CUTTING ERRORS IN ABRASIVE WATER JET CUTTING

2. ABRASIVE CUTTING Pure waterjets are used to cut soft materials, while the abrasive
waterjet cuts hard materials, such as metals, stone, composites and ceramics. The abrasive
waterjet differs from the pure waterjet by means of addition of the abrasive particles which, not
the water, erode the material. Therefore is the abrasive waterjet hundreds or thousands of times
more efficient than a pure waterjet. Abrasive waterjets can cut materials with hardness up to and
slightly beyond aluminium oxide ceramic (AD 99.9). Typical abrasive mesh sizes in abrasive
waterjet cutting are 120, 80, and 50. The different mesh sizes do not have a significant impact on
part accuracy. They have a greater impact on surface finish and overall cut speed. Finer abrasives
(larger mesh number) produce slower cuts and smoother surfaces. In abrasive waterjet cutting it
is often thought that if the abrasive flow rate is reduced, process will save money. On the
contrary, it wastes money. There is a peak performance point where abrasive waterjets will
operate efficiently. Machine operating cost include power, water, air, seals, check valves, orifice,
mixing tube, abrasive, inlet water filters, long term spares (hydraulic pump, high-pressure
cylinders, etc). Basically, these are all the items that need replacement regularly or over the life
of the waterjet system. Abrasive constitutes 2/3 of the machine operating cost of the equipment.
When all overhead are included, the cheapest cutting is always found at the fastest possible
speed. This fact is independent of the material type to be cut, or the power of the system.
Consumption of abrasive is within 0,25 and 0,5 kg/min. Abrasive particles already used has to be
washed and dried what allow recovery of 20 to 60 %. Recycling of abrasive accounts for up to
65% of total operating costs of running the AWJ

3. CUTTING ERRORS ESTIMATION To perform cutting within demanded quality it is of

big importance to have a machine as better accuracy as possible but also to have appropriate
software to control the machine since good software can increase cutting speed. Tolerance ranges
achievable on water jet cutting machines are +0,025 mm to –0,025 mm. Much more realistic is
tolerance range +0,05 mm to –0,05 mm. To avoid cutting errors there are several performances
to be checked: accuracy and repeatability, maximum contour velocity, travel length and
operating pressure. Some results are shown in figure 1. Dynamic accuracy tests are conducted by
either cutting of parts or yet by a device called a Ballbar. Ballbar testing is included in a number
of standards for machine tool accuracy i.e. ISO 230, ASME B5.54 and BS3800. It is accurate to
approximate +/- 0.5 microns, or 20 microinches at 20 °C.
The technique allows pure water to pass through a nozzle of small diameter, 0.18 - 0.35 mm, at
very high pressure, usually 3000 - 4000 bars. The water now travelling about 700-850 m/sec,
holds enough kinetic energy to cut through any material under this jet. As an energy presenting
factor, streaming speed of water can be calculated from Bernoulli equation:

𝑔 1/2
v=(2 ∗ 𝑝 ∗ 𝜌) where is:

p - nominal output pressure, Pa ρ - specific mass, kg/ m3 g - gravity constant, m/s2

Figure 3 shows speed of water stream calculated from Bernoulli equation for different water
pressures. Figure 4 shows influence of nozzle diameter and its distance from material being cut
on water speed for pumping pressure of 3250 bar. One of the bigger costs during abrasive water
jet cutting is the cost of abrasives. Figure 5 shows mass of abrasive influencing streaming speed
of water and abrasives for different nozzle diameters.

4. CONCLUSION Cutting errors of abrasive water jet cutting were shown in this paper. At the
beginning of paper difference between water jet and abrasive water jet was pointed out.
Advantages of abrasive jet cutting were shown, but also the problems with nozzle life, errors on
bottom surface of material. An observation on cutting accuracy and tolerances obtained on
machines of this technology and tolerances influencing factors are discussed. Beside mentioned,
the possible defects, failures and deformations with causes of their appearing are commented.
Especial causes of cut tighten are discussed. The influencing factors on cutting speed and
measuring procedure of abrasive water jet by Laser Transient Anemometer-LTA are explained.
Mixing nozzle and relationship between convenient nozzle and spiral nozzle type are simply
explained. Also, advantages and disadvantages of cutting by abrasive jet according to other
applied processes are specified. 5. REFERENCES [1] D. Saunders, Jet Cutting Technology,
Elsevier Applied Science London and New York [2] Aliko Automation Oy,
http://www.aliko.fi/english /watercut.htm [3] http://www.omax.com [4]
http://www.directindustry.com

Influence of machining parameters on part geometrical error in abrasive waterjet offset-

