Int J Adv Manuf Technol

DOI 10.1007/s00170-017-0581-x

ORIGINAL ARTICLE

A new cutting depth model with rapid calibration in abrasive

water jet machining of titanium alloy
Van Hung Bui 1 & Patrick Gilles 1 & Tarek Sultan 1 & Guillaume Cohen 1 & Walter Rubio 1

Received: 3 March 2017 / Accepted: 23 May 2017

# Springer-Verlag London 2017

Abstract Titanium alloys are widely used in the aeronautical and effective as the accuracy of the models obtained over a
and engineering fields as they show an excellent trade-off number of configurations was to within the order of 5%.
between the mass and mechanical properties, but as hard ma-
terials, they are difficult to machine using cutting tools. The Keywords Machining . Abrasive water jet . Titanium .
abrasive water jet affords a good solution to produce titanium Cutting depth model
parts, especially slim ones. To do so, there is a need to adopt a
modelling approach for the depth milled. However, a general
methodology that takes into account all the parameters leads 1 Introduction
to complex models based on a large number of experiments.
The present article proposes a depth of cut model combined 1.1 Bibliography
with a rapid calibration method. The case addressed is that of
open rectangular pockets on a Ti-6AL-4V titanium alloy. The Titanium alloys are widely used because they are lightweight,
approach introduces the machine configuration notion consid- resistant and have good resistance to corrosion. Thanks to
ering that a given machine, pressure level and abrasive impose these properties, they are used in various industrial sectors
the abrasive flow rate needed in order to obtain an optimal and in particular the aeronautics sector. The considered parts
material removal rate. For a chosen configuration, calibration are structural parts and the grade of titanium is Ti-6Al-4V.
of the model is performed from a series of elementary passes This grade is mainly composed of titanium, aluminium
and just three pocket machining passes. The method is rapid (6%), vanadium (4%) and other constituents such as carbon
(0.08%), iron (0.25%), oxygen (0.13%), nitrogen (0.05%) and
hydrogen (0.012%). The main properties of this alloy are high
tensile strength (860 Mpa), a high yield strength (800 Mpa)
* Patrick Gilles
patrick.gilles@insa-toulouse.fr
and an elongation of more than 10%.
These aeronautical structural parts have to withstand con-
Van Hung Bui siderable forces while being large in size with thin walls.
van-hung.bui@univ-tlse3.fr Machining the latter is difficult using conventional methods
Tarek Sultan such as milling because the cutting forces are high and the thin
tarek.sultan@insa-toulouse.fr walls can be easily deformed. In order to avoid these difficul-
ties and to guarantee the quality of these parts, other machin-
Guillaume Cohen
guillaume.cohen@univ-tlse3.fr ing processes have been studied. Among them, chemical ma-
chining and abrasive water jet machining are the most devel-
Walter Rubio
walter.rubio@univ-tlse3.fr
oped, but chemical machining has major disadvantages be-
cause it uses acids that are dangerous for the environment.
1
Institut Clément Ader (ICA), CNRS-INSA-ISAE-Mines Albi-UPS,
Blind machining using an abrasive water jet is a method
Université de Toulouse, 3 rue Caroline Aigle, that affords a number of advantages. Firstly, the constituents
F-31400 Toulouse, France leading to the removal of material are simply water and
 Int J Adv Manuf Technol

h b a parameter c. The latter is calculated (Eq. 3) from the width

b0.5 b0.5 measured at mid-height. According to this model, the
hmax width corresponding to a height h = 0 is infinite and the stud-
ied height therefore has to be limited [4]. The considered op-
erating parameters are the feed-rate Vf, the stand-off distance
SOD and the abrasive mass flow rate. Their influence on the
Fig. 1 Modelling of the Gaussian curve
maximum depth hmax, the width b0.5 and the material removal
rate (MRR) were studied experimentally on elementary passes
(Fig. 2).
abrasive. In addition, the structure of the machines and tools is The graphs (Fig. 2) show that:
simple and a very fine water jet can generate only low levels of
force. The latter feature makes it possible to limit the defor- – the depth of an elementary pass depends mainly on the
mation of parts during machining. Finally, the abrasive water pressure and the feed rate,
jet allows matter to be removed on all types of materials. For – the width of an elementary pass depends mainly on the
these reasons, this process has been widely studied for the SOD but also the feed rate,
machining of slim parts made of hard materials like titanium – the MRR is in relation to the pressure and flow of
alloys and a number of studies present the parameters abrasive.
influencing machining quality and performance [1, 2]. The – Machining of pockets by the repetition of offset
kerf left by a single pass of the jet was studied and its shape passes (Fig. 3a) then becomes possible [6]. A number
can be modelled using a Gaussian curve [3–5]. When the of studies covering various materials confirm that, in
Gaussian curve is centred on the y axis, the corresponding this case too, the pressure is the main parameter
equation (Eq. 1) is characterised by a depth H and a width influencing the MRR [7] and Fig. 2 shows that, for
factor B. a given pressure, the MRR presents a maximum.

