Flexible asymmetric spinning

Omer Music, Julian M. Allwood (2)*

Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, United Kingdom
1. Introduction
The search for a cost-effective means to produce small batches
of sheet formed parts has driven a wave of innovationin the past 20
years, led by inventors in Japan [1]. Many of these innovations have
been mobile tool (or incremental) processes where the large xed
tools of stamping are replaced by small tools moving in two or
three dimensions. Manual versions of these processes, including
the Power Hammer and the English Wheel, were widely used in
industry until the mid 20th century and when operated by skilled
craftsmen, could create a wide variety of sheet parts with useful
accuracy. Recent developments have aimed to replicate this
achievement, but with computer control replacing the craftsman.
Most prominent in the academic literature has been work on
Incremental Sheet Forming (ISF) with dozens of research groups
world-wide attempting to build on the ideas of Iseki et al. [2] and
Matsubarra [3].
Although spinning is a mobile tool process, it is inexible: the
product geometry is dened by an axisymmetric rigid mandrel. ISF
aims to overcome these constraints and is often described as a
derivative of spinning. However, unlike ISF, spinning is a true net-
shape process: on process completion, the tools apply no force to
the product, and the product perimeter is free, so when the product
is unloaded from the machine it does not change shape. This is in
striking contrast to ISF, where extensive springback on process
completion causes poor geometric accuracy. Furthermore, ISF, like
shear spinning, leads to signicant sheet thinning which prevents
replication of products made by deep-drawing. Thus, despite great
interest in this process among researchers, it has had little take-up
in industry.
There is thus considerable motivation to create a newvariant of
Spinning to preserve the benets of true net-shape production
and allow 908 wall angles without thinning, while overcoming the
constraints of requiring an axisymmetric rigid mandrel. Several
attempts have been made to extend the process design. An early
design by Boldrini, replaced the mandrel with a single roller, but
this is used for adding a short ange to large components, not for
producing whole components. More recent attempts are sum-
marised in Fig. 1.
The processes in Fig. 1bd allow spinning without a mandrel,
and are all forms of shear spinning the outer diameter of the
product does not reduce due to the process, so they are limited by
thinning. Furthermore, the processes in Fig. 1b and c have limited
exibility, and although that of Fig. 1d has more potential, only
experiments making simple cones have been reported. The process
of Fig. 1e is inexible as it retains a rigid mandrel. Can asymmetric
products be spun without a mandrel?
2. Analysis of mandrel contact pressures in spinning
Spinning is inexible because of the mandrel that denes
product geometry, so the key to exploring options for creating a
exible spinning process is to examine the interaction between the
mandrel and the workpiece. In [10] we reviewed previous work on
the mechanics of spinning, and found no analysis of this
interaction. Therefore a nite element simulation of spinning
was set up in Abaqus. The simulation used 20,000 continuumshell-
elements to describe the workpiece. To check the sensitivity of
results to numerical parameters, studies of the element type, mesh
size and number of through-thickness integration points were
performed, and the simulation was validated against results
published in [11].
A case study was setup up for spinning a simple can from a
1 mm thick aluminium 5251-H22 blank of diameter 500 mm with
spinning ratio 2, using a frictionless working roller of diameter
100 mm. The tool path comprised a sequence of involute curves,
following a standard spinning schedule. Rather than simulate the
full duration of the process, which would have taken weeks, a
sequence of simulations was created each starting from an
estimate of the product shape after a given number of passes.
Each simulation was then run for 3 s of process time corresponding
to three revolutions of the spindle.
CIRP Annals - Manufacturing Technology 60 (2011) 319322
A R T I C L E I N F O
Keywords:
Flexibility
Metal forming
Spinning
A B S T R A C T
Metal spinning is used to form shell components, but is constrained by two features: it can only produce
axisymmetric shapes; it requires a dedicated mandrel for each product. Examination of pressures
between product and mandrel revealed that contact is limited to three well dened areas. This suggested
that the full mandrel could be replaced by three rollers. Furthermore, if these rollers could be controlled,
they could represent any symmetric or asymmetric mandrel. A seven-axis machine has been designed,
manufactured, and used to spin trial parts. The machine design is described, and preliminary results give
an indicator of process capability.
2011 CIRP.
* Corresponding author.
E-mail address: jma42@cam.ac.uk (J.M. Allwood).
Contents lists available at ScienceDirect
CIRP Annals - Manufacturing Technology
j ournal homepage: ht t p: / / ees. el sevi er. com/ ci rp/ def aul t . asp
0007-8506/$ see front matter 2011 CIRP.
doi:10.1016/j.cirp.2011.03.136
Fig. 2 shows the key and surprising result of this analysis, for a
simulation representing the point in processing when approxi-
mately half of the can has been formed onto the mandrel, with the
remaining workpiece shaped to form a bell like funnel, and the
working roller currently approximately half way across the bell.
The red arrows normal to the workpiece showthe contact pressure
between mandrel and workpiece, and are clearly limited to three
areas (no further signicant pressure occurs on the opposite side of
the mandrel): Aand B at the limit of where the can has been formed
onto the mandrel, but offset to either side of the working roller; C
at the corner radius of the can where the base turns into the wall.
Intuitively these areas arise because the working roller force tends
to squash the can and bend it at C, so the contacts A and B oppose
squashing and are sufcient to ensure no other contact around the
circumference.
The simulations were repeated for different stages in the
process, and a range of mandrel diameters and the same pattern of
three contact areas from Fig. 2 remained remarkably consistent.
Fig. 3 shows that the angle between the two areas of contact at the
limit of mandrel contact varies between 108 and 308 during the
manufacture of the can, but the pattern of contact remains
consistently as shown in Fig. 2.
In further simulations, a non-axisymmetric mandrel was used,
and Fig. 4 shows an extreme case, with a kidney-bean mandrel.
The contact pressures are at a similar stage to Fig. 2.
The same pattern of three contact areas is still clearly visible,
although the two at the threshold of contact are now not
symmetric, and the contact at the base of the can is slightly
modied. It appears that the effect of the mandrel in spinning is
always limited to three small areas of contact in consistent and
predictable locations.
The implication of this analysis is that the mandrel could be
replaced, in both axisymmetric and asymmetric spinning, by three
rollers: one at the base of the spun product (the blending roll) and
two support rolls placed to either side of the working roll, and
moving along the product as the nal diameter is reached. This
leads to the schematic process design of Fig. 5.
To conrm this design, the simulation was now set up in
reverse with the mandrel replaced by rollers as shown in Fig.
5, and the simulation used to compare the stress state in the
workpiece. Fig. 6 accordingly shows a comparison of equivalent
stress and strain in the workpiece in conventional spinning and
with the conguration of Fig. 5. The gure conrms that the
design based on rollers at the locations where the mandrel
applies pressure to the workpiece in conventional spinning leads
to very similar pattern of deformation in both cases. The Abaqus
simulation was further used to predict the forces on all the