mode turning

Abstract Geometrical error in abrasive waterjet turned parts is an important challenge towards
the abrasive waterjet turning process commercialisation. A Systematic study has not been done
yet to investigate the effects of process parameters on geometrical error in abrasive waterjet
offset-mode turning. In this paper, a comprehensive study has been performed to investigate the
influence of several machining parameters on the geometrical error (part diameter percent error)
in turning AA2011-T4 aluminum alloy round bars. Water pressure, cutting head traverse speed,
workpiece rotational speed, abrasive mass flow rate and depth of cut were considered as the main
machining parameters in a five levels statistical experimental design. Basing on central
composite rotatable design (CCRD), a total of 52 experiments was carried out. The main effects
of the parameters and interactions among them were analyzed based on the analysis of variance
(ANOVA) technique and the response contours for the part geometrical error were obtained
using a quadratic regression model (i.e. RSM). The model predictions were found to be in good
agreement with experimental data. Furthermore, among the significant parameters, pressure,
depth of cut and traverse speed are the most influential parameters, with percent contribution of
almost 25% each. Abrasive mass flow rate places as the least influential parameter with a percent
contribution of 4%. Keywords Abrasive waterjet turning, Erosion, Machining, Response surface
methodology, Geometrical error 1. Introduction Abrasive waterjet turning (AWJT) is an
innovative non-traditional machining technique which enables using advantages of waterjet in
producing axisymmetric parts 1-5 . In the AWJT process, the workpiece revolves while the
cutting head axially moves with a definite depth of cut to produce the required geometry (Figure
1). AWJT has superior benefits in comparison with a conventional turning. Material removal
takes place by means of a flexible cutting tool (abrasive waterjet), so AWJT is less sensitive to
the workpiece shape. It allows to machine at high depths of cut in one pass and offers fairly
higher material removal rates (MRR), especially for hard-to-machine materials 5, 6 . The process
involves low cutting forces, so it is 2 quite independent of the workpiece length-to-diameter ratio
and therefore enables to turn long parts with small diameters 7 . Since abrasives have the
capability to erode almost any materials, this process is ideally suitable for materials with low
machinability such as ceramics, composites, glass, Titanium alloys and so on 6, 8, 9 . However,
this process involves some challenges that limit its rapid growth and use in industries.
Experimental investigations by pioneers of this technique, Ansari and Hashish, show that AWJT
is a near-net shape machining process. 1, 2 . It was reported that the final diameter of the turned
part is usually more than the desired diameter because of the jet deflection and its instability 10.
However, from a visualization study, Ansari et al. 11 pointed out how the abrasive waterjet does
not undergo any significant radial deflection in the region where material removal takes place.
Axinte et al. utilized AWJT as an efficient method to profile and dress grinding wheels and
proved its technological and economical capability 5 . They could turn parts up to ±0.1 mm
accuracy. Studies on precision turning with AWJ showed that the accuracy of turned parts is
affected by the jet trail-back and deflection 10, 12. Machining at high traverse speeds and depths
of cut causes jet instability, which results in rougher surfaces, striation marks, poorer roundness
and inconsistency in achieving the desired diameter. Depending on the position of the nozzle/jet
relative to the workpiece, Li et al. 13 classified AWJT as “radial-mode” or “offset-mode”. They
evidenced the advantages of radial-mode turning over the offset-mode turning including more jet
energy utilization, higher surface speeds, capability of nozzle tilt angle variations and smaller
nozzle stand-off distances. These factors enable the process to provide higher material removal
rates. However, controlling the depth of cut seems to be still an important challenge for radial-
mode turning. Some mathematical models capable to estimate the workpiece diameter
continuous change in AWJT were also presented 2, 7, 14-16. An analytical model suggested by
Ansari and Hashish 2 relates the volume swept by the combined specimen rotation and AWJ
head traverse in the time unit (defined as the volume sweep rate (VSR)) to the material removal
rate. This model could predict the workpiece final diameter for various sets of AWJT parameters.
In spite of the continuous variation of impact angle during the workpiece diameter reduction,
Hashish's analytical model does not consider impact angle modifications. An erosion based
approach considering the varying local impact angle was presented by Manu and Babu 7 to
predict the workpiece final diameter. However, their model does not accurately predict the final
diameter at various traverse speeds. Moreover, when the impact angle tends to zero, their model
over estimates the removed material volume. By applying Hashish erosion model, Zohourkari
and Zohoor presented a model with better estimation in terms of final diameter prediction 15, 17.
Hlavac et al. 16 presented a very comprehensive model, even if all the mentioned models do not
consider the reduction of jet energy utilization at depths of cut lower than the jet diameter, the
exact material flow stress and the focusing nozzle wear. Analytical models are still in their early
stages and must be developed to become practical. Thus, statistical models which are capable to
include the effects of controllable and uncontrollable parameters can be useful to model the
AWJT process. To improve the applicability of AWJT and to improve its accuracy, it is
important to study the effects of operational parameters on the turned parts geometrical error and
look for strategies to reduce it. Up to now, the lack of a systematic experimental study on AWJT
able to show the effect of parameters on geometrical error is sensible, therefore, the effects of
several machining parameters on the part geometrical error in abrasive waterjet offset-mode
turning of AA2011-T4 are investigated in this paper. Five major machining parameters such as
water pressure, cutting head traverse speed, workpiece rotational speed, abrasive mass flow rate
and depth of cut were considered in a five-levels statistical experimental design. Based on central
composite rotatable design (CCRD), a total of 52 experiments was carried out. The main effects
of parameters and interactions among them were analyzed based on the 3 analysis of variance
(ANOVA) technique by Minitab 16® software. The response contours for the geometrical error,
defined as the part diameter percent error, were obtained using a quadratic regression. All
parameters investigated in this research can be adjusted to the desired levels and their continual
change is possible. It must be noted that there are other effective parameters such as orifice and
focusing tube diameter, focusing tube length, abrasive material, size and shape, but the present
study is devoted to act on the most controllable parameters in a standard waterjet machining
center, keeping other parameters at a representative and significant level. This work aims to
obtain a valuable understanding of parameters effects in the abrasive waterjet offset-mode
turning process and presents a statistical model suitable to improve its accuracy.

2. Abrasive waterjet turning strategy Based on the relative position of jet and workpiece, AWJT
can be classified as radial-mode or offset-mode turning. Advantages of offset-mode turning
compared to radial-mode turning are the ability of controlling the depth of cut 13 and better
surface quality 1, 4, 18. Hence, AWJ offset-mode turning has been chosen for this study since it
seems to be more suitable to meet industrial requirements. The schematic of the AWJ offset-
mode turning is shown in Figure 1. In this process, the abrasive waterjet is adjusted in a desired
position defining the nominal depth of cut (DOC) and minimum distance between the focusing
tube tip and the workpiece surface (Lc). The workpiece rotates at the rotational speed N, while
the jet moves along the workpiece rotation axis at the traverse speed u and erodes the workpiece
surface in one pass to the final diameter (Df). Because of low cutting forces and no cutting tool
wear, since the abrasive waterjet is the tool, it is possible to turn parts at high depths of cut in one
pass 6 . Ideally, at the end of the part exposure to the jet, the jet becomes tangent to the surface of
the workpiece. But, in practice, the final part diameter is more than the target diameter (Dt),
which introduces a geometrical error.

3. Experiments The AWJT experimental apparatus was prepared by applying an AWJ machine
(Tecnocut 5- axis handling system with a Flow 9XV-S 380 MPa pump) that is equipped with a
custombuilt lathe with maximum rotational speed equal to 1000 rpm (Figure 2).
30 mm diameter AA2011-T4 circular bars were selected for this study. All the parts were cut to a
10 cm length and carefully cleaned with ethanol alcohol. The material composition of AA2011-
T4 is given in Table 1.

