Comput. Methods Appl. Mech. Engrg.

190 (2000) 149±178

www.elsevier.com/locate/cma

The structure of positional tolerance evaluation:

I. Constructive geometric approach
Mark J. Kaiser a,*, S. Hossein Cheraghi a, Shuhe Li b
a
Department of Industrial and Manufacturing Engineering, Wichita State University, Wichita, KS 67260-0035, USA
b
College of Precision Instrument & Opto-Electronics Engineering, Tianjin University, Tianjin 300072, People's Republic of China
Received 22 December 1997

Abstract

A general geometric framework is developed to quantify the structure of positional tolerance evaluation. The data graph, overlay
graph, and con®guration framework of a feature set is de®ned and used in the evaluation of positional tolerance. Two functionals
based on the con®guration framework are introduced and lead to the development of optimal matings. The con®guration functional
quanti®es the amount of ``turning'' the framework will allow while satisfying the mating condition, while the mating functional is used
to assess optimal mating conditions. A constructive geometric procedure is used to develop the con®guration and mating functionals
for a given feature vector and tolerance speci®cations. Numerical examples are used to illustrate the form of the functionals and the
procedures involved in the analysis. Ó 2000 Elsevier Science S.A. All rights reserved.

Keywords: Constructive geometry; Con®guration framework; Con®guration functional; Geometric metrology; Mating region

Notation

GD data graph …ˆ f…U; R; k†; …F; n; a†g†

GO overlay gage …ˆ …s; r; u††
U design form vector …ˆ …U0 ; U1 ; . . . ; Un ††
U0 external form
…U1 ; . . . ; Un † internal form vector
R design size vector …ˆ …R0 ; R1 ; . . . ; Rn ††
k design location vector …ˆ …k0 ; k1 ; . . . ; kn ††
k0 location of external form
…k1 ; . . . ; kn † location of the internal forms …U1 ; . . . ; Un †
…F; n; a† feature vector
F feature form vector …ˆ …F0 ; F1 ; . . . ; Fn ††
n feature size vector …ˆ …n0 ; n1 ; . . . ; nn ††
a feature location vector …ˆ …a0 ; a1 ; . . . ; an ††
…s; r; u† design tolerance vector
s tolerance form vector …ˆ …s0 ; s1 ; . . . ; sn ††
r tolerance size vector …ˆ …r0 ; r1 ; . . . ; rn ††
u tolerance location vector …ˆ …u0 ; u1 ; . . . ; un ††

*
Corresponding author.
E-mail address: [email protected] (M.J. Kaiser).

0045-7825/00/$ - see front matter Ó 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 0 4 5 - 7 8 2 5 ( 9 9 ) 0 0 4 1 9 - 3
150 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

A manufactured artefact …ˆ …F; n; a††

B design blueprint …ˆ f…U; R; n
k†; …s; r; u†g† o
1 2 n
S measurement data sets ˆ Sm…1† ; Sm…2† ; . . . ; Sm…n†
n o
Sim…i† Bi1 ; Bi2 ; . . . ; Bim…i† j Bij ˆ …xij ; yji †
CF con®guration framework …ˆ CF ……x; y†; u† ˆ f H ˆ …V ; E† j V ˆ …I; A1 ; . . . ; An †; E ˆ
ƒ! ƒ!
…IA1 ; . . . ; IAn †; I ˆ …x; y†g†
vi …
† indicator
Tn variable
v…
† v
iˆ1 i …
† ˆ system structure function
X con®guration space …ˆ f…x; y† j v……x; y†; u† ˆ 1; u 2 S 1 g†
X optimal con®guration space
T con®guration tube
M mating region …ˆ f…x; y† j v……x; y†; u† ˆ 1 for any u 2 S 1 g†
M optimal mating region
fC con®guration functional
fc mating functional

1. Introduction

1.1. The quality of a manufactured artefact

The concept of quality in manufacturing is primarily based on machine capabilities, material compo-
sition, and operator training. Products with small variations in shape, size, and feature location are con-
sidered high quality, while products with large variations are considered poor quality. If shape, size, or
feature location cannot be maintained within prescribed limits, the part is rejected and categorized as ``out-
of tolerance''. This may indicate that the manufacturing process is ``out-of-control'' or the particular
process is inappropriate for the tolerance levels speci®ed by the design blueprints. Tolerances are assigned
in as coarse a manner as possible while maintaining functional speci®cations, and is normally dictated by
cost, labour, equipment, and time considerations.
The nominal, or (perfect) form, geometry is de®ned by the CAD model or blueprint design, and to
compare the (actual) manufactured artefact A with the (ideal) perfect form U, various gauging techniques
are employed. Computer-controlled coordinate measuring machines (CMMs) are currently the inspection
instrument of choice for dimensional and location measurements, where a CMM obtains data by a touch
probe on a point-by-point basis on the surface of the artefact. Coordinate measuring machines are accurate,
general purpose, easily programmable, readily adopted for process control, and have gained wide accep-
tance due to their ¯exible approach to inspection/veri®cation tasks ± especially in the aeronautical and
automobile industries.
Consider the design blueprint B and manufactured artefact A. The design blueprint B is composed of
nominal geometric shapes …U0 ; U1 ; . . . ; Un † referenced with respect to the position vector k ˆ …k0 ; k1 ; . . . ; kn †,
where ki ˆ …ai ; bi † locates a particular point (normally a ``center'' location) of the shape Ui , i ˆ 0; 1; . . . ; n.
This position vector is referenced with respect to a given xy-Cartesian coordinate system. The design
blueprint B is realized as artefact A upon the completion of the manufacturing process. The design forms
…U0 ; U1 ; . . . ; Un † correspond to the feature forms …F0 ; F1 ; . . . ; Fn † referenced with respect to the position
vector a ˆ …a0 ; a1 ; . . . ; an †, where ai ˆ …xi ; yi † locates the feature Fi , i ˆ 0; 1; . . . ; n. The geometric features
are positioned with respect to the base shape F0 and can also be considered positioned with respect to one
other. The nominal shape U0 corresponds to the ``external'' (or base) shape of the feature F0 , while
…U1 ; . . . ; Un † corresponds to the ``internal'' features …F1 ; . . . ; Fn † located within the base shape. The fea-
tures may be any shape or size and located anywhere within F0 . For the artefact to be considered ``in
tolerance'', however, the features will have to satisfy both form and positional tolerances as speci®ed in the
design blueprint. The physical characteristics of the geometric forms may include ribs, bosses, blind/thru-
holes, slots, etc.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 151

There is a one-to-one correspondence between the features Fi and their nominal design forms Ui ,
i ˆ 0; 1; . . . ; n:
Fi $ U i :

For an ideal manufacturing operation, the correspondence is perfect in the sense that the form of Fi
corresponds precisely to the design shape Ui , and the position ai of Fi corresponds to the exact design
speci®cation location ki of Ui for all i, i ˆ 0; 1; . . . ; n. In a real manufacturing environment, however, the
correspondence is ``fuzzy'' due to the stochastic nature of machining, material composition, operator ca-
pabilities, etc. The form and location vector of the feature set …F0 ; F1 ; . . . ; Fn † do not precisely match the
design requirements speci®ed by …U0 ; U1 ; . . . ; Un † and …k0 ; k1 ; . . . ; kn †, and the manufactured artefact A can
deviate from the perfect form U in many ways.
The quality of the artefact A depends on the nature of the variation
· form of F0 ,
· form of feature Fi , i ˆ 1; . . . ; n, and
· location ai of feature Fi ; i ˆ 1; . . . ; n, relative to a datum or feature set.
To determine the quality of artefact A and if A is produced ``in tolerance'', the form of F0 and the form of
each feature Fi ; i ˆ 1; . . . ; n must be compared against their nominal design forms U0 and Ui ; i ˆ 1; . . . ; n,
respectively:
hFormi …F0 ; F1 ; . . . ; Fn † $ …U0 ; U1 ; . . . ; Un †:

