You are on page 1of 12

Computer-Aided Design 6768 (2015) 112

Contents lists available at ScienceDirect

Computer-Aided Design
journal homepage: www.elsevier.com/locate/cad

Preferred feed direction field: A new tool path generation method for
efficient sculptured surface machining
Guillermo H. Kumazawa a , Hsi-Yung Feng a, , M. Javad Barakchi Fard b
a

Department of Mechanical Engineering, The University of British Columbia, Vancouver, BC, Canada V6T 1Z4

Department of Mechanical and Materials Engineering, The University of Western Ontario, London, Ontario, Canada N6A 5B9

highlights
Tool path generation according to a preferred feed direction field of the surface.
Generated tool paths having comparatively short overall tool path length.
Surface segmentation by identifying degenerate points and forming their separatrices.

article

info

Article history:
Received 29 August 2014
Accepted 26 April 2015
Keywords:
Sculptured surface
Ball-end milling
Machining efficiency
Preferred feed direction
Tensor field
Iso-scallop tool paths

abstract
This paper presents a new method to generate efficient ball-end milling tool paths for three-axis
sculptured surface machining. The fundamental principle of the presented method is to generate the
tool paths according to a preferred feed direction (PFD) field derived from the surface to be machined.
The PFD at any point on the surface is the feed direction that maximizes the machining strip width.
Theoretically, tool paths that always follow the direction of maximum machining strip width at every
cutter contact point on the surface would result in shorter overall tool path length. Unfortunately, overlaps
of adjacent machining strips commonly exist for tool paths that follow the preferred directions exactly.
Such redundant machining can be reduced by iso-scallop tool paths. Nonetheless, iso-scallop tool paths
do not in general follow the preferred feed directions. To improve machining efficiency via generating
short overall tool path length, the presented method analyzes the PFD field of the surface and segments
the surface into distinct regions by identifying the degenerate points and forming their separatrices.
The resulting segmented regions are characterized by similar PFDs and iso-scallop tool paths are then
generated for each region to mitigate redundant machining. The developed method has been validated
with numerous case studies. The results have shown that the generated tool paths consistently have
shorter overall length than those generated by the existing methods.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Sculptured surfaces have seen applications in many fields, especially in the aerospace, automotive, and biomedical industries.
These surfaces are characterized by their smooth shape and may
include features such as valleys, mounts, or blends to fulfill requirements such as an optimized airflow or a desirable ergonomic
shape. In order to manufacture components with sculptured surfaces, three- or five-axis computer numerical control (CNC) machine tools with a variety of end mills are used to machine either

This paper has been recommended for acceptance by Kai Tang.

Corresponding author. Tel.: +1 604 822 1366; fax: +1 604 822 2403.
E-mail address: feng@mech.ubc.ca (H.-Y. Feng).

http://dx.doi.org/10.1016/j.cad.2015.04.011
0010-4485/ 2015 Elsevier Ltd. All rights reserved.

the desired part itself or its corresponding die/mold. Although fiveaxis CNC machines provide higher versatility and efficiency compared to their three-axis counterparts, the latter are still commonly
used due to their lower costs and widespread availability.
In three-axis machining of sculptured surfaces using ballend mills, scallops are formed between adjacent tool paths. The
resulting scallop height depends on the interval between the
tool paths, commonly referred to as the side-step. For a given
ball-end mill, a small scallop height constraint requires a small
side-step between adjacent tool paths and consequently, a large
number of tool paths are required to machine the entire surface,
which adds up to a long overall machining tool path length.
Redundant machining occurs when the chosen side-step results in
a scallop height that is smaller than the specified scallop height
constraint.

