A COMPUTATIONAL DESIGN-OF-EXPERIMENTS STUDY OF ROLLER HEMMING

PROCESS TO CORRELATE THE DESIGN PARAMETERS.

Madhavi B. Raskar1, Prof. P S Rade2, Prof.S.C.Shilwant3
1

PG student of Mechanical Engineering Department, Sinhgad Academy of Engineering,

Pune.madhavi_raskar@yahoo.com
2
Asst Professor Mechanical Engineering Department, Sinhgad Academy of Engineering,
Pune,psrade.sae@sinhgad.edu
3
Professor & HOD Mechanical Engineering Department, Sinhgad Academy of Engineering,
Pune,.scshilwant.sae@sinhgad.edu

Abstract
Hemming is a process by which a metal sheet edge of an outer part is bent around an inner part. The accuracy of the hemming
operation affects significantly the appearance of the cars outer surfaces and has therefore a critical importance in the final perceived
quality of assembled vehicles. The main parameters of the process are also given here. A parameter study is performed based on a
Design of Experiments. This parameter study indicates parameters with a big influence on roll-in, which thereafter are used to create
a roll-in response surface of the model. Different quality optimization methods are also described in this chapter. These methods are
based on a roll-in optimization and a decrease in the forming of wave patterns along the flange during the hemming steps. This can be
used by process engineers and robot programmers to create a stable process window by choosing the optimal process setting ranges.
Keywords-Roller hemming, hemming parameter, Roll-in, wrinkle formation, DOE.

1. Introduction of Hemming
Hemming is used to connect two sheet metal
components by folding the edge of an outer panel around an
inner panel to create a smooth edge. Since hemming is the
final stage of automotive forming operations, the defects
introduced cannot be eliminated in the subsequent operations.
Therefore, the hemming quality on the closure panels such as
doors, hoods and deck lids has a great influence on the overall
quality of a vehicle. Hemming represents the most severe case
of 180 degree bending. It consists of three phases: flanging,
pre-hemming and final hemming. Different hemming
technologies are available. They can be distinguished by
several factors, for instance investment level, process time and
technical concept. The technical differences between them are
described here. Three main process types are available: die,
table top and robot roller hemming. Despite their differences
they all have the similarity that they divide the hemming in
different steps: one or more pre-hemming steps and one final
hemming step to complete the hemming. The amount of prehemming steps is dependent on the opening angle of the
flange and the type of process.
1.1Stages in Hemming
Hemming involves the bending of the edge of a sheet
on to itself or another sheet, i.e. the nominal bend angle
is180.The hemming operation is usually a three step process:
Flanging, Pre-hemming, and Hemming.

Fig. 2: Pre-hemming
Flanging is followed by Pre-hemming. In this, a pre-hemmer
forces further deformation of the outer panel by certain
degrees (e.g. 30, 45, 60).
1.3 Final Hemming:

Fig.3: Final Hemming

Final hemming is the last step. This is completed with a final
hemmer that is designed to produce a finished bend angle of
approximately 180..

Fig.1: Flanging
Flanging takes place during the drawing operation and
consists in bending the sheet edge up to 90to 110.
1.2 Pre-hemming

Press-and-die and table-top hemming processes are well suited

in mass production due to their low cycle time. On the other
hand these two processes are product specific, i.e. switching
between one product to another is costly as well as it increases
production time. Hence the need of flexible hemming
processes is more demanding. In this respect roller hemming
process could be a good alternative for assembly processes.
Although beyond its high flexibility, it has higher cycle time
compared to conventional methods; press-and-die and tabletop. New development in robot roller hemming focuses on
reduction of cycle times, which will make the process more
suitable for higher production volumes as well. Two step roller

hemming and higher forming speeds are in development

nowadays.

2. Roller Hemming
There are several drawbacks associated with die and table top
hemming operation. So a new roller hemming process method
is introduced in market. Roller hemming, as the name suggests
a robot runs a roller around the edge of the sheet so that
hemming takes place as a continuous, sequential, localized
operation. The forces involved are much lower. In this process
roller is guided by robot along the hemmed line progressively
bending the flanged height along the part. Thus a roller is used
for plastically deforming the blank, imposing to tool and
assign trajectory controlled by computer. In this way no
conventional die is required and the final shape of part is made
with only trajectory assigned to tool. The process allows the
advantage of using the robot controlled hemming head to hem
several different components in a single cell. Minor changes
and modifications to panel hemming conditions can also be
accommodated allowing a quick and cost-effective reaction.