2
 There is thus an optimal abrasive flow for which the
− x2
yð xÞ ¼ H  e B ð1Þ MRR is maximal (Fig. 3b).

In their article, Alberdi et al. [4], characterise a profile ob-

tained experimentally using measurement of the maximum The curve (Fig. 3b) shows that below the optimal abra-
depth hmax and the width at mid-height b0.5 (Fig. 1). sive flow rate, there is a share of the kinetic energy in the
From these two single measurements, the shape of the water that is not used to propel the abrasive. Similarly,
Gaussian curve is defined by the authors using equations beyond the optimal abrasive flow, part of the abrasive re-
(Eq. 2) and (Eq. 3): mains unable to receive the kinetic energy needed to pro-

2 duce a removal of material. The optimal abrasive flow is
b
h ¼ hmax  e − 2c2 ð2Þ thus set once the type of abrasive has been chosen and the
b0:5 pressure set. Other parameters can influence the MRR. This
c ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3Þ
2  lnð2Þ concerns the tilt angle of the jet in relation to the surface [8]
or the number of machining passes superimposed. It has
Equation (Eq. 2) allows the depth h corresponding to a been observed that the MRR is smaller during a second
width b to be calculated using the maximum depth hmax and pass due to the residual stresses induced during the first

Fig. 2 Influence of operating hmax (µm) b0.5 (µm) MRR (mm3/min)

parameters on elementary passes
7600
44000
36000

3600
800 3600
80 P(bar) 90 Vf(mm/min) 650 P(bar)
600 80 50 600
Vf (mm/min) 800 SOD (mm) 2 Mass flow rate (g/min)
Int J Adv Manuf Technol

Fig. 3 Offset passes and optimal

abrasive flow rate a) b)

MRR(g/min)
Abrasive flow rate
(g/min)

Pitch Optimal abrasive flow rate

pass and the shape onto which the jet impacts [9]. In the controls the MRR [4, 7]. When this pressure is set for a
present study, only a jet perpendicular to the surface and given type of abrasive, there will be an optimal abrasive
machining in a single pass are considered. flow (Fig. 3b) enabling the kinetic energy of the water to
It is possible to model an elementary pass and predict the be used most effectively. The setting parameters thus de-
profile of a pocket bottom by superimposing several elemen- fine the optimal configuration of the machine for the ma-
tary passes [11]. Some works propose to model the removal chining considered.
of material using an abrasive water jet by finite elements The control parameters are those that can be controlled
[10] on a single pass of the jet and simulate successive using the NC program. These parameters are the SOD, the
passes [11, 12]. In order to take machining parameters into feed rate Vf and the pitch offset. For a given configuration,
account directly, depth models using a potential function are these parameters will be those to be varied in the machining
proposed [6]. They show the influence of the offset pitch on program. It should be noted that the SOD determines the width
the depth milled (Fig. 4) in relation to a magnitude N ex- of the elementary passes [4] and that to obtain a pocket with
pressing the overlapping of elementary passes. N = 0 means constant depth, it should not vary during machining. This
that the passes do not overlap, N = 2 means that there are parameter can then be considered to be equivalent to a setting
two passes over a width b0 and N = 4 that there are four parameter.
passes. In what follows, a configuration is defined for a given
Various sweep strategies can be adopted: one-way [13] material, a given pressure, a defined abrasive type and flow
(although this limits productivity), zigzag or spiral [2, 6]. of abrasive and a constant SOD. The control parameters will
then be the feed rate and the offset pitch.
1.2 Conclusion on the literature

Abrasive water jet machining depends on a large number of 2 Model of the depth milled in pocket machining
parameters. Among them, a distinction can be made between
machine setting parameters and control parameters. 2.1 Modelling elementary passes
The setting parameters are those that cannot be modi-
fied during machining. This involves the pressure, the Considering the model of a Gaussian curve (Eq. 1) and
type of abrasive and the abrasive flow rate. The pressure the feed rate Vf as the only one parameter when a

Fig. 4 Lateral offset and depths a) b)

machined Depth (mm)
b0.5
hmax

N=0

pitch pitch

b0 b0
N=2 N=4
 Int J Adv Manuf Technol

configuration is considered, the chosen model used to In order to perform effective modelling, there is a need to
characterise an elementary pass profile is defined by consider the dimensions of the grains of abrasive (Fig. 5b) and
equation (Eq. 4): only retain passes with a depth greater than the mean of the
  maximum dimensions observed on 20 grains of abrasive.
2
− x 2 These measurements can be conducted using the profile pro-
BðVf Þ
yðxÞ ¼ H ðVf Þ  e ð4Þ jector or again with a simple micrometre. When the profile has
considerable depth as compared with the roughness (Fig. 5c),
Each elementary pass is characterised by its maximum modelling will be effective. The curves of the profiles mea-
depth H(Vf) and a width factor B(Vf). It has been shown that sured and modelled then clearly show the depth H(Vf) of the
for a given configuration (§1.2), the width and depth param- elementary pass obtained and it is to be noted that the interval
eters only depend on the feed rate (Fig. 2). This influence will [−2 × B(Vf); 2 × B(Vf)] contains all the points that have a depth
be taken into account using models (Eqs. 5, 6). These potential greater than a hundredth of the maximum depth (Fig. 5c). It
models can fit experimental trials with an accuracy of around can therefore be considered that that interval contains the to-
7% on both the depth factor H(Vf) and the width factor B(Vf) tality of the elementary pass.
as shown in Fig. 13.
H ðVf Þ ¼ Ho  Vf Hv ð5Þ 2.2 Principle for summing of elementary passes