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

In practice, the specified scallop height constraint for a finished


surface is quite small in order to lessen the burden of the subsequent polishing work. As a constant feed rate is commonly used
for a ball-end mill during the finishing machining of a sculptured
surface, a longer overall tool path length correspond to a longer
machining time. Methods to reduce the overall tool path length
have thus been an active research topic in the field of ComputerAided Manufacturing (CAM) in the past two decades. These existing methods aimed to generate the milling tool paths needed to
obtain the desired sculptured surface shape under the specified
scallop height constraint. Conventionally, these methods would require the user to input some a priori tool path generation information from which the tool paths are generated on the surface.
More recently, researchers start to formulate machining strategies
which attempt to determine the user-input information automatically from the given surface shape. In this paper, the related developments are referred to as the surface topology based methods.
1.1. Conventional tool path generation methods
Conventional tool path generation methods can be roughly categorized into iso-parametric, iso-planar and iso-scallop methods.
The iso-parametric method is the earliest method of the three and
generates the required cutter contact (CC) tool paths on the design surface S(u, v) by fixing the value of one of the parametric
variables, say u, while varying the value of v within its domain [1].
The 1u between adjacent iso-parametric tool paths is chosen such
that the resulting scallop height will not violate the specified limit.
This method is mathematically convenient and ensures that the
entire surface is covered by the tool paths. However, depending
on the surface shape, the generated tool paths may be quite dense
in some regions. To improve the iso-parametric method, Elber and
Cohen [2] proposed to eliminate some iso-parametric tool path
segments when the generated tool paths become too dense. The
latest iso-parametric method attempted to optimize the generated
tool paths via an adaptive grid [3]. Despite all these efforts, redundant machining still exists in the generated tool paths.
The iso-planar method was initially proposed as an improvement over the iso-parametric method. This method employs a series of parallel planes whose intersections with the design surface
correspond to the CC tool paths [4]. The separation between adjacent intersection planes provides better control on the resulting
scallop height than the iso-parametric method. The main advantage of the iso-planar method is that it avoids regions of dense tool
paths observed in the iso-parametric tool paths. However, redundant machining still exists for the iso-planar tool paths. Improvements to the conventional iso-planar method have been proposed.
For example, Ding et al. [5] implemented an adaptive iso-planar
method which segmented the surface into regions according to the
surface slope. Consequently, the method required an input from
the user regarding the limits on the slope used for the surface segmentation.
The first method to generate iso-scallop tool paths on a given
surface appeared to be proposed by Suresh and Yang [6]. The isoscallop method sequentially generates an adjacent tool path based
on the current tool path. It aims at ensuring that the resulting scallop height between adjacent tool paths is the same as the specified limit. Redundant machining thus does not exist, resulting in
a smaller overall tool path length than that for the iso-parametric
or iso-planar tool paths. Later, quite a few methods have been proposed, which improve on the calculation efficiency [7] and accuracy [8]. The methods proposed by Feng and Li [9] and Tournier
and Duc [10] improved the calculation accuracy further by carrying out all calculations three-dimensionally. And a method that attempted to improve the speed of three-dimensional calculations

has also been proposed [11]. Since the iso-scallop method is fundamentally a sequential tool path generation method, it requires
setting an initial tool path from which the subsequent tool paths
are generated. This gives room for further improvement and inspires a more recent research direction that attempts to analyze
the topology of the given surface with the goal of minimizing the
overall tool path length.
1.2. Surface topology based methods
Recent developments by Quinsat and Sabourin [12] and
Vijayaraghavan et al. [13] sought to find an optimal orientation for
the parallel intersection planes in order to generate iso-planar tool
paths with minimal overall tool path length for a given sculptured
surface. The optimal orientation was determined by analyzing local
parameters of the surface which influenced the material removal
rate or by a metric of the mean scallop height resulting from a
particular plane orientation. These methods, however, employ a
single iso-planar direction across the entire surface, which cannot
guarantee to minimize the overall tool path length as redundant
machining still exists. For iso-scallop tool paths, attempts have also
been made to find the optimal initial tool path by analyzing the
complete surface geometry. The method proposed by Giri et al. [14]
used loci of maximum local convex curvatures to determine the
initial tool path from which the subsequent iso-scallop tool paths
were generated. User input was required in their work to select and
compare the curvature loci in order to finalize the initial tool path.
A notable work by Kim and Sarma [15] treated the optimal
machining directions at sampled points on the given surface as a
vector field. The vector field was analyzed and then segmented to
generate the tool paths. The employed vector field was unidirectional and this limited its application in tool path generation. This
is because for the problem of generating efficient tool paths, there
are in fact two optimal directions at any point on the surface, which
are opposite to each other. A vector field cannot adequately represent such a bi-directional field, as will be discussed in the subsequent sections of this paper.
This paper proposes a new strategy to generate efficient tool
paths for three-axis sculptured surface machining. It aims at minimizing the overall tool path length for a given ball-end mill and
scallop-height limit without any user input for the initial tool path.
The strategy starts with creating a discrete bi-directional field, or
tensor field, which maps the preferred machining directions across
the surface. The field is then analyzed and segmented into regions
with similar preferred machining directions. For each segmented
region, a principal tool path is determined and used as the initial
tool path from which a series of iso-scallop tool paths are generated. This strategy will result in tool paths with shorter overall
length compared with those generated by the existing tool path
generation methods.
The employed region-based tool path generation strategy has in
fact already been applied in the multi-axis machining of complexshaped parts such as turbine and impeller blades. To reduce cutting
tool orientation changes, Chen et al. [16] introduced a method to
subdivide a complex sculptured surface into a number of easyto-machine surface patches. Similarly, Dai et al. [17] proposed to
divide a complex surface into different regions to facilitate flank
milling of the approximated piecewise ruled surface. Very recently
for three-axis machining, Zhu et al. [18] subdivided a sculptured
surface according to the surface features using the technique
of square lattice subdivision. It was concluded that dividing the
surface into various surface patches was indeed helpful to generate
shorter overall tool path length, thereby increasing the machining
efficiency.
It should be noted here that many of the existing regionbased tool path generation studies, if not all, have evaluated the

