Jo~,l of

p
Materials
°
roeess~g
Teehnellegy
ELSEVIER Journa~ of Materians Processing roc~oiogy 58 (1996) 385-389

A simplified post process for wire cut EDM

Ivano Bel~ranfia, Axel Bertholds b, Dkk Dauw~
Research & Development, AGIE LTD, 6616 Losone, Switzerland
b Microtechnology Section, AGIE LTD. 6616 Losone, Switzerland
~I & I Leuven, AGIE LTD, Heverlee, Leuven, Belgium

Abstract

The paper deals with a technical realization to hnprove the Wire EDM accuracy while cutting at full speed on virtually any
contour. The system which is readily available on commercial wke EDM maclfines is based on the on-line monitoring vnd control
of the wire position by means of an optical sensor. The deviation of the wire position relative to the programmed wire path position
is continuously measured and corrections are made during the machine cutting. Compared to conventional wire EDM machines,
this technique allows to cut complex shapes, arc paths and contours at much faster cutting speed. Practical examples are discussed
and the economical relevance is emphasized.

Key Words: Wire EDM, precision machining, wire lag correction.

Introduction Wire bending and deflection during wire cutting

A constant driving force for EDM maclfine tool builders is One of the major geometrical inaccuracies of a wke EDM
the challenge to design and manufacture wire EDM machines eroded workpieces at high cutting speed is due to the wire
such that various market requirements, in particular those bending and deflection during the cutting. Fig. 1 illustrates
concerning part precision, surface finish and cutting time, can the wire deformation during a rough cut. The wire is fed
be met. Although WEDM is often the sole method to finish through the gap by means of an unwinding system. The wire
the right part in the right time and at a justified customer cost, guides assure that the wire position is maintained properly:
it is permanently challenged by alternative technologies, for ideally file wire should remain on a straight line between
exanlple high speed milling. Therefore the need to improve them but various process threes acting on the wire m ~ e the
constantly the overall performance of the machines, ha fack wire bend.
since the introduction of wire EDM on the market in 1969, The wire can be matilematically modeled by the standard
the cutting speed has been doubled every four years and the vibration equation of motion [10]. It is assumed that the wire
surface finish has been improved by a factor of fifteen [2], [51. mass is uafifotmly spread along its considered length, i.e.
Wire bending, a major cause of cutting imprecision at high between filewire guides. An axial force F (wire tension) is
cutting speed, does not allow to take full advantage of the applied to the wire.
performances reached so far and is frustrating the interest for During machining, mechanical (pressure from the gas
even higher cutting rates for the future. The fact that the wire bubbles formed by the plasma of the erosion mcclmnism),
itself is behaving like a metal string, straightened by two hydraulic (flushing), electrostatic and electrodynamics forces
axial pulling forces and deformed by a sum of (small) forces are acting on the wire perpendicular to the axial force [l], [3],
from the erosion mechanism, makes the wire lose its ideal [6], [7], [8], [9],[10]. Assuming such forces per unit length
straight position (Fig. 1). Hence, when cutting out a q(z,t)to vary as a function of time and space, then the general
curvature, the lag effect of the wire creates a geometrical error differentialequation of motion for a stretchedstring of length l
on the workpiece to be machined. This error can be of the in a plane (along the z axis) can be written as:
order of a hun&ed microns, which for most applications
becomes unacceptable. 32y 34 32y ~y
The traditional cure to such a problem is reducing the F ~-~-- E I -~y pS--+c- +q(z,t) (1)
machine cutting speed as soon as a curvature is approaching, Bz4 = ~t 2 ~t
therefore loosing the advantage of a fast cutting machine.
Examples of time losses are given at the end of this article. A s with:
a remedy for this situation, we propose a hardware system y: wire deflection (m)
with a proper control algorithm that allows to cut virtually F: wire tension (N)
any contour at the maximum cutting speed. The description of E: Young's modulus (N/m2)
such a system is given in the next paragraphs. I: moment of inertia= ur4/4 (m 4)

