26

Integration of reverse engineering and

rapid prototyping with data reduction

Y. H Chen, C. T. Ng, Y. Z. Wang

Dept. ofMechanical Engineering, The University of Hong Kong
Pokfulam Road, Hong Kong,
Tel: 852-28597910, Fax: 852-28585415,
E-mail: YHCHEN@hkucc.hku. hk

Abstract
Reverse engineering is a methodology for constructing CAD models of physical
parts by digitizing an existing part, creating a computer model and then using it to
manufacture the component. When a digitized part is to be manufactured by means
of rapid prototyping machines such as stereolithography apparatus (SLA) and
selective laser sintering equipment (SLS), etc., it is not necessary to construct the
CAD model of a digitized part. This will be described by the proposed novel
method which can construct a STL file (the de facto file format for rapid
prototyping machines) directly from digitized part data. Further more, the STL file
can even be constructed in a way that significant data reduction can be achieved at
the users' discretion.

Keywords
Reverse engineering, rapid prototyping, triangulation, STL

1 INTRODUCTION

Reverse engineering is an important process in the design and manufacturing

context. This can be illustrated by two examples. First, a product that initially
exists in a designer's medium such as clay and wood, must have its surfaces

Computer Applications in Production and Engineering. F. Plonka and G. Oiling (Eds.)

C 1997 IFIP. Published by Cbapman & Hall
290

digitized in order to convert the model to a computer-based

representation(compatible with CAD/CAM systems). Secondly, if part drawings
are not available in CAD form such as some proven old designs or antiques, CAD
models of such parts can be constructed through reverse engineering techniques.
Given a CAD model of a part, traditional methods such as CNC machining can be
used to produce the parts. However, it is sometimes impossible for CNC machines
to machine complicated surfaces both rapidly and cost-effectively. Over the past a
few years, rapid prototyping machines have been widely used in industry. A
variety of rapid prototyping technologies have emerged(Yan and Gu, 1989, Miller
and Grote, 1995, Wallet et al., 1992). They include Stereolithography (SLA),
Selective Laser Sintering (SLS), Fused Deposition Manufacturing (FDM),
Laminated Object Manufacturing (LOM), Ballistic Particle Manufacturing
(BPM), and Three Dimensional Printing (30 Printing). These technologies are
capable of directly generating physical objects from a CAD model. They have an
important common feature: physical parts are produced by adding materials layer
by layer. This is opposite to traditional machining methods which make physical
parts by removing material.
In rapid prototyping, STL file format has become the de facto standard(3D
Systems, 1988). An STL file consists of a list of triangular facet data. Each facet
is uniquely identified by a unit normal and three vertices. The normal and each
vertex are specified by three coordinates each, so there is a total of 12 numbers
stored for each facet. Apart from this, triangular facets in an STL file must also
obey the following two rules.

Facet orientation
The facets define the surfaces of a 3-dimensional object. As such, each facet is
part of the boundary between the interior and the exterior of the object. The
orientation of the facet is specified redundantly in two ways which must be
consistent. First, the direction of the normal is outward. Second, the vertices are
listed in counter-clockwise order when looking at the object from the outside. This
rule is illustrated in Figure 1.

Nonnal

pt3 ~:..-------- ptl

.t/!.

Figure 1 Orientation of a triangular facet.

 Integration of reverse engineering 291

Vertex-to-vertex
Each triangular facet must share two vertices with each of its adjacent triangles.
In other words, a vertex of one triangle cannot lie on the side of another. This rule
is illustrated in Figure 2.

(a) Violation of vertex-to-vertex rule (b) Correct triangulation

Figure 2 The vertex-to-vertex rule.

Most recent commercial CAD/CAM software systems are capable of generating

STL files directly from a surface model. The majority of triangulation methods are
based on a known surface model(Schumaker, 1993, Sheng and Hirrch, 1992).
Triangulation of scattered data in 3D space often has an objective of constructing
smooth surfaces(Ox1ey, 1985, Park and Kim, 1995). Literature about optimized
STL file generation from reverse engineering data cannot be found. The most
relevant publication is found in Hamann paper(Humann, 1994). His proposal
removes a triangle based on curvatures at the three vertices of a triangle. Again,
his research has smooth surface fitting in mind. This paper describes a data
reduction method for automatic STL file generation directly from reverse
engineering data. The proposed method is based on surface normals of triangles
that share a data point.