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

effectiveness their developed methods using the overall milling


tool path length. This practice is generally acceptable because the
milling feed rate is considered fixed for finish machining; thus,
the overall machining time is simply proportional to the milling
tool path length. Also, although extra tool travel is needed to
connect the much increased number of tool path segments (due
to the segmented surface regions), the resulting extra tool travel
time is comparatively small because rapid tool positioning is to be
used between milling tool path segments that are characterized by
much smaller feed rates.
2. Preferred feed direction
To determine the preferred machining or feed direction of a
ball-end mill at a specific point on the design surface, a metric
is required. The applicable metric is in fact a subject of active
research [19,20]. It has been commonly assumed in the research
literature that tool paths following the direction of maximum
material removal will have shorter overall tool path length. In
this paper, the widely used metric of machining strip width is
employed as the metric to determine the preferred feed direction
at a point on the design surface.
2.1. Machining strip width
The notion of machining strip width (W ) appeared to be first
introduced by Lee and Ji [21] to mathematically analyze five-axis
machining and find the optimal cutter feed direction [22,23]. W
can also be calculated for three-axis machining to determine the
preferred feed direction of the cutter [24,25] and there are different
ways to approximate W [26]. In this paper, W is to be evaluated
based on the concept of the scallop surface [9]. The scallop surface
Sscallop (u, v) is a normal offset surface from the design surface
S(u, v) with the offset distance equal to the scallop height h. The
cutting tool motion produces a tool swept volume which intersects
the volume enclosed in-between the design and scallop surface
to create the machined volume. The tool swept volumes contact
with the design surface is a curve on the design surface which
corresponds to the cutter contact (CC) tool path. Points on the CC
tool path are referred to as the CC points (PCC ). A CC tool path
has its corresponding cutter location (CL) tool path, consisting of
CL points (PCL ), and every PCL has a corresponding PCC . In the case
of three-axis machining, the CL point PCL for a given CC point PCC
is calculated by multiplying the normal vector nCC of the design
surface at PCC by the tool radius r. At each CL point, three vectors
can be determined: nCL , which is an extension of and effectively
the same as nCC ; fCL , the tangent vector of the CL path at PCL ; and
tCL , calculated as the cross product tCL = nCL fCL .
The tool swept volume is the volume enclosed by the surface
constituted by the union of all the tool swept profiles along a tool
path. The procedure to derive the tool swept profile at each CL point
is briefly outlined here. The volume of a ball-end mill is made up
of two volumes: a hemisphere for the ball end and a cylinder for
the tool flank. The hemisphere is intersected by the plane created
by tCL and nCL at the CL point. The generated intersection curve
constitutes the bottom tool swept profile, which intersects the
circle of the tool flank cylinders bottom at two points Sa and Sb
(Fig. 1). The vector from Sa to Sb and the tool axis zT create a
plane which intersects the tool flank cylinder and the resulting
intersection curve is the flank tool swept profile. The union of the
bottom tool swept profile and the flank tool swept profile results in
the tool swept profile. The tool swept profile intersects the scallop
surface at two points, Pa and Pb . The projected distance from Pa to
Pb along tCL gives W :
W = Pa Pb tCL .

(1)

Fig. 1. Tool swept profile and machining strip width.

2.2. Preferred feed direction field


It is important to note that, at a given CC point, the value of
W will be the same if the machining feed direction is reversed.
For every CC point sampled on the design surface, there will be
two opposing vectors, +fCL and fCL , which have the same W
magnitude. By sampling a series of points on the design surface and
determining the directions with the largest W , a field of preferred
directions across the entire design surface can be established. In
three-axis machining, if the tool axis is along the z-axis of a global
Cartesian coordinate system, the pair of opposing vectors at each
CC point is to be projected onto the xy plane, as illustrated in
Fig. 2. The resulting field, referred to as the Preferred Feed Direction
(PFD) field in this paper, shows two vectors at every sampled point.
It is because of this reason that a PFD field cannot be analyzed
as a vector field, which is applicable to only one vector at every
point. In the present work, the PFD field will be analyzed as a
tensor field, where every point is characterized by a tensor, which
mathematically describes the two vectors as well as how the two
vectors relate to each other.
Although not essential, it is useful in some situations to
represent the feed direction of the ball-end mill as a scalar. For this
purpose, the feed angle is adopted. The definition and calculation
procedure of the feed angle are outlined here. First, the tangent
plane at the CC point PCC is employed and named as the feed plane
at PCC . It is evident that the feed vector of the ball-end mill fCL
lies on the feed plane. Next, a reference feed vector fref on the
feed plane is taken as the cross product of the tool axis zT and
nCC . In the event that nCC is equal or very close to zT when the
ball-end mill is machining a flat or near flat region of the design
surface, the reference feed vector fref is taken as the x-axis of the
global Cartesian coordinate system of the design surface. The angle
between fCL and fref on the feed plane is then the feed angle . Since
a feed direction is composed of two opposing vectors, a feed angle
can have a value of or + . If , then is subtracted
from .

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

Fig. 3. Machining strip width overlaps for pathline tool paths.

Fig. 2. Preferred feed direction field of a typical surface.