Based on previous researches by Hlavac et al. 19-22, it is possible to assume that abrasives shape
and size change due to their fragmentation while mixing and accelerating in the AWJ cutting
head. This phenomenon depends on several parameters such as pressure, abrasive mass flow rate,
orifice size, mixing chamber inner shape and focusing tube length and internal diameter. Since
the abrasives size and shapes involved with the machining process have been kept constant at a
significant industrial level, their effects were not investigated in this study. Mesh #80 GMA
Australian Garnet was used for all the experiments (Figure 3). A standard 0.3 mm diameter
orifice and a standard 1.02 mm diameter focusing tube were used for all the tests.

3.1. DOC adjustment To obtain the required workpiece geometry, it is important to accurately
adjust the DOC. The reference system applied to set the DOC at each experimental run has been
defined by carrying out an accurate workpiece alignment procedure allowing the jet traverse line
to be parallel to the workpiece axis and the jet to be tangent to the workpiece surface. Low water
pressure has been used during such a procedure in order to obtain a very thin and coherent jet.
After the DOC adjustment, the distance between the focusing tube tip and the workpiece surface
(Lc) was set to 1 mm to avoid collision between. 3.2. Experimental design The ranges of the
selected factors, i.e. water pressure (P), cutting head traverse speed (u), workpiece rotational
speed (N), abrasive mass flow rate (m.a) and depth of cut (DOC), were identified during
preliminary experiments using a “one factor at a time” methodology with respect to acceptable
geometrical error, material removal rate and surface quality. Then the selected factors were
organised in central composite rotatable design a CCRD scheme in five levels each. CCRD is an
effective design able to handle linear, quadratic, and interaction terms in statistical modeling of
processes 23. CCRDs schemes include three sets of design points that are corner points (nF = 2k
), axial points (na = 2k) and center points (nc), where k is the number of process parameters. k =
5 in the present case, so nF and na are equal to respectively 32 and 10. In order to consider the
experiments pure error, it is common to perform some replicated experiments at the center point
(nc). 10 center points were added to the experiments in this case. The distance of the axial points
from the center point is determined by the α value (put a reference here). For a CCRD, α = nF
1/4. According to the number of factors in this study (k = 5), α = 2.378 and a total of 52
experiments was designed to perform. The experimental factors are given in Table 2, where their
coded and uncoded values are reported. Coding factors is an important step in response-surface
analysis to allow a direct comparison of the factors weight on the process response 23, 24. Thus,
higher and lower levels of the corner points were respectively coded with +1 and -1; the center
points were coded with 0 and higher and lower levels of axial points were coded to + α and - α
correspondingly. The linear relationships between the coded and the actual factors values are
given in Eqs. from (1-a) to (1-e).

Subscripts "high" and "low" represent the higher and lower levels of corner points and subscript
"0" indicates the center point. The initial and final diameters, respectively Di and Df, were
measured by means of a Zeiss Prismo 5 HTG VAST coordinate measuring machine (Figure 4)
and the geometrical error GE, defined as the part diameter percent error, was calculated
according to Eq. 2
3.2.1. Response surface methodology Response Surface Methodology (RSM) is a statistical
approach to find a mathematical form of the relationship among the process responses and the
process parameters using statistical and mathematical techniques23, 25-27. The mathematical
equation stating the relationship between the AWJT process parameters and the geometrical error
response can be expressed as:GE=f(x1,x2,x3,x4,x5)+𝜀 (3) where f is the response function and
x1,x2,x3,x4,x5 are respectively water pressure (P), cutting head traverse speed (u), workpiece
rotational speed (N), abrasive mass flow rate (m.a) and depth of cut (DOC) and ε is the error
term. The response function (f) is unknown and RSM has the ability to approximate it by a
suitable polynomial. A second-order polynomial has been chosen due to its ability to model
curvatures in the response surfaces as expressed in Eq. 4.

where 𝜂 is the approximated response and 𝛽ij are the regression coefficients. The coefficients are
found by using the least squares method and the linear regression analysis 23. To obtain
significant parameters, an analysis of variance is carried out. It must be concluded that 7
confidence intervals for selecting significant parameters are typically fixed at the 95%
confidence level 28. 4. Results and discussions 4.1. Statistical modeling of part geometrical
error Abrasive waterjet turned parts obtained from planned experiment (i.e. 32 corner points (nF
= 32), 10 center points (nc = 10) and 10 axial points (na = 10)) are shown in Figure 4.
According to the model defined by Eq. 2, a statistical analysis was accomplished considering the
part geometrical error as process response. The ANOVA results for the geometrical error have
been shown in Table 3. As shown in this table, it is possible to conclude that the secondorder
regression model is significant, since its respective p-value is sufficiently less than 0.05.
Moreover, the null hypothesis of no lack of fit cannot be rejected (p-value higher than 0.05),
which shows that no other predictors are required. It has been found that, among the input
process parameters, water pressure, cutting head traverse speed, abrasive mass flow rate and
depth of cut are significant and workpiece rotational speed is insignificant. In addition, second-
order terms of water pressure (P2 ) and depth of cut (DOC2 ), interaction between water pressure
and traverse speed (P×u), water pressure and depth of cut (P×DOC), abrasive mass flow rate and
depth of cut (m!×DOC) and between traverse speed and depth of cut (u×DOC) are significant.
The higher the pvalue, the less significant the parameter is, hence, the interaction effect of
abrasive mass flow rate and depth of cut places as the last of the significant effects. The other
terms (p-value > 0.05) can be assumed to be insignificant.