09244)136/96/$15.00 © 1996 Elsevier Science S.A. All fights reserved

1'I10924-0136 (95) 02212-9
386 I. Beltramiet al. / Journal of Materials Processing Technology58 (1996) 385-389

r: wire radius (m) further displacement due to the bending in the middle of the
p: wire density (kg/m3) workpiece (parameter d). Solving equation (2) for z = 1/2,
S: wire section (In2) where 1 is the distance between the upper and lower wire
c: specific damping coefficient (Ns/m2) guides, h is the workpieee thickness, H is the distance
q: specific external load fNhn) between the wire guide and the workpiece one get:

Equation (1)can be strongly simplified by considering qo h2

only the static deflection of the wire. In fact, if the erosion y(1)= D + d with D - q°hH , d- (3)
parameters are properly chosen, no significant wire vibrations 2 2F 8F
are present. The equation can be further simplified by noting
that the wire stiffness produces only small deviations from the Even if the wire deflection can be influenced to some
ideal straight line near the wire guides. Hence: extent by increasing the wire tension F, but limited to the
~um yield strength of the wire electrode: 350 to 790
N/ram 2, it is clear from equation (3) that wire lag is inherent to
F D2y~= q(z) (2) the erosion mechanism, since qo cannot be avoided at high
Dz2 cutting speed.

The solution y = f(z) of this equation is a parabola

"within the workpiece" if one assumes that the external load
q(z) = qo is constant over the workpiece height (and time k Z [mm] Z [mm/
independent) while the solution is a straight line "out of the
90 = for h=5Omm for h=150ram
workpiece", namely between the respective wire guides and =
80
the upper and lower workpiece faces respectively since there 175
q(z)=O. 70
145
,,on ! 60
50
g 115
85
40
55
30
20
10
0
t 25

0 200 400 600 800

Fig. 2 Wire deflection measurements for varying workpiece
thickness.
I I hI q(z,t)
The total, pull back force acting on the wire Q = qoh is
roughly proportional to the energy invested in the erosion
mechanism. Thus, for a given maximtnn generator power, D is
almost independent of the workpiece height, whereas d

zl/,,I I increases linearly with it. Hence the correction mechanism

proposed here tends to work better with thin workpieces than
with thick ones, the practical limit being roughly at 100 nun.
Results of test cuts in order to verify the theoretical model
are shown in Fig. 2 for workpiece thickness 50 and 150 ran.
Note that the z-scale for the h = 150 trm curve has been
I D Idl reduced by a factor of 3 in order to compare the wire deflection
y for both workpiece thickness values. From the measured
Fig. 1. Schematic representation of an EDM wire during wire deviations one can deduce the total gap load Q acting on
cutting. the wire. A maximum force of 0.5-0.7 N (depending on the
cutting conditions and on the material cut) was found at
maximum cutting speed. This is higher than the values found
on the literature [1], [7], [8]. This reflects the fact that today's
The maximum wire deflection can be described as the sum wire EDM features more powerful pulse currents and gap
of two components: the wire displacement at the upper and flushing pressures (- 18 bar), yielding much more significant
lower border of ti~ workpiece (parameter D on Fig. 1) and the gap loads on the wire [51.
 L Be#rami el aL / dournal o f Materials Processing Technology 58 (1996) 385-389 387

Wire deflection and its h~p~c, on work#ece precision This/s schematica!l~ represented in Fig. 5. Ar~ optical
rare position sensor, placed between ~ke upper wire guide
As long as a straight cu~ is being made, the wire lag, and and die workpiece, continuously measures Ne wire deflectio~
hence the geomelrical error that yields from it, have hardly any in both X and Y directions relative to its theoretical position
negative el'feet on the output results. However, in practice, at a typical sampling period of 10 ms. The guide path is then
contour paths are very often executed. In such cases, the wire modified by the machine's Numerical Con~'ol according to the
lag causes the wire m "cut the curve" and the geometrical instantaneously measured wire lag in such a way Chat ~ e wire
accuracy will be hampered. Fig. (3) depicts schematically is constantly located on the nominal contour.
what is being obtained when the real wire position does not
match with the wire guide position.

