ARTICLE IN PRESS

International Journal of Machine Tools & Manufacture 48 (2008) 1420– 1428

Contents lists available at ScienceDirect

International Journal of Machine Tools & Manufacture

journal homepage: www.elsevier.com/locate/ijmactool

Experimental and numerical study of angular error in wire-EDM

taper-cutting
J.A. Sanchez a,, S. Plaza a, N. Ortega a, M. Marcos b, J. Albizuri a
a
Department of Mechanical Engineering, Faculty of Engineering of Bilbao, Alameda de Urquijo s/n, Bilbao 48013, Spain
b
Department of Mechanical Engineering and Industrial Design, Faculty of Engineering of Cadiz, c/ Chile s/n, Cadiz 11003, Spain

a r t i c l e in f o a b s t r a c t

Article history: Wire electro discharge machining (WEDM) has become one of the most popular processes for producing
Received 20 November 2007 precise geometries in hard materials, such as those used in the tooling industry. The so-called taper-
Received in revised form cutting involves the generation of inclined ruled surfaces, and it is especially important in the
24 April 2008
manufacturing of tooling requiring draft angles. In this paper a new approach to the prediction of
Accepted 28 April 2008
angular error in wire-EDM taper-cutting is presented. A systematic analysis of the influence of process
Available online 10 May 2008
parameters on angular error is carried out using Design of Experiments (DoE) techniques. A quadratic
Keywords: equation for the prediction of angular error that takes into account electrical parameters and part
Wire-EDM geometry is derived. Validation results reveal a dominant influence of the mechanical behaviour of the
Taper-cutting
wire, rather than that of EDM regime. Following this assertion an original finite element model (FEM) to
Angular error
describe the mechanical behaviour of soft wires, typically used in taper-cutting operations, has been
Finite element modelling
developed taking into account non-linear phenomena such as contact mechanics, plastic behaviour,
stress-stiffening and large displacements. Both the results of DoE techniques and FEM simulation have
been validated through experimental tests in industrial conditions. Deviations of accuracy of WEDM-
tapered parts can be reduced to values below 40 using these models.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction to WEDM taper-cutting Although the sinking electro discharge machining (SEDM)
process was the first one to appear in industry, the last decade has
The electro discharge machining (EDM) process is, by far, the seen a large growth in the development of the wire electro
most popular amongst the non-conventional machining methods. discharge machining process (WEDM). In this case a small-
The EDM process removes material by a series of discrete diameter wire (below 0.33 mm dia) is used as electrode to cut a
electrical discharges that cause localised temperatures high narrow channel in the work. The workpiece is fed continuously
enough to melt or vaporise the metal in the immediate vicinity and slowly past the wire in order to achieve the desired cutting
of the discharge. The only requirement is that both electrode and path. Numerical control is used to control the relative motion
workpiece materials are electrically conductive. The discharges between wire and workpiece during cutting. As it cuts, the wire is
occur between the tool (electrode) and the workpiece in a continuously advanced between spools to present a constant-
dielectric medium under voltage drops over 20 V. This means diameter electrode to the work. Amongst the typical applications
that there is no contact between electrode and workpiece. The of the WEDM process one can find high precision stamping dies,
distance between both (gap) is occupied by the dielectric fluid. extrusion dies, wire drawing dies, etc.
Since material removal is not carried out by the mechanical action The term taper-cutting is commonly used for WEDM opera-
of a cutting or abrasive tool, the EDM process can be used tions aiming at generating parts with tapered profiles, for
independently on the mechanical properties of the workpiece instance, tooling in which a draft angle is present. In this case
material; for instance, no matter how high its hardness is. the angle is achieved by applying a relative movement between
Therefore, the process has become very popular in the tool- the upper and the lower guides, as shown in Fig. 1. The
making industry, in which complex geometries with tight maximum angle that can be obtained is a function of
tolerances must be generated on difficult-to-machine materials. the workpiece thickness and the mechanical behaviour of the
wire. Depending on the machine angles as high as 7301 can be
cut in a workpiece of thickness 400 mm, but this is an upper
 Corresponding author. Tel.: +34 94 601 40 68; fax: +34 94 601 42 15. limit and the angle is normally limited for parts of high
E-mail address: joseantonio.sanchez@bi.ehu.es (J.A. Sanchez). thickness.

0890-6955/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmachtools.2008.04.011
 ARTICLE IN PRESS

J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428 1421

Upper guide
TW

Wire guide (upper)

Dielectric flushing
Y

FA

A
Lower guide

MA
Fig. 1. Relative displacement of upper and lower guides in WEDM taper-cutting.

