Including Geometric Feature Variations in Tolerance Analysis of

Mechanical Assemblies
Kenneth W. Chase
Jinsong Gao
Spencer P. Magleby
Carl D. Sorensen
Mechanical Engineering Department
Brigham Young University

Abstract
Geometric feature variations are the result of variations in the shape, orientation or
location of part features as defined in ANSI Y14.5M-1982 tolerance standard [ANSI
1982]. When such feature variations occur on the mating surfaces between components
of an assembly, they affect the variation of the completed assembly. The geometric
feature variations accumulate statistically and propagate kinematically in a similar manner
to the dimensional variations of the components in the assembly.
The Direct Linearization Method (DLM) for assembly tolerance analysis provides a
method for estimating variations and assembly rejects, caused by the dimensional
variations of the components in an assembly. So far, no generalized approach has been
developed to include all geometric feature variations in a computer-aided tolerance
analysis system.
This paper introduces a new, generalized approach for including all the geometric feature
variations in the tolerance analysis of mechanical assemblies. It focuses on how to
characterize geometric feature variations in vector-loop-based assembly tolerance models.
The characterization will be used to help combine the effects of all variations within an
assembly in order to predict assembly rejects using the DLM.

1. Introduction
Tolerance analysis of mechanical assemblies is an essential step in the design and
manufacturing of high quality products. The appropriate allocation of tolerances among
the separate parts in an assembly can result in lower costs per assembly and higher
probability of fit, reducing the number of rejects or the amount of rework required on
components. Analyzing the cumulative effects of component tolerances on critical
clearances or fits in the assembly is necessary to guarantee that the product will function
properly.
Unfortunately, tolerance analysis generally involves complex and tedious calculations
which are time-consuming and prone to error. In an attempt to solve this problem, efforts
are being made to automate the tolerance analysis and allocation process as much as
possible. With the increasing use of CAD/CAM/CAE systems for mechanical design, it is
desirable to use a CAD model of the assembly to perform the tolerance analysis at the
design stage.
The authors previous paper [Chase, Gao & Magleby 1994] developed a comprehensive
system which is capable of performing assembly tolerance analysis by the Direct
Linearization Method (DLM), using a vector-loop-based assembly tolerance model. It is a
generalized approach employing vector loops, kinematic joints and component tolerances.
However, in the previous paper, only component dimensional tolerances were included in
evaluating the assembly variations and predicting assembly rejects. The geometric feature
tolerances of components in the assembly were not addressed.
Besides dimensional variations, the fluctuation in manufacturing conditions can also cause
geometric feature variations, such as the variation of the form of a feature as compared to
perfect form, for example. The geometric feature variations of a part can affect the
position and orientation of mating parts, and therefore, have the possibility of greatly
affecting the final assembly due to the accumulation of individual geometric feature
variations. Since this variation is inevitable in manufacturing, it must be carefully
controlled in order to produce assemblies which function properly.
The two-component assembly shown in Figure 1 demonstrates the relationship between
dimensional variations in an assembly and the small kinematic adjustments which occur at
assembly time. The assembly has three component dimensions that vary, two on the
tapered groove and one on the cylinder, as shown. The variations in the three dimensions

3
have an effect on the distance U. U is important to the function of the assembly and will
be referred to as an assembly resultant.
The parts are assembled by inserting the cylinder into the groove until it makes contact on
the two mating surfaces. For each set of parts, the distance U will adjust to accommodate
the current value of dimensions A, R, and . The assembly resultant U1 represents the
nominal position of the cylinder, while U2 represents the position of the cylinder when the
variations are present. This adjustability of the assembly describes a kinematic constraint,
or a closure constraint on the assembly.

R + R

U1
U2
Figure 1. Kinematic adjustment due to component dimensional variations
Figure 2 illustrates the same assembly with exaggerated geometric feature variations. For
production parts, the contact surfaces are not really flat and the cylinder is not perfectly
round. The pattern of surface waviness will differ from one part to the next. In this

U
Figure 2. Adjustment due to geometric shape variations

4
assembly, the cylinder makes contact on a peak of the lower contact surface, while the
next assembly may make contact in a valley. Similarly, the lower surface is in contact with
a lobe of the cylinder, while the next assembly may make contact between lobes.
Local surface variations such as these can propagate through an assembly and accumulate
just as dimensional variations. Thus, in a complete assembly model all three sources of
variation, that is, dimensional and geometric feature variations and kinematic adjustments,
must be accounted for to assure realistic and accurate results.
The objective of the research described in this paper is to include geometric feature
variations of components in a vector-loop-based assembly tolerance model. The general
approach includes the characterization of geometric feature variation in 2-D and 3-D
assemblies, and using the DLM to calculate the variation of assembly dimensions or
kinematic variables caused by geometric feature tolerances.

2. Definitions
The geometric feature tolerances defined by ANSI Y14.5M-1982 fall into five main
groups, according to Foster [1992]:
1. FORM

A form tolerance states how far an actual surface or feature is

permitted to vary from the desired form implied by the drawing. It includes
flatness, straightness, circularity and cylindricity.
2. PROFILE A profile tolerance states how far an actual surface or feature is
permitted to vary from the desired form on the drawing and/or vary relative to a
datum or datums. Profile of a line and profile of a surface are the only two
types of profile tolerance.
3. ORIENTATION An orientation tolerance states how far an actual surface or
feature is permitted to vary relative to a datum or datums. It consists of
perpendicularity, angularity and parallelism.
4. LOCATION
A location tolerance states how far an actual size feature is
permitted to vary from the perfect location implied by the drawing as related to
a datum, or datums, or other features. This category includes position and
concentricity.

5
5. RUNOUT

A runout tolerance states how far an actual surface or feature is

permitted to vary from the desired form implied by the drawing during full
(360) rotation of the part on a datum axis. A runout can be either a circular
runout or a total runout.
Figure 3 illustrates the symbols which represent the geometric feature controls defined in
the ANSI standard.