BðVf Þ ¼ Bo  Vf Bv ð6Þ The machining considered is performed in a single pass mak-

ing elementary passes modelled by equation (Eq. 4) offset by
the pitch. The pocket profile obtained (Fig. 6) will thus be the
In these expressions (Eqs. 5, 6) Ho, Hv, Bo and Bv are
sum of elementary passes.
coefficients that are determined experimentally. To do so, it
The pocket bottom profile is thus calculated as the sum of
is necessary to perform elementary passes varying the feed
the elementary profiles offset by the pitch (Eq. 7).
rate and noting the profile of each pass in a plane perpendic-
"  #
ular to the direction of feed movement. The values (H, B) for ðx−ipitchÞ2
n −
each elementary pass are then calcuated using the least squares Y ðxÞ ¼ ∑ H ðVf Þ  e BðVf Þ2
ð7Þ
method. It should be noted that not all elementary passes can i¼0
be used as when the feed rate is high, the depth obtained will
be reduced and the roughness (Fig. 5a) will lead to a substan- In this expression, the pitch is the offset pitch. The magni-
tial error in modelling (Fig. 5a). tudes H(Vf) and B(Vf) are respectively the maximum depth

Fig. 5 Profiles for elementary 0 2 4 6 8

passes and abrasive size 220# 0.002

-0.002

-0.006 a) b)

-0.01

-0.014 Measured Modeled

0 1 2 3 4 5 6 7 8 9 10
0
-0.05
-0.1
H(Vf )
-0.15
-0.2
-0.25
c)
4×B(Vf)
-0.3
-0.35 Measured Modeled
Int J Adv Manuf Technol

and the width parameter of a kerf left by a single pass and are concentrated on the kerf it is in the process of hollowing
modelled by equations (Eqs. 5–6). This expression allows a out while in the second case, the front of material tends to
first value of the depth Ymax of the pocket to be calculated orient the escape of the jet on the machined surface
(Fig. 6). (Fig. 7b) generating removal of additional material.
Preliminary experiments show that the depth measured is
always greater than the depth calculated using equation (Eq. 7)
2.3 Taking the erosion mechanism into account due to the additional abrasion mechanism that appears during
machining of the pocket. Now, the direction of the jet’s escape
A number of studies describe the action of a grain of depends mainly on the jet’s width, meaning that of an elemen-
abrasive on a given material in order to define the mech- tary pass, and the sweep pitch. It will be detailed in the following
anisms for removal of material. The global objective of paragraphs that the sweep pitch must always be smaller than the
these studies is to understand the action of a grain of width parameter B of the jet to guarantee an overlap and thus
abrasive, to model it and generalise it to simulate removal obtain a flat bottom. For a given configuration defined in § 1.2,
of material by the water jet as a whole. The mechanisms the direction of the jet’s escape will remain relatively constant as
for removal of material can be observed on various appli- also the abrasion phenomenon. It is thus possible, for a given
cations [14, 15] but the main ones are cutting and fatigue configuration, to define a coefficient of erosion Ke in order to
related to abrasion, the brittle fracture and fusion associ- take into account the erosion due to the escape of the jet (Eq. 8).
ated with the impact of grains of abrasive on the surface.
The main models for removal of material associated with "  #
n ðx−ipitchÞ2
these mechanisms [16–18] consider the following −
Y ðxÞ ¼ Ke  ∑ H ðVf Þ  e BðVf Þ2
ð8Þ
hypotheses: i¼0

– removal of material is due to the cutting action associated

with a plastic deformation, 2.4 Simplification of the cutting depth model
– on impact of a particle, no propagation of cracks appears
ahead of the particle, The maximum depth will be constant as soon as a flat
– a particle does not fragment on impact and is driven by a bottom appears (Fig. 6). The principle of summing is
plane movement. equivalent to stating that on a section of pocket of length
D containing n elementary passes:
This last hypothesis is important as it highlights the
fact that a particle will exert an action on the matter dur- – the sum of the areas of elementary passes,
ing the path it follows after impact on the surface. During – the area of the machined pocket section,
this displacement, it will erode the surface that has been
machined until its kinetic energy becomes attenuated and are equal. The area Su (Fig. 6) characterising the material
the erosion effect cancels out. As a result, there exists a removed by an elementary pass made at speed Vf is (Eq. 9):
difference between a single elementary pass and a succes-  
sion of elementary passes. In the first case, the jet impacts þ∞
−x2 pffiffiffi
Su ¼ ∫−∞ H ðVf Þ  e BðVf Þ ¼ π  H ðVf Þ  BðVf Þ ð9Þ
2
a plane surface (Fig. 7a) and escapes remaining

