM Ncmtm Ncmta naj Nca naj
j 1 j 1
where cm is the hourly cost for the use of the
measuring system (direct costs and, if applicable,
overheads), ta is the average time required by an
activation of the sensor, naj is the number of activations
to measure the jth feature, and ca is the cost of a single
activation.
In contrast, in a continuous measuring instrument,
measurements are taken with a given frequency while
the sensor moves over the measured feature. Most
continuous measuring instruments scan lines on the
surface (the only notable exception are laser stripe
scanners), so line scanning instruments will be
considered. Examples of continuous measuring
instruments include CMMs equipped with measuring
probes, profilometers, atomic force microscopes and
laser interferometry-based scanners. The CMM for a
scanning instrument is evaluated as:
S lj S
j n pj
CMM Ncmtm Ncm Ncm
j 1 vj j 1 vj
where lj is the length of the scan for the jth feature, vj
is the speed at which the scan is performed, j is the
spatial sampling period, and npj is the number of points
scanned on the jth feature. This formulation is identical
to the one proposed in Eq. (1) if the substitution ta= j/vj
is made. However, vj has a significant influence on the
uncertainty, and, therefore, the formulation in Eq. (2) is
preferable for continuous measuring instruments.
2.2. Inspection Error cost, CE
The evaluation of CE is more complicated because it
involves the probability that type A and/or B errors
happen (it is the sum of the costs related to those
possibilities). Define fAi (fBi ) as the event in which a type
A (B) error is committed when checking tolerance i. In
the following, it will be assumed that the conformance
test suggested in the ISO 14253-1 standard [23] applies
(in case of a non-conformance test the solution would be
similar). This standard requires that, to label a part as
conforming, the measured value of the geometric error
must fall into the conformance zone (i.e., between LSLi
and USLi, respectively, the lower and the upper
specification limits for tolerance i) reduced by the
expanded uncertainty Ui. In the case in which only an
upper tolerance limit is present, as in the case of
geometric tolerances, the probability that one of these
errors occurs is
P f Ai P yi USLi U i | xi USLi (3)
(1)
P f Bi P yi USLi U i | xi USLi (4)
where yi is the measurement result for the geometric
error i and xi is the real value of the geometric error i.
Both yi and xi are random variables, whose statistical
distributions depend on the process and measuring
instrument. Of course, Ui is affected by the measurement
procedure and instrument, thus justifying the search for a
tradeoff between CM and CE. Moreover, Ui is affected by
what is actually measured, such that in coordinate
metrology, one speaks of “task specific uncertainty” [3].
Unfortunately, these probabilities are only marginal
probabilities. In general, the various error probabilities
will be strongly correlated, in particular, when more
tolerances involve a single surface (e.g., when a flatness
tolerance is defined for a plane and that same plane is a
datum feature for another tolerance, or more orientation
and position tolerances share a single datum system).
The overall cost CEB of type B errors can be
expressed as:
(2)
T
CEB NcEB P f Bi (5)
i 1
where cEB is the costT of declaring a non-conforming
part conforming. P f Bi represents the probability
that a type B error is icommitted
1
while verifying any of
the T tolerances.
The evaluation of the cost related to type A errors CEA
is similar. In general, a non-conforming part can be
either discarded or subjected to deeper inspection to
 Giovanni Moroni and Stefano Petrò / Procedia CIRP 10 (2013) 54 – 60
evaluate if it is convenient to rework it. In either case,
the expression for the evaluation of CEA is
T
CEA NcEA P f Ai
i 1
where cEA is the cost for declaring non-conforming a
conforming part, which can be either the value of the
part, if a non-conforming part is always discarded, or the
cost for deeply inspecting it, if reworking is considered
as possible.
Even if computing these probabilities is complicated,
in most production processes, only a few characteristics
are critical. If those are cases in which the six sigma
condition is met, and if the uncertainty is not too large,
the probabilities in Eqs. (3) and (4) will be small.
Therefore, if the critical tolerances can be identified
[24], Eqs. (5) and (6) can be reasonably solved.
In general, P( fBi) is small if a conformance test is
performed, and vice versa, that is, P( fAi) is small in the
case of a non-conformance test [5], which can lead to a
simplified expression of these probabilities:
USL1 U1 x1 USL1
P f Ai P USL2 U 2 x2 USL2
USLT UT xT USLT
P f Bi 0 where k is the expansion factor. The ISO/TS 15530-3
when the conformance test is performed, and
USL1 x1 USL1 U1
P f Bi P USL2 x2 USL2 U 2
USLT xT USLT UT
P f Ai 0 manufacturing
when the non-conformance test is performed. These
probabilities may be interpreted as the fraction of
wrongly rejected (conformance test) or accepted (non-
conformance test) parts (Fig. 1). This formulation is
easier to manage in general.
2.3. Uncertainty evaluation
To evaluate errors cost as described in the present
paragraph, an evaluation of the measurement uncertainty
is required. The ISO/TS 15530-3 [21] technical
specification proposes a procedure to evaluate
measurement uncertainty based on raw data obtained
from repeated measurements of a calibrated artifact.
From raw data, some terms are estimated, such as: ucal
(uncertainty contribution due to calibration uncertainty),
up (uncertainty contribution due to measurement
57
(6)
