Computer-Aided Design 45 (2013) 796–808

Contents lists available at SciVerse ScienceDirect

Computer-Aided Design
journal homepage: www.elsevier.com/locate/cad

Review
5-axis flank milling: A state-of-the-art review
Ramy F. Harik a,∗ , Hu Gong b , Alain Bernard c
a
Department of Industrial and Mechanical Engineering, School of Engineering, Lebanese American University, 36, Lebanon
b
State Key Laboratory of Precision Measuring Technology & Instruments, Centre of MicroNano Manufacturing Technology, Tianjin University, 300072, PR China
c
IRCCyN, UMR CNRS 6597, Ecole Centrale de Nantes, France

article info abstract

Keywords: Flank milling is of importance to machining aircraft structural parts, turbines, blades and several other
Flank milling
mechanical parts. It decreases manufacturing time, enhances quality and reduces cost. Since flank milling
5-axis
developable ruled surfaces do not contain geometrical errors, research on flank milling focuses on
Literature review
the generation of optimal tool trajectory for non-developable ruled surfaces, even generic free-form
surfaces. This includes: envelope surfaces, geometrical errors (overcut, undercut), energy optimization
in tool movement, surface deviations, tool geometry adaptation, tool wear and temperature, and surface
roughness. In this article we present a survey on flank milling as well as suggesting guidelines for future
considerations in solving flank milling tool trajectory optimization.
© 2012 Elsevier Ltd. All rights reserved.

Contents

1. Introduction........................................................................................................................................................................................................................796
2. Preamble .............................................................................................................................................................................................................................797
2.1. Manufacturing strategies ......................................................................................................................................................................................797
2.2. Ruled surfaces ........................................................................................................................................................................................................797
2.3. Definitions ..............................................................................................................................................................................................................798
3. Tool trajectory optimization..............................................................................................................................................................................................798
3.1. Chronological review of tool trajectory optimization related to 5-axis flank milling ......................................................................................799
3.2. Summary map for toolpath generation ................................................................................................................................................................802
4. Manufacturing tool adaptation .........................................................................................................................................................................................803
4.1. Chronological review.............................................................................................................................................................................................803
4.2. Further considerations ..........................................................................................................................................................................................804
5. Other related works ...........................................................................................................................................................................................................804
5.1. Tool flank wear and temperatures........................................................................................................................................................................804
5.2. Sweep milling ........................................................................................................................................................................................................804
5.3. Further considerations ..........................................................................................................................................................................................805
6. Summary ............................................................................................................................................................................................................................805
6.1. Conclusions ............................................................................................................................................................................................................805
6.2. Test surfaces ...........................................................................................................................................................................................................806
Acknowledgment ...............................................................................................................................................................................................................806
References...........................................................................................................................................................................................................................806

1. Introduction manufacturing time, quality enhancement and cost reduction. It

is mainly applicable to ruled surfaces [3], a particular geometrical
Flank milling is of importance to machine complex surfaces type of surface obtained by the motion of a rule on guiding rails.
found mainly in aircraft structural parts, turbines, and blades [1]. It should also be developable [4], where surface normals along the
Tönshoff et al. [2] highlights the importance of flank milling same rule are collinear, even though few attempts have been made
to override the constraints [5]. This survey of the literature re-
calls the advancement in the field of flank milling research as well
∗
Corresponding author. Tel.: +961 70183819.
as laying the foundation for future work on solving flank milling
E-mail addresses: ramy.harik@gmail.com (R.F. Harik), gonghu2012@gmail.com tool trajectory optimization, or at least defining the important pa-
(H. Gong), Alain.Bernard@irccyn.ec-nantes.fr (A. Bernard). rameters affecting the quality of the finished surface. Research on
0010-4485/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cad.2012.08.004
 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808 797
五轴侧铣主要分为两大块：最优刀具轨迹和刀具几何形状
flank milling can be split into two main categories: Optimal tool
trajectories and tool geometry.
The article first presents a Preamble, then in Section 2.1, man-
ufacturing strategies as well as an introduction to the mathemat-
ics of ruled surfaces. The preamble also includes some definitions
that might be useful in consolidating the terms used throughout
the various reviewed publications. Following on from this, we at-
tempt to review the literature with respect to the earlier defined
two main categories (see Fig. 1). Section 3 investigates category 1:
最优刀具轨迹 the generation of the optimal tool trajectory through error reduc-
分为以下四个部分
tion. This can be divided into the following:
计算误差 – Calculation of errors, the ability to rectify the tool path of
twisted surfaces to obtain a machined surface faithful to the
最小化过切、 intended design.
欠切 – Minimization of pure geometrical errors such as overcut (over-
indentation of the tool in the surface) and undercut (falling
short of removing all excess material).
识别包络面(加工面)
– Identification of the envelope surface, sometimes referred to as
machining surface, serving as tool trajectory guides. It is often
Fig. 1. Flank milling literature review and article sections.
seen as a transformation of the original surface along the local
normals.
预测表面粗糙度
– Prediction of surface roughness and comparison of the toler-
ance with acceptable design limits.
Section 4 explores category 2: tool adaptation and suitability for
the manufacturing mode at hand. This branch can mainly be split
into three different problems:
– Study of surface deviations due to tool deflections. In fact,
刀具挠度引起表面偏差
tools are subject to higher flank forces leading to potentially
considerable tool deflection affecting the final quality of the
machined surface. An algorithm not accounting for potential
tool deflection will present sizable differences between the
simulation and the real values. The difference is proportional
to the rule length.
使用不同形状刀具– Usage of different tool geometries to overcome the manufac-
turing problem. Traditionally a cylindrical flat tool is used, but
recently conical tools as well as different geometries have been
used.
设计刀具形状 – Regarding tool geometry, some have considered adapting the
Fig. 2. Manufacturing program subdivision.
tool geometry, as in manufacturing special tools to obtain preset
manufacturing surfaces.
program (Fig. 2). It requires the identification of the optimal
刀具碰撞 – Global tool collision with the surrounding environment—other
than the surface under investigation. machinery, fixture selection, tooling (Fig. 4) and manufacturing
– Minimization of tool energy expenditures by carefully ob- modes. The latter requires the identification of the tool positions
serving that algorithms do not propose sudden tool direction and orientations with regard to the machined surface.
changes to suit geometrical error optimization. The manufacturing mode of interest in our analysis is flank
milling. Manufacturing modes (Fig. 3) can be classified into:
Section 5 explores additional research relevant to the topic:
– Assessing tool flank wear, as well as the analysis of tool temper- – Flank milling (sometimes referred to as side or peripheral
atures in flank milling and its effect on surface quality. milling): chips formed along the flank side of the tool.
扫铣：侧铣的替代– Sweep milling—the alternative when flank milling is not appli- – End milling: particular sweep milling where the contact point
cable. The latter investigation exposes new ideas that can be is the midst of the tool flat end.
translated and adapted within flank milling.
The article ends with a summary of the literature review findings 2.2. Ruled surfaces
and with a discussion of future directions to undertake in resolving
flank milling tool trajectory computation accounting for realistic Ruled surfaces are the results of the movement of a line along
flank force deflections.
guiding curves. This section presents the mathematics of ruled
2. Preamble surfaces required for the following paragraphs. For a detailed
insight on ruled surfaces, readers can refer to [6,7,3].
In this section we introduce the topic parameters and restric- A ruled surface is generated by joining corresponding points
tions, develop an understanding of the nature of the problem, as on two guiding rails – S (u) and T (u) – linearly. The parametric
well as its applicability limitation. We also propose a set of def- equation of the ruled surface given in Fig. 5 is:
initions that would serve to standardize the terms used over the
diverse reviewed publications. P (u, v) = (1 − v)S (u) + v T (u) u ∈ [0; 1] & v ∈ [0; 1] . (1)

