Vous êtes sur la page 1sur 10

J Mar Sci Technol (2009) 14:275284

DOI 10.1007/s00773-008-0033-2

ORIGINAL ARTICLE

Generating cutter paths for marine propellers without


interference and gouging
Yoo-Chul Kim Youn-Mo Lee Myeong-jo Son
Tae-wan Kim Jung-Chun Suh

Received: 21 April 2008 / Accepted: 27 September 2008 / Published online: 14 November 2008
JASNAOE 2008

Abstract We generate cutter paths free of interference


and gouging for manufacturing model propellers using a
five-axis numerical control (NC) machine. Our approach is
faster than using a general-purpose computer-aided design
(CAD)/computer-aided manufacturing (CAM) system.
A roughing cut is made using only three axes for efficiency,
and the finishing cut is made using all five axes to avoid
collisions. Elements of the cutter path that might produce
gouging are eliminated and the pose of the cutter is
adjusted to eliminate interference. A number of models,
including surface-piercing propellers, have been
manufactured.
Keywords Cutter-path generation 
Propeller CAD/CAM  Numerical control (NC) 
B-spline surface

1 Introduction
Five-axis numerical control (NC) machine tools are widely
employed for machining model propellers and ship hulls at
research institutes, using standard CAD/CAM systems to
generate the cutter paths. In the case of ship hulls, it is
comparatively easy to generate cutter paths because there is
little chance of interference between the cutter and the hull
surface, and it is easy to position the workpiece satisfactorily. However, in the case of propellers, there are

Y.-C. Kim  Y.-M. Lee  M.-j. Son  T.-w. Kim (&)  J.-C. Suh
Department of Naval Architecture and Ocean Engineering
and Research Institute of Marine Systems Engineering,
Seoul National University,
Seoul 151-744, South Korea
e-mail: taewan@snu.ac.kr

workholding problems and many opportunities for interference, which means that a lot of expertise is required to
generate the required cutter paths using a standard CAD/
CAM system (Fig. 1). This suggests that a system specifically for marine propellers may be advantageous. There
are already a few dedicated systems, but they are not
widely used. Kuo and Dzan [1] and Youn et al. [2]
investigated the five-axis machining of a marine propeller,
but an analysis of all the steps required to machine a propeller is daunting because of the complicated geometry
involved.
We present a method of generating cutter paths which
makes it easy to machine model propellers on a five-axis
NC machine tool. We start with surface models of the
propeller blades and the root fillets. Then we generate
cutter paths for rough and finish machining from which
interference and gouging are eliminated. We show some
machined results and then give suggestions for further
work.

2 Propeller geometry
It is necessary to understand the sophisticated geometry of
a propeller before going on to generate cutter paths.
A propeller is defined in Cartesian coordinates oriented so
that the x-axis is in the downstream direction, the y-axis is
vertically up, and the z-axis goes off to the port side of the
ship, as shown in Fig. 2.
2.1 Fundamental geometry of a propeller
The gross geometric description of a propeller can be
separated from any coordinate system. The quantities
shown in Fig. 4 define the overall shape of a propeller, and

123

276

J Mar Sci Technol (2009) 14:275284


Definition
of a propeller
geometry

Plan
for
cutter-path

Plan
for
cutter-pose

General purpose CAD/CAM system


Input
fundamental quantities
of a propeller

Generation
of
cutter-path

Specified system

Fig. 1 Comparison between general and propeller-specific CAD/


CAM systems

Fig. 3 Cylindrical blade section definition (source: p. 34 of [3])

Fig. 2 Reference frame of a propeller

are the basis of more detailed blade geometry, which starts


in helical coordinates and proceeds through cylindrical
coordinates to Cartesian coordinates. The quantities that
define the overall geometry of the propeller are the radius,
r, the number of blades, Z, the chord length, c(r), the
maximum blade thickness, tmax(r), the camber, fmax(r), the
skew angle, hs(r), the total axial displacement, iT(r), and
the pitch, p(r) as shown in Figs. 3, 4, 5, and 6. The station,
s, the thickness distribution, t/tmax(s), and the meanline
distribution, f/fmax(s) define the blade section. These fundamental geometry of a propeller are described well in
reference books [3, 4].