This is referred to as the hFO R M i constraints. hFO R M i constraints are important and have been examined
extensively in the literature. The tolerance of the forms is usually assessed by the minimum zone criteria;
i.e., determine the minimum zone which encompasses the geometry of the form, although a number of
methodologies can be applied. For an overview of the minimum zone methodology and further reference to
the literature, see [9].
With respect to a feature, set of features, or datum d, the location ai of each feature Fi , as well as its
position relative to the other features Fi , i 6ˆ j, must be veri®ed to be within acceptable tolerance. A
tolerance zone si is thus associated with the form vector Ui and is used to control the location and orien-
tation of the pattern of features. This is referred to as the hPO S I T I O N i constraints, which the primary
subject of this paper. The hPO S I T I O N i constraints are represented formally as
hPositioni …a0 ; a1 ; . . . ; an † $ …U0 ; U1 ; . . . ; Un †:

and in brief, the location of the feature elements ai must be contained in the tolerance zones si centred at ui .
The characteristics of the tolerance zones, namely their size and location, are assigned by the design
blueprints, and the feature vector a can be moved as a ®xed unit to satisfy positional tolerance.
Artefact A is considered ``acceptable'', or ``in tolerance'', if it satis®es the hFO R M i and hPO S I T I O N i
conditions simultaneously. To reject the part, it is clear that if the form tolerance for any Fi is not satis®ed,
or if the location tolerance for any Fi is not met, then the manufactured part will be rejected (or re-
worked). 1 To assess the degree of quality of the manufactured artefact, it is necessary to assign numerical
values to determine how ``far o€'' target artefact A is from the design form U.

1.2. Scope of analysis

The purpose of this paper is to develop a geometric and computational framework to assess the quality
of positional tolerance in manufactured artefacts. Form tolerances are thus assumed to be satis®ed for each
feature of the artefact. The approach taken in this paper to understand the structure of positional tolerance
is geometrically oriented, and thus geometry plays a fundamental role in the intuition, analysis, and
constructions that follow. The computational aspects of this work, including the algorithmic structure,

1
The variety and type of variations that occur in practice tend to make generalized statements concerning form variations and
position deviations dicult to validate, although particular form variations tend to be ``easier'' to adjust for than position deviations.
152 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

program design, and software development is intentionally subjugated to focus on the geometric aspects of
the problem. This ``constructive geometry'' approach allows us to concentrate on the structure of the
positional tolerance problem rather than on the ``computational geometric'' aspects of the algorithmic
procedures.

1.3. Tolerances of location

Tolerances of location include position, concentricity, and symmetry, and control relationships such
as the location of features from data, centre distances between features, and coaxiality or symmetry of
features [5].
A feature is a physical portion of a manufactured artefact which does not have any dimensions asso-
ciated with it such as a pin, tab, hole, etc. If dimensions are added to features they are called features of size.
Position is a term used to describe the (exact) location of a point, line, or plane of a feature in relationship
with a datum reference or other feature. A position tolerance de®nes a zone within which the axis (for
cylindrical features) or centre plane (for slots, tabs, etc.) of a feature of size is permitted to vary from true
position.
In Fig. 1, an example illustrating the interpretation of the veri®cation requirements as speci®ed in ASME
Y14.5M-1994 is shown. The four circles in the top drawing represent the design pattern with the centre of
each hole marked by a cross. The smaller circles in the bottom drawing represent similar (size) feature
tolerance zones centred at the design location of the form. The centres of the tolerance zone circles can be

Fig. 1. An example illustrating the interpretation of the veri®cation requirements as speci®ed in ASME Y14.5M-1994.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 153

located anywhere provided their relative positions form a pattern identical to the true positions. Composite
tolerance requirements are satis®ed for this example if each of the four hole centres determined by mea-
surement lies within (or can be translated/rotated to lie within) its corresponding tolerance zone.

1.4. Geometric variation analysis of aircraft components

The hPO S I T I O N i condition is an important constraint in a number of industries such as aircraft, auto-
mobile, and precision manufacturing. For example, an aircraft consists of thousands of components which
often have complex geometries and require careful consideration in manufacturing and inspection. A
common feature in aircraft sheet metal components are holes (which are the positions for rivets) and one of
the important issues in manufacturing is veri®cation of the relationship between di€erent features produced
on the surfaces. Holes are drilled on both planar and complex surfaces with the normal vector of the holes
(i.e., the hole axis) intersecting, perpendicular, or at angles to one other. Position tolerance is usually used
to de®ne the relationship between holes in a pattern and is called a feature relating tolerance zone (FRTZ)
as shown in Fig. 2. Depending on the design requirements, a maximum material condition (MMC) may or
may not be applied. If MMC is speci®ed, then the size of the tolerance zone increases as the hole's actual
diameter deviates from its MMC value. Regardless of feature size (RFS) speci®cations impose ®xed tol-
erance zone diameters irrespective of the actual hole size [3,5].
Accurate evaluation of FRTZ is important in assembly and process control. The normal procedure for
veri®cation of the relationship between holes is through the use of a functional gage. Functional gages tend
to be expensive, dicult and time consuming to manufacture, and are rather inaccurate devices (although
they are quite easy to use). Functional gages are attribute gages (i.e., they provide a yes/no answer to
tolerance veri®cation) and the results they provide are not especially useful for statistical analysis. Func-
tional gages therefore cannot be used to study and control the process, and are not as accurate as computer-
aided mathematical-based evaluation techniques.

1.5. Manual measurement of position tolerances

In many design applications, there are patterns of features that must be in the correct position in order to
®t the mating part. To evaluate the part, functional gaging techniques or graphical inspection analysis
(paper gaging) are often applied. The stages of a graphical inspection analysis can be summarized as follows
[6, pp. 260±262]:
1. Data graph construction
1.1. On a piece of paper label the positions of each design form with a cross …‡†.
1.2. Place a point …
† for the actual location of each feature.
2. Overlay gage construction. On a piece of clear transparency, overlay the transparency on the data graph
and construct the tolerance zones si centred at the ideal position …‡† with respect to each feature.
3. Evaluation. The overlay gage is translated and/or rotated to position the points …
† within each speci®ed
tolerance zone si . If the overlay can be positioned such that all points fall within the tolerance zones, the
part is acceptable and considered to be ``in'' positional tolerance. Otherwise, the part is unacceptable.
4. Optimization. The overlay gage is shifted to ``best'' ®t the measurement points with the centres of the
tolerance zones.
The ideal positions of each feature …‡† are determined from the design blueprint and are referenced with
respect to a Cartesian coordinate system. With respect to the same coordinate system, the actual positions
…
† of the manufactured features are determined by measurement from a CMM. 2 The CMM computes the
coordinates of the boundary of the form, and this data is then reduced via a ``best form ®t'' to a reference
location. The evaluation procedure is then performed manually by translating and rotating the overlay
transparency in such a way to include each of the feature location points (
† within their respective tolerance
zones. The primary consideration is that all the feature points reside simultaneously within their respective
tolerance zones. Once the overlay gage is positioned such that all the feature points simultaneously fall

2
This normally involves the determination of accurate data to properly orient the artefact.
154 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

Fig. 2. A typical feature relating tolerance zone (FRTZ) for positional tolerance.

within the tolerance zones, the part is considered to be in position tolerance. If the overlay gage cannot be
positioned to include all the feature points within the tolerance zones, the part is considered to be out-of
positional tolerance (and the tolerance zones are re-adjusted, the part is re-worked, or the part is scrapped).
Methodology to ``optimize'' the overlay gage location by hand is not speci®ed in the procedure outlined in
[6], but a ``least-squares'' routine (between the points …‡† and …
†) is probably what the visual system of the
operator unconsciously employs although other ``objective'' functions learned through training (such as
minimax) are conceivable.