Fig. 1: Rejected fraction of conforming produced parts in the presence
of only an USL.
procedure), uW (uncertainty contribution due to
variability of the manufacturing process), and b
(measurement bias). Then these terms are combined to
yield the expanded uncertainty U (see the GUM [25] and
the VIM [26]):
U 2
k ucal up2 uW2 b (9)
(7)
specification allows one to compensate for the b term,
thus reducing its influence; in the following discussion it
is assumed not to be compensated.
Mathematical expressions found in the technical
specification allow for uncertainty evaluation if only a
(8) single calibrated artifact is adopted in the procedure.
However, a single artifact is not sufficient to describe the
process signature variability.
Furthermore, if the sampling pattern optimization
algorithm (described in §2.4) is applied to a single
artifact, probably the resulting strategy will be very
specific for the particular artifact, thus generating a
strategy which completely lacks robustness. Therefore, a
modification of the standard is proposed, so that more
than one calibrated artifact may be used. In particular,
the b term (the average bias) should be evaluated as
m rm
yi , j xcal, j
j 1 i 1
b (10)
mrm
In Eq. (10) m is the number of calibrated artifacts
adopted, rm is the number of measurement repetitions for
each artifact, yij is the measurement result (estimated
geometric error) for the ith measurement repetition of the
jth artifact, and xcal,j is the reference value for the jth
58
artifact (calibrated geometric error). It is supposed that
each calibrated workpiece is measured the same number
of times; to be as similar to the ISO/TS 15330-3
specification as possible, it is suggested rm 10.
Then, to estimate up, a pooled standard deviation
could be used:
m rm
2
yi , j yj
j 1 i 1
up yj
m rm 1
Substituting Eqs. (10) and (11) in Eq. (9) an
evaluation of U which takes into account more than one
calibrated artifact, and then the interaction between the
sampling strategy and the manufacturing signature, is
obtained.
A final note on uW: ISO 15530-3 introduces this
uncertainty contribution to take into account the
“variability of the production”, that is, part to part
differences in local form deviations or thermal
expansion coefficient; however, if more than one
calibrated workpiece is measured, then the term up
should contain this uncertainty contribution, and then uW
= 0, as the international standard itself suggests.
2.4. Strategy optimization
Having identified a raw data based methodology to
evaluate the uncertainty, the next step is the choice a
sampling strategy that minimizes the inspection cost.
The minimization problem may be formulated as
follows:
min CI min CM
h1 , h2 , , hn , n h1 , h2 , , hn , n
s.t.
hi s1 , s2 , , sS i 1,2, , n
where hi denotes the location of the ith sampling point
and sj is the surface of the jth geometric feature to
inspect; please note that the sample size n is also an
optimization parameter.
The solution of the problem is not straightforward.
Suppose that rm dense measurements of m calibrated
parts have been obtained, and that the sampling strategy
is the same for every measurement. To solve the
minimization problem, the sampling points
corresponding to any sampling strategy may be extracted
from these clouds of points. The extracted subsets of
points will be considered in the measurement uncertainty
estimation. If the sampling pattern is effective, i.e., it is
able to detect regions of the feature that deviate the most
from the nominal geometry, then the uncertainty will be
low. The identification of an optimal pattern can be seen
Giovanni Moroni and Stefano Petrò / Procedia CIRP 10 (2013) 54 – 60
as an optimization problem where at most any different
alternative pattern is compared. However, due to the
combinatorial nature of the problem, it is not possible to
consider every strategy. A suitable combinatorial
algorithm (e.g. simulated annealing or genetic) will be
applied.
Please note that, even if the optimization of the
strategy is based on the presence of a manufacturing
signature, explicit knowledge of the signature is not
rm
yi , j
(11) required. Therefore, signature based sampling strategy
rm
i 1
planning may be time consuming (to perform calibrate
artefacts measurements) but easy, because no effort in
manufacturing signature modeling is required.
3. Case study
A case study has been analyzed involving the two
straightness tolerances defined in point (c) and the
parallelism tolerance defined in point (e) of Table 3 in
the ISO 10791-7 standard [27]. Therefore, three
tolerances are considered in this example..
Ten parts were milled and a 0.045 mm parallelism
and two 0.025 mm straightness tolerances were defined.
The manufacturing cost was evaluated in 40 €, which
can be considered as type I error cost cEA, supposed that
a non-conforming part that is discarded . Then tolerances
and datum features were sampled by means of a CMM
adopting a uniformly spaced sampling strategy, with a
point density of 1 point/mm2; a total of 3720 points were
sampled on each part (1395 on the datum feature, and
2325 on the toleranced feature). Measurement was
repeated ten times for each part. Finally, every part was
calibrated with a standard calibration uncertainty ucal =
0.001 mm. From calibrated geometric errors, it was
CE ascertained that a multinormal statistical distribution
characterizes the combined distribution of the two
(12)
straightness and the parallelism tolerance considered.
The average surface of the ten toleranced surfaces for
parallelism is
Fig. 2: Case study manufacturing signature.
 Giovanni Moroni and Stefano Petrò / Procedia CIRP 10 (2013) 54 – 60 59