..... ........~;..~.......`...~;~'~.~;.;.;.;~.~.~.~;~.~.~;~.~;:~.~.~.~+~.~.~.~.~:.~.~.~.~.~.~.~.~.:.~.~:.:.~.~.~
•. ======================
:~:...~.~:::.~:::~.~:~::~:~...:+~..~.:~:~.~.~.~.~.~.~.::+.~...~...~.~....~.~.~...~.~...~..~......~.~

.:..."
o, ~'
+ .......

: ~.,

X Wire guide position

Wire position

Fig. 3 Geometrical errors due to the combinatior, of ~ e

bending and contour EDM cutting. Ng. 5 Principle of the path correction applied in Wire EDM
machining

A part of a real cut workpiece cross section is visualized

in Fig. 4. One can notice that the comer accuracy is not fully
achieved, and that a geometrical error of some 100 pan is Optical wire position sensor
observed. To remedy these errors, various means have been
tried out by many researchers. Apart from the cutting speed The ,sensor has to be small mad able to work under water at
reduction, which leads to thne losses, software models have up to 20 bar pressure. The wire position measarement must be
been applied for correcting the nominal guide contour. These inunune against electro-magnetic perturbations generated by
models require an extensive knowledge about the influence of the EDM process and independent of the dielectric
all machine parameters on the wire lag. In order to avoid these conditions. Iherefore, an optical measming technique was
complications, a wire position measuring device has been selected. The principle of the s o n a r developed is outlined in
developed and a real time correction of the cutting path to be Fig. 6.
executed has been realized, [4]. The displacemetlt of the wire in each X- and Y-direction i s
/.~ ..5 measured by illuminating the wire with a collimated light
beam having a uniform intensity distribution. The shadow is
then projected on two symmetric triangular-shaped light
detectors so that a wire displacement have opposit effect on
the light intensity received by them. Since each triangular
detector detects the light intensity pattern individually, the
difference between the two intensities I~-I2 is linearly
dependent on the wire position whereas the sum It+I2 is
constant for a given wire diameter. The sum It+I2 can be used
to measure the light intensity fluctuations (transparency of
the dielectric) so that the output response is obtained by the
ratio

I 1-12
(4)
Fig. 4 Cross section of a workpiece cut by wire EDM. I l +12
388 1, Beltramt ot al. /Journal of Materials Processing Technology 58 (1996) 385-389

~angular lenses

wh-e

collimating lenses
Fig. 6 Principle of the optical sensor configuration

The collimated light source is made by projecting light Results of tea| time wire position control
front a red LED iato a plastic optical fibre and then placing
the output end in front e r a cylindrical GRIN lens (or Selfoc Fig. 8 outlines three test cuts of a steel 50 nm thick
lens). The length of the Selfoc lens is chosen so that the out workpiece obtained ~.~.a cutting speed of 300 mm2hnin with a
coming light, which illuminates the wire, is parallel. The 0.33 nun wire diameter. As a comparison purpose, a zigozag of
triangular dete~tora are made of two Selfoc lenses polished 3 nun length was cut using different methods and the cutting
into triangular shapes. The collimated light entering the tune between any two points on the contour was compared.
lenses is focalized at the opposite side, where two other fibres The left graph of this picture (8A) shows file results for a
transmit it far enough from the erosion disUtrbance zone. Two rough cut without any intervention. The comer accuracy is
Si-photo detectors me used to transform the optical rather poor. In the middle figure (8B), a cutting comer strategy
intensities into electrical signals. applied on standard machines was selected. This strategy is
The wire position sensor is mounted under the upper wire based on pulse power reduction as a function of the contour
guide, as shown on Fig. 7. geometry. Already a substantial improvement is achieved

