On the Development of Lofts

for Doubly Curved Sheet Metal Components

K.S.R.K. Prasad *
Professor Emeritus and Principal, Institute of Aeronautical Engineering, Hyderabad 500043, INDIA

P. Selvaraj, Praveen V. Ayachit and B.V. Nagamani

Aeronautical Development Agency, P.B.No.1718, Vimanapura Post, Bangalore 560 017, India

Abstract
Practical automated flat pattern generation with inbuilt production features for doubly curved sheet metal
components (SMCs) is addressed here utilizing a new and unique Point Transformation Algorithm (PTA). This is the
third in the series of papers on practical Flat Pattern Development (FPD) [8] and Production Loft Generation
Systems (PLGS) [9] complementing the pioneering work [6,7]. In the first two publications, automated loft
generation programs have addressed sheet metal components having a Principal Flat Surface (PFS) only. The flat
pattern development of 3-D components that do not have the flat surface(termed as Non-PFS components) having
complex features of double curvature in addition to cutouts and nibbled holes typical of aircraft components were
so far not addressed due to lack of relevant published algorithms. This paper traces the evolution of developments
and provides the record of fully illustrated, automated loft generation scheme for aircraft SMCs including the Non-
PFS components which underwent validation through production tests by sponsors. Details of some of the unique
features of the system like simplified surface model generation, termed as topological model and powerful
algorithms deployed with potential for CAD/CAM applications are included.

---------------------------------------------------------------------------------------------------------------------------------------------
Key Words: FPD, SMC Loft Generation, Point Transformation Algorithm, Topological Model

* Corresponding author:
Professor Emeritus and Principal, Institute of Aeronautical Engineering, Dundigal Village, Quthbullapur (Mandal),
Ranga Reddy (Dist.), Hyderabad – 500043, India
Tel: +91……………….
Fax: +91……………
1. Introduction

Loft is a planar drawing (obtained through the development of SMCs) which incorporate production and
inspection features suited to manufacturing, transfer to astrofoil sheets, released for production. In sheet metal
components with PFS, the direction of strain during forming is well defined, whereas in non-PFS components this is
not the case. Mathematically, doubly curved surfaces are non-developable [3]. Approximate Flat Patterns (FP)
within specified tolerances are accepted in practice. Hence, the method of generating FP for a doubly curved surface
depends on the material characteristics of the component. Therefore, different methods have to be used to obtain FP
for diverse fields such as clothing, shoe manufacturing, aircraft, automotive and naval industries.

Further, a brief survey of literature pertinent to development of 3D doubly curved surfaces complementing the
earlier reported survey for 2D surfaces [8,9] revealed very interesting developments. For example, surface
curvatures in terms of mathematical models of splines and patches as well as mapping techniques were utilized by
Hinds et al [4] for FPD which after trimming (to get rid of excess material) and stitching (to take care of cuts) where
needed for cloth industry. Parida and Mudur [5] addressed planar development of complex surfaces using
geometric features (binomial patches) as inputs subdivided into triangular grids and obtained flat pattern without
considering material or thickness characteristics. The resulting FP had cuts and overlaps which they minimized and

1
deployed for tape layouts in composites (utilized by the sponsorer, Aeronautical Development Agency (ADA) for
composite manufacturing concurrently with authors’ SMC developments [6] also for ADA ). Among the grid
techniques, even finite element techniques were also utilized. Azariadis and Aspragathos [1] have shown greater
appreciation for the need of multitude of industry specific technological developments to suit variety of fields like
aircraft, ship, cloth and shoe industry and presented an approximate design of FP for 3D surfaces. The refinements
(based on surface geometry or generalized inverses) of their FP have reduced gaps and overlaps to suit development
of upper 3D surfaces of shoe. Azariadis et al. [2] also presented a genetic algorithmic approach to solve global
minimization problem under constraints to refine the FP. In summary, both the geometrical constituent manipulation
techniques as well as discretization techniques involving elemental grids in deriving FP in conjunction with
optimization techniques for incorporating constraints and minimization processes have no doubt reduced cuts, gaps
and even overlaps. Yet, they have fallen short of the demands of the sheet metal loft development requirements of
aircraft SMCs, which needed integration of material characteristics and bi-directional flow of metal besides
incorporation of geometrical features. Thus, making a case for evolution of aircraft SMC application specific
practical loft generation procedures.

Research and development of automated Sheet Metal Flat Pattern Development is sponsored by ADA for
over a decade. These developments are authenticated through appropriate validation tests, incorporating essential
production features facilitating loft generation for a restricted class of components (having a Principal Flat Surface
termed as PFS SMCs). These, as well as the procedures in vogue at the industrial lofting group, and in commercial
systems together with a modest indigenous three dimensional application activity have already been published [8,9].
Where, aircraft SMCs have been classified into two major groups: (a) Principal Flat Surface (PFS) (which possess
at least one PFS) category and (b) Non-PFS category. The latter, have no reference PFS and possess curvature in
more than one direction. Such components (referred as set III in that classification), constitute approximately 20% of
aircraft components, have a part of Numerical Master Geometry (NMG). All these technologies were utilized in
the prototype development of Light Combat Aircraft due to Aeronautical Development Agency (ADA) in India in
conjunction with production tests carried out at Hindustan Aeronautics Limited (HAL) and their utilitarian summary
since inception in 1990 till date is also due to appear shortly [10]. Based on these, it is evident that Non-PFS
category of aircraft components are the most complex for which reliable techniques of FPD are unavailable. In what
follows we shall address the evolution of unique omni potent technique which paved the way for loft generation of
not only the earlier addressed components but also the components with bi-directional curvatures with production
features.