plotted in Fig. 2: it is apparent that the surface presents a 35

sawtooth profile, probably due to the multiple cutting Hammersley strategy

Manufacturing signature based strategy

steps required to manufacture this surface, but also a 30

trend along the y axis, which will necessarily influence 25

parallelism. A similar behavior is shown by the datum

Inspection cost [Euro]

feature. 20

Finally, a simulated annealing algorithm was applied

15
in order to select an optimal sampling strategy. In order
to generate graphs in Fig. 4 and Fig. 4, the sample size 10

has been kept constant, so that the relationship between

inspection cost and sample size can be pointed out.
5

Please note more than a single geometric feature is being 0

0 50 100 150 200 250

considered: the indicated sample size sums both the Number of Sampling Points

points sampled on the toleranced and datum feature, and,

throughout the optimization process, sampling points are Fig. 4: Effectiveness of sampling strategies: Inspection cost as the
left free to “migrate” from datum to toleranced feature sample size increases.
and vice versa. Moreover, remember that more than a
geometric tolerance and more than a geometric feature uncertainty reduction does not compensate sample size
are considered: the cost in Fig. 4 sums the costs due the increase, so CI tends to increase with the increasing CM.
measurement of two planes, and to the inspection errors Anyway, because of lower uncertainty the
when checking two straightnesses and a parallelism. A manufacturing signature based strategy always shows
signature based strategy was compared a Hammersley lower cost than the Hammersley strategy.
strategy, which is considered very effective for
measuring planes [28]. As an instance of the 4. Conclusions
effectiveness of the proposed approach, Fig. 3 shows the
behavior of expanded uncertainty (coverage factor k=2) In this paper a methodology has been proposed able
for parallelism as the sample size varies. As expected, as to optimize sampling strategies for inspecting multiple
sample size increases uncertainty reduces; it is apparent geometric tolerances defined on a single part. The
that the manufacturing signature based strategy approach considers that, due to the fact that a single
outperforms the Hammersley strategy. Straightnesses geometric feature can be involved in more than a
uncertainty shows a similar behavior. Fig. 4 shows the geometric tolerance, the estimated geometric errors may
relationship between inspection cost and sample size. be correlated, so the strategy has to be optimized
The graph shows a non-monotonic behavior: at the simultaneously for the whole inspection. The generated
beginning, cost decreases, then it starts to increase, so strategy optimizes inspection cost. Inspection cost is
there is an optimal sample size for which cost is mainly linked to the number of points sampled, and to
minimum. Non monotonic behavior depends on the the probability of inspection errors. The optimization of
initially large uncertainty, which causes CE (and CI) to the sampling points locations can be based on the
be large; then uncertainty quickly drops, so cost reduces. presence of a “manufacturing signature”, that is a
However, as the sample size increases any further systematic behavior of the real geometric feature. The
x 10
−3 proposed methodology suggests that an evaluation of
uncertainty taking into account the interaction between
8
Hammersley strategy
Manufacturing signature based strategy
7 the sampling strategy and the manufacturing signature
can lead to a lower inspection cost. Therefore the
Measurement uncertainty [mm]

sampling strategy characterized by the optimal

interaction may be selected through the comparison of
5

4 several possible strategies.