Accuracy and modelling in WEDM are recognised [1] as key

research lines for the next years, since the requirements of Deformed wire
tolerances and surface finish imposed by industries such as tool shape
making or micromechanics are growing day by day.
For a better understanding of the causes of lack of accuracy, the
mechanics of the process must be analysed. With the wire
subjected to tensile stress, the deformation induced on it during
the process depends on factors such as the cutting conditions, Programmed wire
whether it is a roughing or a finishing cut, the geometry being shape
TW
machined and, of course, the mechanical properties of the wire. In
the case of taper-cutting, the deformation caused by the rigidity of
the wire when forced to generate a given angle is also of primary
importance.
As far as cutting regime and conditions are concerned, different
authors [2–5] have identified the nature of the forces that cause
the flexion (static component) and vibration (dynamic compo-
X
nent) of the wire: electrostatic force, electromagnetic force,
electrodynamic force, the force exerted by the pressure of the Fig. 2. Deformation of the wire in taper-cutting due to rigidity. Forces and
dielectric, and of course, the tensile force imposed on the wire by momentum acting on the wire (after Kinoshita et al. [15]).
the machine itself. Much research work has been devoted to the
modelling of these effects. In [6,7] models to study the effect of (see also Fig. 2):
the dynamic components of the forces can be found. Experimental 2
modal analysis has also been used [8]. Research efforts have also d y T W yðxÞ  F A x  M A
¼ (1)
been devoted to the analysis of the effects of wire deformation on dx2 EI
workpiece accuracy [9–11]. where E is the Young modulus of the wire; I, the momentum of
In all the above cases, research work has focused on vertical inertia of the wire; TW, the axial force imposed on the wire by the
WEDM, with little or none attention paid on WEDM taper-cutting. machine; FA, the normal force at point A; and MA, the bending
However, in the last decade a number of research works dealing moment at point A. The boundary conditions and the resolution
with this operation has been published. Attention has been paid at for the second-order ordinary differential equation can be found
the problems that arise in the CAD definition of the generated in [15]. Using this approach an estimation of the angular error
ruled surfaces [12,13]. On-line adjustment of the axial force caused by the deformation of the wire can be obtained, and from
imposed on the wire as a function of the quality of the discharges that theoretical value, corrections to the position of the guides can
during taper-cutting has been recently proposed by Chiu [14]. be introduced in the Numerical Control of the machine in order to
The problem of prediction of angular error in taper-cutting was compensate the error [16]. It must be pointed out again that this
formulated for the first time by Kinoshita et al. [15]. It must be approach does not include the effect of EDM regime on the
taken into account that when programming a tapered geometry prediction of angular error. This fact is also recognised by a recent
the actual wire shape (deformed) does not coincide with the patent [17], in which the position of guides is corrected using the
theoretical (programmed) shape, which would exactly fit to the data obtained from calibration carried out using precision tooling,
geometry of the guide. This effect, shown in Fig. 2, is due to the and performed previous to cutting (i.e., with no EDM forces acting
stiffness of the wire and it is evident even before the cutting on the wire).
operation starts (i.e., it appears when applying the relative The model proposed in [15] has been programmed for the case
displacement between guides, before the WEDM regime is of a soft wire and used in a set of industrial experiments in order
applied). The consequence is the occurrence of an angular error to assess its validity. A wire commonly used nowadays for taper-
in the WEDM’ed part. In the work by Kinoshita, the influence of cutting, the coated Broncocut-Ws (by Bedra), diameter 0.25 mm,
wire stiffness is considered using the differential equation for the has been used. This wire is considered as soft due to its low yield
static deflection of an elastic beam that can be written as follows strength and high elongation. Mechanical characteristics of the
 ARTICLE IN PRESS