Fig. 1. Recent innovations in spinning process design.

Fig. 2. Contact pressure between mandrel and workpiece in spinning.

Fig. 3. Variation in radial separation of contact areas during spinning.

Fig. 4. Contact pressure between mandrel and workpiece during asymmetric can
spinning.
Fig. 5. Schematic process design for exible asymmetric spinning.
O. Music, J.M. Allwood / CIRP Annals - Manufacturing Technology 60 (2011) 319322 320
rollers of Fig. 5 as an input to the next section on machine
design.
3. Design of a exible asymmetric spinning machine
In order to evaluate machine capability, a set of case study parts
was specied, as shown in Fig. 7 demonstrating increasing
complexity, from left to right, with the kidney bean part, with a
combination of stretch and shrink anges, a limiting reference case.
Realising the schematic of Fig. 5requires several automateddegrees
of freedom, which are the main driver of machine cost. In order to
constrain cost, the working roller was given 2 degrees of freedom,
with the option for manual setting of roller angle, the two blending
rolls were given a manually xed separation and assumed to travel
together in the z-direction, but independently in the r-direction, so
requiring 3 degrees of freedom in all. The blending roll need move
only in the r-direction under computer control (for at bottomed
products) but can be withdrawn manually to allowproduct loading
and unloading. Combined with a controllable spindle, this
determined the need for a seven axis machine.
Analysis of forces within the radially moving axes demonstrated
that for non-axisymmetric products, motor forces would be
dominated by inertia rather than the forces required to form the
part, and in turn these are driven by the requirement for tool
acceleration which is a function of spindle rotation speed and
product asymmetry. A pragmatic solution was to specify a
maximum radial acceleration to allow motor selection and
then require that as the product becomes more asymmetric, the
spindle speed is reduced. The complete machine specications are
given in Table 1.
Several options for machine layout were considered, with the
major constraint being the provision of precise radial motion of the
blending and support rolls inside the spun component. The major
choices considered were:
Mounting all three rollers on a single tailstock projecting into
the workpiece, or on three separate stiff arms held outside the
workpiece;
Providing radial motion through a cam-follower mechanism
with a template shape, through axial actuators and a mechanism,
or through radial actuators.
Limited space inside the workpiece determined the need for
remote radial actuation, and hence rollers mounted on stiff arms as
shown in Fig. 8. The gure shows the support rollers on arms held
on one side of the workpiece, with the blending roller arm held on
the opposite side of the workpiece allowing all seven axes of the
machine to be mounted near to a plane through the product axis.
Detailed design of the axes was governed primarily by the specied
positional accuracy of the tools. Mechanical end stops were
provided on each axis for safety, and the whole mounted on a stiff
horizontal table. The system design built around LabViewincludes
x and y load cells to monitor forces on the working roller, and a
comprehensive safety system. The nal design is shown in Fig. 9
and along with the concept of Fig. 5 has been patented [12].
4. Results and discussion
The machine was commissioned in January 2011, and initial
trials have aimed to evaluate the basic design concept of replacing
the mandrel with three rollers. An example trial axisymmetric part
made from a 375 mm diameter blank of 1.4 mm commercially
pure aluminium is shown in Fig. 10. The gure demonstrates that
the exible concept outlined in this paper works, and is able to
form a cup shape with wall angles up to 908 with controlled
thinning. The surface quality of the trial parts is excellent, the top
and bottom of the parts are perfectly at and the circularity of the
part was within 0.05 mm in a 250 mm diameter part. The same