Flatness

Straightness Straightness
Circularity
(of an axis) (surface element) (roundness)

Perpendicularity

Angularity

Runout
(circular)

Runout
(total)

Parallelism

Profile
(of a line)

Concentricity

Cylindricity
Profile
(of a surface)

True
Position

Figure 3. Symbols for geometric feature controls [ANSI Y 14.5M-1982].

Geometric feature controls allow the designers to specify limits on the form or orientation
of a feature on a part, which are not available through the use of size tolerances alone.
Foster [1986] listed the conditions under which it is appropriate to use geometric feature
tolerancing:
1. Whenever part features are critical to function or interchangeability.
2. Whenever functional gaging techniques are desirable.
3. Whenever datum reference frames are desirable in order to ensure consistency
between manufacturing and gaging operations.
4. Whenever computerization techniques in design and manufacturing are desirable.
5. Whenever the standard interpretation or tolerance is not already implied.

3. Research Review
There are two basic issues which must be resolved when including geometric feature
variations in assembly tolerance analysis. First, how to represent or characterize the
geometric feature variations in an assembly tolerance model, and second, how to evaluate

6
the effects of the geometric feature variations when estimating assembly variation and
predicting the percentage of assemblies which will fall outside of the design specifications.
3.1 Tolerance Representation
Schemes for tolerance representations or characterizations in an assembly have been
developed with the increasing use of solid modeling tools in product design. These
schemes can be generally classified into three groups:
1. Set theoretic model
2. Offset zones
3. Parametric zones
The set theoretic model of tolerances describes a variational class of objects (or parts)
which is defined by the tolerances applied to the nominal object. This variational class is
modeled as a set of points in 3-D space, which contains the nominal object but does not
force any part of the objects real boundary to be in an exact position [Shah & Miller
1990]. The set theoretic model has not been implemented because it is difficult to
mathematically describe objects in terms of its theory [Robison 1989].
Offset zones are created by offsetting the nominal boundary of a part by an amount equal
to the tolerance on either side of the nominal [Requicha 1983]. Offsets are obtained for
the maximum material condition (MMC) and for the least material condition (LMC). The
difference between these two zones comprises the tolerance zone, an envelope within
which the boundary of the part must lie [Shah & Miller 1990]. This method seems to lend
itself to the use of go-nogo gages to check the tolerance condition of a part. A
disadvantage of this method is that it assumes that all surfaces remain in the same
orientation as the nominal surface. It has not be used to model variations in the
orientation of a surface, such as angularity [Robison 1989].
A parametric zone or space is composed of a set of parameters or dimensions and
constraints which describe the nominal shape of the geometry [Hillyard & Braid 1978,
Martino & Gabriele 1989]. Tolerances are treated as small variations in these parameters.
This type of tolerance model is closely related to the variational geometry approach for
CAD modelers [Shah & Miller 1990, Guilford & Turner 1993]. The advantage of this
model is that it uses the constraints and parameters of the geometry to create a set of
equations which may be solved to determine any unknown dimensions or variations
[Gupta & Turner 1993].

7
A combination of parametric zone and offset zone for representing tolerances in an
assembly has been recently proposed by Gilbert [1992]. They use the 4x4 homogeneous
transformation matrix to contain the nominal relations between parts and variations
allowed by the tolerances in an assembly tolerance model. Most geometric feature
variations, except for form tolerances, can be represented by this method.
The assembly tolerance analysis model adopted by this paper is a parametric zone type,
proposed by Robison [1989]. It is composed of a vector-based method for modeling 3-D
mechanical assemblies, which utilizes vectors to represent dimensions between critical part
features and includes a set of kinematic joint types to represent mating conditions between
parts at the contact locations [Chase, Gao & Magleby 1994]. This method also includes
guidelines for identifying a valid set of vector loops to ensure that the tolerance model is
complete. This method lays the vector-loop-based assembly tolerance model over the
solid model, and can be connected with a tolerance analysis package to solve for the
variations on the desired dimensions or clearances. This model is also capable of including
component geometric feature tolerances.
3.2 Tolerance Analysis
After specifying the variation of the individual components in an assembly, the
propagation of the variations in the assembly must be determined. This is frequently
referred as tolerance analysis.
The results of a tolerance analysis are estimates of the mean, variance and other statistical
parameters describing the variation of critical assembly features. If engineering design
limits have been specified for a feature, quantitative estimates of the percent rejects may be
made and compared to desired quality levels.
Five methods have been employed for 2-D and 3-D tolerance analysis: 1) Linearized or
Root Sum Square, 2) Method of System Moments, 3) Hasofer-Lind Reliability Index, 4)
Direct Integration and 5) Monte Carlo Simulation. These have been discussed in previous
papers [Chase & Parkinson 1991, Chase, Gao & Magleby 1994]. All five methods have
been primarily limited to problems for which the assembly resultant of interest may be
expressed explicitly as a function of the dimensions of the components in an assembly.
Establishing such an explicit function for a real assembly is very difficult or impossible for
the designer. In addition, the emphasis of these methods is the propagation of dimensional
variations rather than geometric feature variations in an assembly.