1422 J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428

Table 1 A closer look at the problem reveals the causes of lack of

Mechanical characteristics of the Broncocut-Ws (by Bedra) wire used for the accuracy of the model developed in [15]. Using a value of 0.7 mm
study
for the radius of the guides (rg), which is typical now in industry,
Mechanical properties Broncocut-W TM (by Bedra) the bending moment at the point of contact between wire and
guide, can be obtained from:
Wire composition Base material: CuZn20
4
Coating: CuZn50 EI Eðpd =64Þ
Tensile strength, sr (MPa) 450
MA ¼  ¼
rg rg
Yield strength, sf (MPa) 255
Young modulus, E (MPa) 90,000 90; 000  p  0:254
¼  ¼ 24:65 N m (2)
Elongation, er 30 64  0:7
Poisson coefficient, u 0.33
From this value, the maximum tensile stress due to bending can
be obtained using
Z Z R
sflex 2 1
MA ¼ sydA ¼ y dA ¼ sflex pR3 (3)
R R 4
and substituting
Inclined surface for sflex ¼ 16; 069:30 MPa
measurement of angle
Upper and lower This value is clearly over not only the yield strength (255 MPa),
reference (ground) but also the tensile strength (450 MPa) of the wire material.
Moreover, the tensile stress due to the axial force (8.9 N in this
case) imposed by the machine on the wire must be added to this
value. From this simple calculus it can be deduced that the model
erroneously predicts wire breakage, and only 0.071% of the wire
Fig. 3. Geometry of the test part for measuring angular error in WEDM section works still within the elastic region. Analysis of wire’s
experiments. mechanical behaviour leads to the conclusion that important
effects, non-linear in nature, have not been considered in the
elastic model such as the plastic behaviour of the soft wires of low
Table 2 yield strength and high elongation, the local effects imposed by
EDM parameters (as given by machine tables) for cutting tool steel on an ONA the high curvature of the geometry currently used for the guides,
Prima E-250 WEDM machine
the large displacements between the original (vertical) and the
EDM parameters Settings deformed (tapered) situations, the influence of stress-stiffening,
and the effect of friction. In addition, the model for the mechanical
Open-circuit voltage 110 V behaviour of the wire does not take into account the effect of
Peak current 60 A
cutting regime on the results.
Off-time 5 ms
On-time 2 ms At the sight of the above comments it can be said that still
Average pulse energy 3000 mJ important research work must be carried out for a better
Axial force 8.9 N understanding of the sources of angular error in wire-EDM
taper-cutting. An analysis of the influence of process variables
on the deviations is the previous step towards theoretical
prediction of the error that can then be corrected as it has been
wire, obtained from tensile tests, can be found in Table 1. Angular described in this section. In this work an original contribution to
error has been predicted by solving Eq. (1) using the mechanical the analysis of the factors that influence angular error in taper-
properties of the wire, and the value of the error has then been cutting that leads to the development of experimental and
used to compensate the position of the wire guides as described in numerical methods for the prediction of the error is presented.
[16]. Machining tests using the compensated location of the Section 2 introduces an experimental analysis based on Design of
guides have then been carried out on an ONA Prima E-250 WEDM Experiments (DoE) techniques. In Sections 3 and 4, finite element
machine to know the degree of accuracy that can be obtained on a simulation of the mechanical behaviour of a wire typically used
test part using the approach by Kinoshita. Work material is AISI for taper-cutting is presented. The models are validated in
D2 tool steel. The geometry of the test part is shown in Fig. 3, and industrial conditions and from the work carried out, conclusions
the angle of the inclined surface is measured with respect to the are drawn.
ground surfaces using a Zeiss 850 CNC coordinate measuring
machine. WEDM conditions have been established by machine
table look-up, and they are collected in Table 2. In the experiments 2. Experimental analysis of the influence of process
errors over 300 have been measured. This value can be compared parameters on angular error in WEDM taper-cutting
with what can be considered an ‘exact’ experimental method. This
latter involves cutting the test part without any compensation of In this section a systematic approach of analysis based on DoE
the location of the guides and measuring the angular error. The techniques is presented. The Central Composite Rotatable Design
so-measured value of the error, which is considered ‘exact’ method has been selected to obtain the degree of influence of the
because includes both the effect of wire deformation and WEDM different variables and to derive a mathematical expression that
regime, is then used to calculate the corrected position of the relates angular error and process variables.
guides that will be used in the new machining test. Under this Prior to applying DoE techniques, the input variables must be
approach angular errors below 30 can be achieved. If this value is chosen. Five variables were selected: part thickness (H), angle (a),
compared with the one obtained using the linear model (as said off-time (t0), pulse energy (E) and open-circuit voltage (U0). Part
above, 300 ), it becomes obvious that improvements in theoretical thickness and angle are related to the geometrical description of
prediction of angular error are required. the problem; off-time, pulse energy and open-circuit voltage are
 ARTICLE IN PRESS

J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428 1423

Table 3 Table 4
Ranges of values used for the selected variables in the Design of Experiments Complete set of tests for the DoE methodology

Variables for the DoE Range of variation Experiment no. H (mm) a (1) t0 (ms) E (mJ) U0 (V) Measured angle (1)

Part thickness, H (mm) 10–80 1 27.5 111150 23.75 5125 140 111400 4800
Taper angle, a (1) 5–30 2 27.5 111150 61.25 3375 140 111420
Off-time, t0 (ms) 5–80 3 62.5 111150 61.25 5125 140 111300
Pulse energy, E (mJ) 2500–6000 4 62.5 231450 61.25 3375 140 24140 4800
Open-circuit voltage, U0 (V) 110–150 5 62.5 231450 61.25 5125 120 24150 2400
6 62.5 231450 23.75 5125 140 24150 2400
7 27.5 231450 61.25 3375 120 241230 2400
8 62.5 111150 23.75 3375 140 111280 1200
9 27.5 231450 61.25 5125 140 241190 1200
related to the EDM regime and as explained in previously 10 62.5 111150 61.25 3375 120 111290 2400
published literature, to the forces exerted during the process. 11 62.5 231450 23.75 3375 120 241120
Table 3 collects the ranges of maximum and minimum values 12 27.5 231450 23.75 3375 140 241130 4800
used for each variable in the DoE. In the case of the wire-EDM 13 27.5 111150 61.25 5125 120 111390 3600
14 62.5 111150 23.75 5125 120 111270 3600
parameters, the range has been selected wide enough to be able to
15 27.5 231450 23.75 5125 120 2411101400
take into account very different machining cases. For instance, for 16 27.5 111150 23.75 3375 120 111370 1200
a WEDM operator that usually cuts tool steel, values of off-time of 17 45 171300 42.5 4250 110 171490 4800
80 ms can be surprisingly high. However, these values are 18 45 171300 42.5 4250 150 171490 4800
19 45 171300 5 4250 130 171480
compulsory when EDM’ing advanced materials such as low
20 45 171300 80 4250 130 171480 3600
conductivity ceramics (for instance, boron carbide [18]). The same 21 10 171300 42.5 4250 130 181230 2400
applies to the other electrical parameters. Therefore, the aim has 22 80 171300 42.5 4250 130 171420
been to cover a very wide range of possible EDM conditions. 23 45 51 42.5 4250 130 51100 4800
The experiments were carried out using the same work 24 45 301 42.5 4250 130 301220 4800
25 45 171300 42.5 2500 130 171480 3600
material and WEDM machine as described in Section 1. Thus, a
26 45 171300 42.5 6000 130 171520 4800
Hadamard matrix design with 16 trials, 10 star points and six 27 45 171300 42.5 4250 130 171520 4800
centre points must be built, which involves carrying out 32 tests. 28 45 171300 42.5 4250 130 171540
This approach is capable of representing the influence of the 29 45 171300 42.5 4250 130 171520 4800
30 45 171300 42.5 4250 130 171530 2400
variables and the interactions between them. Machining tests
31 45 171300 42.5 4250 130 171520 4800
involve cutting and measuring the test-part geometry shown in 32 45 171300 42.5 4250 130 171530 2400
Fig. 3. The complete set of tests together with the results of the
angle measured in each one is shown in Table 4. The results for each test have also been included in the last column of the table.
A test of significance can be made on each term of the
regression equation using standard errors and t-distribution. The
terms that are not significant at 90% confidence will be removed
10000 α
from the response equation. Fig. 4 shows the degree of influence
of the different process variables on the angle. It can be observed 1000
that for the given wire the parameters of higher influence on H
100
angular error are related to part geometry, i.e. part thickness and H2
angle, with a smaller role of the EDM regime. t0 U0α t0H αE t02 α2
10 U0H
t0α E2
Once the variables whose influence can be neglected have been U0 E Hα
removed, an expression (Eq. (4)) that can be used to predict the 1
t0E
actual taper angle as a function of the variables involved in the 0.1
process can be obtained: U0t0 U0E HΕ U02