The study of the surface normals along the rules identifies whether
2.1. Manufacturing strategies the surface is also developable. Normals are computed as follows:
Using computer numerically controlled machinery (CNC) to ∂P ∂P
manufacture a part requires the setup of the manufacturing N (u, v) = × . (2)
∂u ∂v
798 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808

Fig. 3. (From left to right) End milling, sweep milling and flank milling; red areas
depict cutter contact locations with surface. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.) Fig. 7. Designed, envelope and constraint surfaces.

Fig. 8. Undercut and overcut (amplified for representation). (For interpretation

of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

2.3. Definitions

Fig. 4. (a) Ball nose, (b) flat-end mill cylindrical and (c) conical tools. In this section we present and pictorially define parameters of
interest: designed surface, tool path, envelope surface, constraint
surface, machined surface, local overcut, global overcut, and
undercut. The above parameters are linked each other and most
of the time the computation of one requires the identification of
another. We refer to billet as the original volume of material. The
term billet is similar to stock, workpart and workpiece. We refer
to part as the finished product. A manufacturing program (Fig. 2)
is applied to obtain a part out of a billet. We refer to ‘designed
surface’ (Fig. 7) as the surface undergoing process planning. It is
the surface that we wish to compute the optimal tool path for
its manufacturing. The ‘constraint surfaces’ are other surfaces that
prevent the tool trajectory from obtaining the ‘designed surface’.
The ‘envelope surface’ is the boundary surface formed by the
Fig. 5. A ruled surface. tool path trajectory. It results from the tool sweeping. The ‘ma-
chined surface’ is the actually obtained surface post-manufacture.
The differences between the computed surface and the machined
surface are related to many factors, such as tool deflection, tool
wear, kinematics, feed drives, errors associated to the non-ideal
machine tool and so on. The aim is to generate a machined surface
as close as possible to the designed surface.
For clarity we will refer to the difference α as the difference
between the designed surface and the envelope surface; and to
the difference β as the difference between the envelope surface
and the machined surface. Local overcut is the amount of material
taken out – that should not have been removed – at a specific tool
position. The global overcut (red on Fig. 8) is the union of local
Fig. 6. A developable ruled surface with collinear normals along the same rule. overcuts across the toolpath. The term undercut (Fig. 8) represents
the excess material remaining after the application of the
A developable ruled surface has its normals collinear along the manufacturing process. This excess can only be computed globally
same rule (Fig. 6). The developability of a surface is indicated and should be finished using a different process at a second stage.
through the parameter ‘twisted’. When literature refers to ‘twisted
ruled surface’ it means a non developable ruled surface. The twista- 3. Tool trajectory optimization
bility degree is the angular difference between surface normals at
both ends of a single rule. It is worthwhile mentioning that planar, Tool trajectory optimization attempts to compensate the errors
cylindrical and conical surfaces are particular developable ruled introduced during 5-axis machining (undercut and overcut).
surfaces. More generally, the envelope surface of a family of planes Bohez [8] presents an extensive literature review for error types
is also a developable ruled surface. and their compensation. The literature review has its roots in [9],
 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808 799

Table 1
Comparison between end and flank milling.
Mode Advantages Disadvantages

End – Suitable for large sculpture surfaces – Lesser amount of manufacturing

– Uses flat end cutter
Flank – Uses cylindrical cutters – Difficult gouge avoidance 难以避免过切
– Suitable for small and medium surfaces – Complex tool interference 复杂的刀具干涉
– More tool strength