8
Gilbert [1992] and Whitney, et. al [1994] proposed a system to estimate the propagation
effect of the dimensional and geometric feature tolerances of components in an open loop
assembly using the linearized model of the 4x4 transformation matrix constraint developed
by Veitschegger and Wu [1986]. He studied most geometric feature tolerances, except
for form variations, and characterized the variation types of these geometric feature
variations for some mating conditions between the parts.
Lin and Chen [1994] developed a linearized scheme for tolerance analysis of closed loop
mechanisms in 3-D space using the 4x4 transformation matrix in the Denavit-Hartenberg
(D-H) symbolic notation common to robotics. This method is significant since it has the
capability of assessing the influence of each error component on a machines accuracy, and
it is also capable of including geometric feature variations. It is particularly useful to
allocate a machines tolerances at its design stage and to characterize the performance of
an existing machine. Thus far, the method has not been applied to static assemblies. It
has only been applied to robotic mechanisms with revolute and prismatic joints (see joint
classifications in a later section).
The authors previous paper [Chase, Gao & Magleby 1994] developed the Direct
Linearization Method for treating implicit assembly functions, such as kinematic
constraints, for assembly tolerance analysis in 2-D space. It is a generalized approach,
employing common engineering concepts of vector loops, kinematic joints and geometric
feature tolerances.
The DLM reduces the implicit equations describing the kinematic constraints to a set of
linear equations for small changes about the nominal, which may be solved directly for the
assembly variances. Assembly geometry is reduced dramatically, so it is very efficient
computationally. It is suited for integration with commercial CAD systems, by which the
required geometry may be extracted directly from a solid model. This paper extends the
DLM method to include all geometric feature variations in an assembly.
A comparison of the accuracy of linearized tolerance analysis with Monte Carlo simulation
was performed earlier [Gao, et. al 1995, Gao 1993]. Several assembly problems, covering
a variety of critical assembly features, were analyzed by both methods and the results were
compared. The mean, standard deviation and percent rejects were computed for each
case. The results of the study demonstrated that the linearized method has an accuracy
equal to Monte Carlo. While the linearized method requires a sample size of one, it was
shown to be equivalent to a Monte Carlo analysis with a sample size of 10,000.

4. The Direct Linearization Method

The assembly constraint equations should be able to describe the small kinematic
adjustments of the assembly resultants or assembly variables occurring at assembly time,
due to manufacturing variations on the component dimensions. In an assembly, the
kinematic constraint may appear as a closed loop, that is, the starting and ending point of
the vector loop for the assembly is the same point, or as an open loop, that is, the
assembly vector loop ends with a gap or clearance. This constraint with vector loop-based
assembly models may be expressed as a concatenation of homogeneous transformation
matrices:
[R1][T1][R2][T2]...[Ri][Ti]...[Rn][Tn][Rf] = [H]

(1)

where [Ri] is the rotational transformation matrix or product of rotational transformation

matrices and [Ti] is the translational matrix at node i, [Rf] is the final closure rotation, and
[H] is the resultant matrix. If the assembly is described by closed loop constraints, [H] =
[I], the identity matrix, otherwise [H] represents the final gap or clearance and its
orientation.
For 2-D assemblies:
cos z
R = sin z

sin z
cos z
0

0
0

1 0 T x
T = 0 1 T y

0 0 1

(2)

For 3-D assemblies:

0
1
0 cos
x
Rx =
0 sin x

0
0
cos y
0
Ry =
sin y

0
sin x
cos x
0

0
0

0 sin y
1
0
0 cos y
0
0

0
0

1
0
T=
0

0
1
0
0

0 Tx
0 Ty

1 Tz

0 1

(3)

10
cos z
sin
z
Rz =
0

sin z
cos z
0
0

0
0
1
0

0
0

where x, y and z are relative rotations about their corresponding axes and Tx, Ty and Tz
are components of the translation vector from one node in the loop to the next.
Equations (1) are nonlinear and require nonlinear solver to find solutions. However, for
small variation about the nominal of each component in the assembly, the equations can be
approximated by the linearized Taylor expansion of equation (1). Therefore, only the
derivatives are needed.
Robison [1989] used a perturbation method for evaluating the derivatives of the assembly
constraint equation (1). For a 3-D case, if the translation or rotation at joint i is the
variable with respect to which the derivatives are desired, a small length perturbation L
or angle perturbation is added to the original value, then the matrix multiplication of
equation (1) is performed:
for translation
(4)

[R1][T1]...[Ri][Ti(L+L)]...[Rn][Tn]{0 0 0 1}t = {X Y Z 1}t

for rotation
[R1][T1]...[Ri( + )][Ti]...[Rn][Tn]{0 0 0 1}t = {X Y Z 1}t

the derivatives can then be approximated numerically by:

Translational variable

Rotational variable

H x X

L
L

H x X

H y

H Y X

Y
L

(5)

11
H z Z

L
L

H Z X

H x

=0

=0

=0

L
H

L
H

H Y

H Z

(6)

= COS ( )

= COS ( )

= COS ( )

where Hx, Hy and Hz are the translational constraint, and Hx, Hy and Hz are the
rotational constraint in x, y and z directions respectively, and , and are the global
direction cosine angles of the local axis around which the rotation is made [Gao 1993].
The DLM uses the linearized Taylor expansion of equation (1) to solve for the assembly
variations. The linearized constraint equations for the closed loops in an assembly may be
expressed in matrix form:
{H} = [A]{X} + [B]{U} = {0}

(7)

where {H}: the variations of the clearance,

{X}: the variations of the manufactured variables,
{U}: the variations of the assembly variables,
[A]: the partial derivatives with respect to the manufactured variables,
[B]: the partial derivatives with respect to the assembly variables,
If the [B] is a full-ranked matrix, the variations of the assembly or kinematic variables can
be obtained by solving equation (7).
{U} = -[B]-1[A]{X}

(8)

If the [B] matrix is singular, the least square fit solution is

{U} = -([B]T[B])-1[B]T[A]{X}

(9)

If the assembly has open loop constraints, the first order Taylor series expansion of the
assembly constraints is:

12
{p} = [C]{X} + [D]{U}

(10)

where {p}: the variations of the open loop assembly variables,

[C]: the partial derivatives with respect to the manufactured variables in the
open loop,
[D]: the partial derivatives with respect to the assembly variables in the open loop,
If one substitutes {U}from equation (8) or (9) into equation (10), then the variations of
the open loop variables can be expressed as:
{p} = ([C] - [D][B]-1[A]){X}

(11)

If [B] is singular,
{p} = ([C]-[D]([B]T[B])-1[B]T[A]){X}

(12)

Equations (8) through (12), above, may be modified to obtain estimates of the tolerance
accumulation in the assembly by worst case or statistical methods. Expressions for each
method are presented in a later section. Both will be augmented to include geometric as
well as dimensional variations.