Taper_angle ¼ 17:884 þ 0:013t 0  0:117H þ 6:308a Fig. 4. Degree of influence of each variable on angular error obtained from the
 0:009U 0 H  0:016U 0 a application of the DoE.

þ 0:026U 0 E  0:025t 0 H þ 0:004t 0 a

Relative error, as defined in Eq. (5), is plot in Figs. 5 and 6 for
 0:015aE  0:008U 20  0:015t 20
the cases of part thickness 40 and 80 mm.
þ 0:046H2  0:021a2  0:005E2 (4)
an  ap
Error ð%Þ ¼ 100 (5)
an
The validity of Eq. (4) has been assessed by comparison of
angular errors predicted by the equation with those measured in where an is the nominal angle and ap is the predicted and the
test parts cut on an industrial WEDM machine. Angles from 101 to measured angle.
301, in a range of thickness from 20 to 80 mm have been studied. The results show that Eq. (4) can be used to adequately predict
Actual machining experiments have been carried out using the angular error for a given EDM regime, part thickness and taper
same equipment, part material, part geometry and EDM condi- angle. When comparing DoE predictions with experimental
tions as those used in Section 1. EDM parameters, thickness and results, the maximum deviation is 130 3400 , which corresponds to
taper angle are different to any of the combinations that appear in the case of part thickness 60 mm and taper angle 201. For most of
the DoE (see Table 4). In the machine used in the experiments the cases the error is below 50 . If expressed in terms of relative
angles as high as 301 cannot be effectively cut in part thickness error, in most of the cases this is below 1%, with a maximum value
over 60 mm, since instability appears during cutting due to poor of 1.8% for the case of part thickness 80 mm and taper angle 101.
flushing. This is why 251 has been the maximum angle cut in part In order to analyse the influence of electrical parameters and
thickness 60 mm, and 201 in the case of 80 mm. Results are shown the accuracy of prediction of Eq. (4), parts of a completely
in Table 5. different material that must be cut using very different EDM
 ARTICLE IN PRESS

1424 J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428

Table 5 Table 6
Results of taper angle predicted by the DoE equation, compared with the angle EDM parameters for the tests using boron carbide as part material
obtained from measurement in actual WEDM tests
EDM parameters Settings
Part thickness Nominal taper DoE Measured angle
(mm) angle (1) prediction (experimental) (1) Open-circuit voltage 150 V
(1) Peak current 10 A
Off-time 80 ms
20 10 101150 4000 101170 2000 On-time 18 ms
20 201230 4900 2012103600 Pulse energy 5700 mJ
30 301250 3700 301250 4100 Axial force 8.9 N

40 10 10160 5800 101130 100

20 2011501100 201190 1900
30 301160 5500 301190 5200

Table 7
60 10 10150 3100 101100 1900 Comparison between the angle predicted by the DoE equation and the angle
20 201130 4400 201270 1800 obtained in the WEDM tests for the case of machining boron carbide
25 251150 2500 251280 5500
Part thickness Nominal taper DoE Measured angle
80 10 101110 2000 101220 800 (mm) angle (1) prediction (experimental) (1)
15 151160 1200 151230 200 (1)
20 201190 3000 201220 1900
20 25 251430 1500 251400 1000
30 15 151330 4300 151200 3100

4
Experimental
DoE
3 As well as an efficient tool such as Eq. (4), the DoE analysis
provides valuable information on the most influencing factors of
Error (%)

Part thickness
the problem. It has been shown that the EDM regime has only a
2
limited influence on the problem. This is in accordance with the
40mm results of Kinoshita, and with the industrial practice of using
1 presetting precision tools to estimate angular error. It can
therefore be valuable to develop a model for the mechanical
0 behaviour of the wire that takes into account only the geometry of
0 10 20 30 40 the problem and the mechanical properties of the wire.
Taper Angle (º)