For these components, the Point Transformation Algorithm (PTA) is exclusively developed for the first
time and is illustrated here. To generate the points on the component geometry, different component geometry
specific methods deploying radial, parallel, ortho normal and hybrid sectioning have been systematically
experimented with. It also utilizes a unique highly simplified SMC CAD model, termed as the Topological Model
(TM) consisting of the component geometry defined through a portion of NMG delineated by 3D surface curves
bounding the basic component surface (does not include projections like nibbled holes) While importing faces
(patches derived from parent surface using boundary curves) appearing in a sheet metal surface model or a solid
model - from one CAD system to another, there is a possibility of data loss, as all translators do not support these
entities. Also, as most CAD systems have translators for importing and exporting points, curves and surfaces, the
TM is found to be accurate, reliable and extremely simple to import. It serves to substitute the FPD methods based
on the surface model (even for the components which were successfully addressed earlier) without the need for
generating the complete SMC surface model. This technique utilizes the PTA and the computational efforts are
reduced employing well known recursive procedure, the adaptation of which in the present context is termed as
Recursive Division Algorithm (RDA).

The PTA transforms a point on the defining contour of a three-dimensional surface on to the FP. This
transformation also embeds the Prasad’s Bend Correction (BC) algorithm [6,9] (at every discretization step) which
facilitates accurate determination of SMCs blank’s dimensions incorporating direction dependent plastic flow. It is
now adapted to the 3D surfaces with double curvature utilizing the vectors of contour curves.

Various design and production features such as open and closed cutouts, nibbled holes, reference axis
mapping etc. have been accounted for in the present work. Local radius of curvature and convex/concave profile of
the component geometry for bend correction are adapted and successfully implemented on PC taking care of the
multidimensional nature of curvature.

2
1.1 Nomenclature

ANGLE OF BEND The angle included between inward normals of bent and unbent surfaces.
BC Bend Correction, the correction needed to be applied to true length of the cross-section of
SMC surface model, in order to obtain the bend corrected length of the corresponding
plane unbent sheet.
CONTOUR Boundary curves of a part of drawing or component
CUTOUTS These are holes, with a variety of shapes and sizes, made in the SMC. In general, these
cutouts, no matter what their geometrical form may be, are not permitted to have sharp
corners and invariably all the cutout shapes have filleted corners to ensure smooth flow of
loads around these corners, there by avoiding the stress singularities.
DIRECTION OF BEND This term is used to facilitate the computation of direction dependent bend correction
ENTITY A drawing unit (like line, arc, surface, B-spline, point etc.).
FILLET An operation creating a smooth arc in lieu of a sharp corner joining entities.
FLANGE Flange is a (plane/curved) surface, which is attached to an edge of the PFS with a fillet
connecting them. It is a part of the SMC, which is produced by bending (in the plane of
bending) along the edge contour. The flanges addressed should have upper end contours,
which are not reentrant and would be with monotonically varying flange heights along
the contour where they are attached to PFS.
FLANGE SEGMENT Flange segments are divisions of flange attached to corresponding PFS segments and are
discriminable by their layers / colours in CAD model.
FP Flat Pattern: It is the planar drawing of the blank required to form the desired sheet metal
component
FPD Flat Pattern Development: It is the process of determining flat pattern for a sheet metal
component corresponding to it’s neutral surface. The neutral surface is the surface within
the sheet that is parallel to the inner and outer surfaces and on which the net stress of the
sheet is zero. This is explained in Fig. 1.
Middle
Surface Inner Surface
Outer Surface
Neutral Surface
r t kt

Front Projected View Right Side View

Principal flat surface Bend Angle

Locus of fillet
start point r: Bend radius
80.0 Deg
t: Component thickness
Outer Mold Line
k: Material factor
Developed
Fillet Φ: Bend angle
kt: Distance between the
Developed Flange inner surface and the neutral
surface

Flat Pattern

Fig. 1. Flat pattern development of a simple sheet metal

3
Focal Point (FPt) The point which all section planes pass through in radial sectioning method
GITA Graphical Interactive Three-dimensional Application (an indigenously developed PC
based CAD) software of Aeronautical Development Agency (ADA)
JOGGLE Unique geometry shape incorporated in sheet metal components to enable the adjacent
parts to overlap/butt each other.
NMG Numerical Master Geometry: This is a CAD model of surface geometry, completely
defining the aerodynamic shape of an aircraft or such other objects like automobiles etc.
OFFSET This is the parameter specifying the normal displacement of the FP contour in the contour
augmentation process. This is an operation for increasing the sheet metal blank area
PFS Principal Flat Surface is a plane (Z=0) and this is the part of SMC surface which is held
intact during the forming process. The existence of this PFS in the SMC is a pre-requisite
for accurate FPD of SMCs manufactured through bend forming. However, SMCs exist
without this also (e.g. stretch formed SMCs).
REFERENCE LINES A pair of local axes defined for the production loft generation. These axes are defined
with respect to the aircraft.
TM Topological Model: it is a CAD model represented through NMG surface and component
boundary curves in 3D space. In contrast, a full-fledged surface model is basically made
of faces (patches derived from parent surface using component boundary curves). While
importing faces appearing in a sheet metal surface or solid model - from one CAD system
to another, there is a possibility of data loss, as all translators do not support these entities.