Pathlines in tensor fields are analogous to streamlines in vector


fields. They are created by following the directions specified in
a tensor field. If pathlines following the maximum W directions
are directly used as the machining tool paths, overlapping of the
resulting machining strip widths will generally occur, leading to
redundant machining. Fig. 3 shows the machining strip width
overlaps when pathlines are directly used as tool paths. A pathline
may also be used as the initial tool path from which iso-scallop
tool paths are generated to avoid redundant machining. However,
as it can be seen in Fig. 4, the resulting iso-scallop tool paths will
progressively drift away from the pathlines. It is thus evident that
a new tool path generation methodology needs to be developed to
address the conflicting issues noted above. The new methodology
developed in the present work is to segment the design surface into
distinct regions according to the PFD field of the surface. The details
will be presented in the next section.
3. Methodology
The fundamental principle of the proposed methodology in
this work is to segment the given design surface into regions of
similar pathlines by analyzing the PFD field of the surface and
subsequently identifying the fields degenerate points and forming
their separatrices. For each segmented surface region, a principal
tool path is then determined and used as the initial tool path for
generating iso-scallop tool paths to cover the segmented region.

Fig. 4. Drift of iso-scallop tool paths from the PFD field pathlines.

This is repeated for all segmented regions on the surface and the
associated procedure is depicted in Fig. 5.
3.1. Degenerate points
Tensor fields can be found in many subject areas such as fluids
and stressstrain mechanics. Studies on tensor field topology were

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

To formulate the tensor representation for each point on the design surface according to its preferred feed directions, the preferred
feed directions are to be taken as one of the two eigenvectors for
the 2 2 matrix representation in Eq. (2). The preferred feed directions are thus similar to the principal stresses in a stress-tensor
field. Also, it is known that eigenvectors of real symmetric matrices are orthogonal. So in the present work, for a point (x, y) with
vectors of preferred feed directions [a(x, y), b(x, y)], its secondorder tensor matrix will have (a, b) as well as (b, a) as eigenvectors. It is evident that eigenvector (b, a) corresponds to a zero
eigenvalue for the PDF field since the preferred feed directions at
each point on the design surface is only along (a, b). The resulting
tensor representation can thus be derived and formulated as:
T(x, y) =

a2
ab

ab
b2

(3)

where it should be noted that both a and b are functions of (x, y).
To obtain the tensor matrix T(x, y) for any point on the surface, bilinear interpolation of the tensor components of the surrounding
sampled points is to be used.
To generate the pathline of a PFD field from a given starting
point P0 (x, y), first its two opposing preferred feed directions
are calculated by either interpolating tensor components of
surrounding points or by directly calculating the maximum W .
Then, one of the two opposing vectors is chosen as the initial and
a subsequent point is attained according to a specified small path
increment. The same procedure is repeated with this new point.
It should be noted that from the two resulting PFD vectors, only
the vector that is not opposing to the previous vector is chosen
and the next point will continue to be found until a surface border
is reached. Then, the opposing PFD vector for the original point
P0 (x, y) is chosen with P0 (x, y) as the starting point again. The
pathline is grown in the opposite direction until a surface border is
reached again.
Pathlines in a tensor field, similar to streamlines in a vector
field, do not intersect one another except at degenerate points.
These points are analogous to critical points in vector fields. At a
degenerate point Pdegen (x, y), the eigenvalues of its tensor matrix
are equal and the tensor matrix has the form:
T(x, y) =

reported by Delmarcelle and Hesselink [27,28] for visualizing and


summarizing the information contained in a tensor field. In these
studies, the notions of degenerate points and their associated
separatrices were introduced as the basic constituents of a tensor
field. In the case of a symmetric, second-order (two-dimensional)
tensor field, each point is associated with a symmetric, secondorder tensor (a 2 2 symmetric matrix). The direction information
from the PFD field in the present work can be analyzed as twodimensional because, as pointed out earlier, each point on the
design surface S(u, v), with coordinates [x(u, v), y(u, v), z (u, v)],
has unique x and y coordinates as well as unique PFD directions
when projected onto the xy plane. The z component in a PDF vector
is not required because the z component for any given projected
xy directions will be unique. It is also important to note that the
preferred feed directions do not have (scalar) magnitudes because
only directions are of interest. In summary, each point on the
design surface with coordinates [x(u, v), y(u, v), z (u, v)] is to be
characterized by a symmetric, second-order tensor in the form of:
T(x, y) =

T11 (x, y)
T12 (x, y)

T12 (x, y)
.
T22 (x, y)

(4)

The tensor components in Eq. (2) thus satisfy the following


conditions:

Fig. 5. Proposed tool path generation method.

(2)

T11 (x, y) T22 (x, y) = 0

and T12 (x, y) = 0.

(5)

Mathematically, this means that the tensor matrix is characterized


by an infinite number of eigenvectors. In the case of three-axis
sculptured surface machining, a degenerate point on the design
surface corresponds to a CC point where there is no particular
preferred feed direction because all feed directions will result in
the same W . This happens, for example, when the local surface at
the point is completely planar.
In the present work, degenerate points are detected by using
bilinear interpolation of the tensor matrices from a rectangular
grid of sampled points. The small surface region enclosed by a
grid of four neighboring sampled points is evaluated in very small
discrete steps. If the calculated values from Eq. (5) are both very
close to zero (below a set threshold) at a point, a degenerate
point is detected. For improved accuracy, the grid where a possible
degenerate point exists is further sub-divided.
Degenerate points can be classified into two categories
according to their surrounding pathlines: trisector and wedge.
A trisector degenerate point has three separatrices and a wedge
degenerate point has only one separatrix. A separatrix is a pathline

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

Fig. 7. Illustration of a wedge degenerate point.