Fig. 5. Relative error (%) in the taper angle predicted by the DoE equation and
3. Finite element modelling of wire deformation
obtained from actual cutting experiments on an industrial WEDM machine. Part
thickness is 40 mm.
At the sight of the limitations of linear models like the one
developed by Kinoshita it was decided to use finite element
simulation for the new model using the commercial software
4
Experimental Ansyss V10.0. Non-linear effects and complex boundary condi-
DoE tions that are difficult to account for using analytical methods can
3 be included using numerical simulation. The following assump-
Error (%)

Part thickness tions have been considered when developing the model:
2
80mm  The effect of the forces acting on the wire due to the EDM
1 regime is not included. The only external force is the axial force
imposed by the machine itself.
0  Only the upper-half of the wire has been modelled. It is
0 5 10 15 20 25 assumed that the behaviour in the lower half will be identical.
Taper Angle (º)  Vertical displacement of the upper-end section of the wire is
avoided.
Fig. 6. Relative error (%) in the taper angle predicted by the DoE equation and  The lower-end of the upper-half of the wire is displaced
obtained from actual cutting experiments on an industrial WEDM machine. Part
horizontally to the ideal centre point of the problem, i.e., the
thickness is 80 mm.
point at an equal horizontal distance from both guides.
 Plastic behaviour of the wire is considered.
conditions were cut. In this case, boron carbide was selected as  The Augmented Lagrangian algorithm has been used to solve
part material. Optimum EDM conditions for this material, the problem of contact.
obtained in [18], are shown in Table 6. Due to the high cost of
the material, only two tests corresponding to the geometry of two Since it is expected that the wire works in the plastic region, the
industrial extrusion dies were cut. Table 7 shows the results. The first step consists in defining the constitutive equations of wire
deviation is similar to that obtained in the tests with steel. material, as well as a plasticity criterion. Fig. 7 shows the bilinear
Therefore, it can be concluded that Eq. (4) is valid for a wide range constitutive law used for the wire material. The tangent modulus
of EDM conditions. has been obtained experimentally, yielding a value of 656 MPa. To
 ARTICLE IN PRESS

J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428 1425

σ (stress) 3.00
MPa
450 − 255
n= = 656 MPa

Angular error (º)

0.30 − 0.0028 2.00

450 1.00

255
0.00
0.00 0.05 0.10 0.15 0.20
Shell thickness (mm)

Fig. 8. Discontinuity between hexahedrons and beams as a function of shell

thickness.

ε (elongation) When taking a close look at the deformation induced on the

2.8x10-3 0.30 wire by the displacement of the guides, it becomes apparent that
the equilibrium equations applied in the initial situation (i.e., with
Fig. 7. Bilinear constitutive law used for the material of the wire.
the wire vertical) cannot be applied in the final situation (i.e., with
the wire deformed due to the movement of the guides). This is
adequately represent the combined stress state, a bilinear known as a problem of large displacements, and it must be
isotropic hardening rate-independent plastic model (Von Mises included in the solution. To do so, the force applied (in this case,
plasticity with associative flow rule and work or isotropic the axial force imposed by the machine itself) must keep at each
hardening of the yield surface) has been used. instant of the simulation the relative position with respect to the
Different types of elements have been used for the different deformed wire, which is known as a follower load.
zones of the wire, according to the requirements imposed by the To complete the model the effect of stress-stiffening has to be
problem on each zone. Thus, in the vicinity of the contact considered. Its influence cannot be neglected in those cases in
guide–wire linear hexahedrons have been used. On the one hand, which axial stiffness is much higher than bending stiffness, as it is
the geometry is simple enough not to require higher-order in the case of wires. A stress-stiffening matrix, obtained as a
elements of curved faces. On the other hand, if element size is function of the stress state present in the previous step of the
small enough, the approach given by linear elements is good simulation, is defined by
enough while reducing processing time because of the limited Z
number of nodes. Mesh density has been made higher at the zone ½Si  ¼ ½Gi T ½ti ½Gj dðvolÞ (6)
at which contact wire–guide will occur, so that contact elements
of enough resolution can be applied on that zone. In the zone in and added to the classical stiffness matrix. In Eq. (6), [Si] is the
which plastic behaviour is not expected, that is, far from the zone matrix that represents the contribution of the stress-stiffening
of contact with the guide, 3D elastic beams have been used. These effect, [Gi] is the derivative of the shape function and [ti] is the
elements have two nodes per element and 6 degrees of freedom Cauchy stress matrix.
(dof) per node. Since three of these dofs are rotational in the nodes
of the beam, and only translational dof is present in the 3D plastic
hexahedrons, an additional condition of continuity must be 4. Validation of the FE model and mesh optimisation
implemented in the section where these two types of elements
meet. In the link shell elements whose stiffness is defined by its Validation of the proposed model involves comparison of
thickness are used, so that the objective of meeting the Bernoulli angular error predicted by FE simulation with that measured in
hypothesis can be satisfied. Trials have been made with different test parts cut on an industrial WEDM machine. Again angles from
values of thickness. Fig. 8 shows the degree of discontinuity, 101 to 301 in a range of thickness from 20 to 80 mm have been
measured as angular error between the zones modelled with studied. Actual machining experiments have been carried out
hexahedrons and with beams, as a function of shell thickness. An under the same conditions as those described in Section 1. Inner
optimum value of 0.15 mm for the thickness of the shells has been radius of machine guides is 0.7 mm. The manufacturer of the
obtained. machine recommends using a constant value of 8.9 N in the zone
To study contact between guide and wire the difference of between the guides. Since axial force in the machine is set in
hardness between both materials has to be taken into account. In vertical position of the wire, its value must take friction into
the case of sapphire, which is used as guide material, hardness is account. Thus, applying the well-known Eq. (7) for flexible wires
about 1600–1800 HV, while hardness of brass is within the range the values of axial forces to be applied by the machine (i.e., in
65–135 HBN. Therefore, it can be assumed that most of the vertical position) can be obtained. These values are also included
deformation due to contact is concentrated on the wire. This is the in Fig. 9 as a function of taper angle.
reason to use an approach of pair rigid-flexible. The selected
Tt
contact elements are first order quadrilaterals. The Augmented ¼ ema (7)
Tv
Lagrangian algorithm has been used to solve the problem of
contact, which of course requires knowledge of the friction where as shown in Fig. 9, Tv is the axial force imposed by the
coefficient, m, between the materials involved. The value of m for machine, Tt the axial force in the zone of taper, a the nominal
the combination of the soft wire used and the material of the angle and m the friction coefficient. Of course, actual taper angle
guides has been obtained experimentally following the metho- should be used. However, if the nominal angle is used instead, the
dology described in [19], yielding a value of 0.29. error introduced is very small and can be neglected. For instance,
 ARTICLE IN PRESS