4
 2. Overview of SMCs
An overview of the aircraft Non PFS sheet metal components with relevant illustrative features is
presented in Fig. 2. The sections that follow, deal with the details of such component models and various aspects of
the development of their production loft

S Non–PFS SMCs
M
C PFS SMCs [6-9]
Classification of Non-PFS SMCs with Apex Edge
Curved Surfaces and Special Features SMCs [9]

Curvature Symmetric Cutouts Design / Production

Features

Single Double
Closed Open Nibbled

Joggles Pilot Tooling

Holes Holes

Single Double

Circular Elliptical Trapezoidal etc. Rectangular

Fig. 2. Overview
Fig. 2. Overview of Non-PFS
of Non-PFS Components
Components

5
 3. An Insight into Surface Model Domain

The PFS category of SMC models represent most of the SMCs used in different industries while the
Non PFS category of SMC models are unique to vehicle shell manufacturing. These are further sub-divided into
two categories (a) Apex edge components (with angle type of sections) and (b) the general class. This paper
primarily addresses FPD of the latter class of the most general complex Non-PFS SMCs. These components have
curvatures in one or more directions simultaneously. The design features of SMCs addressed in this system are open
cutouts, closed cutouts, nibbled holes etc. The SMC surface model belonging to this category comprises NMG
surface, boundary curves and reference planes. Further, the SMC surface model chosen for this purpose can be
categorized as the inner, middle or outer surface model depending upon the choice exercised in creating the surface
model.

4. The Non PFS SMC Flat Pattern Development

The generic approach to FPD of non-PFS component involves creation or importing of a CAD model
from other CAD systems together with the reference axes and various features of SMC on to a PC based CAD
software (GITA), with ease. Its validation and verification precedes the evolution of the technique for the FPD. If
we define model standards, which would be met during the preprocessing stage, the FPD becomes a smooth sail if
the models meeting the standards only are supplied to the system. The steps involved in the generic approach are
given below.
1. Creation or importing of CAD model through translators
2. Verification and preprocessing of CAD model
a) Reference axis definition
b) Identification and auto tagging of the boundary curves
c) Identification of nibbled holes and auto layering
3. Sectioning
4. Point Transformation Algorithm
a) Generating an array of 3D points on Non-PFS boundaries
5. Bend correction algorithm for Non PFS components to get accurate flat pattern

4.1 Creation or importing of CAD model through translators

The surface model is either made directly in the PC CAD system or imported through IGES route from
other CAD systems. Since the surface model is complex, having aircraft geometry, surfaces with boundary curves
defining the components, it is found convenient to choose a model comprising boundary curves in 3D space which
lie on the NMG of the 3D object (say aircraft). We shall term this model as the Topological Model, TM . In fact, it is
too simple to believe as viewing this model, without super posing grid or blended NMG, is impossible as the model
consists of hollow curves only. It is chosen because it is simple to create on a PC based system without referring to
the cumbersome process of surface patch assembly. Such models can be easily imported through IGES route.

4.2 Verification and preprocessing of CAD model

The model created on a PC based system could be a full-fledged surface model or it can be the TM
comprising base surface geometry (portion of NMG), with the component definition completed through bounding
space curves. Alternatively, such models can also be imported onto a PC from other platforms through IGES route.
In either case it is essential to verify the integrity of such models to ensure that their boundary curves do not have
gaps and overlaps. Hence filling gaps and avoiding overlaps of the boundary curves constitutes verification of the
CAD model. This process of validation becomes a prerequisite for FPD. These aspects have become part of the
model standards.

The standard TM is made available on PC together with the reference planes with tag names RP1, RP2 and
NMG surface with tag names 1,2,…,n respectively.

6
4.2.1 Reference axis definition

Reference axis is one of the production features required for components manufacture. The CAD models
are made with reference to a global coordinate system (e.g. aircraft coordinate system). Definition of local
coordinate system to enable for manufacturing is referred as ‘reference axis definition’. Reference planes are created
to realize the ‘reference axis definition’ and these planes are either parallel to X,Y, Z axes or inclined at an angle to
the global (e.g. aircraft) coordinate system. These reference planes correspond to aircraft center line, fuselage
reference line etc. and the same are tagged as RP1 and RP2. The following Fig. 3 shows a typical SMC model with
reference planes and curves.

4.2.2 Identification and auto tagging of the boundary curves

The Non PFS SMC boundary geometries are B-spline curves. The open cutout is converted into a closed
one by generating an edge curve in this open region of the component. All the boundary curves are read, sorted, and
stored in an array. The model check system is designed to check the correctness and availability of pre-defined
identifier in the CAD model. System performs continuity checks and tags the segments sequentially (CC1,….,CCn).