Fig. 6. Illustration of a trisector degenerate point.

that intersects a degenerate point, analogous to the asymptotes of


a hyperbola with the center of the hyperbola being the degenerate
point. However, in a tensor field, a trisector point is characterized
by three sectors of pathlines which can be differentiated by their
separatrices (Fig. 6); a wedge point is characterized by a single fold
of pathlines towards the separatrix (Fig. 7). In comparison with
a vector field, the occurrence of the three-sector or fold shape of
streamlines is impossible since that would mean that the vector
field streamlines would have opposing streamline directions.
As presented by Delmarcelle and Hesselink [28] and Delmarcelle [29], the specific type of a degenerate point can be identified
via the following characteristic coefficient:

= AD BC

(6)

where
A=

1 (T11 T22 )
2

T12
C =
x

B=

T12
D=
.
y

1 (T11 T22 )
2

When < 0, the degenerate point is a trisector point. When > 0,


the degenerate point is a wedge point. A special case occurs when
= 0. This indicates that there are two degenerate points that
are very close to each other and would be detected as a single
degenerate point. For example, when two trisector points are very
close, it may result in a pattern of four distinct sectors of pathlines.
In this case, the two trisector points are to be dealt with separately.
When a pair of wedge points is very close or merging with each
other, the surrounding pathlines form concentric loops and are to
be dealt with as a separate type of degenerate points, known as the
merged wedge degenerate point (Fig. 8).
3.2. Segmentation
The separatrices of degenerate points are pathlines themselves
and they can grow until a surface border or another degenerate
point is reached [2830]. The reason for segmenting a surface
according to the separatrices of degenerate points is that the
separatrices form a skeleton of the PFD field which segments the

Fig. 8. Illustration of a merged wedge degenerate point.

entire design surface into regions where the pathlines have a


similar pattern. At locations near a degenerate point, the pathlines
can be described as bifurcating (trisector), folding (wedge), or
looping (merged wedge). In this section, segmentation of the
PFD field according to the identified degenerate points and their
separatrices will be described in detail.
3.2.1. Trisector degenerate points
As mentioned earlier, a trisector degenerate point has three separatrices. Since a separatrix is actually a pathline that intersects
the degenerate point, it can be established by first determining a
seed point close to the degenerate point and then letting the pathline grow. An applicable seed point in the vicinity of a degenerate point will have its preferred feed direction well aligned with
the vector from the degenerate point to the seed point. As shown
in Fig. 9, this essentially means that an applicable seed point will
have its cos = 1. To ensure reliable search results with reasonable computational costs, identification of the three seed points

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

Fig. 10. Determining the single separatrix seed point for a wedge degenerate point.
Fig. 9. Determining the three separatrix seed points for a trisector degenerate
point.

reason, segments of a pseudo-separatrix should be trimmed at the


intersections.

is done via a combined sampling and bisection method. With the


three seed points identified, each of the three separatrices can then
easily grow by a regular marching method until a surface border or
another degenerate point is reached.

3.2.3. Merged wedge degenerate points


In a PFD field where there are only two wedge degenerate
points and the method to obtain their separatrices and pseudoseparatrices has been applied, the result would ideally be a
single separatrix joining these two wedge degenerate points as
well as their corresponding pseudo-separatrices. The surrounding
pathlines will show a pattern of non-intersecting closed loops.
As the distance between the two wedge degenerate points
approaches zero, the pathline loops will take a more circular shape
if projected onto the xy plane. If the distance between the two
wedge points becomes zero and the two wedge points are literally
merged into one degenerate point, the value of cos will then
remain close to zero (Fig. 11). Under this situation, it is evident
that there is no preferred direction to form any separatrix for the
merged wedge degenerate point. So, for a surface whose PFD field
presents only one merged wedge degenerate point, the surface
cannot be segmented and is to be treated as a single region.

3.2.2. Wedge degenerate points


As mentioned earlier, a wedge degenerate point has only one
separatrix. The seed point for the wedge separatrix is determined
in a similar manner as the seed points for the trisector separatrices.
The only difference is that there will be only one feed angle for
cos = 1, corresponding to the single seed point (Fig. 10). The
single separatrix originated from the seed point cannot segment
the design surface into different regions. To segment the surface
into regions of similar pathlines, it is proposed in this work to add
a pseudo-separatrix. The pseudo-separatrix is to be constructed
at the wedge degenerate point. A wedge plane is first created at
the wedge point using the three-dimensional tangent vector of the
separatrix as the plane normal. The wedge plane then intersects the
design surface and the resulting intersection curve is the pseudoseparatrix. If a design surface has only one wedge degenerate point,
its separatrix and pseudo-separatrix will divide the surface into
three regions. It should be noted that, since the pseudo-separatrix
is not a pathline, it can intersect with other separatrices. For this

3.3. Principal tool path determination


Once the design surface has been segmented into regions of
similar pathlines, the next step is to cover these segmented regions
with iso-scallop tool paths. Note that existing iso-scallop tool path

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

Fig. 12. Generated iso-scallop tool paths using: (a) principal tool path; and (b)
border of the segmented region.