5. Characterizing Geometric Feature Variations

The geometric feature variations defined in the ANSI standard must be modeled so that
their effects will be reflected in the tolerance model of the assembly. By analyzing the
assembly constraint equations, the effects of the geometric feature variations on the
assembly or kinematic variables can then be estimated (see section 5). In the vector-loopbased assembly tolerance model, this is done by modeling the geometric feature variations
with zero length vectors having specified variations or tolerances, placed at the contact
point between mating surfaces. These zero length vectors are considered as independent
variation sources to the dimensional variations in the assembly. The direction in which
they introduce variation into an assembly depends on the type of contact which exists
between the surfaces. For this reason, geometric feature tolerances of components in an
assembly are related to the joint types through which the geometric feature variations are
propagated.

13
5.1. Geometric Feature Tolerance Modeling in 2D
The kinematic joint type and geometric feature tolerances on the parts in contact are the
key elements in analyzing the effect of the geometric feature tolerances on assembly
variations. The commonly used kinematic joint types in 2-D space are modeled in Figure
4 [Chase, Gao & Magleby 1994, Chun 1988]. The effect of the geometric feature
tolerances associated with each of the joints may result in translational variation or
rotational variation. This translational or rotational variation is usually smaller than the
size tolerances on the same parts.

Planar

Cylinder Slider

Parallel Cylinders

Edge Slider

Rectangular Datum

Revolute

Center Datum

Figure 4. Kinematic joint and feature datum types in 2-D space

Figure 5 illustrates how a flatness tolerance zone can affect two mating parts differently
when viewed in 2-D. The cylinder on the left illustrates a translational variation, while the
block on the right exhibits the rotational variation, due to the same geometric feature
variation. The translation of the planar joint is reflected in the dimension variations,
except for runout and concentricity. So, the nature of the contact between mating
surfaces determines how feature variations propagate through an assembly.
The rotation variation for the block on a plane surface in Figure 5 is related to the flatness
tolerance zone and the contact length of the block, in this case, the horizontal dimension
of the block. This contact length is called characteristic length.
Flatness Tolerance Zone
= Tan-1( Characteristic Length )

(13)

where is the rotational variation caused by the flatness in a planar joint.

translational variation is /2 where is the width of the tolerance band.

The

14

Cylindricity
Tolerance
Zone

Nominal
Circle

Rotational Variation
Flatness
Tolerance
Zone

Flatness
Tolerance
Zone

Translational
Variation

Cylinder on a plane surface

Block on a plane surface

Figure 5. 2-D effects of geometric feature tolerance

If the translational variation caused by the geometric feature tolerance is represented by T
and rotational variation by R, all the possible combinations of the geometric feature
tolerances with the kinematic joint types can be summarized in Table 1. The empty cells in
the table mean that the corresponding geometric feature tolerance and kinematic joint
combination does not apply.
Table 1 Rotational and translational variations associated with
corresponding geometric feature tolerance-kinematic joint combinations in 2-D
Joints

Geom
Tol

Planar

Cyl Slider

Edge Slider

RT

Revolute
Par Cylind

RT
T

Note: For a rotational variation, a characteristic length should be provided

for each feature tolerance associated with it.
A debatable point is whether or not rotational variations should also include a translation
normal to the surface. That is, does a block on a plane exhibit both rotation and
translation due to surface variations? The answer depends on how the surface is
manufactured and how it will be inspected. If there is a size dimension normal to the
plane, say, describing plate thickness, the size variation will probably include translation

15
variations. If the inspection for size involves area contact between the instrument and the
surface, the highest waviness peak would determine the size. Thus, waviness would be
included in the size.
The exceptions to this interpretation are runout and concentricity. Consider inspecting a
platter of a computer disk drive. The waviness period associated with runout is generally
much larger than the contact area of either the thickness measuring instrument or the readwrite head. Thus, the waviness would introduce a translational variation independent of
size variation.
On the other hand, rotational variation is also dependent on the period of the waviness. If
the surface waviness for the process has a much longer or much shorter period than the
length of contact between mating surfaces (characteristic length), little or no rotation will
occur. Such differences can be taken into account by adjusting the width of tolerance
band or the characteristic length to produce variations in the model which truly represent
the 3 variations exhibited by the assembled parts. Extensive research is needed in this
area to properly characterize geometric variational effects between mating parts.

16
5.2. Geometric Feature Tolerance Modeling in 3-D
The effects of a geometric feature tolerance on a kinematic joint in 3-D are more
complicated than in 2-D, because a three dimensional local joint coordinate reference
frame is required to describe the position and orientation of the joint. The most commonly
used kinematic joints in 3-D are illustrated in Figure 6 [Robison 1989, Goodrich 1991,
Ward 1992].

Rigid (no motion)

Cylindrical

Cylindrical Slider

Prismatic

Revolute

Parallel Cylindrical

Spherical

Planar

Edge Slider

Point Slider

Spherical Slider

Crossed Cylinder

Figure 6. Kinematic joint types in 3-D

Each geometric feature tolerance may produce variations, either translation, rotation or
both, at the joint in different directions. If the geometric feature tolerance of a cylinder
slider joint is extended to 3-D, the position and orientation of the cylinder will be affected
simultaneously. Viewing along the axis of the cylinder (local z axis), the cylinder may rest
on a peak or in a valley of the plane and that will result in a translational variation normal
to the plane. While, looking perpendicular to the cylinder (local x axis), one end of the
cylinder may lie in a valley and the other end on a peak of a surface and this will cause a
rotational variation about the local x axis (see Figure 7).