As shown in Figure 5, the obtained model accurately fits the experimental data. Upon this, the
final model for geometrical error is given in Eq. 5.
To evaluate the fitting adequacy of the model, the coefficient of determination R2 has been
calculated. The R2 value indicates that 94.77 % of the total deviations in the process response 2
4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
Experiment number GE (%) Experimental data Regression model 9 can be explained by the
model. Since the R2 approaches unity, the model fits the experimental data accurately 23. 4.2.
Effects of the process parameters on geometrical error Percent contributions of the model
effects have been calculated from their sequential sum of squares (Seq SS) from Table 3 and are
graphically shown in Figure 6. It illustrates that, among process parameters, water pressure,
cutting head traverse speed and depth of cut are the most influential parameters on controlling
the part geometrical error and the abrasive mass flow rate places as the least influential
parameter with a percent contribution less than 1%. As the workpiece rotational speed did not
have any significant influence on AWJT results in the tested ranges, it was excluded from further
consideration in the present paper.
Figure 7 shows the contour plot of the geometrical error response with respect to the water
pressure and the workpiece traverse speed at constant levels of workpiece rotational speed (400
rpm), abrasive mass flow rate (5.24 g/s) and depth of cut (3 mm). It illustrates that a geometrical
error reduction is generally obtainable by decreasing the workpiece traverse speed and increasing
the water pressure. An increase of the traverse speed causes a decrease of the exposure time to
the jet action. Less abrasive particles impact the surface and the removed volume from the
periphery of the workpiece decreases. According to Bernoulli's law and the momentum transfer
from water to the abrasives, higher pressures produce more energy and more acceleration of
abrasive particles. This condition results in higher erosion rate during the limited exposure time
and reduces the workpiece diameter more.
Figure 8 displays 3D surface and contour plot of the geometrical error response in relation to the
water pressure and depth of cut while other parameters are kept constant at their middle levels.
The opposite effect of depth of cut and pressure on geometrical error is noticeable in the
hyperbolic paraboloid (saddle) response surface. Higher depth of cut results in higher
geometrical error due to more material to erode 2 and jet instability 10, 29. Higher pressures
reduce the geometrical error as discussed for Figure 7. At the saddle point, the effect of each
parameter compensates the opposite effect of the other one. In this condition, keeping the water
pressure constant and increasing or decreasing the depth of cut leads to reduction of the
geometrical error. This fact happens because during the machining, the initial set DOC is
gradually decreased (what do you mean?). In turning at high depths of cut, if the jet energy is
high enough (high pressures), it is possible to turn the workpiece until reaching a similar
condition to turning at low depths of cut. Thus, produced parts at high and low depths of cut have
almost same geometrical error. Instead, if the water pressure decreases, it is not sufficient to
efficiently erode the whole material volume. So, increasing DOC at low pressures increases the
geometrical error.

Effects of abrasive mass flow rate and traverse speed on the geometrical error are illustrated as
contour plot in Figure 9. It is concluded that lower geometrical error is achievable at high
abrasive mass flow rates and low traverse speeds. Additionally, at high traverse speeds, the effect
of abrasive mass flow rate on reducing the geometrical error is higher than when machining at
low workpiece traverse speeds. Similar results were reported by Ansari 2 .
The simultaneous effects of abrasive mass flow rate and depth of cut are depicted in Figure 10.
In agreement with the previous research by Ansari 2 , increasing abrasive mass flow rate and
decreasing depth of cut reduce the geometrical error. It is worth noting that increasing the
abrasive mass flow rate while machining at low depths of cut does not have a meaningful effect
on the geometrical error. It means that the material that must be eroded is low enough that higher
abrasive mass flow rate cannot be effectively used in the material removal.

Effects of pressure and abrasive mass flow rate on the geometrical error, while other parameters
are kept constant, are demonstrated in Figure 11. Increasing pressure and abrasive mass flow rate
leads to a reduction of the geometrical error. At low pressures, variations of abrasive mass flow
rate almost have no effect on the geometrical error while, at high pressures, abrasive particles are
accelerated enough to turn the part with closer tolerances. These findings confirm the previous
investigations by Ansari and Hashish 2, 3, 30.
Figure 12 shows contour plot of the geometrical error response with respect to traverse speed and
depth of cut at constant levels of pressure, abrasive mass flow rate and rotational speed. It
illustrates that for higher depths of cut decreasing traverse speed can highly reduce geometrical
error. However, decreasing traverse speed increases the time of machining. In practice turning
parts at low depth of cut and high traverse speed in multi passes can be a practical way to
increase the process efficiency.

5. Conclusions An experimental study was conducted to investigate the effects of major process
parameters and their interaction on geometrical error in abrasive waterjet turning. A five-level
experimental design was carried out based on central composite rotatable design. The main
effects of parameters and also the interaction between them were analyzed based on an analysis
of variance. It is found that among the input process parameters, pressure, abrasive mass flow
rate, traverse speed and depth of cut are significantly effective while workpiece rotational speed
has not so significant effect on the geometrical error of AWJ turned parts. Pressure, traverse
speed and depth of cut were detected as the most significant parameters with almost the same
percentage of contribution. Workpiece rotational speed is the least significant parameter. In
addition, interactions between pressure and traverse speed, pressure and depth of cut, abrasive
mass flow rate and depth of cut and between traverse speed and depth of cut were found to be
significant. Besides, a mathematical model for relationship between the process parameters and
geometrical error response has been presented based on a quadratic regression. The model could
successfully predict the geometrical error of AWJ turned parts. The presented study activates a
potential to improve the precision of abrasive waterjet turning. This requires further
investigations to examine reducing geometrical error in relation to improvement of material
removal rate and surface quality.

*ADVANCED ERROR CORRECTION METHODOLOGY APPLIED TO ABRASIVE

WATERJET CUTTING*

ABSTRACT First introduced in the last decade, the concept of “compute first, move later” for
motion control optimized the abrasive waterjet cutting process. Using a cutting model that varied
cutting speed around corners and along arcs provided a greater degree of accuracy. Recent
innovations in corner-cutting strategy allow the cutting of external corners without sacrificing
speed or surface quality. The latest development in machine-control hardware and software
advances the error correction methodology in abrasivejet machining to the next generation. The
nozzle automatically tilts along the cutting path, compensating for taper error as well as lag error
at corners. A tool tip tilt mechanism lets the nozzle tilt quickly about the material entry point
without large accelerations and vibrations. This paper provides insight into the development of
these advanced error correction technologies as applied to abrasivejet cutting.