Fig. 11. Surrounding pathlines around a merged wedge degenerate point.

generation methods consistently employ a sequential tool path


generation approach and require the input of an initial tool path. In
this section, a method to determine the initial tool path based on
the regions pathlines is proposed. This initial tool path, referred
to as the principal tool path in this work, is determined in order
to reduce the drift of the generated iso-scallop tool paths from the
regions pathlines so that the resulting machining strip widths are
maximized across the surface region.
A typical iso-scallop tool path generation method generates
adjacent tool paths from both sides of the initial tool path until a
surface border is reached. If the initial tool path is chosen as one
of the PFD pathlines, the more adjacent tool paths are generated
(or grown from the initial tool path), the less similar the generated
tool paths are to the PFD pathlines. So, the principal tool path
should correspond to a pathline whose adjacent iso-scallop tool
paths would deviate as little as possible from the regions pathlines
as a whole. Determining the theoretical solution for the desired
principal tool path is a very challenging task, if at all possible.
For quick implementations, the middle pathline in the segmented
surface region is taken as the principal tool path. The middle
pathline is chosen because it has the evident benefit of a smaller
number of sequential tool paths to be generated from it, thereby
reducing the chance of the generated tool paths being departed
much from the PFD pathlines. The procedure to determine the
middle pathline is summarized here. First, let the surface region
be covered by a set of pathlines generated from a uniform grid of

seed points. The seed point resolution is to be kept high enough


so that the surface region is covered by a sufficiently large number
of pathlines. The minimal applicable seed point resolution varies
with the geometry of the segmented surface region. It has been
found from extensive computational tests that 50 pathlines are
generally sufficient. Then, the Hausdorff distances between all
possible pathline pairs are evaluated. Note that the Hausdorff
distance between two pathlines is the maximum of the distance
from any point on one pathline to the other pathline. For each
pathline, an average Hausdorff distance is calculated. The middle
pathline, or the principal tool path, is then the pathline with the
minimum average Hausdorff distance.
Once the initial (principal) tool path has been established for
the segmented surface region, the iso-scallop tool path generation
method of Feng and Li [9] is employed to generate the iso-scallop
tool paths across the surface region. Fig. 12(a) shows an example of
the principal tool path along with its adjacent iso-scallop tool paths
for a segmented surface region covered by pathlines. Fig. 12(b)
shows the iso-scallop tool paths generated from the border of the
segmented region. It can be seen that the iso-scallop tool paths
generated using the principal tool path as the initial tool path
follow the PFD pathlines more closely than those generated using
the regions border.
4. Case studies
The proposed method was applied to generating tool paths for
a variety of surfaces. A ball-end mill of radius r = 10 mm was
chosen and the scallop height constraint was set as 0.2 mm. Case
study 1 analyzed a cone-shaped surface where no degenerate point
could be identified and therefore, the whole surface was treated

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

Fig. 13. Case study 1: cone-shaped surface with no degenerate point.

as a single region (Fig. 13). Case study 2 involved a surface with


one trisector degenerate point, which segmented the surface into
three regions (Fig. 14). Case study 3 considered a surface with two
wedge degenerate points, each having its own pseudo-separatrix
and sharing the single separatrix with the other (Fig. 15). The
surface was thus segmented into four regions. The surface in case
study 4 contained one merged wedge degenerate point. As a result,
the whole surface was treated as a single region and only one
principal tool path was used (Fig. 16). A particular surface with
all types of degenerate point (three trisector, two wedge, and one
merged wedge points) was analyzed in case study 5. Eight distinct
regions were segmented, each having its own principal tool path
from which the iso-scallop tool paths could be generated (Fig. 17).
The last case study analyzed the corresponding mold surface of a
Bicycle Seat model. One trisector and two wedge degenerate points
were identified, dividing the surface into six regions (Fig. 18).
For each case study, the overall length of the generated tool
paths using the proposed method has been compared with that
using the conventional iso-parametric, iso-planar, and iso-scallop
methods. The results are summarized in Table 1. For the isoscallop method, both the u and v surface patch borders were
used as the initial tool path. In all cases, the proposed method
is seen to generate tool paths with shorter overall length. Large
improvements are expected and confirmed for iso-parametric
and iso-planar methods. It should be pointed out here that the
conventional iso-scallop method has been well accepted as the tool
path generation method to minimize redundant machining. With
the same scallop height constraint, the proposed method, however,

Fig. 14. Case study 2: surface with one trisector point.