1426 J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428

5
Experimental
Tv
4 FE Simulation

Error (%)
wire 3 Part thickness
guide α (º) Tv (N)
2
10 8.46 60mm
α 8.24
1
15
20 8.04 0
0 10 20 30
25 7.84 Taper Angle (º)
30 7.65
Fig. 10. Relative error (%) in the taper angle predicted by the FEM simulation and
obtained from actual cutting experiments on an industrial WEDM machine. Part
thickness is 60 mm.

Tt = 8.9N
Table 9
Fig. 9. Axial forces acting on the wire at the entrance (Tv) and at the exit (Tt) of the Results of taper angle measured in WEDM’ed parts after guide location
guide. compensation is performed using the values of angular error predicted by the
complete FE model

Part thickness (mm) Nominal taper angle (1) Measured angle in

WEDM tests (1)
Table 8
Comparison between the angle predicted by the complete FEM model and the
20 28 281105700
angle measured in experimental WEDM tests
40 14 131590 2800
40 26 26100 1500
Part thickness Nominal taper Finite Element Measured angle
60 11 101580 800
(mm) angle (1) prediction (1) (experimental) (1)
60 20 201105200
80 8 71560 3200
20 10 101290 2400 101170 2000
80 19 19120 2100
20 201280 5600 2012103600
30 301330 1400 301250 4100

40 10 101200 500 101130 100

20 201200 5100 201190 1900 in Table 9. After compensation, angular error in most cases is
30 301190 1700 301190 5200 below 20 , with a maximum value of 30 2800 . If compared with the
results obtained by the experimental method commented in
60 10 101140 4100 101100 1900 Section 1, that exhibited errors about 30 , it can be said that the
20 201140 500 201270 18 numerical method achieves an excellent degree of accuracy. It
25 251130 2900 251280 5500
must be beard in mind that the procedure for correction of the
position of the guides only accounts for the component of
80 10 1011103600 101220 800
15 1511101800 151230 200
deformation of the wire due to its rigidity, neglecting the
20 201100 4900 201220 1900 influence of phenomena such as wire vibration.
Additional information can be obtained from the stress–strain
analysis available from simulation. The case in which the wire is
most heavily loaded corresponds to a taper angle of 301, with part
let us assume a high error between the actual and the nominal thickness 20 mm. In this case, the maximum Von Mises stress is
taper angle of, say, 300 . In this case the error in the estimation of Tv located near the zone of contact with the guide, reaching a value
would be 0.02 N, small enough to be neglected. As it was done in of 378 MPa. At this point, the maximum plastic strain is 18.85%.
Section 2, results obtained from numerical simulation must be Therefore, the wire is working within the plastic region but the
compared with those obtained from experimental tests (see maximum strength is not exceeded in any case, which is a clear
Table 8). Fig. 10 shows, as an example, the relative error for the improvement with respect to the elastic model. Fig. 11 shows the
case of part thickness 60 mm. stress distribution in the vicinity of the contact guide–wire.
Errors are now higher, in general, than those obtained when Results also show that linear elastic behaviour occurs at a certain
using the DoE predictions. In many cases, the error is over 100 , distance from the contact wire–guide, with stresses below the
with a maximum value of 140 3600 for the case of thickness 60 mm yield strength of the material. This fact validates the hypothesis of
and taper angle 250 . If expressed in terms of relative error, the having used elastic beams for the central part of the wire,
maximum is about 2%. The deviation between FEM and DoE resulting in a reduction of processing time.
predictions can be explained if one takes into account that FEM Contact stresses develop in the zone where wire and guide
simulation does not consider the effect of EDM regime on angular meet during deformation. The maximum value is 389 MPa and it
error. Results also show that deviation between the two methods occurs for angle 201, the influence of part thickness being
of prediction are not very high, confirming the fact that most of practically negligible. Fig. 12 shows the zone of the guide where
the error is due to the mechanical behaviour of the wire and, to a those maximum stresses appear, obtained from a simulation
lesser extent, to the EDM regime. made with taper angle 201 and part thickness 60 mm. It is
Prediction of angular error obtained from FE simulation has expected that wear of guides will be maximum at that point.
been used in scattered WED machining tests to correct the The results above exhibit a very good degree of agreement with
location of the guides (following the procedure given in [16]), and the actual mechanical behaviour of the wire. On the one hand, the
thus improve angular accuracy of parts. The results are collected model is very complete and introduces the different sources of
 ARTICLE IN PRESS