Fig. 6 Successive passes and x D

pocket profile obtained by 0
summing -5 -3 -1 1 3 5 7 9 11 13 15
-0.1
-0.2 Su
-0.3
-0.4
Ymax(Vf)
-0.5
Successive passes
Y(x) -0.6
pitch
-0.7
-0.8
-0.9
-1 Pocket profile obtained by summing
 Int J Adv Manuf Technol

a) Jet’s displacement b) Jet’s displacement Ke thus allows the maximum depth in relation to the feed rate
and the pitch to be calculated (Eq. 12):
pffiffiffi
π:H ðVf Þ:BðVf Þ
YmaxðVf Þ ¼ Ke: ð12Þ
pitch

2.5 Formation of the pocket bottom in relation

Jet’s escape Jet’s escape
to the sweep pitch
Fig. 7 Direction of the jet after impact
When the sweep pitch is great as compared with the width
B(Vf) of the elementary pass considered, there will be no over-
The area Ape of the pocket section of length D and lap and the pocket bottom will not be formed (Fig. 8a). When
depth Ymax(Vf) (Fig. 6) containing n elementary passes the pitch diminishes, an overlap appears and a corrugated
is (Eq. 10): pocket bottom will emerge (Fig. 8b). When the sweep pitch
pffiffiffi is of the order of the parameter B(Vf), a flat pocket bottom will
Ape ¼ YmaxðVf Þ  D ¼ n  π  H ðVf Þ  BðVf Þ ð10Þ appear (Fig. 8c). Finally, when the sweep pitch becomes low
as compared with the width of the elementary pass, the over-
The distance D is equal to the number of elementary passes lap will become significant and repeated erosion will cause
n multiplied by the pitch (Eq. 11): degradation of the surface condition (Fig. 8d).
Experimental tests were conducted in order to define
D ¼ n  pitch ð11Þ
conditions that ensure the formation of a flat pocket bot-
tom and that avoid a repeated erosion. A link between the
The equality of areas calculated (Eq. 10) considering the width of a single pass and the sweep pitch was observed.
relation (Eq. 11) as well as consideration for the erosion factor To guarantee a flat pocket bottom and a surface condition

0.1 0.1

0 0
-5 5 15 25 -5 5 15 25
-0.1 -0.1

-0.2 -0.2 Mean depth

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
a) b)
-0.6 -0.6

0.1 0.2
0 0
-5 -0.1 5 15 25 -5 5 15 25
-0.2
-0.2
-0.4
-0.3
-0.6
-0.4
-0.8
-0.5
-1
-0.6
-0.7 -1.2
-0.8 -1.4
-0.9 c) -1.6 d)
Fig. 8 Influence of the sweep pitch on formation of the pocket bottom
Int J Adv Manuf Technol

that is not degraded, it is therefore necessary for the pitch the corrugated bottom shows a regular oscillation and
to be such that it ensures a correct overlap without how- the depth model allows for an extremely good predic-
ever leading to a repeated erosion that will degrade the tion of the mean depth machined (Fig. 8b).
surface. For these reasons, the pitch to be used to produce
a pocket must be defined from the factor B(Vf) of the
elementary pass used (Eq. 13): 2.6 Method to set up the machined depth model

0; 7  BðVf Þ≤ pitch≤ 1  BðVf Þ ð13Þ The method presented (Fig. 9) relies on the given config-
uration, that is a given material, pressure, stand-off dis-
When a pocket is machined using a pitch greater tance, type of abrasive and flow of abrasive. It allows the
than B(Vf) an undulateed bottom will appear and the coefficients for the various models to be defined from
model (Eq. 12) will not apply exactly. Nevertheless, experiments (Fig. 9).

Fig. 9 Method to set up the depth Given Material and Configuration: pressure, stand-off
model
distance, type of abrasive and abrasive flow.

Performance of a series of elementary passes varying

the feed rate Vf

Measurement of each profile obtained with the feed rate Vf and

identification of the corresponding magnitudes H(Vf) and B(Vf)

Elimination of elementary passes whose depth is smaller than the mean

of dimensions measured on 20 grains of abrasive.

Determination of coefficients Ho, Hv, Bo and Bv for models (Eq. 5-6)

using the least squares method.

= ×

= ×

Pocket machining

Determination of the erosion factor Ke as the mean of relations between the depth
measured and the depth calculated for each of the pockets

Constitution of the model:

= .
 Int J Adv Manuf Technol

Table 1 Configurations tested 3.2 The experimental set-up

Configuration Stand-off distance Pressure Size of the
N° SOD (mm) (bars) abrasive (#) The machine used was a FLOW MACH4C (Fig. 10)
equipped with a PASER4 cutting head. The nozzle was
1 100 1000 220 of diameter 0.33 mm and the focus mechanism 1.02 mm
2 100 2250 220 in diameter and 101.6 mm in length. The pressure was
3 100 3500 220 generated using a Hyplex-Prime pump with a maximum
4 40 3500 220 of 4000 bars. Control of the NC machine was ensured by
5 100 3500 120 two software packages (Flowpath and Flowcut) provided
6 40 3500 120 by FLOW. Two abrasives of different brands and sizes
were tested: Barton HPX 220# and Opta Minerals120#.
The flow of abrasive was set using a pierced washer
whose diameter conditions the flow.
The material machined was Ti-6AL-4V titanium alloy.
3 Application of the method The coupons machined were 4 mm thick and the pockets
were rectangular, 15 mm wide and about 15 mm long
3.1 The different configurations tested (Fig. 11). This last dimension is adjusted to obtain a
sufficient number of elementary passes to guarantee that
A number of configurations were defined using three a regular bottom be obtained. Measurement of the pro-
pressure levels, two SOD levels and two types of files of the elementary passes and pocket bottoms was
abrasive. The combinations are shown in the Table 1. performed using an ALICONA IF optical profilometer
Configurations 1, 2 and 3 were intended to determine (Fig. 11). This device allows a surface to be acquired
the influence of pressure on the model. Configurations 3 and 1000 curves, distributed regularly over 2 mm, are
and 4 were used to identify the influence of the SOD. calculated by making the intersection between the sur-
Comparisons between configurations 3 and 5 as well as face and the 1000 planes perpendicular to the feed
4 and 6 allow the influence of the abrasive to be movement.
characterised. As stated previously, an optimal flow of The profiles considered in this study (Fig. 12) are mean
abrasive corresponds to a given abrasive conditioned by profiles calculated over those 1000 curves.
the pressure (Fig. 2). Preliminary tests conducted on
elementary passes for the material studied Ti-6AL-4V
allowed optimal flows to be defined for each configura- 3.3 Results obtained for configuration N° 1
tion. The optimal flow of abrasive was 128 g/min for
configuration 1, 222 g/min for configuration 2, 313 g/ The first configuration was observed making 12 elementary
min for configurations 3 and 4 and 408 g/min for con- passes with feed rates of between 125 and 4300 mm/min.
figurations 5 and 6. Depths H(Vf) and width factors B(Vf) are shown (Fig. 13).

Fig. 10 Machine FLOW

MACH4C

Pierced washer Nozzle 0,33

Int J Adv Manuf Technol

Fig. 11 Measurement device 2 mm

Vf
Elementary passe

2 mm
Milled pocket

Vf

The abrasive used had a grain size of 220# and the mean The erosion coefficient Ke was calculated in accor-
size measured on 20 grains was 0.053 mm. The models dance with the approach presented (Fig. 9). Its value of
obtained (Fig. 13) by equations (Eqs. 5–6) did not consider 1.096 allowed the depth model to be completed
the four last points of each graph for which the depth was (Eq. 12). The depths modelled are shown (Fig. 14).
less than 0.053 mm. The coefficients calculated (Eqs. 5–6) The model set-up has a mean error of 3.6% and a max-
were Ho = 62.333; Hv = −1.028; Bo = 1.429; Bv = −0.001. imum error of 8% and is thus effective. Also, for a
The mean accuracy obtained over all the points by the depth targeted depth level, the pocket machined using
model H(Vf) was 7.2% and that obtained on the width pa- pitch = 1.2 × B(vf) does indeed produce a corrugated
rameter model B(Vf) was 7.6%. Modelling of the elementary bottom (Fig. 15) whose profile corresponds to the
passes is thus effective. A series of 12 pockets was then modelled pocket profile (Eq. 8).
machined aiming at 4 depth levels: 0.60; 0.35; 0.15 and
0.05 mm. For each depth level, two pockets were calculated 3.4 Identification of a rapid calibration procedure
by equation (Eq. 7) using pitch = 0.7 × B(vf) and on configuration N° 1
pitch = 1 × B(Vf) so as to obtain the depths aimed at and
a flat bottom (Eq. 13) and a third pocket was calculated The performance of elementary passes while varying the
setting pitch = 1.2 × B(vf) in order to verify the appearance feed rate is fairly rapid and only takes a few minutes.
of a corrugated bottom. In this latter case, it is the mean However, production of pockets is much longer and
depth (Fig. 8b) that is estimated by the equation (Eq. 7). may take several hours. An investigation was conducted
The depths calculated and measured for the 12 pockets are to determine whether a good estimation of the coeffi-
shown (Fig. 14). cient Ke was possible only producing a limited number

Fig. 12 Mean profiles measured Width(mm) Width(mm)

0 2 4 6 8 10 -2 3 8 13 18 23
0 0

-0.2
Depth(mm)

-0.1
Depth(mm)

-0.4
-0.2
-0.6
-0.3
-0.8
Elementary passe Bottom of the pocket
 Int J Adv Manuf Technol

0.5 1.6
0.45 1.4
0.4
1.2
H(Vf) (mm) 0.35

B(Vf) (mm)
0.3 1
0.25 0.8
0.2
Eliminated points
0.6
0.15 Eliminated points
0.4
0.1
Vf(mm/min) 0.2
0.05 Vf(mm/min)
0 0
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000