is still able to further reduce the overall tool path length by a


notable percentage depending on the specific surface geometry.
As for computational efficiency, the traditional iso-parametric
tool path generation method is the most efficient, followed by the
iso-planar method, as expected. Generating the iso-scallop tool
paths has been known to take longer time due to its sequential
procedure of tool path generation. The proposed method is
evidently the least computationally efficient because of the much
added computational load to segment the design surface into
different regions. Nonetheless, it should be noted that unlike
the machining time, which is recurring for each part being
machined, tool path generation is an off-line computational task
that only needs to be performed once for a given part. Hence, the
much increased computational time is effectively acceptable since
reduction in the machining time is of much more value.
5. Conclusions
A new method has been presented in this paper to generate isoscallop tool paths which aim to follow preferred feed directions
(PFDs) on a given sculptured surface. The generated tool paths feature shorter overall tool path length under the specified scallop
height constraint, compared with tool paths generated by all existing methods. This is essentially achieved by segmenting the PFD
field of the given surface into regions of similar PFDs via identifying the PFD fields degenerate points and forming their separatrices. It is possible that due to the presence of islands/cavities of
specific geometry, the PFDs are similar across the island/cavity

10

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

Table 1
Comparison of overall length of tool paths generated by the various methods.
Tool path
generation
method

Case study 1: no
degenerate point

Iso-parametric
Iso-planar
Iso-scallop, u
Iso-scallop, v
Proposed

Case study 2: one


trisector point

Case study 3: two


wedge points

Case study 4: one


merged wedge point

Case study 5: three


trisector, two wedge,
and one merged
wedge points

Bicycle Seat

Tool
path
length
(mm)

Improvement Tool
path
length
(mm)

Improvement Tool
path
length
(mm)

Improvement Tool
path
length
(mm)

Improvement Tool
path
length
(mm)

Improvement Tool
path
length
(mm)

Improvement

10,424
8,211
6,978
6,461
6,294

39.6%
23.3%
9.8%
2.6%

61.1%
20.4%
6.4%
5.1%

9.0%
6.5%
5.1%
1.9%

16.9%
16.5%
5.4%
5.4%

37.6%
37.2%
7.2%
8.2%

48.9%
42.0%
12.7%
3.3%

28,978
14,165
12,049
11,887
11,277

8580
8352
8227
7964
7810

13,722
13,658
12,050
12,050
11,403

24,934
24,781
16,770
16,949
15,565

18,464
16,269
10,810
9,762
9,436

Fig. 15. Case study 3: surface with two wedge points.

boundaries. In this situation, the presented method will fail


to segment the islands/cavities into individual surface regions.
Nonetheless, islands/cavities of general geometry often produce
distinct PFDs, resulting in successfully segmented individual islands/cavities by the presented method.
During the implementation of the proposed method, a sampling
resolution needs to be chosen to map the PFD field of the
given surface. Computational time will be increased if the chosen
resolution is high. On the other hand, not all degenerate points
could be identified if the resolution is low. Determining the optimal
resolution for a given surface is a challenging task and beyond
the scope of the current work. As for the initial or principal tool
path for generating the iso-scallop tool paths for a segmented
surface region, the middle pathline in the segmented region has
been employed in order to reduce the drift of the generated

Fig. 16. Case study 4: surface with one merged wedge point.

iso-scallop tool paths from the PFD pathlines. The presented


procedure to find the middle pathline in each segmented region
is rather simplistic and the identified middle pathline will vary
slightly with the pathlines that are sampled in the region. A
more elaborated procedure will need to consider and evaluate the
resulting machining strip widths. It may lead to further reduction
of the overall tool path length, but not by much.
A couple of practical issues need to be noted here. As the
design surface is segmented into different regions, the tool entry
and exit motions for each tool path segment at the region
borders have to be carefully planned in order to reduce tool
engagement/disengagement marks. It should also be noted that
current iso-scallop tool path generation methods consistently

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

11

Fig. 18. Bicycle Seat surface with one trisector and two wedge points.

References

Fig. 17. Case study 5: surface with three trisector, two wedge, and one merged
wedge points.

employ a sequential tool path generation approach. Such a


sequential approach has been known to lead to undesirable highcurvature and/or looping tool path curves due to accumulative
changes in the sequentially generated tool paths. The present
work serves to address this issue, to some extent, by segmenting
the design surface into different regions of similar PDF pathlines.
However, seemingly sharp corners in the generated tool paths can
still be found. Such geometric discontinuity in the tool path much
affects the related machining performance. Balancing the tool path
smoothness and overall tool path length is a challenging task and
among the tool path generation research studies to be carried out
in our research group.
Acknowledgments
This work was funded in part by the Natural Sciences and Engineering Research Council (NSERC) of Canada. The scholarship
awards from the Foreign Affairs and International Trade Canada
and the Consejo Nacional de Ciencia y Tecnologa Mxico, which
supported the first author to complete the research reported in
this paper and receive his Masters degree in 2012, are also gratefully acknowledged. The Bicycle Seat model was downloaded from
the GrabCAD website (http://grabcad.com/) and used with the kind
permission from its creator Dusan Tijanic. The authors also wish to
thank the anonymous reviewers for their constructive comments
that have greatly helped to improve the quality of this paper.