1.6. Automated position tolerance procedure

The paper gaging technique is a manual procedure, and it is thus labour intensive, tedious, and
inaccurate. Paper gaging can only provide an approximate solution and a best ®t position of the overlay
gage is dicult to realize in practice. Furthermore, when the tolerance zones are numerous, small, and
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 155

complex-shaped, and if the CMM data yield points near the boundary of the associated zones, a visual
solution methodology may not even be possible.
In most practical applications, the features are usually holes, and so the tolerance zones can be assumed
to be circles of varying diameter. Only when the simultaneous measurement of the holes falls within each of
their zones can the part be considered acceptable. As the number of holes increases, the manual procedure
becomes increasingly complex, and to take full advantage of the automation o€ered by CMMs, it is de-
sirable to automate the position tolerance procedure. Requicha [15] was one of the ®rst researchers to
propose a geometric theory of tolerancing, and a large amount of work on tolerances of features have been
examined (see the references in [7]). The positional tolerance model has been examined to a lesser extent and
in a number of different (but related) contexts, mainly in the analysis of the position uncertainties of parts
in an assembly; e.g., in [12±14] a con®guration space method for analysing position uncertainties is con-
sidered, while in [17] an arti®cial intelligence approach is developed. More closely related to the model
considered in this paper is the work of Gunasena and Lehtihet [10,11] who formulated and solved the
position tolerance problem using nonlinear programming for the case of same size circular zones (RFS).
Carpinetti and Chetwynd [2] applied a genetic algorithm approach for a similar problem. In [1,4,8,16,18],
optimal process plans based on arti®cial intelligence and other expert-based reasoning schemes to realize
positional tolerance speci®cations are described.

2. Geometric framework for feature pattern evaluation

A generalized mathematical framework is developed to evaluate feature patterns. Motivated by the

manual tolerancing procedure, a geometric formulation to the position tolerance problem is developed
which quanti®es the structure of the design and the tolerance vector. The data graph, overlay gage, and
con®guration framework are de®ned and constructed.

2.1. Design blueprint characterization

The design blueprint B is represented by U and is composed of a nominal (external) design shape U0 and
a family of internal geometric shapes …U1 ; U2 ; . . . ; Un †. The nominal shapes Ui ; i ˆ 0; 1; . . . ; n, are located
with respect to an xy-Cartesian coordinate system and a reference position de®ned by the coordinate point
ki ˆ …ai ; bi †; i ˆ 0; 1; . . . ; n. The size of the nominal shapes are denoted by R ˆ …R0 ; R1 ; . . . ; Rn †, where
Ri 2 R; i ˆ 0; 1; . . . ; n. In conjunction with the manual procedure for positional tolerance evaluation, the
coordinate point ki corresponds to the cross …‡† as described in the data graph construction. Fig. 3 depicts

Fig. 3. The design blueprint U ˆ …U0 ; U1 ; U2 ; U3 ; U4 † and design form location vector k ˆ …k1 ; k2 ; k3 ; k4 †, where ki ˆ …ai ; bi †,
i ˆ 1; . . . ; 4; is denoted by a cross …‡†.
156 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

Fig. 4. The design blueprint U and tolerance zone vector s ˆ …C1 ; C2 ; C3 ; C4 † composed of circular elements Ci of radius ri centred at
the design location ki , i ˆ 1; . . . ; 4.

Fig. 5. The design blueprint U, tolerance vector s ˆ …s1 ; s2 ; s3 ; s4 †, and tolerance location vector u ˆ …u1 ; u2 ; u3 ; u4 †, where
ui ˆ …ci ; di †, i ˆ 1; . . . ; 4 is denoted by an  …†.

a speci®c tolerance problem to which we will refer throughout this discussion. Associated with each
nominal form Ui is the tolerance zone si ; i ˆ 0; 1; . . . ; n. The form of the tolerance zone si normally cor-
responds to a circle Ci of radius ri centred at ki ˆ …ai ; bi † as in Fig. 4. The size of the tolerance circles are
presented in vector form as r ˆ …r0 ; r1 ; . . . ; rn †, where ri 2 R; i ˆ 0; 1; . . . ; n. Generally speaking, the toler-
ance zones can be any shape or size and arbitrarily located at ui ˆ …ci ; di †; i ˆ 0; 1; . . . ; n as in Fig. 5. To
distinguish the location of ui from ki , an x …† is used to locate ui .
The data describing the design blueprint, including the form type, size, and location of the design shapes
Ui , and the size, shape, and location of the tolerance zones si are assumed known and provided by the
design blueprint. The design form type (shape and size) is denoted by the aggregate vectors
U ˆ …U0 ; U1 ; . . . ; Un † and R ˆ …R0 ; R1 ; . . . ; Rn †, and the location of the design forms is described by the
vector k ˆ …k0 ; k1 ; . . . ; kn †, where ki ˆ …ai ; bi †; i ˆ 0; 1; . . . ; n. The shape and size of the elements of the
tolerance form vector are denoted by s ˆ …s0 ; s1 ; . . . ; sn † and r ˆ …r0 ; r1 ; . . . ; rn †. The tolerance vector is
associated with the form vector U and is used to control the location and orientation of the pattern of
features. The location of the tolerance forms is described by the vector u ˆ …u0 ; u1 ; . . . ; un †, where
ui ˆ …ci ; di †; i ˆ 0; 1; . . . ; n. The design blueprint is thus characterized by two three-tuples: …U; R; k† and
…s; r; u†, where the elements of the form vector U and tolerance vector s have magnitudes de®ned by R and
r, and are located at k and u, respectively.

2.2. The data graph and overlay gage construction

The manufacturing process P realizes the form vector …U0 ; U1 ; . . . ; Un † of the design blueprint B as the
feature vector F ˆ …F0 ; F1 ; . . . ; Fn † as in Fig. 6. The individual features Fi are stochastic representations
of the design forms Ui and can be considered random shapes in a mathematical sense. The features of the
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 157

Fig. 6. The manufacturing process realizes the form vector U of the design as the feature vector F ˆ …F0 ; F1 ; F2 ; F3 ; F4 † of
artefact A.