Fig. 4 Fundamental quantities describing the overall geometry of a


propeller

2.2 Definition of propeller geometry

meanline distribution, thickness distribution, chord length,


camber, and maximum thickness. We define the suction
side, which is the upper side of a blade section, by adding
half of the thickness to the meanline. The pressure side,
which is the lower side of a blade section, is created by
subtracting half of the thickness from the meanline
(Fig. 7).

Propeller geometry is defined by stacking up the blade


sections radially on concentric cylindrical surfaces. We
start by defining the blade sections in helical coordinates
along the pitch helix, convert these sections to cylindrical
coordinates, and finally into Cartesian coordinates. The
axis n1 is defined along the pitch helix, and n2 is perpendicular to n1 in helical coordinates. The origin of n1 and n2
is located at the mid-chord position. The blade section
geometry is defined in the n1n2-plane by using the station,

s  50
 cr;
100
n2 EC  ET ;
f
EC
s  fmax r;
fmax
t
ET
s  tmax r  0:5:
tmax
The blade sections are expressed in helical coordinates in
terms of the skew angles and axial displacements as follows:

123

n1

J Mar Sci Technol (2009) 14:275284

277

Fig. 5 Definition of pitch: a


helix definition on a cylinder of
radius r, and b development of
helix on the cylinder (source:
p. 35 of [3])

Fig. 6 Definition of maximum thickness, chord length, and camber

Fig. 8 A point on the blade in helical coordinates

Fig. 7 Definition of a blade section

p iT  n2 cos / n1 sin /; rhs n1 cos / n2 sin /;


iT iG is iG rhs tan /;
where iT is the total axial displacement, which is defined by
adding the skew-induced axial displacement to the rake,
and / is the pitch angle. A point p (Fig. 8) on the blade
surface is defined in helical coordinates and converted via

cylindrical to Cartesian coordinates. Helical coordinates


are converted to cylindrical coordinates using the following
equations:
r r;
rhs n1 cos / n2 sin /
h
;
r
x ir n1 sin /  n2 cos /:
It is then simple to convert the cylindrical components to
Cartesian coordinates:

123

278

J Mar Sci Technol (2009) 14:275284

x iT n1 sin /  n2 cos /;
y r cos h
z r sin h:
In the case that a propeller has a tapered hub, there are
three methods that can be used to define a propeller
geometry in non-Cartesian coordinates [5]. These methods
use the same way mentioned above but defining a pitch line
in the tapered cylinder.
2.3 Definition of the root fillet
The root fillet reduces stress at the base of the blade. There
are several ways to define a root fillet. One is to generate a
series of planes that are perpendicular to the root surface
and to the blade, then create a fillet curve in each plane, and
join them up. Another is to find the intersection between
offsets from blade and hub directly. Both of these methods
require intersection curves or new surfaces to be created.
We propose a simpler way to define the root fillet using the
radial distribution of blade thickness.
We can express the cross-section of the hub and the
thickness distribution of the blade on a perpendicular to the
pitch helix of the root section. A new thickness distribution
which includes the root fillet can then be created with
tangential continuity to the original thickness distribution
and at the hub surface. The constraints for finding the new
thickness distribution are as follows:
Condition 1: An arc of radius 3T and the thickness
distribution have a common tangential vector at p1 that
is a distance T from the root.

Fig. 10 Propeller geometry with and without a root fillet

Condition 2: An arc of radius T/3 and hub have a


common tangential vector at p2 on the hub.
Condition 3: The arcs of radius 3T and T/3 have a
common tangential vector at p3.
We can position p1, p2, and p3 on a plane perpendicular to the pitch helix using these conditions to create
a new thickness distribution which includes the root fillet
(Fig. 9). The fillet is then added to the blade definition of
Sect. 2.2. Figure 10 shows a propeller with and without a
fillet. This method creates the whole of the root fillet
using the thickness distribution, but we can create different root fillet geometry at the leading and trailing
edges if a chord length distribution is defined in the same
way.