3. Parameter Study Based On Design of

Experiments

Fig. 4: Displacement in Hemming Line (a) Roll-in (b) Rollout

Before discussing some of the important parameters from
mentioned parameters let us study the concept of roll-in and
roll-out. The displacement of the contour is called roll-in if
the outer part edge moves inward and roll-out if the outer
part edge moves outward as seen in fig 4.Roll-in and roll-out
values affect the visual quality of the parts, so it should be
controlled accurately. These values are preferred to be uniform
and small along hemming line. Roll-in tends to be seen on
convex contours whereas roll-out tends to be seen on concave
edges.
3.1 TCP-RTP Distance
The TCP RTP distance is defined as the perpendicular
distance from the roller tool centre point (TCP) to the robot
target point (RTP) located at the edge of the hemming bed.

Fig.5: The TCP RTP distance between orientations of the

roller
When the TCP RTP distance is 0.5 mm the theoretical roll-in
value is 0.3 mm. This value is taken as the low value of the
parameter. The high value of this parameter is 1.5 mm. Very
little plastic deformation occurs if this distance is increased
more (in combination with a roller orientation of 60)
resulting in zero result values.
3.2 Orientation of the roller:
The orientation of the roller is the angle of the roller with the
horizontal. The orientation angle of the roller defines in how
many steps the product is hemmed. In a standard three step
hemming process the orientations are: 60 -> 30 -> 0.In a

two-step hemming process they are: 45 -> 0. Some of the

angles are product specific.
3.3 Diameter of roller:
The diameter of the roller has an effect on the deformation
area along the flange. By varying the diameter of the roller its
effect on roll-in and wrinkling after pre-hemming is
investigated. The smaller roller shows a wave pattern in the
flange after

Fig.6: (a) Line Contact (b) Point contac

the process, whereas the larger roller does not (fig.6(a) and
(b)). The diameter of the roller also determines the size of the
stress and strain distribution; the larger diameter shows only a
local stress and strain concentration at the top of the flange
(point contact). The smaller roller shows a stress and strain
concentration over the full height of the flange (line
contact).The flange (dotted line) is bent extremer around a
smaller roller. The tangent line of the roller at the flange has a
larger angle with the smaller roller.
3.4 Flanging Length
The flange length (L) is the distance from the bend to the edge
of the panel. For reducing the defects and compact joint the
flange length generally considered is 9 to 9.5 mm. The flange
angle is taken from 90 to 110.The radius (R) of this flange is
generally 3mm.The above discussed are some of the important
parameters. There are many more parameters like prehemming angle, pre-strain, flanging die radius,pre-hammer
path, clearance, material of the panel, rolling direction, path of
roller, etc.

4. Analysis based on factorial design

A design of experiments is a set of experiments where
different factors (parameters) are investigated simultaneously.
This way the influence of different factors can be found in less
time than would be required if each factor was investigated
independently. Also possible interactions between different
factors can be identified. There is an interaction between
factors when a change of one factor influences the effect of
another factor on the results. The principle behind a DOE is
used in this work. In the table the investigated factors are
given (A, B and C). These are all two-level factors. The
minimum of the factor range is the -1 value (low level) and the
maximum is the +1 value (high level). Higher level factors can
also be investigated. The table 2 represents the design of
experiments based on a factorial design. Here all the
experiments which are conducted are listed. The number of
experiments is dependent on the sort of factorial design (full
or fractional).
Parameters
High Level(mm) Low Level(mm)
Flange length (A)
7.8
10.2
Roller diameter(B) 45
60
Flange radius.(C)
2
3
Table 1: Main Parameters for DOE
Three main parameters are considered in the factorial analysis;
flange length, roller diameter and flange radius, all effects are
consider in this study. Full factorial design allows estimation
of all three two factor interaction. (AB, AC and BC).As well
three factors interaction (ABC) ,including main effects (cause

by A,B and C), this bring total seven effect the most can
estimate from the eight-run factorial design, because one
degree of freedom is used to estimate the overall mean. Table
2 shows list of all seven effects. The main effects calculated
are listed in the A, B, and C columns.

Main Effect

Interaction Effect

Response

Standard

AB

AC

BC

ABC

Y1

1.281

1.062

0.805

0.702

0.919

0.763

8
Effect

+
0.1693

0.410

+
+
+
+
+
+
-0.0310
-0.127
0.024
0.3325
0.3807
-0.337
Table 2: Complete matrix including interaction, with effect calculated.