blades can be readily approximated by ruled surfaces, while fan

and axial compressor blades are not, owing to the generally more
severe twist and many strict geometric complexities. [24] demon-
strates the initial approach taken by [25] at Pratt & Whitney Canada
to flank mill axial compressor blade. By imposing the three de-
sign curves on each surface to lie on a highly twisted but nev-
ertheless ruled surface, and then reducing the deviation between
the ruled surface and the machined surface by introducing 2- pass
flank milling, they were able to flank mill two simple integrally
bladed axial rotors. The next phase of major development in flank
Fig. 9. SPO method of [21]. milling at Pratt & Whitney Canada was reported in [26], target-
ing highly complex arbitrary surface blade designs. Complex air-
where errors in modeling and machining of centrifugal impellers foil surfaces would requires a lot more than 3 design curves – typ-
are studied. The authors state that literature (prior to 2000) ically 5 to 25 curves – and one cannot approximate such blade
neglected modeling errors (α ) and restricted itself to measurement surfaces with ruled ones. Thus ruled surface, guiding rails, under-
and compensation (β ). It served in studying 5-axis machining tool cut, overcut, etc. are no longer considered. Instead, cutter paths
kinematic chain design and analysis [10,11]. Also, tool collision and were generated to match the design curves in a weighted man-
interference [12,13] is scarce in the field of 5-axis flank milling. ner. What is important is not the closest geometrical match, but
Several references can be found for the problem in end/sweep the virtually identical match for aerodynamics and structural per-
milling [14,15], but only a couple of Refs. [16–18] are found dealing formance requirements. Generally, the resultant flank mill surface
with constraints while attempting to flank mill ruled surfaces. cannot be approximated by a ruled surface, since the contact curve
The constraints are not internal (i.e., existence of a boss in the between the conical cutter and the blade can deviate very signif-
midst of a surface); they are, rather, boundary limitations. In [18] icantly from a straight line even in a single pass. That is why the
the objective is to improve the kinematic behavior of machine system has been named the ‘‘Arbitrary Surface Flank Milling’’ sys-
tools during milling. In the following, we will first attempt to tem. More recently, in [27], Wu outlines a technique to transform a
chronologically review flank milling literature dealing with tool design blade into an equivalent performance but more flank mill-
trajectory optimization. The review will present the main concept able airfoil. This significantly expands the domain of applicability
and results. For detailed application of the algorithm, the reader of flank milling. Another important achievement is the capability
is encouraged to consult the reference. We will then present a to flank mill with a shaped cutter, i.e., the cutter is a general body
summary map dynamically linking the studied research. of revolution defined according to the design intent airfoil geom-
etry. Based on the same concept and software methodology, they
3.1. Chronological review of tool trajectory optimization related to 5- have also successfully demonstrated the superior hardware qual-
axis flank milling ity and cost effectiveness of Flank Super Abrasive Machining (Flank
SAM) of blades in nickel-based alloy, a very difficult-to- mill ma-
One of the first publications dedicated to error computation in terial. Elber and Fish [28] studies the ability to flank mill surfaces
flank milling [19] (article in German) describes the economic im- by approximating any general surface into a ruled one. The target
portance of flank milling and discusses methods of reducing sur- is the machining of saddle-like (hyperbolic) and convex surfaces
face discrepancies without providing a detailed error computation rather than minimizing errors. The authors do not show an inter-
algorithm. The topic seems to have been attempted in German in- est in exploring time optimization. Attempting to simultaneously
dustries as [20] presents an initial theoretical model to compute remove higher volumes of material is not investigated. Redonnet
flank milling based on another German reference. The first article et al. [29] proposes to ‘share’ the error between the two optimal
to detail flank milling error computation is [21]. It offers a com- tool positions of non-developable ruled surfaces. The optimal po-
parison between end [22] and flank milling detailed in Table 1. The sitions were previously defined as placing the tool tangent at the
main contribution is in the algorithms SPO (single point offset) and guiding rails referencing [30]. This proposition is based on adding
DPO (double point offset). They calculated the cutter location data a third tangency condition: Tangency of the tool lower generating
(referred to as CLDATA) using vectors as presented in [23]. line to the rule (besides the tangency to the two guiding rails on
Algorithms are described analytically and experimented with both sides of the extreme points of the rule). A proposed method
on a test surface. In SPO (Fig. 9), the tool is identified by the position was applied on Table 10. A quantitative analysis comparing
of tool center (O) and tool direction (Ta ). Ta is collinear to the [21–23,26,28,29], gave an error of 0.22 mm using the same tool
surface tangent b at P, while O is vectorially computed. Similarly dimensions (whereas it was 0.585 mm in [21]). Abdel-Malek and
the DPO method is applied on the same test surface. Two points Yeh [31] shows a methodology to compute tool swept volume that
are selected along the rules (in a uniform subdivision of the rule). can be used to approximate errors. Leu et al. [32] argues that previ-
These points are offset at a distance equal to the tool radius along ous methodologies are not suitable for NC real time data machining
the surface normal. The vector joining these two points forms the verification and presents, as an alternative, the sweep-envelope
proposed orientation of the tool. differential equation.
Up to 1982, application of flank milling in the jet engine in- Rubio et al. [30] developed an overall 5-axis machining program
dustry had been limited to centrifugal impellers because impeller for free-form surface. Of the several proposed algorithms, flank
离心叶轮
800 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808

Table 2
Comparison of results between [21,49] (units are in mm).
[21] [49]