17
For example, in Figure 7, if the translational variation in the local joint coordinate
reference frame is represented by Ty and the rotational variation by Rx, the combination of
the cylindrical slider and plane geometric feature tolerance flatness will produce the
combined variations (TyRx).

y
Rotational Variation
x
Flatness
Tolerance
Zone

3D cylindrical slider joint

Nominal
Circle
Translational
Variation

Cylindricity
Tolerance
Zone
Flatness
Tolerance
Zone

View normal to the cylinder axis

The effect of feature variations in 3D

depends upon the joint type and
which joint axis you are looking
down.

View looking down the cylinder axis

Figure 7. 3-D effects of geometric feature tolerance

Goodrich [1991], Ward [1992] and Gao [1993] studied all the possible combinations of
the kinematic joints and geometric feature tolerances in 3-D space (see Table 2). These
variation combinations can then be used to estimate the assembly or kinematic variations
caused by the geometric feature tolerances.
It is important to point out that the x, y and z axis for each of these joints is a local axis
specific to each joint (see Figure 7 and 8). From Figure 8, it can be seen that the degrees
of freedom for kinematic motion (K) and the degrees of freedom for feature variations (F)
are mutually exclusive. Everywhere K is removed by constraints, there is possibility for F.
This fact is true for all of the kinematic joints listed in Figure 6.

18

K Kinematic Motion
F Geometric Feature Variation
F

K
y

Cylindrical Slider Joint

Planar Joint

Figure 8. Degrees of freedom for kinematic motions and geometric feature variations.
Table 2 Rotational and translational variations associated with
corresponding geometric feature tolerance-kinematic joint combinations in 3-D
Joints

Geom
Tol

Planar

Rx Rz R x Rz

Rx Rz

Rx Rz Rx Rz Rx Rz RxRzTy

Revolute

Rx Rz Rx Rz R x Rz Rx Rz

Rx Rz Rx Rz Rx Rz Rx Rz

Tx Tz Tx Tz

Cylindrical

Rx Rz Rx Rz R x Rz Rx Rz

Rx Rz Rx Rz Rx Rz Rx Rz

Tx Tz Tx Tz

Prismatic RxRyRzRxR yRz

Spherical
Crscyl

RxRyR z RxRyR zRxRyRz RxRyRz

Tx TyTz

Ty

Ty

Tx TyTz
Ty

Ty

Ty

Ty

Ty Rx

Ty

Ty Rx Ty Rx Ty Rx Ty Rx
Ty Rx Ty Rx Ty Rx Ty Rx Ty Rx

Ty

Ty Rx Ty Rx Ty Rx Ty Rx Ty Rx Ty Rx Ty Rx Ty Rx
Ty Rx Ty Rx Ty Rx Ty Rx Ty Rx
Ty Rx

Ty

Parcyl

Ty Rx Ty Rx Ty Rx Ty Rx

E
P Ty Rx
C
Cylsli
P Ty Rx
Pt
Pntsli
Ty
P
S
Sphsli
P
Ty

Ty Rx
Ty Rx

Edgsli

Tx TyTz Tx TyTz Tx TyTz

Rx

Rx

Rx

Ty
Ty
Ty
Ty

Ty

Ty

Ty

Ty

Ty

Ty
Ty

Ty

Ty

Ty

Ty
Ty

Ty

19
Note: Crscyl = Crossed cylinders
Parsyl = Parallel cylinders
Edgsli = Edge slider (E for edge and P for plane)
Cylsli = Cylindrical slider (C for cylinder and P for plane)
Pntsli = Point slider (Pt for point and P for plane)
Sphsli = Spherical slider (S for sphere and P for plane)
A rigid joint is a joint with all kinematic degrees of freedom removed. Therefore, all axes
are possible sources for geometric feature variations. Applicable degrees of freedom
depend on the contact existing between mating surfaces. For example, if two plates are
welded or fastened, they still have position and orientation variations.
5.3 Application of the Envelope Rule
The foregoing discussion defined a comprehensive system for modeling geometric
variations in assemblies. A complete discussion must address the interaction of geometric
and size variations.
The geometric tolerance standard ANSI Y14.5M-82 defines Rule #1 as: the surface of a
feature shall not extend beyond a boundary (envelope) of perfect form at MMC. This
rule is also called the Envelope Rule. It states that all points on the surface of a feature
must lie within the limits defined by its MMC envelope (Maximum Material Condition).
This means that any surface variations due to flatness, parallelism, etc. must reduce to zero
as the size approaches MMC.
The envelope rule applies to worst case conditions. It was established to permit part
inspection with go-nogo gages. Go-nogo gages are manufactured to the MMC and
LMC values for the dimension being checked and are used for 100% inspection of parts.
The envelope rule should not be applied to statistical variations. Statistical analysis of
variation is based upon a distribution centered about a nominal or ideal value.
Independent random variations of the component dimensions are the result of process
variations and add as root-sum-squares, with the resulting distribution extending
theoretically to infinity. Statistical analysis of variation always predicts a certain
percentage of parts will fall outside of the design limits. There is no way to apply the
envelope rule to a statistical model. Therefore, this paper proposes that the envelope rule
only be used for worst case analysis and not be applied to statistical analysis of variation.

20
The application of the envelope rule to assemblies needs to be generalized, since the
envelope rule applies to the individual component dimensions used in the tolerance
analysis of an assembly. Procedures for determining when to reduce specific geometric
variations in order to avoid violating the MMC envelope for components must be defined
for various assembly conditions. The envelope rule does not apply to some geometric
variations, such as runout. These cases must be identified and accounted for. The focus
of this paper, however, will be on statistical analysis of assemblies. The generalization to
worst case analysis is left for a future study.