1 INTRODUCTION After two decades of development, abrasive waterjets have become a

mainstream technology that works hand-in-hand with other machining technologies in modern
machine shops. Abrasivejets cut five to ten times faster than EDM. While lasers are limited to
non-reflective materials, abrasivejets are not. Abrasivejets can cut thicker materials than lasers,
and the production cost is two-thirds less. Fixturing is less complicated on an abrasivejet than on
a traditional machining center, and setup time is reduced. As the abrasivejet becomes accepted in
more and varied applications, however, the challenges increase. Greater productivity (faster
machining times) and higher accuracy are two constant and often contradictory challenges.
Consider the challenge of higher accuracy. These errors and defects can be found on parts
machined by abrasivejets: ‰ Striation marks along the bottom half of the cut surface ‰ Jet lag
errors at corners of the path and on small arcs ‰ Taper and barrel errors ‰ Geometrical error
from kerf width change ‰ Lead-in and lead-out defects ‰ Frosting and rounding on top of the
kerf ‰ Burrs at the bottom of the kerf Previous work by Zeng et al. (1999) addressed striation
marks and jet lag errors at corners of path and on small arcs. A jet shape modeling concept has
been described by Henning and Anders, which shed light on real-time taper error compensation.
The feasibility of such compensation was also proven later by Knaupp et al. A work by Groppetti
et al. investigated edge rounding on top and burrs at bottom of the kerf. A study by Anderson and
Johansson was devoted to lead-in and lead-out defects. This paper presents a systematic
approach for error correction starting from a new concept of motion control. 2 THE CONCEPT
OF “COMPUTE FIRST, MOVE LATER” A machine tool controller essentially does three
things: it accepts a tool path as input, which is defined by a relatively small number of
parameters; it provides a user interface; and it generates the physical position of a tool along
multiple axes as a function of time. Most machine tool controllers accept their input in the form
of “G Code,” consisting of sequential movement commands interlaced with On/Off commands
for various machine functions. A three line sample of G Code is shown below:

000 G00 X1 Y2

010 M73

020 G01 X2 Y3 F120

In this example, the lines are sequentially numbered in steps of 10. The first line indicates that
the nozzle should move to coordinates X1, Y2 at a preset rapid traverse speed. The next line
turns on something, such as a cutting jet. The third line moves the nozzle to coordinates X2, Y3
at a speed of 120. The controller executes these commands one at a time, and performs any
calculations necessary to coordinate the multiple axes as the motion proceeds. Some controllers
process a few lines ahead, helping to manage accelerations at sharp turns of the path. This code
can be written by hand or generated automatically with a CAD/CAM program. Note that this
type of controller requires the exact motion path and the desired speed of motion as input. The
part shape and the tool path shape usually have a one-to-one relationship independent of the
speed and accelerations along the path in machining applications. But this assumption is not true
for abrasivejet cutting. In 1996 a patent was issued on a different control mechanism called
“compute first, move later.” This control mechanism was developed specifically for abrasivejet
cutting. The same functions are performed, but by using the large memory and computing power
of modern personal computers, they may be performed in a different order. In this type of
controller, the part shape is first described by a series of lines and arcs, much like the G Code
representation, but without any data regarding speeds of motion. The entire tool path is then
interpolated with speeds set according to a model of the abrasivejet cutting process. At this point,
the path is stored digitally in a large data array with about five kilobytes per centimeter of the
tool path. Here it can be iteratively manipulated to correct for various abrasivejet behaviors,
producing a part of the desired shape and tolerance. The ability to process the entire path
iteratively is a major advantage in abrasivejet machining. First, there is the ability to look ahead
for the entire length of the path. In a thick material, a sharp corner near the end of the path may
require the abrasivejet to slow at the start of the path. In a thin material, high speed on a straight
portion of a path may require tilting of the nozzle to avoid taper on the edge of the part. The error
correction phases described below manipulate the large data array that represents the path by
changing the speeds as a function of path geometry, and by inserting tilting motions to
compensate for the natural taper produced by the abrasivejet. This results in more accurate parts.

3 ERROR CORRECTION PHASE 1: REDUCING GEOMETRICAL AND COSMETIC

ERROR On parts cut with an abrasivejet, the most profound errors are the striation marks and
jet lag errors at corners and arcs. These errors often make the parts either cosmetically
unacceptable or outside of tolerance requirements. An abrasivejet cutting model was published
by Zeng et al. in 1992. The model predicted the cutting speed for a given quality index,
combined with other process parameters. By selecting a proper quality index in the range of one
to five, the striation marks along the cut surface are partially or completely eliminated. Choosing
a quality index of “one” creates a separation cut, a very rough cut without any concern for
surface finish. Choosing a quality index of “five” creates a smooth and striation-free surface.
Employing a similar idea, the amount of permissible jet lag errors at corners and arcs is used to
calculate an effective quality index. This effective quality index depends on the angle of corners,
and the radius of arcs. The result of using the effective quality index at corners and arcs is a
reduced cutting speed at these locations. By using the “compute first, move later” concept and
the PC-based software infrastructure built on this concept, the transition of cutting speed at
corners and arcs can be conveniently handled at increments of one motor step. In practical terms,
this means the cutting speed can be varied every 12 microns or so along the path. Varying the
speed one motor step at a time is a vast improvement over the traditional CNC controller—it
allows higher cut quality at corners. 4 ERROR CORRECTION PHASE 2: BALANCE OF
SPEED AND QUALITY These advances to abrasivejet machining technology have been
available since 1993. Since then, the technology has been improved even further—the cutting
process is faster, and cuts are more precise. Some of the more recent enhancements include: ‰
Automatic setting of lead-in length and speed based on a pierce model, providing the fastest
possible dynamic piercing ‰ A new cornering strategy that treats every possible corner geometry
uniquely ‰ Automatic addition of “corner passing” based on a cutting model built into the
software, speeding up outside corners where there is room to do so.