Maximum undercut 0.582 0.2644

Maximum overcut 0.585 0.2114
Maximum angle change – 0.6426

directed at manufacturing a ruled surface, but rather on describing

a tool path benefiting from ruled surfaces curvature properties. The
real-time control approach uses high-order motion properties and
is suitable for high-precision machining of free-formed surfaces.
Tournier et al. [42] and Duc et al. [43] puts forward the concept
of the machining surface. In this paper, a first machining surface
is calculated representing the tool trajectory (set of tool axes
in flank milling). This machining surface is described as a ruled
Fig. 10. Positioning algorithm of [33]. surface, the guiding curves of which correspond to two B-spline
curves. The envelope surface of the tool movement during milling
is approximated, following a kinematical approach, with the aim 移动b样条曲线
of calculating geometrical errors. Therefore, the control points (代表加工曲 面)的控制点，
of the B-spline curves (standing for the machining surface) are 使几何偏差最
小化
moved so that the geometrical deviations are minimized. The
concept was enlarged in [44] by the introduction of energy
optimization to preserve the smoothness of the trajectory.
This work is completed in [45]. The machining surface concept
was also presented in [42]. It was developed using end milling
with several tool geometries. It was not implemented for flank
milling, but the proposition is highly valuable and can be applied
to flank milling. Lartigue et al. [46] introduced an assessment and
Fig. 11. Maximum errors of different algorithms according to [34]. correction methodology for the tool path while performing a 5-axis
flank milling operation. At first the tool path is simulated using
milling trajectories were investigated. The tool is positioned the envelope surface. A kinematic approach is used and presented
collinear to the rule considered. Two positions are investigated: for the particular case of flank milling. The appendix provides a
placing the tool, respectively, tangent to each guiding rail. detailed calculation that leads to a set of curves, later interpolated
Error calculations generate a three-equation system with four to obtain the envelope surface. At a second stage, geometrical
unknowns; a tolerance is then selected to resolve the system. deviations between the cutting tool and the designed surface to
An example is carried on a test figure (surface data not given). be machined are calculated. Following this, a correction of the
Monies et al. [33–35], Redonnet et al. [29], Senatore et al. [5,36,37] tool path is proposed by displacing points of the envelope surface
are considered as a continuum to the effort initiated by Rubio accounting for the error in geometrical deviations. The proposition
et al. [30]. Monies et al. [33] suggests an improved positioning is verified on a sample ruled surface whose parameters are not
tending towards achieving two goals: interference and process given. The error evaluation is measured to read up to 0.04 mm
time reduction. The tool positioning is based on having one point of and, when corrected, they reached less than 0.02 mm. This method
contact (M2 in Fig. 10) and a two points tangency (M0 and M1 ). The did not mention how to generate an effective initial toolpath (see
milling cutter is rotated around y2 . The algorithm generates seven Fig. 12).
equations with seven unknowns (the positioning parameters) and Bedi et al. [47] opens a series of tool trajectory planning
is solved based on a Newton method. Errors are computed on initiatives [48,49]. The strategy places the tool tangential to the top
surface, defined in Table 10, and are compared to [30]. Results show and bottom curves at equal values of u (the non-ruled direction).
drastic improvements of errors, mainly as the minimum radius The methodology consists of building the Frenet–Serret frame
at the end of the conical milling cutter increases. In [34] the at the rule extremities. Then the tool axis is found through the
authors compare the algorithms of [19,21,30–33], while presenting identification of two points on the (T , N ) Frenet plane. The points
in depth analysis for all of the respective algorithms (see Fig. 11). are an intersection result between the tangential positioning of
Lin and Koren [38] developed a feasibility study for an 5-axis the tool at the guiding curves and the (N , B) planes respectively.
machining interpolator, assuming parts are gouge-free. Further to this proposition, and through a series of geometric
Tsay and Her [39] establishes an analytical model to reduce algebraic computations on vectors, the authors obtain a system of
discrepancies between the designed and machined twisted ruled four equations with four unknowns. The latter is solved through
surfaces. The author uses the term twist to identify the more numerical computations. The error deviation is then approximated
recognized term of ‘non-developable’. The error is computed at the mid-curve of the ruled surfaces (see Table 11). Menzel
based on a series of projections and geometrical analysis. The et al. [49] presents a flank milling positioning methodology that
author concludes that undercutting is automatically related to the is a modification of [47]. The author claims 88% less under-cutting
‘twisting angle’, that is, the degree of non-developability. Several then in [47]. The methodology consists of a 3-step optimization:
numerical examples are given studying the variation of several Initialization of the position, making the tool tangent to a rule line,
parameters, as well as a test case. A comparison of average errors and finally making the tool tangential to two guiding rails (that
is computed and presented. The methodology is used as a base, are not the top/bottom curves) and one rule line. The algorithm
where the authors present in [40] a flank cutting technology efficiency was compared on the surface (Table 9) given in [21].
that can be applied to generate 5-axis cutter paths for machining Results of this comparison are given in Table 2.
centrifugal compressor impellers. Although [41] uses the ruled Chiou [50] suggests using the cutter swept envelope to compute
surfaces theory to generate a toolpath, the application is not the remaining machined part geometry. The swept envelope is
 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808 801

(a) Positioning errors. (b) Optimized ruled surface.

Fig. 12. Original tool path errors on left for study given in [46], optimized ruled surface on right after application of proposition.

Table 3
Comparison of results shown in [48] for surface in Table 12 (units are in mm). Tools axis
Method → Radial Parametric Motion trajectory surface
Error ↓ (u)

0.2 −0.03 0.015 −0.025 0.04 −0.03 0.012

0.5 −0.041 0.04 −0.06 0.1 −0.03 0.015
0.9 −0.05 0.05 −0.12 0.17 −0.05 0.05

Undercut
Table 4
Machining deviations relevant to Fig. 13(a) (units are in mm). Overcut
Designed surface
Maximum deviation Average deviation Standard deviation

Undercut 0.02084 0.00399 0.00329

Overcut 0.01932 0.00357 0.00286 Fig. 13. Distribution of manufacturing errors for [53] (a) with TPO before
optimization (b) with TPO after optimization.