2 AUTOMATED STL FILE GENERATION

There are various methods for part digitization. The presented research utilizes a
Mitutoyo BLN 122 coordinate measuring machine (CMM) to digitize a part. In
order not to miss any detail of part geometry, a large number of measurement
points are normally defined. If all points are used in STL file generation, the file
can easily becomes huge that the whole rapid prototyping process will be slowed
down significantly. In fact, keeping lots of data points in nearly planar regions is
rather unsophisticated. Removal of data points in these regions would not affect
the accuracy for rapid prototyping machines. A criterion based on normals of
neighboring triangles has been defined as to which points can be removed in
accordance with users' requirement.
Point data from CMM measurement is usually very regular and can be arranged
in matrix form. If all measurement points are to be used for STL file construction,
292

the problem becomes quite simple. In fact, the proposed method initially uses all
points to form triangles in the following simple way:

Given a data point matrix M(i,j)

where i = 1, ... rn;J = 1, ... Jt
Forj=1ton
fori= 1 to (m-1)
join M(i,j) to M(i+ 1,});
Forj = 1 to (n- 1)
fori= 1 tom
join M(i,j) to M(i,j+ 1);
fori= 1 to (m-1)
join M(i,j) to M(i+ 1J+ 1);
End
With the above procedure, data points can be triangulated as shown in Figure 3.
It can be seen that triangles formed in this ways conform to the STL file vertex-to-
vertex rule.

• • • • • •
• • • • • •
• • • • • •
• • • • • •
• • • • • •
(a) Data points (b) Triangulation

Figure 3 Initial triangulation.

To generate a complete STL file, normalized surface normals of triangles must

be specified. Given a triangle as shown in Figure 1 (pt~. pt2 and pt3 must be
arranged in counter-clockwise direction), normalized normals can be easily
calculated from the following procedure.

(a) Form two vectors V3t and V32 by

V3t= Ptt- Pt3

Vn = Pt2- Pt3 (1)

(b) Calculate surface normal from the cross product.

 Integration of reverse engineering 293

(2)

(c) Normalize n through the division.

A n
n=- (3)
lnl
With the above information, STL file can be readily generated.

3 DATAREDUCTION

3.1 Data reduction in percentage

If the digitized object has large planar or near planar regions, it is desirable to
remove some of the points in these regions. In order to identify which points can
be removed, a point weighting scheme based on a point surrounding triangles'
normals is developed and described in the following:

(a) Search all triangles A that share the point in concern.

(b) Let one of the surface normal U in ll be the reference vector, calculate
subtractions of U With the rest of surface normals in A.

UV1 =V1 - U
UV2=V2- U
•
•
•
uv.=v.- u (4)

(c) Transform all resulting vectors to first quadrant by taking the absolute values
of all vector components, e.g. uv2 = ai + bj + ck is transformed to uv2 = lali
+ lb~ +!elk.
(d) Sum up all the resulting vectors.

(5)

(e) Calculated the average quantity d by

294

(6)

If d is small for a given point, the point stands a higher chance of being
removed. Which points are to be removed depends on users' specifications. If a
user want 100/o of the point data be removed, then points with the lowest 100/o of d
will be removed.

3.2 Data reduction by bounded error

In computer visualization, polygonal mesh, triangles in particular are usually used

to approximate object surfaces. In the triangulation of an object surface, accuracy
is often controlled by bounded error(Kalvin and Taylor, 1996). The bounded error
concept is also adapted in this research to control errors resulting from data
reduction of an STL file in rapid prototyping.
In data reduction, two strategies can be used. One is to remove the point one at a
time. The other is to remove several points, or a patch at the same time. After
consideration of error accumulation and computation efficiency, patch removal
strategy is implemented in this research.
Suppose a patch has k triangles. The normals and areas of the k triangles are nt
to Ilk and St to St respectively. Let

(7)

The patch direction normal is calculated by:

(8)

" n
n=- (9)
lnl

The direction normal n" is an indication of the direction of maximum projection.

Based on the patch direction normal, a minimum enclosing box of a patch can be
calculated. For illustration purpose, Figure 4(a) shows a patch and its bounding
box. After data reduction (Suppose the three points in the middle are removed),
the new patch and its bounding box are shown in Figure 4(b). The bounded error
Ep is measured along the patch direction normal, that is

Ep=:O., -D. (10)

 Integration of reverse engineering 295

i
Patc:h nonnal diredlon

(A) Before removal

i
Patc:h nonnal diredlon

(B) After removal

Figure 4 Error definition.

This error can be specified by the user so that the accuracy after data reduction is
still within control.
296

4 RE-TRIANGULATION

When a point is to be removed, triangles sharing this point must be removed too.
This leaves a lank region' which must be covered through re-triangulation. The
implemented re-triangulation method is described in the following:

Point to be
dlleted

(a) Point P to be removed (b)Distance Calculation

The shortest
one is found
and use it to
construct one
trian&le

(c) Form a new triangle (d) The rest of triangles are formed in the same way

Figure 5 Re-triangulation.