4.1 Stage 1: Piercing There are many ways to pierce a material. A material can be pre-drilled
mechanically, pierced by a stationary abrasivejet, pierced by moving the jet back and forth over a
fixed distance (“wiggle” piercing), or dynamically pierced by turning on the abrasivejet and then
slowly moving across the material. Each of these has its advantages, and each is appropriate for
different situations. The two most popular methods are dynamic piercing and wiggle piercing,
because they generally offer the fastest piercing with the most convenience for a wide variety of
applications. Wiggle piercing has been a good balance between fast piercing, short pierce
distance, and ease of programming. It is generally faster than most implementations of dynamic
piercing, and is especially good for piercing thick material in small spaces, such as piercing a 6
mm hole in 5 cm thick steel. Dynamic piercing is limited by the difficulty in determining and
setting the ideal lead-in length and speed for each pierce point in the path. This is solved with the
use of a dynamic piercing model that looks at the machining conditions and automatically sets
the pierce length and speed based on the geometry of the part and the pump and nozzle setup.
Using this model, it is possible to greatly speed up piercing performance without any user
intervention or knowledge of the process. If a tiny hole is being pierced, then the lead-in will
shrink, and the piercing speed will be reduced to make up for the loss of room. If the hole is
larger, the pierce length will increase to the optimal length, and the speed will be increased
appropriately. This makes dynamic piercing ideal for most applications, with the exception of
small holes in thick materials, where wiggle piercing may still be the preferred choice. 4.2 Stage
2: Corner Corrections The main drawback of earlier controller technologies was that the
controller treated outside corners with the same cutting model parameters as inside corners. The
controller slowed more than necessary for outside corners. By applying separate cutting models
to inside and outside corners, both speed and precision are improved. Outside corners can be
machined much faster, and inside corners can be machined at precise speeds and accelerations
that minimize blowout and kickback more than previously possible. 4.3 Stage 3: Corner Passing
Corner passing refers to the concept of overshooting a corner, reversing the abrasivejet, and then
continuing along the tool path. The advantages of corner passing are that software can program
the corner passes automatically and determine the best geometry to use, adding corner passes
only when and where needed To use corner passing, software first computes how much jet lag
there will be at each outside corner by using the built-in cutting model. If there is room to do so,
the software automatically adds a short element slightly longer than the length of one jet lag past
the corner. This allows the controller to move the abrasivejet at full speed past the corner. The
controller then accelerates back to the corner, and continues cutting. All this motion occurs at
virtually full cutting speed, allowing for dramatically faster machining of outside corners.
Figures 1 and 2 show the same tool path calculated for 2.5 cm thick mild steel. Lighter areas
show fastest abrasivejet motion while darker areas are slower. In Figure 2, the lead-in length has
been increased for dynamic piercing, outside corners are treated differently than inside corners,
and corner passing allows for even faster jet motion.

Tests showed that for 150 different parts ranging from mechanically simple parts to complex
artworks, an average speed increase of 128.4% was achieved. One part machined slightly slower
(98% of previous speed), 14 parts machined in excess of 150% faster, and one part had a
machining speed increase exceeding 200%. Tests were done with several material and thickness
settings. The parts also come out to slightly higher tolerance. By slowing down less on outside
corners, there is less opportunity for the kerf width to grow. Figure 3 shows a 20 cm thick
aluminum rectangle machined using corner passing.
In Figure 4, notice the high quality inside corners. While the corners are not perfect, the amount
of washout has been reduced to almost nothing. The sharpness of the inside corner, therefore,
becomes primarily a function of the nozzle diameter. Corner passing increases the tolerance of
this part by preventing excess kerf width growth from slowing down around outside corners.

5 ERROR CORRECTION PHASE 3: ELIMINATING TAPER

5.1 Adaptive Cutting with a Tilting Nozzle

The third phase of error correction is the easiest to understand, but the most difficult to
implement. The nozzle needs to be tilted by an amount sufficient to remove the natural taper of
the jet stream. Implementation of this concept requires three things: ‰ Predicting the amount of
taper caused by the jet stream ‰ A mechanism for mechanically tilting the jet ‰ A five-axis
control system capable of producing the required coordinated motion. As discussed previously,
the nozzle speed must be varied along the path to achieve sharp corners and precise radii. As
shown in Figure 5, the shape of the taper produced by the jet is a function of speed. Thus, the
nozzle must tilt to different angles along the path according to the speed of motion.
5.2 Tool Tip Tilting Mechanism The mechanism for tilting the nozzle must provide quick
response to follow the angle required by the X-Y path motion. Typically, the tilting mechanism
will be carried on a large X-Y table used for cutting parts from plates of material. Quick motions
of the tilt head imply high accelerations of the X-Y mechanism, unless the tilting mechanism
pivots about the nozzle tip. A patent application has been recently filed for a mechanism closely
approximating a nozzle tip pivot and reducing the accelerations required by the X-Y mechanism.
The mechanism is easiest to understand by first considering a two-dimensional simplified
version (see Figure 6).

A nozzle is mounted on a plate, supported by two links connected to a stationary top plate. When
the plate moves sideways, the nozzle tilts while the tip remains almost stationary. For the small
angles needed to remove taper produced by the jet, the X-Y axes need to move only a few tens of
microns to compensate for the tip motion. This idea was then extended to a three-dimensional
mechanism where a third link is added and the lines defined by the three-link spherical pivots
join at the nozzle tip. The three-link mechanism has one more degree of freedom than the two-
dimensional example in Figure 6. Adding this degree of freedom also means the movable plate
can twist about a vertical axis. For the device to be useful, this motion must be controlled. The
motion is restrained by replacing one of the links with a drive shaft that has a universal joint at
each end. A photograph of a prototype of this mechanism is shown in Figure 7 where a ball has
been placed at the tool tip location for testing.

The prototype shown in Figure 7 was driven with two linear actuators that moved the tilting plate
attached to the nozzle. After some experimentation with the model, it became obvious that a
simpler and more compact mechanism could be built by actuating the yokes of the universal
joints with rotary actuators. Figure 8 shows a solid model of the final design now in production.