considered as the totality of the points that belongs to the trace 0.60
of the cutter. A methodology is proposed where the explicit swept 0.50
0.40
profile of a taper-end cutter is computed. The author applies the 0.30
proposition on a ruled surface (without referring to its equation). 0.20
0.10 0.5-0.6
Moreover, no comparison between other methodologies with 0.00 0.4-0.5
respect to error enhancement is provided. Weinert et al. [51] and -0.10 0.3-0.4
1 0.2-0.3
Lee and Nestler [52] also proposed other swept volume generation 11 0.1-0.2
methodologies. 21
0-0.1
-0.1-0
Li et al. [48] studied three flank milling error measurements
31
methods: The radial method [9,19,47,21], the parametric method Curve Parameter Value %
and the closest point [33,34]. The authors concluded that the radial 41
v
S1
method seems poor – even though it is simple to compute – as
it is not really calculating the error between the point on the Fig. 14. Distribution of the approximation error applied on a test example of [54].
ruled surface and the point on the machined surface and that the
parametric method will fail to give an accurate measurement of
legitimizes the previous geometrical studies, mainly [29]. The
the shortest distance between the ruled surface and the machined authors further stress that [46] did not introduce the geometrical
surface. Following on from this, two methods are proposed: positioning used and that the surface treated is not twisted.
the tangent plane method and the motion method. In the first Nonetheless, the article proposes a study of the error between the
method (Tangent Plane), the tangent plane to the grazing surface envelope and theoretical surfaces. An example with high values
is intersected with a line normal to the ruled surface. In the second (larger cutter diameter and high twistability) showed accordance
method (Motion), the shorter distance between a ruled point and with the kinematic study (slight difference between real and
a line constructed from the point’s corresponding grazing point in calculated angles of tool rotation). Chu and Chen [54] presented
the direction motion is computed. Examples are carried on two a flank milling manufacturing approach where a general ruled
test surfaces (points given in Tables 12 and 13). Results conclude surface is approximated with consecutive developable patches
that the proposed methodologies are simple and robust with better eliminating tool interference. The algorithm solves through a
results (Table 3). geometric computation while guaranteeing G1 continuity for tool
Gong et al. [53] uses a cylindrical tool to develop a positioning path smoothness optimization. Fig. 14 shows the application test
strategy to minimize error between the intended designed surface error results. The article does not present a smoothness parameter
and the machined one. It is proved that errors are linked to or formulation.
the deviation between the tool axis trajectory surface and the Senatore et al. [36] covered the analysis of the rotation axis
offset surface of the designed surface. The initial tool trajectory influence on the generated errors. Besides the significant error
is calculated using a proposed three-points offset method. Then, enhancement of 10% upon optimization of the rotating axis, the
a least square method was used to fit the offset surfaces of article also offers a detailed identification of the error/radius error
the designed surface using tool trajectory surface. Results of the relations. In [37], a study to provide assistance in choosing the
propositions are found in Fig. 13 and errors are reported in Tables 4 radius for milling cutters was proposed. The maximum errors
and 5. (overcut–undercut) and the maximum radius are determined from
Senatore et al. [5] criticizes efforts to deduce the optimal an estimation of maximum error in relation to a tolerance interval.
positioning using the envelope surface (in reference to [46]) and Gong et al. [55] presents a new methodology named BCELTP (Basic
 802 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808

a b Table 7
Comparison of results on surface in [49] (units are in mm).
D = 20 Tool envelope surface [47] [49] [44]

Undercut 0.2876 0.0061 0.0086

Overcut 0 0.0091 0.016
24 Designed surface
40 P
Table 8
Comparison of results on surface in [5] (units are in mm).
11.31° [21,29] [53,5] [49]

Undercut 0 0 0.007
Overcut 1.5 0.0165 0.07

Fig. 15. Flank milling free form surfaces with conical cutter.
Pechard et al. [44] introduces the concept of energy optimization
while generating the optimal toolpath. The target is to find
Table 5 a pseudo solution that gives minimal geometrical deviations
Machining deviations relevant to Fig. 13(b) (units are in mm).
while preserving a correct smoothness of the trajectory. The
Maximum deviation Average deviation Standard deviation method, labeled Geo5XF, is applied on the surfaces of [49,5,21],
Undercut 0.01770 0.00252 0.00255 as well as on an industrial ‘‘Impeller’’ surface. The results were
Overcut 0.01849 0.00213 0.00253 compared in context (Tables 7 and 8). Zhu et al. [60] proposes
a sphere congruence model to analytically compute the swept
envelope surfaces of general rotary tools. With this model, an
Table 6
elegant approach was presented to efficiently compute the signed
Comparison of results shown in [58] for surface in Table 9 (units are in mm).
distance between a point in space and the swept surface without
[21] [29] [49] [53] [58] constructing the swept surface itself. The first-order differential
Undercut 0.582 0.220 0.264 0.093 0.068 increment of the signed point-to-surface distance with respect
Overcut 0.585 0.220 0.211 0.119 0.067 to the differential deformation of the tool axis trajectory surface
was derived. It characterizes quantitatively the change of the
geometric error of the machined surface under the change of
Curvature Equation of Locally Tool Positioning). The method relies
the tool trajectory. The distance function theory is innovative
on local optimization of tool position and unify the calculation of and exciting. It provides a natural way to extend the approach
tool positions into a group equations for all kinds of cutting tools, presented in [58] to deal with generic rotary cutters [61,62]. Zheng
which can handle both end milling and flank milling. This local et al. [63] recently used multi-objective programming: attempting
applicability enables the user to adjust tool positions individually to optimize simultaneously trajectory smoothness and machining
until the relative normal curvature between the envelope surface errors. The tool path that yields the minimal geometric deviation
and the designed surface is minimized. The algorithm is applied was used to check the existence of the feasible solutions to the
and verified on several examples; Fig. 15 shows one of those optimization problem. The proposed method was developed for a
examples relevant to flank milling. Sprott and Ravani [56] uses the chosen cutter with a fixed type and size without consideration to
geometric presentation of a surface to derive an offset method that global interference.
intersects all of the normals along a ruling. The obtained line is
constructed using the generator trihedron. The cutter is cylindrical. 3.2. Summary map for toolpath generation
Test surface data are not given but rather a geometric figure.
假设全局 The review of tool trajectory optimization contains a handful
优化等于 Wu et al. [57] uses dynamic programming to solve the toolpath
局部优化 problem. Previous methodologies were classified as good for of publications. To facilitate the reviewed concepts, Fig. 16 depicts
并不是 the main publications:
正确的 the local optimums, and that the assumption where the global
optimum equals the sum of local optimums is not generally true. – Early Works (1979–2001): Surveyed by Monies et al. [34]
The attempt is to transform the geometrical program of tool path in 2001, it includes preliminary simple positioning by Stute
generation into a mathematical programming problem. A Z -buffer et al. [19], the introduction of SPO & DPO methods by Liu [21],
is used for a quick estimation of the machining error. Ding and the investigation of dual positions by Rubio et al. [30], the
Zhu [58] proposes a global optimization methodology that is novel enhancement of the latter by Redonnet et al. [29] as well as by
in terms of the perspective of approximating the tool envelope Monies et al. [33].
surface to the data points on the designed surface following the – Optimization Work (2001–2011): Two main publications high-
minimum zone criterion recommended by ANSI and ISO standards light this part: Lartigue et al. [46] and Bedi et al. [47]. The first
for tolerance evaluation. The interchangeability principle was introduced the concept of envelope surfaces on which [5,36,37]
developed to optimize the tool path. The method was compared built different enhancements: Error computation between en-
velope and theoretical surfaces, optimization of rotating axis
with the main methodologies to date. Results are shown in
and linking the cutter radius to error. The second proposed tan-
Table 6.
gential positioning and was enhanced by Menzel et al. [49] and
Gong and Wang [4] presents an extension of [53] based on the
Li and Jerard [22]. Other optimization techniques were intro-
proposed analytical calculation method of envelope surfaces using
duced by Gong et al. [53]. The latter was extended in [4], and by
a moving frame [59]. It provides a new definition for the error Ding and Zhu [58].
computation, where it uses the latter to establish an optimization – Novel Concepts (2008–2011): Recently, more novel concepts
model to obtain the global optimized tool axis trajectory surface. have been studied to manufacture non-developable surfaces
The method is applied on the Archimedes’ helicoid test surface outside of the regular optimization track. Wu et al. [57]
of [5] (Table 14). The authors provided supplementary material introduced an attempt to use dynamic programming to solve
to verify the algorithm in different CAD/CAM systems. They the problem, Pechard et al. [44] energy minimization, Zhu
conclude on the universality of the methodology on both ruled et al. [60] tolerances and Gong and Wang [16] constraint
and non-ruled surfaces as well as the usability of generic cutters. surfaces.
 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808 803