Fig. 6. Comparison of conventional and mandrel free spinning with the
conguration of Fig. 5.
Table 1
Specications for the asymmetric exible spinning machine.
Maximum blank diameter 500mm
Maximum blank thickness 2mm
Typical materials Al. 5251 H22, Mild steel
Maximum spinning ratio (hence product diam.) 2 (250mm)
Maximum spindle rotation speed 200rpm
Maximum radial acceleration of tools 2250mm/s
2
Maximum radial speed of tools: 350mm/s
Maximum force on tools 3750N
Maximum tool deection 0.1mm
Smallest corner radius on an asymmetric part: 75mm

Fig. 8. Selected layout design for support and blending rollers.

Fig. 7. Target case-study product geometries.
O. Music, J.M. Allwood / CIRP Annals - Manufacturing Technology 60 (2011) 319322 321
failure modes as encountered in conventional spinning have been
observed in this setup, and there is no indication of difference in
forming limits compared to conventional spinning. A programme of
extensive trials is now being planned, to include analysis of tool path
design, investigation of tool forces and forming limits, and to work
towards the more complex asymmetric shapes in Fig. 7. Future
machine development will include the addition of a line-camera, to
allowclosed loop control of geometry, following the approach in [13],
and could include addition of further degrees of freedom for
instance to allow controllable separation of the support rollers.
This paper was motivated by the need for exible net-shape
sheet forming. Analysis of contact pressures between mandrel and
workpiece showed a predictable set of three contacts, leading to
the novel machine design. Initial trials of the machine have
conrmed the design concept, and potentially this approach could
lead to a radical expansion in the applicability of spinning.
Acknowledgements
The rst author, and the costs of the machine, were supported
by the EPSRC through the Dorothy Hodgkins Scheme, Ford through
its University Research Programme, Novelis, Metal Spinners Ltd.,
Cummins Generators and Siemens VAI.
References
[1] Allwood JM, Utsunomiya H (2006) A Survey of Flexible Forming Processes in
Japan. International Journal of Machine Tools & Manufacture 46/15:19391960.
[2] Iseki H, Kato K, Sakamoto S (1992) Flexible and Incremental Sheet Metal
Bulging Using a Path-Controlled Spherical Roller. Transactions of JSME-C
58:31473155.
[3] Matsubara S (2001) A Computer Numerically Controlled Dieless Incremental
Forming of a Sheet Metal. Proceedings of IMechE Part B Journal of Engineering
Manufacture 215/7:959966.
[4] Aihara Y, Matsubara S (2002) Incremental Forming of Sheet Metal into
Spherical Shapes which have an Undercut-Study of a Numerical Control
Forming System Driven in a Cylindrical Coordinate System. Journal of JSTP
43/495:341345.
[5] Kitazawa K, Wakabayashi A, Murata K, Seino J (1994) A CNC Incremental Sheet
Metal Forming Method for Producing the Shell Components Having Sharp
Corners. Journal of JSTP 35/406:13481353.
[6] Shima S, Kotera H, Murakami H (1997) Development of Flexible Spin-Forming
Method. Journal of JSTP 38/440:814818.
[7] Amano T, Tamura K (1984) The Study of an Elliptical Cone Spinning by the Trial
Equipment. Proceedings of 3rd International Conference on Rotary Metalworking,
213224.
[8] Gao XC, et al, (1999) Experimental Research on a New Technology Ellipse
Spinning. Journal of Materials Processing Technology 94(23):197200.
[9] Awiszus B, Meyer F (2005) Metal Spinning of Non-Circular Hollow Parts. 8th
International Conferences on Technology of Plasticity, October, Italy, 913.
[10] Music O, Allwood JM, Kawai K (2010) A Review of the Mechanics of Metal
Spinning. Journal of Materials Processing Technology 210:323.
[11] Wang Q, Wang T, Wang ZR (1989) A Study of the Working Force in Conven-
tional Spinning. Proceedings of 4th International Conference on Rotary Metal-
working, 103108.
[12] Allwood JM, Music O (2010) Spin Forming Process and Apparatus for Man-
ufacturing Articles by Spin Forming, UK Patent Application GB1016611.4.
[13] Allwood JM, Music O, Raithathna A, Duncan SR (2009) Closed-Loop Feedback
Control of Product Properties in Flexible Metal Forming Processes with Mobile
Tools. CIRP Annals 58/1:287290.

Fig. 9. Completed machine design.

Fig. 10. A trial product from the new machine.
O. Music, J.M. Allwood / CIRP Annals - Manufacturing Technology 60 (2011) 319322 322