3 Surface modeling

Fig. 9 Generating a root fillet using the thickness distribution

123

A surface model is essential if we are to generate closely


spaced cutter paths. We generate a large number of radial
sections (around 200) by spline interpolation of the fundamental geometric quantities of a propeller described in
Sect. 2.2 and fit these cross-sections with a cubic nonuniform rational B-spline (NURBS) surface [6].

J Mar Sci Technol (2009) 14:275284

279

Fig. 13 Definition of a hub surface

Fig. 11 Direction of the parameter u along the chords of two blades:


National Advisory Committee for Aeronautics (NACA) and surfacepiercing propellers

Fig. 14 Five-axis milling machine

Fig. 12 Interpolated blade surface

3.1 Blade modeling


The u parameter of our NURBS model runs along the
chord of the blade, and the v parameter is in the radial
direction, so that a blade surface can be modeled by laying
sections radially. Figure 12 shows an interpolated blade
section. A blade generally has a tangential discontinuity at
its trailing edge. We therefore make the trailing edge on the
suction side the starting point of the u parameter, and the
trailing edge on the pressure side its end point. However, a
surface-piercing propeller also has a tangential discontinuity at its leading edge. In this case we model the suction
and pressure sides separately, as shown in Fig. 11.
3.2 Hub modeling
The hub surface is the cylindrical surface between the
pressure side of a blade and the suction side of the
neighboring blade. We develop the cylindrical surface into

a plane and define the suction and pressure sides of the root
section on that plane, and then transform it back to a cylinder. We can use the same approach for a surface-piercing
propeller, even though it does not have a single trailing
edge. The planar development of the hub surface of a
surface-piercing propeller, and the same surface after
transformation back into three dimensions, are shown in
Fig. 13. In the case of a tapered hub, it is impossible to
develop the tapered hub surface into a plane. However, on
the hub surface, the suction side point and the pressure side
point that have the same x component have the same
radius, which means that a curve between two points is a
part of a circle. This makes it possible to define hub surface
as a B-spline surface in three-dimensional (3D) space.

4 Cutter-path generation
Machining a model propeller requires a roughing and a
finishing cut. Youn et al. [2] suggested an additional
semifinish cut, but this is not necessary for propellers with
six or fewer blades, and we only describe two cuts. Our
five-axis NC machine tool has three Cartesian axes (x, y, z)
and a table with two rotational axes (a and b), as shown in
Fig. 14. b is the tilt angle, which is a rotation about the yaxis of the machine, and a is the rotation angle, which is a
rotation about the propeller axis. A five-axis machine is
used for machining sophisticated geometry because it
allows the cutter to adopt a wide range of positions.

123

280

J Mar Sci Technol (2009) 14:275284

Fig. 15 Overview of the rough cutting process

4.1 Roughing cut


The roughing cut creates the approximate shape of a propeller from a cylindrical blank. Only the Cartesian axes of
the machine tool are used at this stage. The pressure side of
the propeller is machined with one setup and the suction
side with another. The material between the blades is
removed first, and then the blade surfaces are formed.
Figure 15 shows the overall procedure to generate the
roughing cutter path. A coordinate transformation is necessary because the coordinate system in which the
propeller is defined is different from that used by the
machine tool. The axis of rotation of the propeller is made
to correspond with the z-axis of the machine tool. Contours
are generated at successive z-coordinates to give an
appropriate depth of cut.
To find the intersection curve between the blade surface
and each roughing plane, which is normal to the propeller
axis, we use Bezier subdivision. First, we convert the Bspline surface to Bezier patches by inserting knots. Then
we look for intersections, starting at the patch nearest to the
hub and proceeding to the patch at the tip. If the intersection passes through a Bezier patch, we obtain a curve in
the parameter space of the patch by fixing one parameter,
and then finding the intersection between the corresponding Bezier curve and the roughing plane by subdividing the
curve. We can decide quickly whether there is an intersection using the convex-hull property, and the solution
always converges. The blade silhouettes are obtained by
projecting the leading and trailing edges onto to a plane
perpendicular to the axis of rotation. The silhouette and
each contour on the blade surface are then offset and their
intersections are computed, as shown in Fig. 16. The gap
between each pair of adjacent blades is bounded by the two
blade silhouettes, and these regions are cut in the same
way. Any self-intersection in the resulting contours must be
removed as it might cause gouging. We then generate a
zigzag pattern, running from tip to hub, and clip it to the
contour. An appropriate allowance for finishing must also
be made in the x, y, and z directions.
Figure 17 shows a schematic representation of a roughing
cut, and Fig. 18 shows some actual roughing cutter paths.