artefact are measured to detect their actual form, size, and location. Each feature is sampled (normally with
j
a CMM) and the result of the measurement process is a set of data points Sm…j† ˆ fBj1 ; Bj2 ; . . . ; Bjm…j† g, where
j j j j
Bi ˆ …xi ; yi †; i ˆ 1; . . . ; m; j ˆ 0; 1; . . . ; n. The superscript notation j on the sampled data set Sm…j† is the
index that corresponds to the number …n† of form features, while the subscript m…j† depends upon the
speci®c form under consideration and corresponds to the size of the measurement set. The size of the
sampled data sets change with the feature size and type, machine capabilities, user preference, and time
constraints on the measurement process, and this dependence 3 is denoted by m…j†.
j
The sample sets Sm…j† for each feature are then processed to ascertain form and positional information.
The output of the analysis is a determination of the closeness of the shape of the feature to its intended
form, as well as the determination of a single ``centre'' point ai ˆ Ai ˆ …xi ; yi †; i ˆ 0; 1; . . . ; n, that locates
the feature (and corresponds to the point (
† in the manual procedure). The size of each form feature Fi is
given by ni , and the size vector is denoted by n ˆ …n0 ; n1 ; . . . ; nn †, where ni 2 R; i ˆ 0; 1; . . . ; n. As discussed
earlier, this data is assumed given. The measurement process introduces additional error in the realization
of the artefact, and the extent of the error depends upon the accuracy of the CMM and the sampling
strategy used to collect the data points. The position of each feature is denoted by the feature location
vector a ˆ …a0 ; a1 ; . . . ; an †, where ai ˆ Ai ˆ …xi ; yi †; i ˆ 0; 1; . . . ; n (see Fig. 7). Due to the stochastic nature
of the processes involved to realize F, the values of ai and ni are random variables which (under normal
measurement conditions) correspond to their expected value.
In summary and in accord with the manual measurement of positional tolerance, the data graph GD is
composed of the form and feature vector:
GD ˆ f…U; R; k†; …F; n; a†g
and the overlay gage graph GO is composed of the tolerance vector:
GO ˆ …s; r; u†:
The design blueprint B is thus characterized by the form and tolerance vectors
B $ f…U; R; k†; …s; r; u†g;
which are assumed known and provided by the design engineer. The output of the manufacturing process P
(i.e., artefact A) is represented by the feature vector …F; n; a†:
A $ …F; n; a†:

3
For instance, in the assessment of roundness the normal sample size is to take three, four, or ®ve points on the feature (not 100 or
1000). It would be highly unusual to take 10 or more sample points to test for circularity for example (unless the form is very large in
size), but for more complicated shapes a large data set is absolutely necessary for an accurate evaluation of the part.
158 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

Fig. 7. The feature vector F, measurement sample sets S4 ˆ …S31 ; S42 ; S53 ; S10 4
†, and feature location vector a ˆ …a1 ; a2 ; a3 ; a4 †, where
ai ˆ Ai ˆ …xi ; yi †, i ˆ 1; . . . ; 4 and the sample data are denoted by points …
†.

The geometric form of the feature vector F, including the magnitude vector n and location vector a is
assessed by measurement equipment and the individual elements of the feature vector are considered
stochastic variables.

2.3. Positional tolerance evaluation

In the manual procedure of positional tolerance evaluation, the design tolerance vector …s; r; u† of the
overlay gage GO is graphed on a transparency sheet and (manually) translated and rotated with respect to
the feature vector …F; n; a† described through the data graph GD . The objective of the operator who is
assessing positional tolerance is to simultaneously cover the feature vector locations described by the points
…
† positioned at ai ; i ˆ 1; . . . ; n, with the design tolerance elements which normally correspond to circular
regions. The evaluation procedure for positional tolerance is as follows:
If GO covers the feature vector …F; n; a†, then A is in positional tolerance.
If GO does not cover …F; n; a†, then A is out-of positional tolerance.
To ``cover'' the feature …F; n; a†, the tolerance zones si of the overlay gage are allowed to move under rigid
structure motion and must cover all the points de®ned by ai , i ˆ 1; . . . ; n. The design vector …U; R; k† usually
plays an indirect role in the analysis through the association ui ˆ ki , while the relative size of the features
(ni ) are not considered in the standard evaluation of positional tolerance (since the data reduction of the
sample sets yield point estimates ai for the centre location).
For computational purposes it is convenient to consider the overlay gage graph GO as ®xed in the plane,
and the feature vector …F; n; a† as a composite and rigid moving structure. The positional tolerance
evaluation procedure can thus be expressed:
If …F; n; a† is contained within GO , then A is in positional tolerance.
If …F; n; a† can not be contained within GO , then A is out-of positional tolerance.
This later formulation leads to the de®nition of a con®guration framework and an analysis of the structure
of positional tolerance.

2.4. The con®guration framework

In the manual evaluation of positional tolerance, the operator shifts the overlay transparency such that
the tolerance zones de®ned by GO cover the feature vector …F; n; a†. In a computational environment,
however, it is convenient to reverse the situation and consider the overlay gage GO as ®xed and to de®ne a
movable con®guration framework based on the feature(less) location vector a. Positional tolerance is then
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 159

ƒ! ƒ! ƒ! ƒ!
Fig. 8. Con®guration frames for (a) reference point I: CF ˆ fIA1 ; IA2 ; IA3 ; IA4 g, and (b) reference point I 0 : CF0 ˆ
ƒƒ! ƒƒ! ƒ! ƒƒ!
fI 0 A1 ; I 0 A2 ; IA3 ; I 0 A4 g.

evaluated by translating and rotating the framework relative to the ®xed tolerance zone in such a manner
that the elements of the framework fall within the zones de®ned by …s; r; u†. The advantage of this ap-
proach is that the structure 4 of the con®guration framework allows the positional tolerance problem to be
quanti®ed and analysed in ways that are dicult using standard algebraic techniques. The structure of the
con®guration framework induces ``structure'' associated with the positional tolerance problem.
The con®guration framework CF of the feature set F ˆ …F1 ; . . . ; Fn † is constructed with respect to
point I ˆ …x; y† and is de®ned as follows. Choose point I ˆ …x; y† within the base shape F0 such that I 62 Fi ,
ƒ!
i ˆ 1; . . . ; n. The con®guration ``framework'' CF is de®ned by the collection of vectors IAi , i ˆ 1; . . . ; n,
which are ``joined'' at point I:
ƒ! ƒ!
CF ˆ fIA1 ; . . . ; IAn g;

where Ai ˆ ai ˆ …xi ; yi †, i ˆ 1; . . . ; n, locates the centre of each (measured) feature. The con®guration
framework changes its structure as a function of position I as shown in Fig. 8, but is considered a composite
object (rigid structure) which can only be translated and rotated as a ®xed unit after the form of the
framework is determined from the selection of point I (see Fig. 9). Thus, regardless of the location of point
I, after I is selected, the con®guration framework is ``frozen'' in form. An arbitrary element of the con-
ƒ!
®guration framework, say vector IAj , forms angle u with respect to the xy-coordinate system, and this angle
is used to de®ne the orientation of the framework (see Fig. 10). The angle u rotates through 2p-radians and
is an element of the unit circle S 1 , u 2 S 1 .

De®nition (Configuration framework). The con®guration framework CF ……x; y†; u† de®ned with respect to
point I ˆ …x; y† and the feature(less) point set a is a graph H ˆ …V ; E† with n ‡ 1 vertices and n edges, where
V ˆ …I; A1 ; A2 ; . . . ; An † and E ˆ …IA1 ; IA2 ; . . . ; IAn †:
n ƒ! ƒ! o
CF ……x; y†; u† ˆ H ˆ …V ; E† j V ˆ …I; A1 ; . . . ; An †; E ˆ …IA1 ; . . . ; IAn †; I ˆ …x; y† :

The con®guration framework is written CF ……x; y†; u† to denote the translation and rotation dependence
explicitly.

4
The methodology is also consistent with the goals of geometric metrology [9] where the structures of metrology are considered the
fundamental elements of analysis.
160 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

ƒ! ƒ! ƒ! ƒ!
Fig. 9. Translation and rotation of the con®guration framework CF ˆ fIA1 ; IA2 ; IA3 ; IA4 g.