Reference Plane, RP1

RP2

• Reference Axes

Cutout
Component boundary

NMG surface

Fig. 3. A topological model with NMG surface and reference axes

4.2.3 Identification of nibbled holes and auto layering

SMCs with closed profiles may contain fillet and flange surfaces attached to it termed as ‘Nibbled Hole’.
Automatic extraction of data (involving parsing of input data and geometric reasoning) from a given model is
carried out for all geometric entities as detailed in [9]. In order to identify the complex geometric features of the
components and perform flat pattern development of the projections, which are protruding out of the NMG surface,
the data corresponding to each of such features are organized in different layers. The algorithm of the process
involved is as follows

Step1: Store the sequence number of all CCVs (Composite Curves CC1,…. ,CCn bounding the domain of the model)
in an array (viz. ccv_array[ ])

Step2: Store the sequence number of NMG surfaces in layer 0 in another dynamic array with variable size (viz. surf
array [xxx ]), where xxx has to be filled based on the number of surfaces present.

7
Step3: Get the list of curves for each of the CCVs (ie. CC1,,….,CCn)

Step4: Locate the corner points of surfaces and compare them with end points of CCVs.

Step5: Compare the cutout contour with corner points of the surface to find fillet surfaces attached to it.

Step6: If fillet surfaces are found, identify the flange surfaces attached to these.

Step7: Put all the untagged surfaces obtained through above mentioned steps and attached to CCVs sequentially in
layers ‘1’ to ‘n’ respectively.

Layer Management is an automated process for organizing various parts of the model in different layers during
preprocessing. It is designed for the ease of identification and processing of different CAD model geometry features
utilizing the layering characteristics of CAD.

Layers 0 and 252 are reserved for TM together with NMG surface and FPD output with essential production features
respectively whereas layers 1through 251 sequentially stores the fillet and flange surfaces attached to curves
CC1,….,CCn.

4.3 Evolution of various techniques for FPD

A sheet metal component has significant thickness. As can be seen from equation (1), (refer section 4.5.3)
the material factor ‘k’ and the component thickness are important parameters in determining the bend corrected
length and hence the flat pattern.

Since there is no unique way of developing FP for SMCs with Non-PFS geometry, different methods are
conceived and researched to evolve suitable techniques. Hence, some candidate methods, which suit the task are
conceived, designed, developed, tested and evaluated before zeroing onto a reliable technique. One such method is
adopted from earlier success with Normal Cross Section Development (NCSD) for PFS components; namely
sectioning the CAD model and decomposing it into an equivalent 2D problem to derive solution. Thus, different
candidate sectioning methods were designed and experimented with to meet the accuracy demand of a given
geometry to evolve feasible techniques. Some of these techniques are summarized in the following sections.

4.4 Sectioning methods

These are basically transitional adaptations of sectioning techniques [8,9], which were successfully utilized
for FPD of PFS class of SMCs. All length computations in these methods for segments of the curve contained within
the model interior only need be bend corrected.

4.4.1 Radial sectioning

Radial sectioning is a method of making angular divisions on a surface. This is most suitable for fully
curved SMCs with low radius of curvature bounding curves as well as for axi-symmetric components. Appropriate
number of divisions is made based on geometry of the surface. The sectioning method is explained below.

Though this method is ideally suited for ellipsoidal shells, its limitations (listed below) are dominant for majority of
components such as those shown in Fig. 4(b). Actually the shell shown is a very special case (axi-symmetric). It can
in fact be developed considering a single generator curve. The resulting FP is a circle whose dimension is the bend
corrected length of the generator curve.

Where as, more complex components such as those shown in Fig. 4(b), one has to adopt a combination of techniques.
In the example problem, one can divide the component into three regions A,B and C. The middle region B can be
addressed by the normal sectioning procedure detailed in [8,9]. For the regions A and C, radial sectioning procedure
can be adopted as detailed in the following steps.

8
 B
Traces of Sectioning Planes I

II
A FPT C

In YZ Plane III
NMG Surface
(a)

4(a) 4(b)
Radial Divisions

Fig. 4. Radial sectioning method

Step1: Identify the Focal Point (FPt) shown in Fig. 4(b).

Step2: Generate the boundary curves for NMG surface.

Step3: Divide regions containing bends (viz., A and C) into 3 segments designated as I, II and III. Their extent can
vary depending on the curvature.

Step4: Create different set of angular planes at focal point in all the three segments depending on the geometry.
For example, Fig. 4 (b) shows more planes (angular divisions) in region ‘I’ and ‘III’ (corresponding to
curved regions).

Step5: Section surface with these planes and store the bend corrected 3D lengths and angles in a two dimensional
array.

Step6: Transfer the focal point on to a 2D plane.

Step7: Map the angular lengths with reference to the focal point on the 2D plane.

Step8: Fit a smooth curve passing through the end points of angular lines.

Limitations

(i) The user should identify the focal point in the CAD model interactively.

(ii) Surfaces should be independent and discrete entities to apply radial sections.

(iii) To plot angular lines, reference planes are required.

9
(iv) Determination of sub regional ranges (a,b,c) and divisions are difficult as these are component dependent.

Thus, radial sectioning can not be used as an independent technique due to its limitation in its applicability.

4.4.2 Parallel sectioning

Parallel sectioning technique is another method of FPD process. The procedure is an adaptation of the CSD
(Cross Section Development) method [8].

NMG

Region A

Reference Curve R2
a
•
• •b
• c O
• d
e

Region B Reference Curve R1

Fig. 5. Parallel sectioning method

Step1: Identify the reference curve and origin ‘O’ (the intersection point of the curves).