(l) Form a point list L that are affected by the removal of a point. In figure 5(a),
if point Pis to be removed, the affected point list will beL= {Pt. P2, P3, P4,
P5, P6}.
(2) Take three consecutive points from the point list L and check whether the
triangle formed by the three points is inside or outside the region formed by
L. If not. take another three consecutive points and check again.
(3) Calculate the length of the line formed by the first and the last point of the
three points as shown in Figure 5(b) in dashed lines.
(4) Repeat (2) and (3) until all combinations of consecutive three points have
been exhausted.
(5) Find the shortest length and use the corresponding three points to form a real
triangle as shown in Figure 5(c). Calculate the normal of the newly formed
triangle and delete the middle point of the three points from L have Lt.
(6) Repeat from (2) with point list L until no more than four points left in Lt.
 Integration of reverse engineering 297

Figure 5(d) shows are-triangulated region after a point is removed. Point removal
and re-triangulation will go on iteratively until certain criteria are met. Two
criteria are given for users' specification. One is the percentage of points to be
removed and the other is the bounded error caused by points removal.

5 EXPERIMENTAL RESULTS

With the implemented method, a number of examples have been tested. One of
the example is a human face. It was digitized using a Mitutoyo BLN 122
coordinate measuring machine as shown in Figure 6. As some parts of the human
face has very low curvature change, some data points can be removed. In the
experiment, data reduction for 20%, 50% and 80% have been specified for

Figure 6 STL file without data reduction.

298

Figure 7 STL file after 50% data reduction.

prototype fabrication. Figure 7 shows the STL file after 50% of data reduction. It
can be seen that regions with small curvature changes have large data reduction.
This is exactly what is desired.

6 CONCLUSION

This paper has presented a novel method for STL file construction directly from
CMM measurement data. When desired. the user can specify either the percentage
of points to be removed or the bounded error caused by points removal in order to
speed up computation and to reduce storage space. Several prototypes with
different amounts of data reduction have been built using an SLS equipment
which shows satisfactory results.
 Integration of reverse engineering 299

7 ACKNOWLEDGEMENT

This research is carried out with a CRCG grant (code:337/064/0023) from the
University of Hong Kong.

8 REFERENCES

3D Systems (1988) Stereolithography interface specification. USA.

Hamann, B. (1994) A data reduction scheme for triangulated surfaces. Computer
Aided Geometric Design, 11, 197-214.
Kalvin, A.D. and Taylor, R.H. (1996) Superfaces: polygonal mesh simplification
with bounded error.IEEE Computer Graphics & Applications, May, 64-77
Miller, J.L. and Grote, K.H. (1995) Solid freeform manufacturing technologies as
an important step in the product development process. Computers in Industry,
28, 11-16.
Oxley, A. (1985) Surface fitting by triangulation. The Computer Journal, 28(3),
335-339.
Park, H. and Kim, K. (1995) An adaptive method for smooth surface
approximation to scattered 3D points. Computer-Aided Design, 27, 929-939.
Schumaker, L.L. (1993) Triangulations in CAGD. IEEE Computer Graphics &
Applications, January, 47-52.
Sheng, X. and Hirrch, B.E. (1992) Triangulation of trimmed surfaces in
parametric space. Computer-Aided Design, 24(8), 437-444.
Wall, M.B., Ulrich, K.T. and Flowers, W.C. (1992) Evaluating prototyping
technologies for product design. Research in Engineering Design, 3, 163-177.
Yan, X. and Gu, P. (1996) A review of rapid prototyping technologies and
systems. Computer-Aided Design, 28(4), 307-318.

9 BIOGRAPH

Dr. Y.H. Chen is an assistant professor in The Department of Mechanical

Engineering, The University of Hong Kong since 1993. Before joining The
University of Hong Kong, Dr. Chen has worked in Motorola Electronics Pte, Ltd
(Singapore), Asia Matshushita Electronics Pte, Led (Singapore), and Swire
Technologies Pte, Ltd (Hong Kong) as senior automation engineer, senior
research engineer and automation manager respectively. Dr. Chen research
interest include CAD/CAM, reverse engineering and rapid prototyping.
Mr. C.T. Ng is an M.Phil research student under Dr. Chen.
Dr. Y.Z. Wang is currently a research associate working with Dr. Chen. He is a
lecturer in The Department of Mechanical Engineering, Tianjin University, P.R.
China