ƒ!
Fig. 10. The structure of the con®guration framework CF ……x; y†; u†. An arbitrary element of the framework (in this case, IA2 ) forms
1
the angle u 2 S with respect to the coordinate system and is used to describe the orientation of CF ……x; y†; u†.

3. Analytic structure for feature pattern evaluation

Based on the de®nition of the con®guration framework, the notion of mating is characterized through
the con®guration space and mating region which allows a formal de®nition of positional tolerance to be
described. Con®guration and mating functionals are then employed which leads to the notion of an optimal
mating, optimal mating region, and optimal con®guration space.

3.1. Formal de®nition of mating

For a given design form U and manufactured artefact A, essentially two questions arise in determining
the tolerance quality of A:
1. Does the form of the features F0 ; F1 ; . . . ; Fn satisfy the hFO R M i requirements as speci®ed by the design
blueprints?
2. Does the pattern of features F1 ; . . . ; Fn satisfy the hPO S I T I O N i requirements as speci®ed by the toler-
ance speci®cations?
The con®guration framework
n ƒ! ƒ! o
CF ……x; y†; u† ˆ H ˆ …V ; E† j V ˆ …I; A1 ; . . . ; An †; E ˆ …IA1 ; . . . ; IAn †; I ˆ …x; y†

is the tool used to determine whether the feature forms ``®t'' within the tolerance zones de®ned by the
overlay gage GO . ``Fit'' (or mating) is de®ned as the condition where all the ``heads'' of the vectors of the
con®guration framework CF ……x; y†; u† are located on or within each associated tolerance zone
ƒ!
…s1 ; s2 ; . . . ; sn †, i.e., IAi 2 si , i ˆ 1; . . . ; n. This condition is formalized by de®ning the indicator variable
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 161

 ƒ!
vi …
† ˆ 1 if IAi 2 si ;
0 otherwise;
ƒ!
and then requiring all the vector heads IAi to simultaneously reside within their corresponding tolerance
zones, i.e.,

\
n
vˆ vi …
† ˆ 1:
iˆ1

The value of the indicator variables vi …
† are determined jointly by the data graph
GD ˆ f…U; R; k†; …F; n; a†g and the overlay gage GO ˆ …s; r; u†. The system structure function v requires
each of the constraints to be satis®ed simultaneously. Inclusion is determined using the functional relation
ƒ!
that de®nes si as a system constraint, and then verifying that the vector IAi is an element of the zone.

De®nition (Mating). Con®guration framework CF ……x; y†; u† is said to mate with the overlay data graph GO
if
\
n
vˆ vi …
† ˆ 1;
iˆ1

where
 ƒ!
vi …
† ˆ 1 IAi 2 si ;
0 otherwise:

This provides a formal de®nition of mating.

3.2. The con®guration space and mating region

A mating m is de®ned by both a position …x; y† and orientation u 2 S 1 of the con®guration framework
CF ……x; y†; u†, and thus a mating corresponds to an element of the product space R2  S 1 :

m $ ……x; y†; u† 2 R2  S 1 :

The product space can be realized in various ways. If the position …x; y† of the con®guration framework is
®xed, three possibilities arise: mating may not occur for any orientation, one orientation may induce
mating, or an interval of angular values u ˆ ‰a; bŠ 2 S 1 may satisfy the mating condition. For an illustration
of the last condition see Fig. 11. Similarly, for a given orientation u 2 S 1 of the con®guration framework,
there may be several, one, or no locations within the plane that satisfy mating. In brief, a mating m is
de®ned by both the position and orientation through the vector ……x; y†; u†, and the collection of mating
vectors form the con®guration space X.

De®nition (Configuration space). The con®guration space X is a subset of the product space R2  S 1 which
represents the collection of vectors ……x; y†; u† that de®ne the location and orientation of CF ……x; y†; u† such
that mating with GO occurs:

X ˆ fm j CF ……x; y†; u† mates with GO g  R2  S 1 :

The con®guration space can be realized in a number of ways, and perhaps the simplest realization of X
utilizes three-dimensional space. For a given orientation u 2 S 1 the domain where mating occurs is com-
puted:
Xu ˆ f…x; y† j CF ……x; y†; u† mates with GO for u 2 S 1 g  R2 :
162 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

Fig. 11. The con®guration framework CF ……x; y†; u† mates with artefact A at point I ˆ …x; y† and orientations u ˆ a and u ˆ b.
Artefact A is in positional tolerance for orientation a 6 u 6 b.

Each domain Xu is then ``stacked'' vertically (or lifted) in the z-direction (one upon another) as an in-
creasing function of u for 0 6 u 6 2p. The con®guration space X is the collection of the slices Xu stacked in
this manner:

X ˆ Xu  S 1 ;

where the z-axis is used for the variation of the angular parameter u. The con®guration space is thus
realized as a generalized tubular structure, and in this context is labelled the con®guration tube.

De®nition (Configuration tube). The con®guration tube T provides a realization of the con®guration space
X, where mating for a ®xed orientation u is represented by Xu , and the mating region is lifted in the
z-direction as an increasing function of u; 0 6 u 6 2p:

T ˆ …x; y†; u† j X ˆ Xu  S 1 ; 0 6 u 6 2p :

For the point …x; y† and orientation u, the con®guration framework either mates or does not mate as
shown in Fig. 12. The collection of points in the plane where mating does occur (for some orientation of the
con®guration framework) is called the mating region, M. The projection of the con®guration tube or-
thogonally onto the xy-plane yields the mating region M:
[
M ˆ T? ˆ Xu :
u2S 1

The mating region M represents the collection of points in the plane where a mating occurs, and thus, can
also be considered as the union of the slices Xu for all u, 0 6 u 6 2p.

Fig. 12. The con®guration framework CF ……x; y†; u† does not mate with artefact A at point J and orientation u.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 163

De®nition (Mating region). The mating region M is a planar subset which represents the collection of
points such that the con®guration framework CF ……x; y†; u† mates with GO for some orientation u:
M ˆ f…x; y† j CF ……x; y†; u† mates with GO for some u 2 S 1 g  R2 :

Artefact A is said to be in positional tolerance if there exists a mating of the con®guration framework
CF ……x; y†; u† with GO . In other words, if the mating region is non-empty then the artefact is in positional
tolerance. Conversely, artefact A is out-of (positional) tolerance if M is the empty set; i.e., if there does not
exist a location where mating occurs for any orientation. This leads to a characterization of positional
tolerance de®ned in terms of the mating region.

De®nition (Positional tolerance). Artefact A is in positional tolerance if M 6ˆ ;. Conversely, artefact A is

out-of positional tolerance if the mating region M is empty.

3.3. The con®guration and mating functional

Two functionals based on the con®guration framework are de®ned and lead to the development of
optimal matings. The con®guration functional is used to quantify the amount of ``turning'' the framework
will allow while satisfying the mating condition, while the mating functional is used to determine optimal
matings. A direct construction of the con®guration and mating functionals is described in Sections 5 and 6.
For the point …x; y† 2 M which references CF ……x; y†; u†, more than one mating for the con®guration
framework may occur if there is a range of angular values u ˆ ‰a; bŠ that satisfy the mating condition. The
con®guration functional is used to quantify the amount of ``turning'' the framework will allow (while
satisfying the mating condition).