Fig. 16. Timeline of main flank milling toolpath generation references.

4. Manufacturing tool adaptation used for chatter suppression optimization, thus canceling β . Chou
and Yan [69] investigates the importance of relating machine tool
dynamics and control in a CAM process to the geometrical data
4.1. Chronological review
in a CAD model: Geometric properties of curves are related to
dynamic properties of the coordinated cutting process. Ramaraj
Studies involving manufacturing tools can be mainly catego- and Eleftheriou [70] presented a model for the mechanics of ma-
rized into: (1) Manufacturing tool geometrical change and adap- chining, using shear strain properties and variations of shear stress
tation in the calculation of the computed surface, and (2) tool with applied normal stress. Wang et al. [71] presents a study of
chatter suppression and flank deflection generating β . Many stud- curve interproximation with different energy forms and param-
ies have been carried out on chatter suppression in end or sweep eterization techniques. The surface energy form might be used
milling—i.e. in [64–68], where tool cutting forces and wear were as an indicator of high non-developability. Erkorkmaz and Altin-
804 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808

tas [72] highlighted the important point that the generated tra-
jectories must not only describe the desired tool path accurately,
but must also have smooth kinematic profiles in order to maintain
0.18
high tracking accuracy and avoid exciting the natural modes of the
0.16
mechanical structure or servo control system. Monies et al. [35] Vf
0.14
examines optimal cutter selection based on [33,34]. It indicated 0.12
that the choice of the cutter is a tradeoff: while small cutters Y (mm) 0.1
produce greater theoretical accuracy, large cutters deflect less 0.08
and remove material more quickly. Chiou and Lee [73] offers a 0.06
comprehensive understanding of tool geometries as well as an 0.04
enhanced G-Buffer model by the constructed swept envelope. Sev- 0.02
Y 0
eral techniques are offered to find the 3D shape-generating pro- 10
15 6
files. The studies do not involve flank milling, but rather end 20 12
25 18 Z (mm)
milling. Dugas et al. [74] presented a machining simulation for X 30
Z X (mm) 35 24
NC application. The simulation tool was developed locally at the
IRCCyN laboratory. The main benefits of the tool are to analyze Fig. 17. Deformed surface of [77].
and localize errors obtained by tool deflection. The latter should
be taken into account when computing the optimal tool trajec- 4.2. Further considerations
tory. It would function as a corrective measure while manufac-
turing. The real feedrate and tool engagement are calculated, then Literature integrating optimal toolpath planning with force
the deflection is estimated and accounted for. Larue and Ansel- mechanistic models estimating deflections and variation on flank
metti [75] presented the deviation of a machined surface in flank milling ruled surfaces has not been undertaken. Larue and Ansel-
milling due to tool deflection. The authors predicted the defects metti [75] presents the most viable model to be further considered
in undertaking the above integration. The model is based on the
of the tool during cutting by applying an identification proce-
particular planar surface; however a generalization should be pos-
dure of force model coefficients. The study was carried out on
sible. The prediction model allowed the determination of the vari-
a planar surface and not a general ruled surface. The tool con- ations of the form and position of the surface points with a margin
sidered was assimilated to a fixed beam. The initial assumptions of 5%.
were further experimentally validated. The error between the the-
oretical calculated deflection and the digitally measured profile 5. Other related works
fell in a 0.1 mm tolerance. At a second stage, the author in [76]
generalized the methodology and presented simulation of flank 5.1. Tool flank wear and temperatures
processes using multiple tool geometries and on a general ruled
Analyzing cutting tool temperature [84,85] and measuring tool
surface—instead of applying the force prediction model on a pla- flank wear [86] provides additional insight – above tool deflection
nar surface. The machining forces are predicted in the virtual analysis – to understand the difference between the computed
environment by modeling the cutter/part intersection along the and machined surfaces. Several means of measuring the flank
toolpath. For the purposes of this study, the authors used geomet- wear have been developed. Methods based on real time vision
rical modeling to predict the forces. One of the main results would technology [87] are of particular interest for automated trajectory
be the feedrate scheduling capacity of the model in such a way as rectification. Estimating flank wear is of great concern since the
amount of flank wear is often used in estimating the tool life [88].
to keep cutting forces at desirable ranges. Landon et al. [77] proves
Bhattacharyya and Ham [89] presented one of the first wear
the possibility of predicting tool deflection with the use of a cut-
model analyses on the flank sides of the tool. These sides are the
ting force model. Models are experimentally driven. Fig. 17 shows ones used to perform a flank milling operation. The model did
the reconstruction of the deformed surface of a milled workpiece not determine interaction coefficients but rather had them given,
in side milling. Zhu et al. [61] provides a methodology to inte- as well as the force data. The paper also presented a two-flank
grate the cutting force calculations in the generation of the tool- wear model split within two areas defined as the insensitive and
path. Two methodologies are presented distinctively for 3-axis and sensitive regions. Karthik et al. [90] delivered a method to measure
5-axis machining. The cutting forces were calculated following a wear parameters to predict a potential breakage or failure of the
tool. The parameters that are detected are area, average depth and
mechanistic model. Ferry and Yip-Hoi [78] and Ferry and Altintas
volume of crater wear.
[79,80] developed a complete understanding of 5-axis flank milling
mechanics as well as a feedrate optimization technique. In [81], a 5.2. Sweep milling
predictive model integrating NC functions with axis capacities is
presented. The formalism is applicable to a multitude of tool archi- Numerical control machining of free form surfaces studies was
tectures. Quinsat et al. [82] proposed a method to characterize 3D initiated using the more general sweep milling at first in [91],
topographies of a machined surface, integrating the effects of high where interpolation techniques [92,93] were developed to ensure
velocities in the 5-axis machining setup. Chaves-Jacob et al. [83] safe, concise and accurate machining of curved objects [94]. Studies
aimed at improving the trajectories based on the geometric inputs
suggested the adaptation of the tool geometry to eliminate the er-
(scallop heights and chordal deviation) rather than on abstract
ror. Instead of optimizing the tool path, a tool profile will be as- algebraic quantities. Some attempted to define the optimal tool
sociated with the distance profiles between the tool axes and the geometry – i.e. [95] comparing ball-mills versus end-mills – while
surface at the preliminary toolpath (generated by a simple posi- most concentrated on the optimal tool trajectory computation.
tioning). Choi et al. [96] is amongst the first research published where
 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808 805