Modeled Mesured Modeled Mesured

Fig. 13 Measurement and modelling of elementary profiles

of pockets. To this purpose, the points corresponding to N° 1 and N° 8 leads to an accuracy of the depth model
the lowest and the highest feed rate used to establish of 5.8%, a calibration relying on lines N° 1 and N° 5
models H(Vf) and B(Vf) (Eqs. 5–6) were retained as has an accuracy of 4.8% and a calibration relying on
well as a third value situated between the first two lines N° 5 and N° 8 has an accuracy of 4.4%. It is thus
(Table 2). possible to limit calibration just machining two pockets,
The extreme values are those of line N° 1 and N° 8, or even only one, but in such cases, the approach fails
while line N° 5 was chosen as an intermediate value. to benefit from a mean effect and may lead to impaired
The erosion factor Ke is determined as being the mean precision.
of the relations between the depth measured and the
depth calculated for lines N° 1, 5 and 8. This factor 3.5 Modelling with rapid calibration for configurations N°
Ke = 1.137 then allows the model to be finalised and 2 to N° 6
calibration from just these three pockets allows a mean
accuracy of 5% to be obtained, which is only just great- Having been set up on configuration N°1, the rapid
er than that of 3.4% obtained considering all the calibration method was used on configurations N° 2 to
pockets. The rapid calibration method thus applies ex- N° 6 that involve setting parameter changes for the
tremely well to modelling of the depth of the pockets pressure, SOD and flow and type of abrasive. As for
and it should be noted that even if that calibration relies handling of configuration N° 1, a series of 12 elemen-
on the results obtained up to line N° 8, it allows for a tary passes was performed and the shallower passes
good prediction for the depths obtained on lines N° 9, were eliminated to establish models of H(Vf) and
10, 11 and 12 since the mean of deviations on these B(Vf). Following this, three pockets were produced to
lines is 8.1%. The model thus extrapolates beyond the define the erosion coefficient Ke. To do so, the size of
domain that was used to define it. the grains of abrasive considered were 0.053 mm for
Calibration is possible using just two pockets. In the the abrasive of size 220 # and 0.092 mm for the abra-
case presented (Table 2), a calibration relying on lines sive of size 120 #. The results are presented in Table 3

Fig. 14 Depths of pockets for 0.700

configuration 1 0.600
Depth of pockets (mm)

0.500

0.400

0.300

0.200

0.100

0.000
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Modeled Mesured Depth (mm) Calculated Depth (mm) Vf(mm/min)
Int J Adv Manuf Technol

Fig. 15 Pocket with a corrugated Width (mm)

bottom 0
0 5 10 15 20
-0.05

Depth (mm)
-0.1
-0.15
-0.2
-0.25
Measured Modeled

and the shaded columns concern the values used to 3.6 Experimental validation of the models established
determine the erosion coefficient Ke. by rapid calibration
The various coefficients calculated and presented Table 3
show specific features: Identification (Table 3) of the parameters for depth
(Eq. 5) and width (Eq. 6) models of the elementary
– the values of coefficients Hv for the depth model of an passes as well as the coefficient of erosion Ke allows
elementary pass (Eq. 5) vary between −1.061 and −0.983. the depth machined for configurations 2 to 6 (Table 1)
The depth of an elementary pass can thus be considered to be predicted using equation (Eq. 12). As for configu-
as inversely proportional to the feed rate Vf. ration 1, for each configuration 2 to 6, four depth levels
– the value of the coefficient Bv for the width model of an were considered, eight pockets calculated to obtain a flat
elementary pass (Eq. 6) varies between −0.112 and bottom and four to verify the appearance of a corrugated
−0.010, which confirms that the width of an elementary bottom. The corresponding machining tasks were per-
pass is quasi-independent of the feed rate Vf. formed and the results obtained are presented (Fig. 16).
– the coefficient of erosion Ke varies very little since it
remains between 1.066 and 1.147. The ratio between
the sweep pitch and the width of an elementary pass 4 Results and discussion
(Eq. 13) gives a constant orientation to the jet’s escape
and thus produces identical additional erosion on the var- The results presented show an extremely good correla-
ious machined pockets. tion between the depths obtained using the model and

Table 2 Rapid calibration on configuration 1

N° Vf (mm/min) H(Vf) (mm) Calculated Ymax Mesured Mesured depth/ Modelled Ymax Precision
(mm) by (Eq. 7) Ymax (mm) Calculated depth (mm) by (Eq. 12)

1 125 0.575 0.589 0.666 1.160 0.653 2.0%

2 187 0.580 0.586 0.614 0.659 7.3%
3 208 0.349 0.351 0.377 0.396 5.2%
4 301 0.582 0.577 0.644 0.662 2.7%
5 314 0.348 0.344 0.392 1.107 0.396 1.0%
6 423 0.174 0.170 0.189 0.197 4.6%
7 514 0.344 0.334 0.407 0.391 3.9%
8 655 0.169 0.163 0.194 1.144 0.193 0.6%
9 1095 0.154 0.177 0.186 5.1%
10 1433 0.049 0.053 0.060 12.1%
11 2336 0.045 0.052 0.055 7.5%
12 4309 0.038 0.045 0.049 7.6%
Ke = 1.137
 Int J Adv Manuf Technol