De®nition (Configuration functional). The con®guration functional fC …x; y† is a continuous function

fC : M ! R ‡
de®ned in terms of the con®guration framework, overlay gage GO ˆ …s; r; u†, and feature vector …F; n; a† as
follows:
Z
1
fC …x; y† ˆ v……x; y†; u† du;
2p
where the indicator variable

1 if CF ……x; y†; u† mates with GO at …x; y† 2 R2 and u 2 S 1 ;
v……x; y†; u† ˆ
0 otherwise;

and the 1=2p factor is used to normalize the integral value.

A measure of ®t between the design tolerance vector s ˆ …s1 ; . . . ; sn † and the feature set
F ˆ …F1 ; . . . ; Fn † of the artefact is de®ned in terms of the distances between the nominal location of the
tolerance elements si (i.e., ui ) and actual location of the feature elements Fi (i.e., ai ). The minimum value of
the sum of these distances d…ui ; ai † is determined as a function of u, and then the distance functional
X 2
min d…ui ; ai †
u2S 1

is minimized over the plane. This leads to the notion of the mating functional.

De®nition (Mating functional). The mating functional fc …x; y† is a continuous function

fc : M ! R‡
164 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

de®ned in terms of the con®guration framework, overlay gage GO ˆ …s; r; u†, and feature vector …F; n; a† as
follows:
fc …x; y† ˆ min c…s; F†;
u2S 1

where
!1=p
X
n
c…s; F† ˆ c…u; a† ˆ dim
iˆ1

and
q
2 2
di ˆ d…ui ; ai † ˆ d……ai ; bi †; …xi ; yi †† ˆ …ai ÿ xi † ‡ …bi ÿ yi † :

3.4. Optimal mating and optimal con®guration space

Mating functionals induce a natural optimization problem that seeks to ®nd the position and orientation
of CF ……x; y†; u† to minimize the ®t measure c…s; F† over M. An optimal mating can be de®ned in a variety
of ways, and depending upon the application di€erent measures may be employed. The only requirement of
the measure is that it be well-de®ned and computationally tractable.

De®nition (Optimal mating). A mating m is said to be optimal if it satis®es minM fc …x; y†.

De®nition (Optimal mating region). The optimal mating region M  M is the set of optimal matings:
 
M ˆ …x ; y  † j …x ; y  † satisfies min fc …x; y† :
M

Optimal matings also induce an optimal con®guration space X . For a given orientation u of the con-
®guration framework, the solution to
min c…s; F†
M

yields an optimal subset Xu  Xu  M which may be composed of one or more position vectors. The
collection of these subsets con®gured as the Cartesian product determines the optimal con®guration space
X :
X ˆ Xu  S 1 :

De®nition (Optimal configuration space). The optimal con®guration space X  X is the collection of
optimal mating vectors m ˆ ……x ; y  †; u †, i.e.,
 
  
X ˆ m j m satisfies min fc …x; y† :
M

The projection of the optimal con®guration space onto the base xy-plane determines the optimal mating
region M ˆ …X †? .

4. Induced con®guration functionals and associated parameters

The con®guration framework CF ……x; y†; u† induces functionals which realize the structure of positional
tolerance. The framework CF ……x; y†; u† is a function of position …x; y† and orientation u 2 S 1 , and can be
decomposed according to the interaction of these variables as follows:
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 165

I. framework translation without rotation,

II. framework rotation without translation, and
III. framework translation and rotation.
Each case induces a speci®c con®guration functional and associated parameter class.
Case I: Framework translation without rotation. If the orientation of the con®guration framework
CF ……x; y†; u† is held ®xed at u 2 S 1 while it is translated over the plane, then the region Xu where mating
occurs forms a subset of Xu , Xu  Xu . Framework translation without rotation induces a ``position''
con®guration functional v……x; y†; u† de®ned by

1 if CF ……x; y†; u† mates with GO ;
v……x; y†; u† ˆ
0 otherwise:

The functional v……x; y†; u† is stepwise constant and the region where v……x; y†; u† ˆ 1 determines Xu:
Xu ˆ f…x; y† j v……x; y†; u† ˆ 1g:
Xu represents the region in the plane where mating occurs for the ®xed orientation u, and there is a unique
correspondence between the orientation u 2 S 1 and the region Xu:
u 2 S 1 $ Xu  M:
The size of the mating region for a given u 2 S 1 is expressed by the Lebesgue measure A…Xu†:
Z Z
1
A…Xu† ˆ v……x; y†; u† dx dy;
A…F0 †
where division by the area of F0 , A…F0 †, serves as a normalization factor.
Case II: Framework rotation without translation. The reference position of the con®guration framework
CF ……x; y†; u† is held ®xed at …x; y†, and the framework is twisted to determine the amount of angular ro-
tation such that the condition of mating is satis®ed. Framework rotation without translation induces an
``angular orientation'' con®guration functional v……x; y†; u† de®ned by

1 if CF ……x; y†; u† mates with GO ;
v……x; y†; u† ˆ
0 otherwise:

The orientation con®guration functional measures the allowable amount of rotation of the framework at a
given point in the plane. The amount of ``turning'' that the framework can undergo while satisfying the
mating condition is determined by the functional value
Z
1
fC …x; y† ˆ v……x; y†; u† du;
2p

where in this case division by 2p serves as a normalization factor.

Case III: Framework translation and rotation. Case III represents the general case. If the con®guration
framework is allowed to vary in orientation and position (or in other words, if the indicator variable
v……x; y†; u† as de®ned in Case I is considered a function of orientation, or the indicator variable v……x; y†; u† as
de®ned in Case II is considered a function of position), then the con®guration functional is determined. As
previously de®ned,

1 if CF ……x; y†; u† mates with GO ;
v……x; y†; u† ˆ
0 otherwise:

The functional v……x; y†; u† has value one if the con®guration framework CF ……x; y†; u† mates with GO at the
point …x; y† and orientation u 2 S 1 .
The mating region M and con®guration space X are de®ned in terms of this functional as follows:

M ˆ f…x; y† j v……x; y†; u† ˆ 1 for any u 2 S 1 g;

166 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

X ˆ f…x; y; u† j v……x; y†; u† ˆ 1; u 2 S 1 g:

It is clear that M  X, and the area of the mating region, A…M†, and volume of the con®guration space,
V …X†, indicate the relative sizes of the induced space.
The con®guration functional indicates the amount of turning the con®guration framework can undergo
while satisfying the mating condition and is de®ned by
Z
1
fC …x; y† ˆ v……x; y†; u† du:
2p
The con®guration functional fC …x; y† is zero at all points outside the mating region; i.e., fC …x; y† ˆ 0 if
…x; y† 62 M.
An ``extremal'' position …x ; y  † in the plane is de®ned as the solution to
max fC …x; y†
M

f
and the collection of points that satisfy this condition forms the extremal set M:
 
f ˆ …x ; y  † j …x ; y  † satisfies max fC …x; y†  M:
M
M

Level sections of the function fC …x; y† create equipotential contours of the con®guration functional which
map onto the region M. The average value of the con®guration functional as a function of position over the
mating region determines the average amount the framework can turn:
Z Z
1
hfC …x; y†i ˆ fC …x; y† dx dy
A…M†
and the variance of the con®guration functional indicates the extent of the variation with respect to the
mean:

V …fC …x; y†† ˆ hfC …x; y†2 i ÿ hfC …x; y†i2 :

An ``extremal'' orientation u is de®ned as the solution to
max A…Xu †
u2S 1

Fig. 13. The con®guration framework CF ……1:58; 1:67†; u† for the tolerance design …s; r; u†, where s ˆ …C1 ; C2 ; C3 ; C4 ; C5 ; C6 †,
r ˆ …0:7; 1:2; 2:0; 0:5; 1:7; 0:7†, and u ˆ ……5; 1†; …3; ÿ2†; …ÿ5; ÿ2†; …ÿ4; 3†; …5; 7†; …5:5; 3††.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 167

Fig. 14. Evaluation of the con®guration functional fC …x; y† at two points in the plane. The valid angular range shown is equivalent to
the amount of angular twisting the con®guration framework CF ……x; y†; u† can undergo while satisfying the mating condition.