Step2: Create parallel planes normal to a reference curve.

Step3: Section the NMG surfaces with these planes as shown in Fig. .5

Step4: Measure the section curve lengths from the reference curve R1 starting with the first curve (R2). For e.g.,
store the lengths oa, ob, oc ,od and oe (effecting bend correction where relevant).
Step5: Plot the curve lengths in a FP plane and store the end points.

Step6: Generate smooth B-spline curves passing through the end points corresponding to component model
boundary to obtain FP contour.

Here the limitation of this technique lies in choosing the number of sections in regions such as A and B which need
very high number of sections, yet with no guarantee for accurate FP determination.

4.4.3 Ortho-normal sectioning

This is another method of FP development wherein the given component is sectioned by sets of parallel
planes normal to both reference axes respectively. This is a simple adaptation of cross sectional development for a
pair of orthogonal directions. This procedure of FP development is explained below.

10
 L1r (1)*
•
• •
•

L2s (1)* d1r

•
Region A R2

• Segment
••
L2s (2)*• • length of the
O intersection
Reference Curve R2 d2s curve
•
••
L1r (2)*

R1

Normal Planes
R1
L2s (2)* •
Reference Curve R1 L1r (1)
• • •
d1r L2s (1)*
Region B • • R2
O’
* Cumulative bend corrected length of L1r (2)
intersection curves between reference •
curve and corresponding end point d2s

o:
• Points on boundary of FP
Fig. 6. Ortho normal sectioning method

Step1: Identify two reference curves (R1, R2) and the and the origin.

Step2: Create planes (such as (P11 , …,P1r,…; and P12, …, P2s, …) on both the reference curves and store the
distances from the origin of these planes (viz., d1r and d2s measured along the reference lines R1 and R2)
respectively in two different arrays applying bend correction where relevant.
Step3: Intersect the surfaces with these planes and obtain and mark the lines normal to the reference curves

Step4: Calculate cumulative segmental section curve lengths corresponding to each of the sectioning plane and store
it (e.g. L1r(1) and L1r(2) are the lengths corresponding to a typical sectioning plane P1r parallel to R1 and L2s(1)
and L2s(2) and the lengths corresponding to a typical sectioning plane P2s parallel to R2 as shown in the inset.
Step5: Transfer the 3D distances of the sectioning planes from the origin (such as d1r and d2s) along the reference
curves R1 and R2 on to a 2D plane and mark them.
(Note.: All the cumulative distances are obtained by summing up segmental lengths bend corrected where
relevant).
Step6: Mark the end points of the 3 dimensional sectioning lines corresponding to sectioning planes P1r and P2s (viz.,
L1s(1), L1s(2); L2r(1), L2r(2) ).
Step9: Fit smooth B-spline passing through all the end points of the normal lines.

This is ideally suited for rectangular type of components; however, its limitations for general components are
significant.

11
Limitations

(i) Region A and B shown in Fig. 6. require more sections.

(ii) In these regions, interpolations would be needed as it might not be possible to obtain required intersections
without unduly increasing the divisions.
(iii) Sorting of these intersection points is tedious, time consuming and hence inefficient.

Thus, the above methods are found to be efficient for a specific set of components while the same are
highly inefficient for the rest. Thus, these developments could not result in a unique, reliable and efficient technique
suitable for all the categories of general Non-PFS SMCs. Hence a need arose to look beyond the sectioning
techniques that became the integral part of FPD system developments since 1989 till date. In what follows we shall
go down from line entities to point entities and try to evolve techniques, which would be able to transform each
point on the boundary of the TM of SMC on to a plane facilitating creation of FP. Hence, the task consists of an
accurately transforming each point on the TM to a corresponding point on the FP. This proposition looks tall and
infinitely time consuming at first glance. However, systematic research has yielded selection of points, which would
permit accurate determination of FPs with reliability and economy consistent with specified accuracy limits.

4.5 The Point Transformation Technique

This new technique applied for SMC flat pattern development of a wide range of components including the
components with bi-directional curvature comprises the following steps.

a) Creation of topological model as detailed in (Fig. 7) section 4.5.1.

b) Finding a suitable method of dividing each of the CCVs of TM for obtaining 3D set of points needed to
uniquely define these curves to facilitate FPD. Fig. 8 and section 4.5.2 describe the method of generating an
array of 3D points on component boundaries of NMG needed for efficient processing.

c) The process of obtaining FP corresponding to topological model together with associated nomenclature (vide
Fig. 9), termed as the Point Transformation Algorithm PTA are given in section 4.5.3

d) Nibbled holes etc are then processed as offsets to this basic part of FP as illustrated in Fig 10 and described in
section 4.5.4.
Thus the above four steps constitute the Point Transformation Technique.
4.5.1 Algorithm for point generation on TM

The TM with face boundary curves is discritized during pre processing of FPD. Features such as open
cutout, closed cutout and nibbled holes etc. are recognized as boundary curves on NMG. These features are
identified, tagged and layered automatically. The Fig 8. shows three closed curves with corresponding tag names, in
which, the first defines the component and the others define the cutouts.