It provided 95% confidence for all result. This p-value is

compared to a level of significance (-level). The correlation
is different from zero if the p-value is less than or equal to this
-level. A default value is 5 %.In the table plus level is high
and minus level is low. By taking simply highs and the lows
we determine the difference or contrast. This is the effect of
factors. Mathematically, the effect of factors:

EFFECT
-0.34

AB

SSmodel= SSBC+SSAC +SSABC

SSmodel=0.7289
SS residual= SSA +SSB +SSC +SSABC
SS residual =0.092663

0.33
0.02

C -0.13
B -0.03

For A Effect={[1.062+0.702+0.919+1.253]/4}{[1.281+0.805.0.410+0.763]/4}
=0.984-0.81475

Largest effect is SSBC, SSAC and SSABC

0.38

AC

Where n refer to the number of data points collected at each

level. Ys refer to the associated responses. The effects are-

The effects of all factors can be compared on relative

magnitudes of the effects and the statistical significance of
both main and interaction effects. This can be done by a
normal probability plot and a Pareto chart of the effects.
Roller diameter (B) and Flange radius. (C) These two
parameters have a significant influence on the roll-in. Their
effects are bigger than the significance level of the results It
has a decreasing effect on the roll-in: The orientation of the
roller (factor B) also has a significant influence on the roll-in.
The sign of the coefficient indicates the direction of the
relationship. A negative sign means that increasing one factor
reduces the result value. The effects of A are not influenced by
a significant interaction. The interaction of parameters B and
C is the biggest interaction.
SS=N/4(Effect) 2

ABC
BC

Effect=[Y+/n+] - [Y-/n-]

A Effect = 0.1693

1.253
0.899375

0.17

Fig.7: Pareto chart of effect of the investigated parameters

on the roll-in

5. Conclusions
Based on the validated model, a 8-run fractional factorial DOE
is conducted to study the responses of the hem roll-in/roll-out
and the maximum surface wrinkle on the exposed hem
surface. Among three input variables, flanging length,
flanging radius and roller diameter the following variable
effects are identified to be significant:
(a) The flanging radius and roller diameter having maximum
impact on the roll-in/roll-out.
(b) The diameter of the roller determines the deformed shape
locally during the pre-hemming and has therefore an
influence on the forming of wave patterns along the
flange.
These quantitative results can serve as guidelines for the
hemming process and product design. Particularly, the above
variables are found to affect the hemming quality and
formability measures nearly independently, which provides
the insights for process parameter selections to achieve
desired hem dimensions while minimizing the risk of hem
surface wrinkling, roll-in and cracking.
6. SCOPE FOR FUTURE DEVELOPMENT
Considerable research effort has been made to improve final
perceived quality of hemmed products. However, there are
still many problems to overcome. The major efforts on
research and development should be focused on the following
topics:
1. To expand robot roller assembly capacities and
capabilities.

2.

3.

Perform a parameter study: analyze the influence of

different parameters, investigate which are important
for the final product quality and process time, and
finally optimize these parameters.
Reduction of cycle time required for the process by
using three principles as follows:
a) Task separation: Necessary tasks should be
defined and allocated to different resources.
b) Robot coordination: Coordinate robot to
work simultaneously.
c) Cell Configuration: Configure robot cell
layout to minimize idle time.

REFERENCES
1.Selim Gurgen,Mustafa Ilhan Gokler, Haluk Darendeliler,
Cetin Cengiz elikkaya,Kemal ErdenAnalysis of Roller
Hemming Process for a Vehicle Tailgate ClosureThe 11th
International Conference On Numerical Method In Industrial
Forming
Processes,
2013
2.N. Le Maout, P.Y. Manach, S. Thuillier Influence of
prestrain on the numerical simulation of the roller hemming
process Journal of Materials Processing Technology 212,
450
457
,
2012
3.IgorBurchitz,DavidFritsche,Grundmann,MatthiasHillmann
Efficient Planning and Numerical Analysis Of Industrial
Hemming Processes8th International Conference and
Workshop on National Simulation of 3D Sheet metal forming
processes
,2011
4.Urs
Eisele,
Karl
Roll
,Daimler
AG,
Sindelfingen,Development Of An Empirical Model To
Determine Results From FEA Roller Hemming Processes
LS-DYNA Forum, Bamberg Metallumformung,20105.

5.XingHu,Z.Q.lin,S.H.Li,Y.X.ZhaoFracture limit prediction

for roller hemming of aluminum alloy sheet Materials and
Design 31,1410-1416,2010
6.N.LeMaout, S.Thuillier, P.Y.Manach Drawing, Flanging
and hemming of metallic thin sheets: A multi-step process
Materials and Design 31, 2725-2736,2010
1.
2.