A B C

iiiiiiiilliii Fig. 8 Test cuts with different strategies applied. Relative

cutting time:
A) without intervention:
B) with comet" strategy:
40 %
100 %
Fig. 7 Illustration of the assembled wire position sensor. C) with AGIEPILOT: 42 %
 1. Beltram~ et aL / dournal of Mager~als Processing Technology 58 (~996) 385-38? 389

subs~fiaL For example, a conto~r having ~2% of

connec~g radi~ ~s cut twice as fast using AGIEPILOT.

o A wh-e deflection analysis was presented. It was

shown that the wke deflection ha wke EDM can be described
as a bending string, featuring a pambo~c deflection.
A 8 C - An o f i g ~ l solution was discussed about the real time
m~~ent mad correction of the wire position during wire
Fig. 9 Test cuts ~th different strategies applied. Relative EDM cutting. The description of an optical sensor was given
cutting time: as well as the strategy applied to remedy from comer
A) without intervention: 51% imprecision.
B) with comer strategy: 100 %
, Cutting results with either ~ne of the strategies
C) with AGIEPILOT: 53 %
applied were compared. Substantial rnachi~fing time gains and
comer accuracy improvements w~e demonsUated, when
optical wire position and control was applied.

Relative cutting time

References

[1] Balleys F., 1977, ~,Removal Rate versus Accuracy in

Wke EDM)), Proceedings of the tSEM 5, Switzerland,
Wolfsberg, pp. 157-159.
[2] Beltranfi I., <(AGIEPILOT)), 1992, Internal report AGIE.
~ 40, [3] Beltmmi I., 1989, ~Il Sensore O~co)), Internal report
AGIE, pp. l-g.
[4] Beltr~ui I. et al., 1991, ~Electro erosion machine
o featuring photoelectric sensing means for measuring wire
o t0 20 3o 40 electrode deflection)), US Patent # 5,057,662.
% of connection radii <0.7 ram [5] Dauw D.F., Albert L., 1992, (~About the Evolution of the
Wile Tool performance ha Wire EDM,, Annals of the
CIRP Vol. 41/1/1992, pp. 221-225.
Pig. l0 Gain in cutting time as a ftmct.~on of workpiece [6] Dauw D.F., Sflfioul H., Delpretti R., Tfic~co, <(Wire
complexity (steel, workpiece flfickness 50 ~,,un, roughing). Amlysis and Control for Precision EDM Cutting)),
Trondhehn 1989, Annals of the CIRP, Vol. 38/1/|989,
pp. 191-194.
though the cutting speed was substantially reduced. The last [7] Dekeyser W., 1988, ~(Knowledge-Based System for
picture 8C is the result yielding when the real tune wire Wke-EDM)), Ph.D. Thesis, K.U. Leuven, Belgium,
position control technique is applied. A remarkable 88D2, UDC 681.3"12:621.9.048.4. pp. 8°25.
improvement of com~ precision is obtained while [8] Hensgen G., 1984, (~Werkzeugspezifische Einfl~se beim
maintaining the ~ t m a cutting speed. Indeed, a cutting funkenerosiven Schneiden mit ablaufender
time reduction of 58 % was obtained in comparison with the Drahtelektmde,, Ph.D. Thesis, T.H. Aachen, pp. 36-52.
standard comer strategy. Fig. 9 is a similar test cut and gives [9] Kdnoshita N., Fukui lvl., Khnura Y., 1984, <~Study on the
similar remits. Wire-EDM: In Process Measurement of Mechanical
As already mentioned before, the real time wire position Belmviour of Electrode-Wire)), Annals of the ClRP Vol.
control technique offers a major contribution to cutting time 331111984, pp. 89-92.
reduction when contours are being cut. Fig. 10 shows the [10] Par.~chow IL, 1984 .LPoer die Kr'&tte und ihre
evolution of the cutting time reduction versus the relative Wirktmgen beim elektroerosiven Sclmeiden mit
amount of connection radii. With increasing contour ablaufender Drahtelektrode,, Ph.D Thesis, T.H. Aachen,
complexity, one can observe that the cutting lime reduction is Germany.