NMG

Boundary Curves CC3

CC1
CC2
Fig. 7 Component showing boundary curves and CCVs

12
4.5.2 Generating an array of 3D points on non-PFS boundaries

Dividing the TM boundary and identifying essential number of boundary points is a task linked to the
efficiency of this process. Number of options such as divisions based on a number of criteria like equi-divisions,
divisions linked to geometric parameters of the curves etc. are tried resulting in the evolution of a reliable and
foolproof algorithm. Since the objective of this task is - minimal effort and maximum efficiency consistent within
the required accuracy, the division process is linked to minimal error. The quantum of difference between curve and
arc lengths is utilized for optimizing the division control.
Points are generated on the boundary curves such that the curve length between any pair of adjacent points
and the corresponding chord length are varying within a predefined tolerance limit Δ. This Δ is the factor that
controls number of points and hence the elapsed time and efficiency.

Fig. 8 Illustration of sequence of point division

1. Consider a typical boundary curve shown in Fig. 8(a) to fit minimum number of points on it.

2. Measure the curve length and chord length between the end points p1p2 shown in Fig. 8(a). Check if the
difference between these lengths (Δd) is within the prescribed tolerance Δ.

3. If Δd is more, divide the curve into two equal segments with the point p3 at the middle of the curve as shown in
Fig. 8(b). Find Δd1 corresponding to p1p3 and p3p2.

4. Repeat the process of division until difference Δd1 is within the tolerance limit. Fig. 8(c) shows the curve divided
into three segments p1p3, p3p4 and p4p2. Find Δd for each of these curves. At this stage, it is found that curve
p3p4 has attained the tolerance Δ. Now the process of division is required only for p1p3 and p4p2.

5. The process is repeated till all segments of division attain the tolerance limit. (the current example required 7
divisions and 8 points)

The following diagram (Fig. 9) illustrates minimum number of points generated on boundary curves of a typical
Non-PFS component.

13
 Open Cutout

Closed Cutout

RP2

Boundary
BoundaryCurves
Curves
RP1
NMG

Fig. 9 Points generated on boundary curves

4.5.3 Bend correction algorithm for non PFS components

This algorithm utilizes the topological model together with NMG and reference axes. The geometrical
constructional details, which are needed to explain the algorithm, are first detailed in Fig. 10. Since the
nomenclatures are all relevant to this algorithm only, they are given in this figure itself. The detailed steps of the
algorithm and the sequence they are performed are then explained using this diagram.

Step1: Utilizing the reference planes (i.e. XZ and YZ), a pair of sections is made on SMC. These are termed as
principal reference curves (R1 and R2).
Step2: The point of intersection of these curves ‘O’ is termed as the origin of the PTA and is located on the NMG.

Step3: A pair of normal lines corresponding to these curves is plotted on the FP (parallel to reference planes). Their
intersection point O’ (Point corresponding to the origin above) serves as the origin for FP diagram. Pair of
normal lines are termed as axes of FPD, viz. O’X and O’Y (corresponding to the reference curves R1 and R2).

Step4: The curves and surfaces (fillets and flanges) attached to boundaries are identified and shifted to another layer.
Step5: Generate minimum number of points (say P1 to Pn)on the boundary curve such that the resultant curve
maintains sufficient accuracy .
Step6: At each point of the boundary curve (say at P), a characteristic pair of orthogonal set of ‘sectioning planes’ is
created parallel to reference planes (section curves Cx and Cy corresponding to it).
Step7: Obtain curve lengths as curvilinear coordinates (Sx and Sy) of points with respect to reference planes by
sectioning NMG surface with these sectioning planes. Apply BC to these curve lengths to get accurate lengths
at each of the chosen boundary points (P1 to Pn)to obtain Sxf and Syf as defined in the diagram.
Step8: Transfer the lengths onto the reference axes in FP drawing.
Step9: Repeat this for all the boundary points of the curve (P1 to Pn). Generate a smooth B-spline curve, which
corresponds to the face boundary curve on the NMG.

14
 X-Z Plane
Y-Z Plane
Φcxi
R1 : Reference curve obtained by Cxi
sectioning the NMG by XZ plane rcxi
R2 : Reference curve obtained by
Z
sectioning the NMG by YZ plane
Cy1
O•
• Cy Cx : Surface curve on NMG obtained
Cx3 Cy2 R1
Cx1 Cx2 • • by sectioning through point P parallel to XZ plane
• P
Cx Cy : Surface curve on NMG obtained
X’
by sectioning through point P parallel to YZ plane
R2 Sx : Length of space curve Cx; Sxi are lengths -
Sy Syf of curves Cxi (i = 1-3)
y
O’ P’ Sy : Length of space curve Cy; Sy1 and Sy2 are –
x Ps
Sx Pf lengths of Cy1 and Cy2

FP Sxf Sxf = Sx + B.C for the curve length of Cx

diagram Syf = Sy + B.C for the curve length of Cy Φcy1
P’ = Point projected on XY plane
Cy1

Ps = Intersection of Sx and Sy rcy1

Y’ Pf = Intersection of Sxf and Syf

Fig. 10. Point transformation algorithm

Bend correction = (k t φ) – ½ t φ - ½ (t φ) s1 s2 ….. (1)