Table 3 Rapid calibration for configurations 2 to 6

Configuration N°2
P = 2250 bar, SOD = 100 mm, abrasive size 220 #
Vf (mm/min) 691 1035 1150 1666 1742 2343 2846 3629 6066 7941 12,941 23,872
Coefficients for (Eq. 5) and (Eq. 6)
Ho = 407,337; Hv = −1061; Bo = 1947; Bv = −0.061
Pitch (mm) 1834 0.727 1112
Calculated depth (mm) by (Eq. 7) 0.497 0.467 0.127
Measured depth (mm) 0.560 0.521 0.152
Ke = 1.145 1.126 1.115 1.195
Configuration N°3
P = 3500 bar, SOD = 100 mm, abrasive 220 #
Vf (mm/min) 1470 1958 2099 2522 2803 2946 3623 3963 5145 5375 7305 9602
Coefficients for (Eq. 5) and (Eq. 6)
Ho = 469,310; Hv = −0.983; Bo = 2350; Bv = −0.083
Pitch (mm) 1834 1763 0.961
Calculated depth (mm) by (Eq. 7) 0.447 0.222 0.115
Measured depth (mm) 0.474 0.232 0.127
Ke = 1.068 1.060 1.046 1.099
Configuration N°4
P = 3500 bar, SOD = 40 mm, abrasive 220 #
Vf (mm/min) 2816 3750 4020 4831 5368 5642 6939 7590 9853 10.294 13.991 18.390
Coefficients for (Eq. 5) and (Eq. 6)
Ho = 1644,825; Hv = −1048; Bo = 1151; Bv = −0.072
Pitch (mm) 1195 1148 0.626
Calculated depth (mm) by (Eq. 7) 0.384 0.184 0.090
Measured depth (mm) 0.425 0.205 0.110
Ke = 1.147 1.107 1.113 1.221
Configuration N°5
P = 3500 bar, SOD = 100 mm, abrasive size 120 #
Vf (mm/min) 2371 3164 3394 4085 4558 4794 5919 6459 8428 8814 12.083 16.048
Coefficients for (Eq. 5) and (Eq. 6)
Ho = 931,979; Hv = −1037; Bo = 2787; Bv = −0.071
Pitch (mm) 1834 1037 1300
Calculated depth (mm) by (Eq. 7) 0.454 0.441 0.212
Measured depth (mm) 0.492 0.483 0.232
Ke = 1.088 1.077 1.094 1.092
Configuration N°6
P = 3500 bar, SOD = 40 mm, abrasive size 120 #
Vf (mm/min) 4542 6060 6500 7824 8730 9183 11.336 12.371 16.142 16.881 23.142 30.736
Coefficients for (Eq. 5) and (Eq. 6)
Ho = 2005,306; Hv = −1037; Bo = 1912; Bv = −0.112
Pitch (mm) 1195 0.676 0.847
Calculated depth (mm) by (Eq. 7) 0.357 0.338 0.159
Measured depth (mm) 0.385 0.319 0.187
Ke = 1.066 1.078 0.944 1.176

the depths measured. For the various configurations, the close to that mean since the maximum error calculated
mean for errors obtained between the calculated results for each configuration remains less than 13% except for
and the measured results (Table 4) is of the order of 5%. configuration N° 4 for which a maximum error of 22.6%
The error calculated for each machined pocket remains resulting from an anomaly is observed. This value is
Int J Adv Manuf Technol

Fig. 16 Results obtained— Configuration 2 Configuration 3

comparison of the depths 0.60 0.60
modelled and measured Measured Measured
Modeled Modeled

Depth (mm)

Depth (mm)
0.40 0.40

0.20 0.20

0.00 0.00
0 10000 20000 30000 0 5000 10000 15000
Vf (mm/min) Vf (mm/min)

Configuration 4 Configuration 5
0.60 0.60
Measured Measured
Modeled Modeled
Depth (mm)

Depth (mm)
0.40 0.40

Anomaly
0.20 0.20

0.00 0.00
0 5000 10000 15000 20000 0 5000 10000 15000 20000
Vf (mm/min) Vf (mm/min)

Configuration 6
0.60
Measured
Modeled
Depth (mm)

0.40

0.20

0.00
0 10000 20000 30000 40000
Vf (mm/min)

only present once and is identified on the graph be 8.9%. The method adopted applies uniformly over the
(Fig. 16). For that pocket, the depth machined is consid- various configurations tested and over the feed rate
erably shallower than that calculated and derives from ranges used. The mean and maximum errors also show
clogging of the abrasive supply system. If this value is that this method is effective since it combines rapid and
ignored, the maximum error for the configuration 4 will simplified calibration with a very high level of accuracy.
For each configuration, the four pockets with corru-
gated bottoms are effectively obtained but a corrugated
Table 4 Errors for the various configurations tested bottom also appears for pitch = 1 × B(Vf) for the
greatest depths when the pressure exceeds 1000 bars
Configuration N° Max error (%) Mean error (%)
(Fig. 17).
1 12.1% 5.0% The ratio between the width factor B(Vf) and the
2 12.4% 5.2% sweep pitch (Eq. 13) thus seems to be slightly influenced
3 9.7% 2.4% by the pressure. For configurations 2 to 6, the corrugated
4 22.6% 5.1% bottom appears only once abnormally for pitch = 1 ×
5 5.8% 2.4% B(Vf) when the greatest depth is achieved. A pitch of
6 5.8% 1.9% between 0.6 × B(Vf) and 0.9 × B(Vf) would thus be more
appropriate.
 Int J Adv Manuf Technol