123

Fig. 16 Blade and outline contours at one cutting depth

Fig. 17 Schematic of a roughing cut

4.2 Finishing cut


The finishing cut follows the roughing cut and creates the
final shape of the propeller. Finishing is performed in a
single setup to ensure accuracy. We use a ball endmill and
all five axes of the machine tool. An overview of the finishing cut is shown in Fig. 19. Initially, we locate the
propeller so that its axis is aligned with the z-axis of the
machine tool and locate the coordinate origin at the center
of the tilting axis b, as shown in Fig. 20.
Cutter-location (CL) points are calculated by offsetting
the cutter-contact points along the normal from the surface

J Mar Sci Technol (2009) 14:275284

281

Fig. 21 Cutter-contact (CC) and cutter-location (CL) point


Fig. 18 Roughing cutter paths

Fig. 19 Overview of finishing cut

Fig. 22 Gouging

Fig. 20 Machine coordinates for finishing cut

(Fig. 21). Finally, the cutter pose is determined to avoid


interference between the stem of the cutter and the blades.
The cutter-contact points are defined in terms of the (u, v)
parameters because the propeller surface is modeled as a Bspline surface and the cutter paths run along isoparametric
curves. The parametric interval between cutting paths is
determined using the out-of-tolerance method [7] to ensure
that the required accuracy is achieved.

Gouging will occur if the offset radius is larger than the


curvature of the surface, but it can happen in other situations, as shown in Fig. 22. Many offset schemes [8, 9]
cannot deal with the more complicated cases.
Most of a propeller blade is convex and smooth, which
makes it relatively easy to machine. There is a higher
possibility of gouging in the root region because of the root
fillet. We calculate the length between each offset point
and a number of points on the surface in order to eliminate
gouging. We create a set of inspection points distributed
over the surface, and find which is the nearest to the offset
point. If the length between the nearest point and the offset
point is less than the cutter radius, gouging is detected and
this offset point is excluded from the cutter path. We need
enough inspection points to ensure accuracy, and we

123

282

J Mar Sci Technol (2009) 14:275284

Fig. 23 Gouging-free region in the parameter domain

determine how many to use from the second derivatives of


the surface [10] as follows:
 pow r
1
M1 2M2 M3
;
n

8

Fig. 24 Gouging-free cutter path

where
pow 0:5;
M1 ksuu u; vk;
M2 ksuv u; vk;
M3 ksvv u; vk;
 tolerance;
sxy x

and

y directional derivatives:

Since the root fillet is defined by two arcs of radius 3T and


T/3, we can limit the search for gouging to a region less
than 3T from the root. Figure 23 shows the gouging-free
region in the parameter domain, and Fig. 24 shows the
resulting cutter paths.
We also need to avoid interference between the stem of
the cutter and the workpiece. Choi and Jerard [7] suggested
that the cutter pose can be determined by checking for
collisions at sample points with a range of angles of rotation and tilt. An alternative method using check vectors on
each blade was introduced by Youn et al. [2]. The former
method requires a lot of computing time and the latter can
miss interferences because the cutter is regarded as a single
vector without a volume. We use a modified version of the
first method to do some checks, but once a collision-free
situation is established, we assume that it will perpetuate
because there are no sudden changes in cutter position,
since the blade is smooth.