Though this algorithm and the utility was earlier presented in references [6,9], it is being explained now for the
completeness and to clarify its application to doubly curved 3D surfaces while the choice options of creating ‘inner’,
‘middle’ and ‘outer’ surface models are exercised.
Here: k is a material factor which quantifies the fraction of the distance of neutral axis location relative to the inner
surface in terms of the sheet metal thickness (t). This is obtained for different materials at different hardness levels
basing on practical test data and is stored in the data base of the software system for ready retrieval during the
process of FPD

t: Component thickness
φ: Angle of bend
s1: sign convention for SMC surface model (-1 for inner surface, 0 for middle surface and +1 for outer surface),
s2: sign convention for direction of bend (+1 for clockwise bend and –1 for counter-clockwise bend)

The clockwise bend refers to the direction of angle which the normal of the bend surface makes with the unbent
reference surface at each location.
As the process of bend correction is truly complex for components with bi-directional curvature (as they utilized
local vector manipulations), this is best illustrated through the example of bend segments described in PTA into
which this correction process is integrated.

Illustration (Refer Fig. 10)

Section Curve Cx is divided into minimum number of arcs, say into three arcs Cx1,Cx2 and Cx3 as shown in the
example given in Fig. 10. Determine the radius ‘r‘ of any segment of 3 dimensional section curve and the angular
bend ‘Φ’ between the end normals of the segments and calculate the bend corrected length as under.

15
In the illustrative example the curves Cx and Cy are divided into two segments each and their curve lengths Sx1, Sx2,
Sx3,Sy1, Sy2 ; and the radii of their bends rx1, rx2, rx3 ; ry1, ry2 as well as the angles of the bend Φx1,Φx2,Φx3; Φy1,Φy2 are
determined for each of these four segments The insets given in Fig. 10 clearly explain how these parameters are
determined from the geometric entities of segmental curves. Hence the final the bend corrected segment lengths Sxf
and Syf are obtained as under:
Sxf = Sx1 + Sx2 + Sx3 + (BC for Cx1 , Cx2 and Cx3)………..(2)
Similarly Syf. = Sy1 + Sy2 + (BC for Cy1 and Cy2)……...(3)
Note : each bend correction formula corresponding to the four segments use the segment specific parameters which
are identified through the suffice utilized for specifying each of them. The angles between the normals as well as
between the tangents are identical.

4.5.4 Nibbled Hole Processing using offset method

Open and closed cutout boundary having fillet and flange surfaces attached to it is called nibbled hole. Fig.
11(a). illustrates a nibbled hole attached to a closed cutout boundary on Non-PFS 3D surface.

Basic component boundary

where flange is attached
NMG
Nibbled Hole

P'3 P3
• p'3 •
. P3

P1 P2

Lp1p3
.
P1

Lp1p3

(a) (b)

Fig. 11 Nibbled hole development process

Following algorithm illustrates the development of nibbled hole

1. The fillet and flange surfaces attached to the boundaries of the holes are appropriately placed in different layers.

2. Mark a point P1 on the boundary curve corresponding to the nibbled hole as shown in Fig11(b)

3. Insert a plane normal to the NMG as well as to the basic component boundary at point P1 and section the

projection to obtain the cross section P1P3 as shown in the inset.

4. In these normal planes, vectors normal to the edge curves at P1 and P3 would become tangents to the sectional
curve P1P3 intersecting at point P2. A vector joining P1P2 gives the direction of unfolding.
5. Measure the bend corrected length of the section Lp1p3 of the flange at the location of the attachment of the
flange (viz. P1) to the basic component boundary.

Details of the processing of a nibbled hole portion:

6. Transfer the coordinate point P1 to P1’(onto FP plane) using PTA.

16
7. Mark the FP point corresponding to P1 by offsetting the contour curve in the direction of trace of P2 by a

distance equal to bend corrected length of ‘FLp1p3’ (ie.,P1’P3’).

8. Repeat the process for all the boundary curves of the hole. Fill the gaps by appropriately extending the curve
contours and trimming the curves in case of overlapping contours.

The following Fig. 12. illustrates the Nibbled hole processing in case of fillet surface attached to an open cutout.
The offset curve and the direction of offset p1’ is also marked in the figure, whereas Fig. 13 shows the process in
case of closed cutouts.

FP
contour
P1 P1’
Flange
surface

(a) Component (b) FP Nibbled Hole excluded (c) FP Nibbled Hole included

Fig. 12 Two stage processing of nibbled hole of open cutout components

P2 P1 P8
P2′ P1′ P8′
Flange
P3 P3′ P7 P7′ Surface
P4′ P5′ P6
P4 P6′
P5

(a) Component (b) FP nibbled hole excluded (c) FP nibbled hole included

Fig. 13 Two stage processing of nibbled hole of closed cutout components

4.5.5 FPD automation using point transforming technique PTA

Automation of FP development utilizing point transforming technique which actually transfers a point on
NMG to a corresponding point on FP PTA is a generic approach to the development process. It has been designed
for components ranging from simple sheet metal geometry to complex doubly curved surface. Various features of
SMCs addressed in this automatic process are open cutouts, closed cutouts and nibbled holes etc. In order to develop
the FP of this class of SMCs, a two-stage process is utilized. Fig. 15 illustrates one such actual component.