6. Derivative Evaluation for Geometric Feature Variables

In order to calculate the derivatives for the linearized assembly equations, small variations
corresponding to the geometric variations must be introduced at appropriate points in the
loop. The effect on loop closure may be calculated and used to calculate the sensitivities,
as was done for length and angle sensitivities.
Although geometric variations may affect an entire surface, they can only introduce
variation into an assembly at the contact surfaces between mating parts. Geometric
disturbances propagate through the joints along the constrained directions, that is, in the
direction of the geometric degrees of freedom (F axes in figure 8).
In the matrix loop equation, as the transformation matrices traverse the loop, they
transform a local coordinate system from node to node. At each joint, the rotating
coordinate system must be oriented to coincide with the local joint reference frame.
Appropriate geometric feature variations may be added at the joint as permitted (see Table
1 and Table 2). Each geometric feature variation, is treated as a zero length vector or
rotation with a plus or minus variation in the permitted direction.
If the variation caused by the geometric feature tolerance-kinematic joint combination is a
rotation, the geometric feature variable can be represented in the assembly kinematic
constraints by a rotational matrix or a combination of matrices, as defined in equations (2)
and (3). Rotation angles x, y and z are now geometric feature variables with nominal
value zero and variation equal to the corresponding rotation tolerance about the given
local joint axis.
If the variation caused by the geometric feature tolerance-kinematic joint combination is a
translation, a translation matrix, as defined in equations (2) and (3), may be inserted at the
appropriate node in the assembly kinematic constraint equation. The translation

21
components Tx, Ty and Tz are the geometric variation components along the local joint
axes. They have zero nominal length, but can still introduce a variation along the
corresponding joint axes.
The combination of translational and rotational variations caused by the geometric feature
tolerance can be handled by combining the above operators together as directed in Table 1
and Table 2 and inserting them into the constraint equations. For the general case, if a
geometric feature tolerance is to be added at joint i, the assembly constraint equation (1)
becomes:
[R1][T1]...[Ro][Rix][Riy][Riz][Ti][R-o]...[Rn+1][Tn+1][Rf] = [I]

(14)

where [Ro] represents the rotation to orient the rotating coordinate system to coincide
with the local joint reference frame and [R-o] is the reverse operation, that is, returning to
the orientation just before the geometric feature tolerance was added. Figure 9 illustrates
such transformations for the flatness-edge slider combination in 2-D. This process can be
repeated throughout the vector loop of the assembly.
Direction for
translational
variation
Out vector
Ri+1

R-o

Ro

In vector

Figure 9. Adding flatness variation to 2-D edge slider joint

After the vector loop is traversed by the above process, the kinematic constraint equation
will have a form similar to equation (1), with a few more matrix operations. The
numerical derivative evaluation procedure discussed for length and angle variables is also
applicable, that is,
for translation
[R1][T1]...[Ri][Ti(0+L)]...[Rn][Tn]{0 0 0 1}t = {X Y Z 1}t

(15)

for rotation
[R1][T1]...[Ri(0 + )][Ti]...[Rn][Tn]{0 0 0 1}t = {X Y Z 1}t

(16)

22
where L is a small change in length in the translation direction and is a small change in
the rotation direction. The derivatives can then be approximated numerically by equations
(6), as before.

7. Including Geometric Feature Variations in the DLM

The above section described the scheme for characterizing or representing geometric
feature variation in 2-D and 3-D assemblies. This enables us to include the effect of
geometric feature variation in evaluating the assembly variations.
The DLM uses the linearized Taylor expansion of equation (1) to solve for the assembly
variations. If the geometric feature variations of the components in the assembly are
included, the linearized constraint equations for an assembly may be modified to include
geometric variations:
Closed loops:
{H} = [A]{X} + [B]{U} + [F]{} = {0}

(17)

Open loops:
{p} = [C]{X} + [D]{U} +[G]{}

(18)

where {}: the variations of the geometric feature variables,

[F], [G]: the partial derivatives with respect to the geometric feature variables,
{H}, {p}, {X}, {U}, [A], [B], [C] and [D] are defined in equations (7)
and (10).
Solving equation (17) for {U}, the closed loop assembly variations:
{U} = -[B]-1[A]{X} - [B]-1[F]{}

(19)

If one substitutes {U}from equation (19) into equation (18), then the variations of the
open loop variables can be expressed as:
{p} = ([C] - [D][B]-1[A]){X} + ([G] - [D][B]-1[F]){}

(20)

8. Estimation of Kinematic Variations and Assembly Rejects

The accumulation of variations in the assembly can then be estimated by applying a worst
case or statistical model [Goodrich 1991, Robison 1989] to equations (19) and (20).
Worst case:

23
n

j=1

j=1

d
d
U i = Sij Tolij + Sij Tolij

T ASM

(21)

Statistical model:
n

Ui =

Here

j=1

Sdij Toldij

+ Sij Tolij
j=1

T ASM

(22)

Sd is the tolerance sensitivity matrix for the dimensional variables,

S is the tolerance sensitivity matrix for the geometric feature variables,
Told is the tolerance vector for the dimensional variables,
Tol is the tolerance vector for the geometric feature variable,
n is the number of dimensional variables,
m is the number of geometric feature variables,
TASM is the design limit for assembly variation Ui.