J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428 1427

Table 10
Comparison between the angles predicted by the complete and the simplified
MPa finite element models

378 Part thickness Nominal taper FEM complete FEM simplified

(mm) angle (1) model (1) model (1)

338 20 10 101290 2400 101280 5100

20 201280 5600 201280 3900
30 301330 1400 301320 1400

298
40 10 101200 500 101190 4200
20 201200 5100 201190 4700
258 30 301190 1700 301180 4000

60 10 101140 4100 101140 2400

218 20 201140 500 201130 5800
25 251130 2900 251130 1900

178 80 10 1011103600 101110 2400

15 1511101800 151110 2400
20 201100 4900 201100 4800
138

98 cases, such as taper angle 301 and part thickness 20 mm. This is
due to the high number of elements required: 1248 hexahedrons,
10 beams and 21 shell elements. Therefore, the possibility of using
58.1
a simplified model in which processing time can be dramatically
reduced while keeping the accuracy of the results was studied,
18.1 bearing in mind that, of course, using a simpler model involves
loosing some information.
-5
5x10 m Three-dimensional plastic beams were tested for the simplified
model, so that transition shells can be eliminated. This new type
of element is quadratic, with three nodes and 6 dofs per node.
Fig. 11. Stress distribution predicted by the FE model for the case of part thickness Boundary conditions are kept similar to those used in the
20 mm and taper angle 301. complete model, and mesh density is higher in the zone of
contact wire–guide. The total number of elements in this case is
MPa 100. Modelling of contact has also been made simpler: while first
389 order quadrilaterals are used again on the surface of the guide,
modelling of the wire has been greatly simplified. Instead of using
the same type of elements, the nodes of the plastic beams have
345
been used to define the contact, which, keeping a similar
approach, is defined as rigid-flexible.
302 The accuracy of the simplified model was examined by
comparing the prediction of angular error with that provided by
259 the complete model. Simulations involving part thickness
20–80 mm and taper angle 10–301 were carried out and the
evolution of angular error with respect to the complete model was
216 studied. From the results, collected in Table 10, it becomes clear
that although the new model is clearly simpler, accuracy is not at
173 all lost. The deviation between both models in the prediction of
the angle is always below 10. However, processing time with the
simplified model is drastically cut, with longest simulation runs of
130 about 15 m.
Stress analysis can also be used for comparison between
86.3 models. The most severely stressed situation, that is, the
simulation run for part thickness 20 mm and taper angle 301 is
shown in Fig. 13, so that it can be compared with the results
43.2
included in Fig. 11. Again, with the wire working within the plastic
5x10-5 m region, maximum strength is not exceeded. The maximum value
0 of the Von Mises stress is 378 MPa for the complete model and
394 MPa for the simplified model, the difference being about 4.2%.
Fig. 12. Distribution of contact stresses between wire and guide as predicted by As far as to maximum plastic strain concerns, its value is 18.85%
the FE model for the case of part thickness 60 mm and taper angle 201. for the complete model and 21.2% for the simplified model.
Probably the most important limitation of the simplified
non-linearity that have to be considered; but on the other hand, model is the analysis of contact stresses that it provides. Since
computational cost is high. Processing time may range from 20 m contact is only defined via the nodes on the beams used to model
in the less severely stressed cases, up to 4 h in the most difficult the wire, only contact forces can be evaluated, but not contact
 ARTICLE IN PRESS

1428 J.A. Sanchez et al. / International Journal of Machine Tools & Manufacture 48 (2008) 1420–1428

MPa the EDM process parameters. Finite element modelling was

used to include non-linearities such as the plastic behaviour of
394 the wire, contact mechanics, stress-stiffening and large
displacements. Errors in the prediction are higher than those
350 predicted by the DoE, but they are still in most cases below 100 .
When using these results to compensate the location of the
307 guides, angular deviation can be reduced down to about 30 .
 In order to reduce computational cost a simplified FEM model
based on the use of three-dimensional plastic beams has been
263
proposed and validated. Thus, simulation time has been
reduced down to 150 in the worst of the cases but accuracy is
219 not lost, since the deviation between both models in the
prediction of the angle is below 10.
175

Acknowledgements
131
The authors wish to thank the support given by the Spanish
87.7 Ministry of Education (MEC) to the project ‘‘Research on common
models for Micro-Wire Electro Discharge Machining (m-WEDM)
43.9 and conventional Wire Electro Discharge Machining (WEDM)
accuracy optimisation’’ (DPI2004-06581-C02-01).