[1] Loney GC, Ozsoy TM. NC machining of free form surfaces. Comput-Aided Des
1987;19:8590.
[2] Elber G, Cohen E. Toolpath generation for freeform surface models. ComputAided Des 1994;26:4906.
[3] He W, Lei M, Bin H. Iso-parametric CNC tool path optimization based on
adaptive grid generation. Int J Adv Manuf Technol 2009;41:53848.
[4] Huang Y, Oliver JH. Non-constant parameter NC tool path generation on
sculptured surfaces. Int J Adv Manuf Technol 1994;9:28190.
[5] Ding S, Mannan MA, Poo AN, Yang DCH, Han Z. Adaptive iso-planar tool path
generation for machining of free-form surfaces. Comput-Aided Des 2003;35:
14153.
[6] Suresh K, Yang DCH. Constant scallop-height machining of free-form surfaces.
ASME J Eng Ind 1994;116:2539.
[7] Lin RS, Koren Y. Efficient tool-path planning for machining free-form surfaces.
ASME J Eng Ind 1996;118:208.
[8] Sarma R, Dutta D. The geometry and generation of NC tool paths. ASME J Mech
Des 1997;119:2538.
[9] Feng HY, Li H. Constant scallop-height tool path generation for three-axis
sculptured surface machining. Comput-Aided Des 2002;34:64754.
[10] Tournier C, Duc E. A surface based approach for constant scallop height toolpath generation. Int J Adv Manuf Technol 2002;19:31824.
[11] Yoon JH. Fast tool path generation by the iso-scallop height method for ballend milling of sculptured surfaces. Int J Prod Res 2005;43:498998.
[12] Quinsat Y, Sabourin L. Optimal selection of machining direction for three-axis
milling of sculptured parts. Int J Adv Manuf Technol 2006;27:11329.
[13] Vijayaraghavan A, Hoover AM, Hartnett J, Dornfeld DA. Improving endmilling
surface finish by workpiece rotation and adaptive toolpath spacing. Int J Mach
Tools Manuf 2009;49:8998.
[14] Giri V, Bezbaruah D, Bubna P, Choudhury AR. Selection of master cutter paths
in sculptured surface machining by employing curvature principle. Int J Mach
Tools Manuf 2005;45:12029.
[15] Kim T, Sarma SE. Toolpath generation along directions of maximum kinematic
performance; a first cut at machine-optimal paths. Comput-Aided Des 2002;
34:45368.
[16] Chen ZC, Dong Z, Vickers GW. Automated surface subdivision and tool path
generation for 3 21 12 -axis CNC machining of sculptured parts. Comput Ind 2003;
50:31931.
[17] Dai Y, Zhang X, Ding H, Mu H, Wang C. Sub-regional flank milling method.
In: Proceedings of the 5th international conference on intelligent robotics and
applications. 2012. p. 22734.
[18] Zhu J, Hozumi A, Tanaka T, Saito Y. High efficiency tool path generation for
freeform surface machining based on NURBS subdivision. Key Eng Mater 2015;
625:3727.

12

G.H. Kumazawa et al. / Computer-Aided Design 6768 (2015) 112

[19] Barakchi Fard MJ, Feng HY. Effective determination of feed direction and tool
orientation in five-axis flat-end milling. ASME J Manuf Sci Eng 2010;132:
061011. 110.
[20] Johanna Senatore J, Segonds S, Rubio W, Dessein G. Correlation between
machining direction, cutter geometry and step-over distance in 3-axis milling:
Application to milling by zones. Comput-Aided Des 2012;44:115160.
[21] Lee YS, Ji H. Surface interrogation and machining strip evaluation for 5-axis
CNC die and mold machining. Int J Prod Res 1997;35:22552.
[22] Lee YS. Non-isoparametric tool path planning by machining strip evaluation
for 5-axis sculptured surface machining. Comput-Aided Des 1998;30:55970.
[23] Chiou CJ, Lee YS. A machining potential field approach to tool path generation
for multi-axis sculptured surface machining. Comput-Aided Des 2002;34:
35771.
[24] Chen Z, Dong Z, Vickers GW. Most efficient tool feed direction in three-axis
CNC machining. J Integr Manuf Syst 2003;14:55466.

[25] Chen Z, Dong Z, Vickers GW. A new principle of CNC tool path planning for
three-axis sculptured part machininga steepest-ascending tool path. ASME J
Manuf Sci Eng 2004;126:51523.
[26] Yoon JH. Two-dimensional representation of machining geometry and tool
path generation for ball-end milling of sculptured surfaces. Int J Prod Res 2007;
45:315164.
[27] Delmarcelle T, Hesselink L. Visualizing second-order tensor fields with
hyperstreamlines. IEEE Comput Graph Appl 1993;13:2533.
[28] Delmarcelle T, Hesselink L. The topology of symmetric, second-order tensor
fields. In: Proceedings of the IEEE conference on visualization. 1994. p. 1407.
[29] Delmarcelle T. The visualization of second-order tensor fields (Ph.D. thesis),
Stanford University; 1994.
[30] Tricoche X, Scheuermannn G. Topology simplification of symmetric, secondorder 2D tensor fields. In: Geometric modeling for scientific visualization.
Springer-Verlag; 2003. p. 17184.

You might also like