3
A main drawback of a signature based sampling
strategy is that if the signature itself changes then the
2
measurement uncertainty will probably increase;
1 however, it is possible to adopt efficient “statistical
0
process control” techniques to identify signature
modifications. Similar methodologies have been
0 50 100 150 200 250
Number of Sampling Points

proposed by Colosimo et al. [29], and the application of

Fig. 3: Effectiveness of sampling strategies: Parallelism uncertainty as these techniques is useful because a modification in the
the sample size increases. signature usually causes a worsening of product quality
60 Giovanni Moroni and Stefano Petrò / Procedia CIRP 10 (2013) 54 – 60
and is linked to undetected failures in the manufacturing
process.
Acknowledgements
Financial support to this work has been provided by
Italian Ministry of Education, University and Research,
as part of the project PRIN2008 (prot. 200853ZT3Z).
References
[1] Kunzmann, H., Pfeifer, T., Schmitt, R., Schwenke, H.,
Weckenmann, A., 2005. Productive metrology - adding value to
manufacture. CIRP Annals - Manufacturing Technology 54, p.
155. doi: 10.1016/S0007-8506(07)60024-9.
[2] Burdick, R. K., Borror, C. M., Montgomery, D. C., 2003. A
review of methods for measurement systems capability analysis.
Journal of Quality Technology 35, p. 342.
[3] Wilhelm, R. G., Hocken, R., Schwenke, H., 2001. Task specific
uncertainty in coordinate measurement. CIRP Annals -
Manufacturing Technology 50, p. 553. doi: 10.1016/S0007-
8506(07)62995-3.
[4] Weckenmann, A., Knauer, M., Kunzmann, H., 1998. The
influence of measurement strategy on the uncertainty of CMM-
measurements. CIRP Annals - Manufacturing Technology 47, p.
451 . doi: 10.1016/S0007-8506(07)62872-8.
[5] Kruth, J. P., Van Gestel, N., Bleys, P., Welkenhuyzen, F., 2009.
Uncertainty determination for CMMs by monte carlo simulation
integrating feature form deviations. CIRP Annals - Manufacturing
Technology 58, p. 463. doi: 10.1016/j.cirp.2009.03.028.
[6] Sprauel, J. M., Linares, J. M., Bachmann, J., Bourdet, P., 2003.
Uncertainties in CMM measurements, control of ISO
specifications. CIRP Annals - Manufacturing Technology 52, p.
423. doi: 10.1016/S0007-8506(07)60616-7.
[7] ISO, 2004. ISO 1101: Geometrical product specifications (GPS) -
tolerances of form, orientation, location and run out.
[8] Dowling, M. M., Griffin, P. M., Tsui, K.-L., Zhou, C., 1997.
Statistical issues in geometric feature inspection using coordinate
measuring machines. Technometrics 39, p. 3. doi:
10.2307/1270764.
[9] Summerhays, K. D., Henke, R. P., Baldwin, J. M., Cassou, R. M.,
Brown, C. W., 2002. Optimizing discrete point sample patterns
and measurement data analysis on internal cylindrical surfaces
with systematic form deviations. Precision Engineering 26, p. 105
. doi: 10.1016/S0141-6359(01)00106-4.
[10] Colosimo, B. M., Gutierrez Moya, E., Moroni, G., Petrò, S., 2008.
Statistical sampling strategies for geometric tolerance inspection
by CMM. Economic Quality Control 23, p. 109. doi:
10.1515/EQC.2008.109.
[11] Colosimo, B. M., Moroni, G., Petrò, S., 2010. A tolerance interval
based criterion for optimizing discrete point sampling strategies.
Precision Engineering 34, p. 745 . doi:
10.1016/j.precisioneng.2010.04.004.
[12] Colosimo, B. M., Gutierrez Moya, E., Moroni, G., Petrò, S., Mar.
2007. “Measuring strategy definition for flatness: a case study”.
10th CIRP International Conference on Computer Aided
Tolerancing. Erlangen, Germany, p. CD–ROM.