123

Fig. 25 Interference map

We conduct interference checks at sample points on each


isoparametric curve of the cutter path. These sample points
are created using half-cosine spacing, so that they are closer
together at the leading edge. We make an interference map
for each isoparametric curve of the cutter path, as shown in
Fig. 25. Then we calculate the centroid of the feasible
region of rotation and tilt at each sample point and interpolate these centroids using a cubic spline in ab space.
This curve is used to determine the pose of the cutter.
If the axis of the cutter is initially aligned along the
vector (0, 0, 1), it is modified by the tilt and rotation angle,
a and b, as follows:

J Mar Sci Technol (2009) 14:275284

283

Fig. 26 Calculation of distance between inspection points p1, p2, and


p3 and the cutter

Fig. 28 Cutter path for the fishtail region of a surface-piercing


propeller

Fig. 27 Finishing cutter path

Fig. 29 Finishing cutter path for the hub surface

inspection point is less than the cutter radius, interference


will occur (Fig. 26).
The finishing cutter paths traverse the blades in zigzag
curves of constant v. The starting value of v is not at the tip
itself because the chord length at the tip is generally very
small so that starting there would result in diverging tool
paths. Figure 27 shows an example of a finishing cutter path.
In the case of a surface-piercing propeller of the type
shown in Fig. 11, an uncut region remains after the pressure and suction sides of the blades have been machined
because there are two trailing edges. An additional cutting
process is needed to machine this so-called fishtail region,

1
cos a sin b
@  sin a cos b A:
cos b
The pose of the cutter is verified by calculating the
distance between the inspection points on the surface and
the cutter. A ball endmill is hemispherical below its center
point and cylindrical above. Therefore, the distance
between an inspection point and the cutter center is used
when the inspection point is below the cutter center;
otherwise the required distance is that between the
inspection point and the cutter axis. If the distance to the

123

284

J Mar Sci Technol (2009) 14:275284

5 Conclusions and discussion


We have proposed a simple method to generate cutter path
free of interference and gouging to machine models of
conventional and surface-piercing propellers. Figure 30
shows the overall process of manufacturing a model propeller, and Fig. 31 shows model DTNSRDC P4381 and
P4382 propellers, and a modified 841-B propeller of
150 mm diameter, which were all manufactured using the
techniques described above. Factors such as choice of
material, machining conditions, and cutter vibration
determine the machining accuracy and have not been dealt
with herein. Research on these aspects is essential to
guarantee accuracy.
Acknowledgments This work was supported by grant nos. R012004-000-10518-0 and R01-2005-000-11257-0 from the Basic
Research Program of the Korea Science and Engineering Foundation,
and in part by the RIMSE R&BD Program.
Fig. 30 Process of manufacturing a model propeller

References

Fig. 31 Completed model propellers

which is a ruled surface that connects the two trailing


edges. The cutter path for this region is generated in a
similar manner to those for the sides of the blade, as shown
in Fig. 28.
The hub surface is a cylinder interrupted by the blade
surfaces. Interference can occur at the border between the
different surfaces, and so the cutter path for the cylinder
omits the regions close to the blade surfaces. In this case
the parameter u is in the direction of the propeller axis and
the parameter v is in the direction of an arc of the hub
surface. We generate a zigzag cutter path in the direction of
v for successive increments of u, as shown in Fig. 29.

123

1. Kuo H-C, Dzan W-Y (2002) The analysis of nc machining efficiency for marine propellers. J Mater Process Technol 124:389
395
2. Youn JW, Jun YT, Park SH (2003) Interference-free tool path
generation in five-axis machining of a marine propeller. Int J Prod
Res 41(18):43834402
3. Carlton J (2007) Marine propellers and propulsion, 2nd edn.
Butterworth-Heinemann, London
4. Kim YC (2006) Design of propeller geometry using blade section
adapted to surface streamlines. PhD thesis, Seoul National University, South Korea
5. Neely SK (1997) Non-cylindrical blade geometry definition.
Propellers/Shafting97 Symposium 31:1311312
6. Farin G, Hansford D (2000) The essentials of CAGD. A. K.
Peters, USA
7. Choi BK, Jerard RB (1998) Sculptured surface machining. Kluwer, Dordrecht
8. Aomura S, Uehara T (1990) Self-intersection of an offset surface.
Comput Aided Des 22:417422
9. Ravi Kumar GVV, Shastry KG, Prakash BG (2002) Computing
non-self-intersecting offsets of nurbs surfaces. Comput Aided
Des 34:209228
10. Piegl LA, Tiller W (1999) Computing offsets of nurbs curves and
surfaces. Comput Aided Des 31:147156

Vous aimerez peut-être aussi