The software for implementing the tilt begins with the tool path corrected for the other jet errors.
The speed at each point in the path is known, so the taper angle normal to the path motion can be
calculated. A calculation is then performed to find the required positions for the two rotary
actuators that drive the universal joint yokes. For all but the smallest angles, the tilt requires a
slight correction to the positions of the X, Y, and Z actuators. This correction is added, and the
software performs the same operations at the next point on the path. At present, this calculation
is performed at 790 points per centimeter of the tool path.

5.3 Taper Modeling

Experience shows that taper is affected by process parameters such as material thickness, type of
material, size of nozzle, and size of abrasive particles. An experimental pilot study on
quantifying taper was done a few years ago. This screening experiment evaluated 11 variables:
material machinability, thickness, water pressure, orifice diameter, mixing tube diameter, mixing
tube wear condition, abrasive material, abrasive mesh size, abrasive flow rate, stand-off distance,
and quality index.
The results are illustrated in Figure 9. The main effect of each variable is showed as a single
letter, e.g., letter “D” stands for the main effect of the orifice inside diameter (ID). The
interaction effect of two variables is showed as a two-letter word, e.g., “AB” stands for the
interaction effect of machinability and thickness. In some cases, a main effect may be
confounded with an interaction effect such as “H+CK.” Since interaction effects are usually
small compared to main effects, this graph is still useful as a tool to sort out the main effects.
Those effects on or close to the straight line that passes through “0.0000, 50.00” are considered
noises and are ignored. Those far away from the straight line are important effects. The pilot
study showed that seven of the variables were most important in determining taper and therefore,
an experiment with seven variables could be implemented to develop the taper model. The seven
variables were abrasive size, orifice inside diameter, water pressure, quality index, abrasive flow
rate, workpiece thickness, and machinability. Even though the mixing tube inside diameter has a
strong effect on taper, it was not considered as a variable in this experiment because of the fixed
ratio between the inside diameter of the mixing tube and the orifice. The resulting model is a
quadratic function of main effects and interactions with a total of 28 terms. Figure 10 shows the
correlation between the model and the observed taper errors, which shows the model predictions
agree quite well with the observed values.

5.4 Experimental Results Figure 11 shows two 50 by 50 mm squares (6.4 mm thick aluminum)
machined with and without the tool tip tilting mechanism. Taper measurements were done using
a dial indicator accurate to 2.5 microns, on seven spots evenly spaced across the 50 mm length of
two opposite sides. The results are shown in Table 1.
Without tilting, the sample shows a taper error of 0.197 mm on one side and –0.012 mm on the
other. The difference in taper error on these two sides indicates a perpendicularity error between
the nozzle and the sample surface. The average taper for both sides is 0.092 mm (0.0036 inch).
Using the tool tip tilting mechanism, the taper errors on the sample are 0.007 and 0.012 mm on
the two opposite sides. The average is 0.009 mm (0.0004 inch) and the difference between the
two sides is 0.005 mm (0.0002 inch). Both the taper error and the perpendicularity error were
corrected. To demonstrate that the tool tip tilting mechanism is able to handle more sophisticated
patterns, two parts of a gear pattern were cut as shown in Figure 12.

6 CONCLUSIONS The method of “compute first, move later” allows sophisticated calculations
and adjustments for error correction at every motor step. The use of a cutting speed model
combined with the ability to vary the tool path speed one step at a time provides a convenient
way to minimize the striation marks and most jet lag errors at corners and arcs. Innovative
techniques such as corner passing and multiple-acceleration schemes enhance productivity and
quality at the same time. Finally, the development of a patent-pending tool tip tilt mechanism
and a taper model improves the quality of abrasivejet machining by eliminating taper.

A key parameter to characterize the kerf profile error generated by abrasive water-jet
Abstract
As the only cold high-energy beam machining technology, abrasive water-jet (AWJ) is one of
the most rapidly developed techniques in material manufacturing industry. However, the
application of AWJ is limited by the cutting accuracy it can achieve. Kerf profile generated by
AWJ is different as the cutting parameters change. As a result, it has become a major factor
which affects the cutting accuracy when AWJ is used as a machining tool. Researchers used
taper error to characterize kerf profile error generated by AWJ in the past years. And many
efforts have been put on how to eliminate taper error by using a tilting cutting head of a 5-axis
AWJ machine. However, using taper error to characterize the kerf profile error generated by
AWJ is not accurate since kerf profile error might appear in different styles. And using a 5-axis
AWJ machine to eliminate taper error is only effective in some special cases. To effectively
eliminate taper error, the first thing needs to do is to find out whether the kerf profile error can be
compensated or not. Based on research, a key parameter, named kerf profile coefficient Ω, which
can be used to characterize kerf profile error and further to guide people to use different ways to
compensate kerf profile error, has been defined in this paper. To further illustrate the efficiency
of this coefficient, a series of cutting experiments have been carried out and the experimental
results have been discussed.
Keywords Abrasive water-jet · Precision cutting · Kerf profile error · Kerf profile coefficient.
1 Introduction Abrasive water-jet (AWJ) is one of the most recently developed manufacturing
processes. In this process, clean water is pressurized to a very high pressure, which is as high as
420 MPa. The high-pressure water is then forced to come out from a very small nozzle. After it
comes out from that nozzle, a high-speed water-jet beam is formed. This high-speed water-jet
beam could be used to cut some soft materials, such as paperboard, sponge, etc. If abrasive
particles are added into that water-jet beam, a high-speed abrasive jet beam is formed. This high-
speed abrasive jet beam could be used to cut all kinds of materials. As a very promising
manufacturing method, AWJ has been used extensively in industry currently. Now, parts with
tolerance less than 0.1 mm can be cut by AWJ directly. However, the further application of this
technology has been limited by its cutting accuracy.