[13] Moroni, G., Petrò, S., 2010. Virtual CMM based sampling
strategy optimization. In “Product Life-Cycle Management:
Geometric Variations”, M. Giordano, L. Mathieu, F. Villeneuve,
Editors, Wiley-ISTE, London, UK, chap. 22, p. 385–404.
[14] Moroni, G., Petrò, S., Jun. 2010. “Optimal sampling strategy for
orientation tolerance verification”. 7th CIRP International
Conference on Intelligent Computation in Manufacturing
Engineering. Capri, Italy.
[15] Ceglarek, D., Moroni, G., Petrò, S., Oct. 2010. “Manufacturing
process error signature and CMM uncertainty costs”. 5th Annual
Meeting of the American Society for Precision Engineering.
Atlanta, GA, USA, p. 321–324.
[16] Lee, J., Johnson, G., Sep. 1993. Optimal tolerance allotment using
a genetic algorithm and truncated monte carlo simulation.
Computer-Aided Design 25, p. 601. doi: 10.1016/0010-
4485(93)90075-Y.
[17] Jeang, A., 1999. Optimal tolerance design by response surface
methodology. International Journal of Production Reserch 37, p.
3275. doi: 10.1080/002075499190284.
[18] Diplaris, S. C., Sfantsikopoulos, M. M., 2000. Cost - tolerance
function. a new approach for cost optimum machining accuracy.
The International Journal of Advanced Manufacturing
Technology 16, p. 32. doi: 10.1007/PL00013129.
[19] Shan, A., Roth, R., Wilson, R., Dec. 2003. Genetic algorithms in
statistical tolerancing. Mathematical and Computer Modelling 38,
p. 1427. doi: 10.1016/S0895-7177(03)90146-4.
[20] Loof, J., Hermansson, T., Soderberg, R., 2007. An efficient
solution to the discrete least-cost tolerance allocation problem
with general loss functions: Models for computer aided
tolerancing in design and manufacturing. In “Models for
Computer Aided Tolerancing in Design and Manufacturing”,
J. K. Davidson, Editor, Springer Netherlands, p. 115–124. doi:
10.1007/1-4020-5438-6_13.
[21] ISO, 2004. ISO/TS 15530-3: Geometrical product specifications
(GPS) – coordinate measuring machines (CMM): Technique for
determining the uncertainty of measurement – part 3: Use of
calibrated workpieces or standards.
[22] Moroni, G., Petrò, S., Tolio, T., 2011. Early cost estimation for
tolerance verification. CIRP Annals - Manufacturing Technology
60, p. 195. doi: 10.1016/j.cirp.2011.03.010.
[23] ISO, 1998. ISO 14253-1: Geometrical product specifications
(GPS) - inspection by measurement of workpieces and measuring
equipment - part 1: Decision rules for proving conformance or
nonconformance with specifications.
[24] Fiorentini, F., Moroni, G., Palezzato, P., Semeraro, Q., Jun. 1992.
“Feature selection for an automatic inspection system”. 24th CIRP
International Seminar of Manufacturing Systems. Copenhagen,
Denmark, p. 199–208.
[25] ISO/IEC, 2008. ISO/IEC GUIDE 98-3: Uncertainty of
measurement - Part 3: Guide to the expression of uncertainty in
measurement (GUM:1995).
[26] ISO/IEC, 2007. ISO/IEC GUIDE 99:2007(E/F): International
vocabulary of basic and general terms in metrology (VIM).
[27] ISO, 1998. ISO 10791-7: Test conditions for machining centres -
part 7: Accuracy of a finished test piece.
[28] Lee, G., Mou, J., Shen, Y., 1997. Sampling strategy design for
dimensional measurement of geometric features using coordinate
measuring machine. International Journal of Machine Tools and
Manufacture 37, p. 917. doi: 10.1016/S0890-6955(96)00096-X.
[29] Colosimo, B. M., Pacella, M., 2009. A comparison study of
control charts for statistical monitoring of functional data.
International Journal of Production Reserch 47, p. 1. doi:
10.1080/00207540802662888.