0.001
5x10-5 m References

[1] K.H. Ho, S.T. Newman, S. Rahimifard, R.D. Allen, State of the art in wire
Fig. 13. Stress distribution predicted by the simplified FE model for the case of part electrical machining (WEDM), International Journal of Machine Tools and
thickness 20 mm and taper angle 301. Manufacture 22 (2004) 1247–1259.
[2] F. Van Dijck, Physico-mathematical analysis of the electro-discharge process,
stresses. In a simple approach, the total force acting on the wire Ph.D. Thesis, K.U. Leuven, 1973.
can be obtained by summing up the forces on all the nodes that [3] D.F. Dauw, I. Beltrami, High-precision wire EDM by on-line wire position
control, Annals of the CIRP 43/1 (1994) 193–197.
are in contact with the guide. For instance, for the case of part [4] W.L. Dekeyser, R. Snoeys, Geometrical accuracy of wire EDM, International
thickness 60 mm and taper angle 201, the contact force in the Symposium on ElectroMachining (ISEM) IX (1989) 226–232.
simplified model evaluated using this approach reaches a value of [5] F. Han, M. Kunieda, T. Sendai, Y. Imai, Simulation on influence of electrostatic
force on machining characteristics in WEDM, International Journal of Electro
0.530 N. For the same case, the complete model predicts a total Machining 7 (2002) 31–36.
contact force of 0.576 N. Therefore, a non-negligible error of about [6] D.F. Dauw, H. Sthioul, R. Delpretti, C. Tricarico, Wire analysis and control for
8% is made in this case. precision EDM cutting, Annals of the CIRP 38/1 (1989) 191–194.
[7] N. Kinoshita, M. Fukui, T. Kimura, Study on the wire-EDM: in process
measurement of mechanical behaviour of electrode wire, Annals of the CIRP
33/1 (1984) 89–92.
5. Conclusions [8] H. Yamada, N. Mohri, N. Saito, T. Magara, K. Furutani, Modal analysis of wire
electrode vibration in wire-EDM, International Journal of Electro Machining 2
(1997) 19–24.
In this work a study of angular error in WEDM taper-cutting
[9] T. Magara, H. Yamada, T. Yatomi, K. Kobayashi, Study on improvement of
has been presented. From the research carried out the following surface straightness in wire EDM, Precision Engineering 16 (2) (1994)
conclusions can be drawn: 156–165.
[10] M. Yan, P. Huang, Accuracy improvement of wire-EDM by real-time wire
tension control, International Journal of Machine Tools and Manufacture 44
 Previously published linear models exhibit unacceptable errors (2004) 807–814.
in the prediction of the angle due to the fact that they do not [11] A.B. Puri, B. Bhattacharyya, An analysis and optimisation of the geometrical
consider neither the effect of process parameters, nor non- inaccuracy due to the wire lag phenomenon in WEDM, International Journal
of Machine Tools and Manufacture 43 (2003) 151–159.
linear effects related to the mechanical behaviour of the wire. [12] J. Wang, B. Ravani (Eds.), Computer aided contouring operation for travelling
 A systematic study of the effect of process parameters on wire electric discharge machining (EDM), Computer Aided Design 35 (2003)
angular error has been carried out following the methodology 925–934.
[13] J. Richard, V.K. Nguyen, I. Stroud, Standardisation of the manufacturing
Central Composite Rotatable Design. Part thickness, taper process: IMS/EU STEP-NC project on the wire EDM process, IMS Forum, Part 2
angle, pulse off-time, pulse energy and open-circuit voltage (2004) 1197–1204.
have been considered in the study. From the analysis it was [14] Y.Y. Chiu, Y.S. Liao, H.C. Li, P.C. Sue, Study of taper-cut machining of WEDM
machine, in: Second Manufacturing Engineering Society International Con-
concluded that most of the influence is due to part thickness ference, Madrid, 2007.
and angle, which determine wire’s mechanical behaviour, the [15] N. Kinoshita, M. Fukui, T. Fujii, Study on wire-EDM: accuracy in taper-cut,
influence of the EDM parameters on angular error being Annals of the CIRP 36/1 (1987) 119–122.
[16] M. Kinoshita, Control device for wire electric discharge machining with taper
smaller. working correction function, Patent JP11165219A, 1997.
 A quadratic equation that expresses actual taper angle as a [17] F. Martin, F. Altpeter, Method and device for measuring and adjusting the
function of part geometry and EDM process variables has been electrode for taper machining on an electrical discharge machine, Patent
KR200601105616, 2006.
proposed. The maximum deviation observed in the prediction
[18] J.A. Sanchez, I. Cabanes, L.N. Lopez de Lacalle, A. Lamikiz, Development of
with respect to experimental data is 130 3400 , but for most of the optimum electro discharge machining technology for advanced ceramics,
cases is below 50 . The validity of the equation has been tested International Journal of Advanced Manufacturing Technology 18 (12) (2001)
in tool steel and in boron carbide with similar results. 897–905.
[19] J.A. Sanchez, S. Plaza, L.N. Lopez de Lacalle, A. Lamikiz, Computer simulation
 At the sight of the results of the DoE it was decided to model of wire EDM taper-cutting, International Journal of Computer Integrated
the mechanical behaviour of the wire neglecting the effect of Manufacturing 19 (7) (2006) 727–735.