the cutter location optimization problem is translated into a 2D show that error computations are sometimes computed linearly
constrained minimization problem. The algorithm computed the whereas, due to tool rotation, the real cutting errors are higher than
cutter location data, checked for gouging, computed joint values, theoretical values. Affouard et al. [108] presented a topical method
checked the over-limit as well as the collision before submitting for avoiding the tool traversing singular positions in 5-axis sweep
the optimal join values. If a set of cutter orientation angles is milling. The approach deforms the tool path until the maximum
not found to be adequate, then new angles are proposed to be displacement of the tool is enough to leave the singular cone. This
checked. The algorithm was applied to 5-axis face milling of marine will also improve the performance dynamics. Langeron et al. [109]
propellers. You and Chu [97] presents a systematic scheme for generated a new format of tool path polynomial interpolation
the verification of tool paths in 5-axis machining of sculptured in 5-axis machining using bspline curves to reduce the usually
surfaces. The method is mainly applied to sweep milling and produced tangency discontinuities. Lamikiz et al. [110] presented a
introduces an interesting concept of surface decomposition in methodology of estimating the precision of complex 5-axis milling
order to estimate the error. The subdivision into discrete sample centers based on the Denavit and Hartenberg formulation. Even
points allows a clear geometrical computation of the error. The though the work was applied to multi-axis drilling rather than our
methodology verification is presented and demonstrated. Even subject of interest – flank milling – it clearly shows the importance
though the application is in sweep milling, the methodology as well as the complexity of kinematics when manipulating 5-axis
of error detection might still be applicable for flank milling. manufacturing operations. The methodology proposed estimating
assembly errors and then reducing those with more weight on the
Another sweep milling methodology based on the envelope
tool tip error.
differential equation to characterize the tool swept volume is
found in [98]. The algorithm uses tool grazing points to reduce
computation complexity. Warkentin et al. [99] proposes an 5.3. Further considerations
intersection approach to multi-point machining of sculptured
In [75], 5% of the unpredicted tool deviation was attributed
surfaces. The authors present the methodology using a ball nose
to a phenomenon not yet identified. We believe that further
tool performing sweep milling. The concept is to have the tool
investigation into tool wear and cutting temperatures might
in contact with the surface at multiple locations (the tip of the
optimize and further reduce this error. Moreover, we have covered
tool). The scallop left by the tool path is to be minimized. The
sweep milling, mainly the references that might be projected for
work presents a list of commonly used cutting tools as well as
our needs, i.e. error verification and swept volume generation.
the multi-point contact between a toroidal cutter and a surface.
Results are computed through geometrical transformations and
6. Summary
calculations. There is an abundant set of examples, mainly one
considering machining a cubic Bezier surface. Duc et al. [43]
In the following section we present a summary of our findings
presents a concept of a ‘machining surface’ (not to be confused
surveying flank milling literature. We also gather all the test
with our definition) that embeds all toolpath information. The surfaces for future comparisons.
application of the ‘machining surface’ ensures a good concordance
between the first design intents and the machined part. Rao and
6.1. Conclusions
Sarma [100] suggests a method for the removal of excess material
in the vicinity of the cutter contact point due to the mismatch
We can draw the following conclusions from our review of the
in curvatures between the tool and the designed surface. This literature:
article applies to sweep milling where tool is considered a flat-end
tool. The method finds the curvatures of the tool swept surface (1) Most of the papers focus on the development of optimal tool
and then gouging is detected by comparing it with the designed path of flank milling, the objective of which is to minimize
surface. Roth et al. [101] presents a method to determine the the geometrical machining errors between the tool envelope
imprint curve of tools as it executes a 5-axis tool movement. It surfaces and the designed surface. Great improvement has
been achieved based on academic and industrial colleges’
is applied on sweep milling. The obtained surface can be used to
efforts.
rectify the tool trajectory to reduce errors. The method, assessed
(2) Flank milling technology is mostly used on ruled surfaces,
by two tests, makes the approximation by assuming piecewise
because cutting tools are cylindrical surfaces or conical
linear motion of the tool. Roth et al. [102] complements [101]
surfaces, which are all ruled surfaces. Moreover, some methods
by proposing a volumetric force model of the milling process
have also tried to flank mill general free-form surfaces, which
based on the depth buffer of a high end rendering engine. The are close to ruled surfaces. Currently, machining errors from
method allows for increased simulation accuracy while reducing toolpath planning may be negligible for these surfaces.
memory requirements. It allows force magnitudes and variations (3) The basic objective of flank milling is to improve machining
of a full 3-axis cut to be determined. Chiou and Lee [103] presents strip width and reduce machining error simultaneously. More
a closed-form solution of the swept profile of a generalized cutter than ruled surfaces or quasi ruled surfaces, we may extend
in 5-axis NC machining. The approach is of interest because it this idea to more general free-form surfaces. Our concern is
is based on the machine configurations as well as the exact how to determine whether these surfaces can be machined
tool movements recorded in NC part programs. The authors had using flank milling technology. It will be meaningful to extend
previously presented [104] another approach to generate 5-axis them to more general cutting tools for more general surfaces
tool paths, called machining potential field. The latter searches to improve machining efficiency.
for a preliminary optimal tool path of which adjacent toolpaths (4) Compared with end milling, flank milling has higher cutting
are proposed. Fussell et al. [105] uses a Z -buffer approach to forces, which may lead to more deformation of cutting tools
model the part geometry during a 5-axis sculptured surface and generate extra errors, even the roughness get worse.
machining. The application is concerned with end milling and the Apart from the cutting force, there are some other relevant
intersection between the tool path and part, which is evaluated factors, such as tool geometry and the parameters of flank
using a 3-axis approximation. Experimentation showed that the milling. In future work, more attention may be paid to the
model is accurate except under large transient conditions. Yoon design of cutting tools for flank milling and optimization of
et al. [106] and Ho et al. [107], respectively, present sweep milling the machining parameters according to the cutting force or
optimization and trajectory smoothing algorithms. The latter [107] roughness of machined surface.
806 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808