In equation (22), Ui2 is an estimate of the statistical variance of assembly variable Ui. If
all the component tolerances Told and Tol represent 3 variations, Ui will represent
the 3 variation of Ui.
The tolerance sensitivity matrices can be calculated by:
For a well-determined system:
Sd = -[B]-1[A]

(23)

S = -[B]-1[F]

(24)

For an over-determined system:

Sd = -([B]T[B])-1[B]T[A]

(25)

S = -([B]T[B])-1[B]T[F]

(26)

Equation (21) or (22) can also be used to estimate the variations of the open loop
variables. Here, the sensitivity matrices are:
For a well-determined system:
Sd = [C] - [D][B]-1[A]

(27)

24
S = [G] - [D][B]-1[F]

(28)

For an over-determined system:

Sd = [C] - [D]([B]T[B])-1[B]T[A]

(29)

S = [G] - [D]([B]T[B])-1[B]T[F]

(30)

The estimation of the assembly rejects is based on the assumption that the distribution of
the assembly variable(s) is Normal, which is a reasonable assumption for most assemblies.
The estimate of a kinematic or assembly variation is treated as three standard deviations,
and this deviation together with the mean value of the kinematic or assembly variable can
be used to calculate by integration or table the assembly rejects for a given assembly batch
if the limits for the assembly variable are specified.

25

9. Example
As an example to demonstrate how to include geometric feature tolerances in the variation
estimation of kinematic or assembly variables and prediction of assembly rejects, the tape
hub locking assembly is examined. Figure 10 shows the assembly with geometric feature
tolerances applied.

0.010

e
a

Plunger

d
c

0.001
0.001

Arm

Base
k

r
0.001

Figure 10. Tape hub locking assembly with geometric feature tolerances
In this case study, effort will be concentrated on the complete locking assembly model
which is used to mount and hold a magnetic tape reel in place on a tape drive hub. The
assembly consists of a plunger which can slide vertically against a guide mounted on the
base and an arm which can slide horizontally on the base and which makes contact with
the beveled surface of the plunger. The position of the arm is determined by the vertical
position of the plunger, its level angle, the height of the arm, arm corner radius and the
step on the base. As the plunger moves down, it forces the arm and pad outward against
the inner surface of the tape reel, locking it in place.
The design requirement for the tape reel assembly is to have adequate interference
between the pad and the tape reel. Then, the critical feature is the magnitude of the
variation on dimension r when the tolerances of the components including geometric
feature variations are considered.

26
Table 3 shows the geometric information for the assembly. Dimensions with a given
tolerance are the manufactured variables, while those without specified tolerances are the
assembly or kinematic variables. Table 4 lists all the geometric feature tolerances applied.

27
Table 3 Dimensional data of the tape hub locking mechanism
Part Name
Height I of base
Width of base
Length of plunger
Contact distance
Radius of arm
Width of arm
Thickness of pad
Height of arm
Position of pad
Height II of base

Transformation

Nominal Dim

Tolerance()

Variation

Rotation

90

Translation k

0.398

0.006

Independent

Rotation

-90

Translation a

0.355

0.005

Independent

Rotation

Translation b

1.000

0.01

Independent

Rotation

75

0.5

Independent

Translation c

0.3194

Kinematic

Rotation

-90

Translation d

0.06

0.002

Independent

Rotation

15

Kinematic

Translation e

0.309

0.003

Independent

Rotation

Translation f

0.05

0.002

Independent

Rotation

-90

Translation g

0.488

0.004

Independent

Rotation

-90

Translation r

1.854626

Kinematic

Rotation

90

Translation h

0.203

0.002

Independent

Table 4 Geometric feature tolerance data of the tape hub locking mechanism
Name
1
2
3
4

Joint type
Cylinder slider
Planar
Planar
Planar

Feature
Tolerance band
Profile, arm corner radius
0.010
Flatness of arm
0.001
Flatness of base
0.001
Perpendicularity of arm
0.001

The derivative matrices can be obtained according their definitions, i.e.,

Character length
N/A
0.3690
0.3690
0.3690

28
H x
k

H y
[A] =
k
H

k

H x
a
H y

H x
b
H y

H x

H y

H x
d
H y

H x
e
H y

H x
f
H y

a
H
a

b
H
b

d
H
d

e
H
e

f
H
f

H x
g
H
g
H
g

H x
h

H y
h
H

0.9659 1 1 0
0
0 1 1 0.3980

= 1 0 0 1.3550 0.2588 0 0 1 1

0 0 0
1
0
0 0 0
0
H x
c

H y
[ B ] =
c
H

c

H x

H y

H x
r 0.2588 0.6910 1.000

H y
= 0.9659 1.4957 0.0000

0
1
0

H x
a
1
H y
[F]=
a1
H

a1

H x
a 2
H y

H x
a 3
H y

a 2
H
a 2

a 3
H
a 3

(26)

(27)

H x
a 4

H y
a 4
H

a 4

0.2030
0.2030
0.2030
0.9659

= 0.2588 1.8547 1.8547 1.8547

0
1
1
1

(28)

Sd = -[ B ]-1[ A ]
1.0353 0 0 0.1457 0.2679 0 0 1.0353 1.0353
=
0
0 0
1
0
0 0
0
0

0.2679 1 1 0.3307 1.0352 1 1 0.2679 0.2679

S = -[ B ]-1[ F ]

(29)

29
0.3717
0.3717
0.2679 0.3717

=
0
1
1
1

1.0353 0.3918 0.3918 0.3918

(30)

With the sensitivity matrix known, the variations of the kinematic or assembly variables
can then be calculated by applying equation (9).
c
0.0082

= 0.5659
r
0.0136

(31)

The variations of the kinematic or assembly variables without geometric feature tolerances
is presented here for comparison.
0.0079
c

= 0.5000
0.0124
r

(32)

In this assembly, dimension r is the one which has the design specification since its value
and variation will affect the performance of the mechanism. If the design spec for r is set
to be TASM and the estimated variation r is equal to 3 standard deviations, then the
design spec can be used to calculate the assembly reject rate by either standard normal
distribution table or integration or empirical methods.