Fig. 17 Pocket with a slightly Width (mm)

corrugated bottom 0
-1 4 9 14 19
-0.1

Depth (mm)
-0.2
-0.3
-0.4
-0.5
-0.6
Measured Modeled

5 Conclusion of jet–workpiece traverse speed and abrasive grit size on the character-
istics of the milled material. J Mater Process Technol 161(3):407–414
2. Goutham U, Hasu BS, Chakraverti G, Kanthababu M (2016)
The study presents a model for pocket depth in blind machin- Experimental investigation of pocket milling on Inconel 825 using
ing using an abrasive water jet in a context where a machine, a abrasive water jet machining. Int J Curr Eng Technol 6(1):295–302
pressure and an abrasive are given. The material considered 3. Srinivasu DS, Axinte DA, Shipway PH, Folkes J (2009) Influence
of kinematic operating parameters on kerf geometry in abrasive
was a Ti-6Al-4V titanium alloy. In order to achieve productive
waterjet machining of silicon carbide ceramics. Int J Mach Tools
machining, the optimal abrasive flow rate, i.e., that corre- Manuf 49(14):1077–1088
sponding to the greatest MRR, is considered. This machine, 4. Alberdi A, Rivero A, López de Lacalle LN, Etxeberria I, Suárez A
material, pressure, abrasive and optimal abrasive flow rate (2010) Effect of process parameter on the kerf geometry in abrasive
water jet milling. Int J Adv Manuf Technol 51(5):467–480
combination plus the stand-off distance defines the machining
5. Carrascal A, Alberdi A (2010) Evolutionary Industrial Physical
configuration concept. For a configuration thus defined, the Model Generation. Proceeding of the International Conference
feed rate influences the width and depth of an elementary pass HAIS. San Sebastian, Part I, June 2010. 327–334
and the depth of the pockets will only depend on the feed rate 6. Alberdi A, Rivero A, López de Lacalle LN (2011) Experimental
study of the slot overlapping and tool path variation effect in abra-
and the sweep pitch.
sive waterjet milling. J Manuf Sci Eng 133(3):034502-1–034502-4
The width and depth of a series of elementary passes were 7. Dittrich M, Dix M, Kuhl M, Palumbo B, Tagliaferri F (2014)
modelled. These two parameters allow a simple pocket depth Process analysis of water abrasive fine jet structuring of ceramic
model to be defined that takes into account the erosion specific surfaces via design of experiment. Procedia CIRP 14:442–447
8. Nguyen T, Wang J, Li W (2015) Process models for controlled-
to the milling of pockets. So as to proceed with rapid imple-
depth abrasive waterjet milling of amorphous glasses. Int J Adv
mentation of the model, a rapid calibration method is pro- Manuf Technol 77(5):1177–1189
posed that allow accuracy to be maintained. The entire ap- 9. Boud F, Loo LF, Kinnell PK (2014) The impact of plain waterjet
proach is validated experimentally since the models devel- machining on the surface integrity of Aluminium 7475. Procedia
oped varying the pressure, stand-off distance and type of abra- CIRP 13:382–386
10. Kowsari K, Nouraeia H, Samarehb B, Papini M, Spelt JK (2016)
sive are accurate to within 5%. Furthermore, the study details CFD-aided prediction of the shape of abrasive slurry jet micro-
the mechanisms for formation of pocket bottoms and shows machined channels in sintered ceramics. Ceram Int 42(6):7030–7042
that there exists an interval of the sweep pitch over the width 11. Tamannaeea N, Spelt JK, Papini M (2016) Abrasive slurry jet
of an elementary pass ratio within which the pocket bottom micro-machining of edges, planar areas and transitional slopes in
a talc-filled co-polymer. Precis Eng 43:52–62
can be considered to be flat and with a satisfactory surface 12. Anwar S, Axinte DA, Becker AA (2013) Finite element modelling
condition. To round off this approach, a further study will be of overlapping abrasive waterjet milled footprints. Wear 303(1–2):
proposed in order to establish a connection between the ge- 426–436
ometry of an elementary pass, the shape of the pocket bottom 13. Escobar-Palafox G, Gault RS, Ridgway K (2012) Characterisation
of abrasive waterjet process for pocket milling in Inconel 718.
generated by that elementary pass and a given sweep pitch, as Procedia CIRP 1:404–408
well as characterisation of the surface condition taking the 14. Paul S, Hoogstrate AM, van Luttervelt CA, Kals HJJ (1998) An
erosion mechanism into account. experimental investigation of rectangular pocket milling with abra-
sive water jet. J Mater Process Technol 73(1–3):179–188
15. Konga MC, Axinte D, Voice W (2010) Aspects of material removal
mechanism in plain waterjet milling on gamma titanium aluminide.
J Mater Process Technol 210(3):573–584
16. Finnie I (1960) Erosion of surface by solid particles. Wear 3(2):87–103
References 17. Bitter JGA (1963) A study of erosion phenomena-Part2. Wear 6(3):
169–190
1. Fowler G, Shipway PH, Pashby IR (2005) Abrasive water-jet con- 18. Hashish M (1987) Milling with abrasive-waterjets: a preliminary in-
trolled depth milling of Ti6Al4V alloy—an investigation of the role vestigation. In Proceeding of the fourth U.S. waterjet conference. 1–20