Fig. 15. A birds-eye view of the con®guration functional fC …x; y†.

168 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

Fig. 16. The con®guration framework C……1:53; 1:72†; u† corresponding to the maximum value of the con®guration functional.

Fig. 17. The con®guration function fC …x; y† depicted with and without rendering. The mating region M corresponds to the region
M ˆ f…x; y† j fC …x; y† > 0g.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 169

e:
and the collection of angles that satisfy this condition forms the extremal set U
 
Ue ˆ u j u satisfies max A…Xu †  U :
u2S 1

The collection of orientations u 2 S 1 such that mating does not occur for any position forms the set
N …u† ˆ fujA…Xu † ˆ 0g;
which represent the null space of orientations ± orientations which do not admit any location for mating.
The complement of this set is described by
U …u† ˆ fu j A…Xu † 6ˆ 0g
and represent the orientation set where mating occurs. It is clear that the one-dimensional ``size'' of these
sets ± the length L…
† ± are such that
L…N …u†† ‡ L…U …u†† ˆ 2p:
The moments of Xu are determined in the usual manner. The ``average'' size of the region X is determined
from

Fig. 18. The mating functional fc …x; y† for the positional tolerance problem de®ned by cp for p ˆ 1 and …s; r; u†, where
s ˆ …C1 ; C2 ; C3 ; C4 ; C5 ; C6 †, r ˆ …0:7; 1:2; 2; 0:5; 1:7; 0:7†, and u ˆ ……5; 1†; …3; ÿ2†; …ÿ5; ÿ2†; …ÿ4; 3†; …5; 7†; …5:5; 3††. The functional
fc …x; y† is depicted with and without rendering.
170 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

Z
hA…Xu †i ˆ A…Xu† du

and the variance of this measure is computed from

2 2
V …A…Xu †† ˆ hA…Xu † i ÿ hA…Xu †i :

5. Construction of the con®guration functional

The con®guration functional depends directly upon the structure and geometric relationships between
the feature and tolerance vector. The functional is suciently complicated, however, such that it cannot be
computed in closed form, and to illustrate the form of the functional it must be constructed.
A procedure to construct the con®guration functional is outlined as follows.

Procedure
Input: Ai ˆ …xi ; yi †, ki ˆ …ai ; bi †, si ˆ si ……xi ; yi †; ri †, i ˆ 1; . . . ; n; / 2 S 1 ; I ˆ …x0 ; y0 †.
Output: fC …x; y†.

Fig. 19. A birds-eye view of the mating functional for p ˆ 1 and the optimal location and orientation of the con®guration framework.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 171

1. Construct the con®guration framework CF ……x0 ; y0 †; u† with respect to point I.

2. Check the ``®t'' of the con®guration framework with the tolerance vector s ˆ …s1 ; . . . ; sn † located at
k ˆ …k1 ; . . . ; kn † of size r ˆ …r1 ; . . . ; rn †.
2.1. If Ai ˆ …xi ; yi † 2 si , then
2 2
…xi ÿ ai † ‡ …yi ÿ bi † 6 r2i …1†
and if condition (1) holds for all i, i ˆ 1; . . . ; n, then store the value of v……x; y†; u† ˆ 1 for orientation
u.
2.2. Rotate the framework CF ……x; y†; u† about point I by Du ˆ /: u ˆ u ‡ Du.
2.3. Update the coordinate points of Ai ˆ …xi ; yi † after rotation using the transformation
x0i ˆ xi cos / ‡ yi sin /; yi0 ˆ yi cos / ÿ xi sin /;
and go to Step 2.1.
2.4. Continue for 0 6 u 6 2p.
3. Construct the con®guration functional fC …x0 ; y0 † at …x0 ; y0 †:
Z
1
fC …x0 ; y0 † ˆ v……x0 ; y0 †; u† du:
2p

Fig. 20. The mating functional fc …x; y† for the positional tolerance problem de®ned by cp for p ˆ 2 and …s; r; u†. The functional fc …x; y†
is depicted with and without rendering.
172 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

4. The con®guration functional fC …x; y† is determined by repeating Steps 2 and 3 and computing the orien-
tation functional over a grid de®ning F0 .
The blueprint B is de®ned by f…U; R; k†; …s; r; u†g and the manufactured artefact A is realized by the
feature vector …F; n; a†. The design form U, size R, and location vector k are prescribed by the design
blueprints along with the tolerance form s, size r, and location vector u.
In a typical design blueprint u ˆ k which connects the design vector with the tolerance vector. The input
parameters in the following example include the tolerance vector and feature vector. The output of the
construction is the con®guration functional.

Example. The design tolerance form …s; r; u† is described by

s ˆ …C1 ; C2 ; C3 ; C4 ; C5 ; C6 †;
r ˆ …0:7; 1:2; 2:0; 0:5; 1:7; 0:7†;
u ˆ ……5; 1†; …3; ÿ2†; …ÿ5; ÿ2†; …ÿ4; 3†; …5; 7†; …5:5; 3††

and is illustrated in Fig. 13.

The feature vector …F; n; a† is dimensionless (n ˆ 0) and located at the position a:
a ˆ u ˆ ……5; 1†; …3; ÿ2†; …ÿ5; ÿ2†; …ÿ4; 3†; …5; 7†; …5:5; 3††;

Fig. 21. A birds-eye view of the mating functional for p ˆ 2 and the optimal location and orientation of the con®guration framework.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 173

i.e., the feature location vector a is assumed to coincide with the tolerance location vector u (and so the
feature is initially in positional tolerance).
The con®guration framework CF ……x0 ; y0 †; u† for I ˆ …x0 ; y0 † ˆ …1:58; 1:67† is shown in Fig. 14, and the
amount of angular twist that allows mating is computed for two locations in the plane as shown. The valid
angular range is equivalent to the amount of angular twisting the con®guration framework can undergo
while satisfying the mating condition. The amount of angular twist for the positions shown in 4.5 and 7.5
radians.
A birds-eye view of the con®guration functional is depicted in Fig. 15, and the maximum value of the
con®guration framework is determined to reside at (1.53, 1.72) as shown in Fig. 16.
The con®guration functional is then constructed in Fig. 17 with and without rendering for the grid size
0.15 and rotation angle Du ˆ / ˆ 0:05°.