17
 Closed NMG
Cutouts

CC2

Nibbled Hole (CC5)

CC1

CC3

Fig. 14 Component with closed cutouts and nibble hole

Stage I: The TM corresponding to the boundary curves lying on the NMG is first transferred to FP drawing plane.
CC1’, CC2’, CC3’ and CC4’ are the basic FP contour.

Stage II: The features of SMCs, which are appendages, are addressed by reconstructing or modifying the basic FP
contour shown in Fig. 15. Finally, CC1’, CC2’, CC3’ and CC5’ only represent FP, while CC4’ which was an
intermediate stage gets merged into the interior of FP.

Reference Axis2

CC5’
CC4’

[ Insert figure 15 about here ] Reference Axis1

CC3’

CC2’
CC1’

Fig. 15. FP output after processing nibbled hole

18
The block diagram for illustrating the input and the corresponding FP output processing details are shown in Fig. 16.

the nomenclatures that are used in this flow chart pertain to the PLGS [9].

FPD PROCESS

Material and
Hardness Production Features
Pilot Hole Processor
Tooling Hole Processor
Geometric
Parameters
FPD Processing Methods
Apex Edge System [9]
Process Control
Parameters

Non PFS Component Point Transformation Algorithm

Features Modules
Single/ Doubly Curved Reference Axis Definition,
Open Cutout Contour Discretization Techniques,
Closed Cutout Tagging and Auto Layering,
Nibbled Holes Generation of 3D Array Points,
and Curvature Computation

Fig. 16 Block diagram for input and output process of non- PFS SMCs

5. Conclusion

This paper has brought out unique, platform independent, well-defined algorithms for FPD of Non-PFS
SMCs utilizing the PTA. Any production loft generation system could utilize this technique to cater to various
design features such as cutouts, nibbled holes and production features common to aircraft SMCs including transfer
of reference axis to FPD. Even as a technique, this PTA for FPD is new and is also integrated into a full-fledged
FPD system. The resulting software is termed as GITA Flat Pattern Development -II (GFPD-II) as it utilizes GITA
CAD platform of ADA.

The software including the new algorithms is proven with testing and validation on actual aircraft SMCs
(some of these are illustrated in this paper), in collaboration with an aircraft factory (Hindustan Aeronautics Limited
- HAL), engaged in the concurrent production of new aircrafts for ADA. It is developed using ‘C’ language and
SDK (Software Development Kit) tools on VC++ 6.0 platform and works in WINDOWS environment. This is now
extensively used for loft generation of production series of Light Combat Aircrafts by HAL.

19
 6. Acknowledgements
The authors wish to acknowledge the initiative of the Aeronautical Development Agency, Bangalore
for sponsoring this research and IIT Bombay for sustained support and constant encouragement for the successful
completion of the task. In particular directors of ADA and distinguished scientists Drs. Kota Hari Narayana, T.G.Pai
and R.K.Ramanathan deserve special thanks for rigorously pursuing these indigenous advanced developments of
FPD activities, by providing continuous feed back on validation through production tests carried out in their
directorates.
ADA’s GITA-CAD group in general and Mrs. V.Vani in particular deserve a special mention for GITA
activities. At IITB the editorial assistance rendered by Mr. G. Aravind Rao and secretarial assistance rendered by
Mrs. Kunju Kunju are appreciated.

References

[1] Azariadis P. N and Aspragathos N. A. (1997), Design of plane patterns of doubly curved surfaces,
Computer Aided Design 29(10), 675-685

[2] Azariadis P. N, Nearchou A. C. and Aspragathos N. A. (2002), An evolutionary algorithm for

generating planar developments of arbitrarily curved surfaces, Computers in Industry 47(3), 357 – 368.

[3] Faux, I D and Pratt, M J (1983), Computation Geometry for Design and Manufacture, Ellis Horwood,
UK.

[4] Hinds B.K., McCartney L. and Woods G. (1991), Pattern development for 3D surfaces, Computer
Aided Design 23(8), 583-592.

[5] Parida L., Mudur S.P. (1993), Constraint-satisfying planar development of complex surfaces,
Computer Aided Design 25(4), 225-232.

[6] Prasad.K.S.R.K, Lalitesh Kumar.K, Pavan Kumar.D.K, Selvaraj.P (1991), Computer Aided Flat
Pattern Development CAFPAD, System Document, No. IITB/AE/CASD/091-CAM-02, Dept. of
Aerospace Engg., IIT Bombay, India.

[7] Prasad.K.S.R.K, Nilesh Parmar and Pavan Kumar (1993), Computer Integrated Flat Pattern Package
CIFPAC, System Document, No. AE/CASD/093-CIM-02, Dept. of Aerospace Engg. IIT Bombay,
India.

[8] Prasad.K.S.R.K, Selvaraj.P (2005), On the Considerations for the Design of an Automated Aircraft
Sheet Metal Component Production Loft Generation System (PLGS), International Journal of
Production Research 43(14), 3045-3067.

[9] Prasad.K.S.R.K, Selvaraj.P (2004), Practical Methods of Computer Aided Flat Pattern Development
(FPD) for Sheet Metal Components (SMC) – An overview, International Journal of Production
Research 42(15), 3011-3039.

[10] Prasad K.S.R.K, Selvaraj P, Praveen V Ayachit,, Nagamani B.V. and Ramanathan, R.K., ‘Automated
Aircraft Sheet Metal Component Production Technologies’ to appear in Journal of Aircraft

20