Fig. 1 Kerf profiles in AWJ cutting [1]

Unlike those traditional manufacturing methods, in which the kerf profile is a fixed one which
matches to the shape of the tool, AWJ’s kerf profile changes when any cutting parameter
changes. Many researchers put lots of effort to investigate AWJ’s kerf profile. As early as 1990,
Matsui et al. demonstrated that the kerf profile is a function of cutting speed [1]. According to
Matsui et al., the kerf profile curvature changes from convex to concave as the cutting speed
decreases, the results he got are shown in Fig. 1. In 2009, Zeng et al. reported that the kerf profile
is convex in thin workpiece and concave in thick workpiece [2]. As a result, AWJ’s kerf profiles
might appear in different styles. Some of them appear as a positive trapezoid on cross section,
and some of them appear as an inverted trapezoid. Others appear as a barrel on cross section,
etc., as shown in Fig. 2 [3]. As any of the cutting parameters change, the kerf profile might
change in appearance. Therefore, it’s a big challenge to get accurate kerf profile by using AWJ
as a cutting tool.
To get precision cutting, some other researchers used taper error to characterize kerf profile
generated by AWJ. Taper error was defined as half of the difference between the top and bottom
kerf width. Taper angle was defined as the arctangent of taper error over thickness. In 2007,
Hashish reported that taper increases with the increasing of the cutting speed [4]. In 2008,
Maccarini et al. revealed that the taper increases with the increasing of the hardness of the
workpiece [5]. Furthermore, many researchers have investigated the taper of different materials
machined by AWJ and revealed the relations between taper and cutting parameters, such as
cutting speed, water pressure, etc. [6–9]. Except that, some researchers tried to use mathematical
models to characterize taper error. In 1992, Chung et al. derived a linear regression equation
which showed that the taper is related to mixing tube diameter, abrasive flow rate, stand-off
distance, etc. [10]. Also in 1998, Groppetti et al. assumed that the input energy is dissipated
along the thickness, and then he derived a mathematical model of taper [11]. In 2000, Annoni et
al. derived a relatively simple taper model based on the analysis of a multiple linear regression
[12]. After knowing the factors which affect taper, a method, which could eliminate taper error,
has been proposed and applied in AWJ cutting process [4, 13]. In this method, a cutting head was
tilted a small angle along the direction, which is perpendicular to the cutting head moving
direction, to compensate taper error, as shown in Fig. 3. Since then, this method has been used
extensively in industry to eliminate taper error. However, not each taper error can be eliminated
by using the above method because the kerf profile might present in different styles under
different cutting conditions.
CONTROLLING TAPER
Types of Taper Inherent to Waterjet Machining

In waterjet machining, taper refers to a difference in the kerf width at the top of a cut and the kerf
width at the bottom of a cut. A typical waterjet cut has a slight taper, the most common type
being V-shaped taper. This type of taper is generated because the jet stream loses some of its
cutting energy as it cuts deeper into the material. Slightly more material is removed at the top of
a cut, where the jet stream enters the material, than at the bottom of the cut, where the jet stream
exits. V-shaped taper is usually associated with fast cutting. In general, the greater the nozzle
speed, the more pronounced the taper will be. Slow cutting speeds can produce reverse
taper where the kerf width is wider at the bottom of the cut than at the top. This is the result of
the jet stream removing more material at the bottom of the cut than at the top of the cut. Reverse
taper can also occur when cutting very soft materials. A third type of taper is barrel taper,
where the kerf width is widest in the middle of the cut. This type of taper can occur when cutting
very thick materials. There are other, less common types of taper that could be produced, but
those are caused by calibration or set-up errors rather than the natural physics of water.
V-shaped taper is the most common type of taper and is caused by the fact that the jet has more
cutting energy at the top of the cut than at the bottom and so it tends to make a wider cut there. In
general, the faster the cut the more pronounced the taper will be. The reverse taper happens
during very slow cuts or in soft materials where the material is rapidly eroded. Because the jet
stream expands the farther away from the nozzle it gets, it removes more material from the
bottom than from the top. The barrel taper on the other hand usually only occurs in very thick
materials. The taper is never worth worrying about when working on parts 1/8” (3 mm) thick or
below as the taper is barely noticeable. There are methods however that can be used to reduce the
amount of taper on larger parts. The primary method as mentioned is through controlling the
cutting speed and is usually calculated by the cutting program after the material data is inserted.
Another method is through the use of a tilting cutter head to compensate for the spread. However
this requires an upgrade in cutting eqipment to be able to follw a curve and tilt the head
accordingly to maintain a taper free edge and thus cost will more.

Why Taper Matters For many cutting jobs, reducing the small amount of taper inherent in
waterjet cutting is not necessary. You can even use taper to your advantage in certain instances,
for example alternating the top side of two parts so the taper in one cancels out the taper in the
other for a very tight fit. But for high precision jobs, or when cutting thick parts that must
interface with other pieces along the machined edge, eliminating taper is critical. Slowing down
machining speeds can usually reduce V-shaped taper, but won’t always reduce it enough to meet
specified tolerances. And slow cutting speeds are not optimal for making complete parts with the
highest possible accuracy in the least amount of time.

How to Eliminate Taper from PartsAn articulated cutting head, such as the Tilt-A-Jet, can tilt the
cutting nozzle in a direction perpendicular to the direction of cut to offset the naturally occurring
taper and produce parts with taper-free edges. The taper is moved to the scrap portion of the
material. With a tilt-only cutting head and predictive software, you can eliminate taper from parts
without slowing down the cutting speed. Predictive software uses an advanced cutting model to
calculate the type and amount of taper that will occur at each entity along a tool path. The
software calculation takes into account the material type, thickness and desired edge quality.
During the cutting process, the software rapidly adjusts the nozzle position so the angle of the jet
stream offsets the natural taper.

Methods to Reduce Taper without a Tilting Head

 To reduce V-shaped taper, slow down the cutting speed.

 To reduce reverse taper, increase the cutting speed.
 Stack thin materials. Taper is usually most pronounced in materials less than 1/8” (3 mm)
thick. If available, use the stack height calculator in the controller software to determine
the optimum number of sheets to stack to make the most parts in the least amount of time.
 Use a low stand-off. The closer the nozzle is to the material, the less the jet stream will
spread and the less taper it will produce.
 Use the best quality abrasive available. Lower quality abrasives will tend to have less
consistent particle sizes. Different particle sizes will increase taper.
 Ensure the Z-axis is perpendicular to the material in both X and Y directions. If the Z-
axis is at a slight angle, it can produce a rhombus taper.