6. Construction of the mating functional

A con®guration functional measures the amount of angular rotation the con®guration framework can
undergo, while a mating functional is used to determine the ``optimal'' location for mating. Thus, while
con®guration functionals and con®guration spaces are interesting constructs, it is the mating functional,

Fig. 22. The mating functional fc …x; y† for the positional tolerance problem de®ned by cp for p ˆ 1 and …s; r; u†. The functional
fc …x; y† is depicted with and without rendering.
174 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

mating region, and optimal mating region that provide the analytic foundation for the geometric structure
of positional tolerance evaluation.
A mating functional de®nes a global measure of ®t between the feature set F ˆ …F1 ; . . . ; Fn † and the
nominal design tolerance s ˆ …s1 ; . . . ; sn †. A mating functional is denoted fc …x; y† and induces optimal
mating con®gurations over the mating region M. The mating functional is simply the distance measure
between the point feature Fi and tolerance form element si over the elements of the vector. The tolerance
form is assumed ®xed and the feature element is determined by ai ˆ …xi ; yi †, and so
q
2 2
di ˆ d…si ; Fi † ˆ d……xi ; yi †; …ai ; bi †† ˆ …xi ÿ ai † ‡ …yi ÿ bi † :

A class of mating functionals can be derived from the distance measure di and include
!1=p
X
n
p
c …s; F† ˆ dip ; p ˆ 1; 2; . . . ;
iˆ1

c1 …s; F† ˆ max di ; c0 …s; F† ˆ min di :

16i6n 16i6n

Fig. 23. A birds-eye view of the mating functional for p ˆ 1 and the optimal location and orientation of the con®guration framework.
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 175

The mating functional is determined by computing the minimum of cp as a function of orientation:

fc …x; y† ˆ min c…s; F†:
u2S 1

The con®guration functional fC …x; y† is de®ned over the mating region M and measures the amount the
framework CF can twist (an angular measure) while maintaining ®t within the tolerance zones. The mating
functional fc …x; y† on the other hand determines the location and orientation of the optimal ®t between the
feature and tolerance vector (in the sense described by c…s; F†). By minimizing the mating functional over
M, a global measure of best ®t (optimal con®guration location and orientation) is determined.
The construction of the mating functional closely follows the development of the con®guration func-
tional.

Procedure
Input: Ai ˆ …xi ; yi †, ki ˆ …ai ; bi †, si ˆ si ……xi ; yi †; ri †, i ˆ 1; . . . ; n; / 2 S 1 ; I ˆ …x0 ; y0 †.
Output: fC …x; y†.
1. Construct the con®guration framework CF ……x0 ; y0 †; u† with respect to point I.
2. Compute the distance functional

Fig. 24. The mating functional fc …x; y† for the positional tolerance problem de®ned by cp for p ˆ 0 and …s; r; u†. The functional fc …x; y†
is depicted with and without rendering.
176 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

cp …s; F†; p ˆ 0; 1; 2; . . . ; 1
as a function of u 2 S 1 .
3. Determine min cp …s; F†.
u2S 1
4. The mating functional fc …x; y† is determined by repeating Steps 2 and 3 for the points of a grid de®ning
M.
fc …x; y† ˆ min cp …s; F†:
u2S 1

Fig. 25. A birds-eye view of the mating functional for p ˆ 0 and the optimal location and orientation of the con®guration framework.

Table 1
Optimal mating con®guration parameters

p c…X† …x; y†

0 0.477 …1:033; 0:478†

1 3.535 …1:783; 2:217†
2 2.35 …1:933; 2:167†
1 0.884 …1:383; 1:867†
 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178 177

Example (continued). The mating functional fc …x; y† for cp for p ˆ 1 is depicted in Fig. 18 with and without
rendering, and the optimal location and orientation of the con®guration framework is shown in Fig. 19.
The mating functionals for p ˆ 2, p ˆ 1, and p ˆ 0 and optimal con®guration framework are depicted in
Figs. 20±25, respectively. Notice the shift of the con®guration framework and the variable structure of the
functional as the value of p changes. Table 1 summarizes the value of the mating functional and the optimal
locations.

7. Conclusions

A general geometric framework for positional tolerance evaluation is developed. The framework is
created by introducing a set of fundamental relations and then understanding the structure of the relations
through functional entities.
The con®guration framework of a measurement data set allows the geometric structure inherent to the
positional tolerance problem to be realized. The con®guration functional determines the amount of rota-
tion the con®guration framework can undergo while satisfying the condition of mating. The con®guration
functional fC is associated with feature position evaluation and uniquely quanti®es and describes the po-
sitional tolerance aspects of the manufactured artefact. The structure of the con®guration functional is
complicated and is generally not expected to admit a unique extremal point. Mating functionals quantify
the ``optimal'' locations and orientation for mating, and provide measures that can be used for quality
control purposes.

Acknowledgements

This work was supported in part by the National Science Foundation under grant DMI-9713707,
Wichita State University under grant URCA, GM-3753, and a State/Industry/University ADMRC grant.

References

[1] H.R. Berenji, B. Khoshnevis, Use of arti®cial intelligence in automated process planning, Computers in Mechanical Engineering
(1986) 47±55.
[2] L.C.R. Carpinetti, D.G. Chetwynd, Geometric search methods for assessing geometric tolerances, Computer Methods in Applied
Mechanics Engineering 122 (1995) 193±204.
[3] C.M. Creveling, Tolerance Design, Addison-Wesley, Reading, MA, 1997.
[4] C.H. Fortin, J.F. Chatelain, A soft gaging approach for complex cases including datum shift analysis of geometrical tolerances, in:
Geometric Tolerancing, Scienti®c Press, Singapore, pp. 312±327.
[5] L.W. Foster, Geo-Metrics IIIm, Addison-Wesley, Reading, MA, 1993.
[6] G.K. Grith, Measuring and Gaging Geometric Tolerances, Prentice-Hall, Englewood Cli€s, NJ, 1994.
[7] C.C. Hayes, A manufacturing planner that reasons about ecient plans to achieve positional tolerance speci®cations, Concurrent
Product and Process Engineering 85 (1995) 279±288.
[8] C.C. Hayes, P.K. Wright, Automated process planning: Using feature interaction to guide search, Journal of Manufacturing
System 8 (1) (1989) 1±16.
[9] M.J. Kaiser, Geometric metrology I. Separation functionals in tolerance assessment, Journal of the Franklin Institute:
Engineering and Applied Mathematics B 335 (1) (1998) 129±162.
[10] E.A. Lehtihet, N.U. Gunasena, On the composite position tolerance for patterns of holes, Ann. CIRP (1991).
[11] E.A. Lehtihet, N.U. Gunasena, Models for the position and size tolerance of a single hole, in: Proceedings of the ASME Winter
Annual Conference on Manufacturing Metrology Symposium, 1988.
[12] M. Inui, M. Miura, Con®guration space based analysis of position uncertainties of parts in an assembly, in: Geometric
Tolerancing, Scienti®c Press, Singapore, 1997, pp. 132±146.
[13] M. Inui, M. Miura, F. Kimura, Relative positioning of assembled parts with small geometric deviations by using hierarchically
approximated con®guration space, in: Proceedings of the IEEE Conference on Robotics Automation, 1995, pp. 1605±1612.
[14] M. Inui M. Miura, F. Kimura, Analysis of position uncertainties of parts in an assembly using con®guration spaces in octree
representation, in: Proceedings of the ACM Symposium on Solid Modeling and Applications, 1995, pp. 73±81.
[15] A.A.G. Requicha, Toward a theory of geometric tolerancing, International Journal of Robotics Research 2 (4) (1989) 45±60.
178 M.J. Kaiser et al. / Comput. Methods Appl. Mech. Engrg. 190 (2000) 149±178

[16] R. Sodhi, J.U. Turner, Relative positioning of variational part models for design analysis, Computer-Aided Design 26 (5) (1994)
366±3786.
[17] R.H. Wilson, J.C. Latombe, Geometric reasoning about mechanical assembly, Arti®cial Intelligence 71 (2) (1994) 371±396.
[18] J. Xiao, Automatic determination of topological contacts in the presence of sensing uncertainties, in: Proceedings of the IEEE
Conference on Robotics and Automation, 1993, pp. 65±70.