9. Conclusions
A generalized approach has been presented for characterizing geometric feature tolerances
in vector-loop-based assembly tolerance models and including such tolerances in the
estimation of variations and assembly rejects for 2-D or 3-D assemblies. The method is
simple and gives satisfactory results if the tolerances are relatively small compared with
the associated nominal dimensions of the parts [Gao 1993].
In the example problem, the geometric feature tolerances increased the estimated assembly
variations by about 10 percent (see equation (31) and (32)). This illustrates the
importance of including geometric feature tolerances if an accurate estimation of assembly
variation is desired.

30
Table 1 and Table 2 listed the most commonly occurring combinations of geometric
feature variations and kinematic joint types in 2-D and 3-D assemblies. These tables are
subject to change as new assembly conditions are considered or after more thorough
studies are made.
Clearly, there are several problem areas requiring further study:
1. Characterizing geometric variations and their affects on mating part variations
2. Defining the interaction of geometric variations with size dimensions
3. Generalizing Rule #1 and its application to assemblies.
It is felt that vector models of assemblies constitute an efficient and effective tool for
tolerance analysis of mechanical assemblies, which can be adapted as further model
refinements are made.

Acknowledgments
This work was sponsored by ADCATS, the Association for the Development of
Computer-Aided Tolerancing Software, a consortium of twelve industrial sponsors and
the Brigham Young University, including Allied Signal Aerospace, Boeing, Cummins,
FMC, Ford, HP, Hughes, IBM, Motorola, Sandia Labs, Texas Instruments and the U.S.
Navy.

31

References
ANSI (American National Standard Institute), 1982, Dimensioning and Tolerancing,
ANSI Y14.5M - 1982, American Society of Mechanical Engineers (ASME), New York,
NY.
Chase, K. W. and Parkinson, A. R., 1991, A Survey of Research in the Application of
Tolerance Analysis to the Design of Mechanical Assemblies, Research in Engineering
Design, v 3, pp.23-37.
Chase, K. W., Gao, J. and Magleby, S. P., 1994, Generalized 2-D Assembly Tolerance
Analysis Method with Small Kinematic Adjustments, Submitted for publication.
Chun, K. S., 1988, Development of Two-Dimensional Modeling Method for CAD
Systems, M.S. Thesis, Mechanical Engineering department, Brigham Young University,
Dec.
Foster, L. W., 1986, GEO-METRICS, Addison-Wesley Publishing Company.
Foster, L. W., 1992, GEO-METRICS II, Addison-Wesley Publishing Company.
Gao, J., 1993, Nonlinear Tolerance Analysis of Mechanical Assemblies, Ph.D.
Dissertation, Mechanical Engineering department, Brigham Young University, Aug.
Gao, J., Chase, K. W. and Magleby, S. P., 1994, Generalized 3-D Assembly Tolerance
Analysis Method with Small Kinematic Adjustments, Submitted for publication.
Gilbert, O. L., 1992, Representation of Geometric Variations Using Matrix Transforms
for Statistical Tolerance Analysis in Assemblies, M.S. Thesis, Massachusetts Institute of
Technology, Department of Mechanical Engineering, Cambridge, June.
Goodrich, C. G., 1991 Representation and Modeling of Geometric Form Variations for
3-D Tolerance Analysis, M.S. Thesis, Mechanical Engineering department, Brigham
Young University, June.
Guilford, J. and Turner, J., 1993,
Representational Primitives for Geometric
Tolerancing, Computer-Aided Design, Sept.
Gupta, S. and Turner, J., 1993, Variational Solid Modeling for Tolerance Analysis,
IEEE Computer Graphics and Applications, May.
Hillyard, R. C. and Braid, I. C., 1978, Analysis of Dimensions and Tolerances in
Computer-Aided Design, Computer-Aided Design, vol. 10, no. 3, pp. 209-214.
Larsen, D. V., 1989, An efficient Method for Iterative Tolerance Design Using Monte
Carlo Simulation, M.S. Thesis, Mechanical Engineering department, Brigham Young
University, Jan.

32
Larsen, G. C., 1991, A Generalized Approach to Kinematic Modeling for Tolerance
Analysis of Mechanical Assemblies, M.S. Thesis, Mechanical Engineering Department,
Brigham Young University.
Lin, P. D., and Chen, J. F., 1994, Analysis of Errors in Precision for Closed Loop
Mechanisms, Journal of Mechanical Design, vol 116, pp. 197-203.
Martino, P. M. and Gabriele, G. A., 1989, Application of Variational Geometry to the
Analysis of Mechanical Tolerances, Failure Prevention and Reliability, ASME Paper No.
DE-Volume 16, pp. 19-27.
Requicha, A. G., 1983, Toward a Theory of Geometric Tolerancing, International
Journal of Robotics Research, vol. 2, no. 4, pp. 45-60, Winter.
Robison, R. H., 1989, A Practical Method for Three-Dimensional Tolerance Analysis
Using a Solid Modeler, M.S. Thesis, Mechanical Engineering Department, Brigham
Young University.
Shah, J. J. and Miller, D. W., 1990, A Structure for Supporting Geometric Tolerances in
Product Definition System for CIM, ASME Manufacturing Review, vol. 3, no.1, pp. 2331, March.
Veitschegger, W. K. and Wu, C., 1986, Robot Accuracy Analysis Based on
Kinematics, IEEE Journal of Robotics and Automation, vol. RA-2, no. 3, pp. 171-179,
September.
Ward, K., 1992 Integrating Geometric Form Variations into Tolerance Analysis of 3-D
Assemblies, M.S. Thesis, Mechanical Engineering department, Brigham Young
University, Aug.
Whitney, D. E., Gilbert, O. L. and Jastrzebski, M., 1994 Representation of Geometric
Variations Using Matrix Transforms for Statistical Tolerance Analysis in Assemblies,
Research in Engineering Design, vol. 6, pp. 191-210.