Table 9 References
Test surface of [21].
T0 T1 T2 B0 B1 B2 [1] Young HT, Chuang LC, Gerschwiler K, Kamps S. A five-axis rough machining
approach for a centrifugal impeller. International Journal of Advanced
x 0 11.507 23.014 0 11.507 23.014 Manufacturing Technology 2004;23:233–9.
y 0 0 20.2324 20.429 20.429 20.429 [2] Tönshoff H, Gey C, Rackow N. Flank milling optimization—the Flamingo
z 33.995 33.995 33.995 0 0 0 project. Air and Space Europe 2001;3:60–3.
[3] Peternell M, Pottmann H, Ravani B. On the computational geometry of ruled
Table 10 surfaces. Computer-Aided Design 1999;31:17–32.
Test surface of [29]. [4] Gong H, Wang N. Optimize tool path of flank milling with generic cutter based
on approximation using envelope surface. Computer-Aided Design 2009;
30u + 10u + 30 + 20v u − 10v u − 10v 
 2 2
 41(12):981–9.
√ √ [5] Senatore J, Monies F, Redonnet JM, Rubio W. Analysis of improved positioning

1

S (u, v) = 100 u − 50 − 100v u + 90v − 60v sin πu in five-axis ruled surface milling using envelope surface. Computer-Aided
2
Design 2005;37(10):989–98.

70 − 30u2 − 130v + 50v u2


[6] Pottmann H, Wallner J. Approximation algorithms for developable surfaces.
Computer-Aided Design 2005;3:1459–68.
Table 11 [7] Chen HY, Pottmann H. Approximation by ruled surfaces. Journal of
Test surface of [47]. computational and Applied Mathematics 1999;102:143–56.
[8] Bohez E. Compensating for systematic errors in 5-axis NC machining.
T0 T1 T2 B0 B1 B2 Computer-Aided Design 2001;34:391–403.
[9] Bohez E, Ranjith Senadhera SD, Pole K, Duflou JR, Tar T. A geometric modeling
x 1 0 0 1 −0.25 0
and five-axis machining algorithm for centrifugal impellers. Journal of
y 0 0 1 0 −0.25 1
Manufacturing Systems 1997;16(6):422–36.
z 1 1 1 0 −1 0
[10] Bohez E, Makhanov SS, Sonthipermpoon K. Adaptive nonlinear tool path
optimization for five-axis machining. International Journal of Production
Table 12 Research 2000;38(17):4329–43.
First test surface of [48]. [11] Bohez E. Five-axis milling machine tool kinematic chain design and analysis.
International Journal of Machine Tools and Manufacture 2002;42:505–20.
T0 T1 T2 B0 B1 B2 [12] Suh S, Lee K. Avoiding tool interference in four-axis NC machining of
x 65 30 0 60 30 15 rotationally-free-surfaces. IEEE Transactions on Robotics & Automation
y 15 30 60 9 30 75 1992;v8:718–29.
[13] Takeuchi Y, Watanabe T. Generation of 5-axis control collision free tool
z −5 −5 −5 −35 −35 −35
path planning and postprocessing for NC data. Annals of the CIRP 1992;41:
539–42.
Table 13 [14] Lee YS. Admissible tool orientation control of gouging avoidance for 5-axis
complex surface machining. Computer-Aided Design 1997;29(7):507–21.
Second test surface of [48].
[15] Lee YS. Non-isoparametric tool path planning by machining stip evaluation
T0 T1 T2 T3 B0 B1 B2 B3 for 5-axis sculptured surface machining. Computer-Aided Design 1998;
30(7):559–70.
x 5 40 75 110 5 40 75 110 [16] Gong H, Wang N. 5-axis flank milling free-form surfaces considering
y 65 35 35 65 50 80 20 50 constraints. Computer-Aided Design 2011;43(6):563–72.
z −35 −35 −35 −35 −5 −5 −5 −5 [17] Zheng G, Zhu LM, Bi QZ. Cutter size optimisation and interference-free
tool path generation for five-axis flank milling of centrifugal impellers.
International Journal of Production Research. 2011. on line. http://dx.doi.org/
Table 14 10.1080/00207543.2011.611631.
Test surface of [5]. [18] Castagnetti C, Duc E, Ray P. The domain of admissible orientation concept:
 π  a new method for five-axis tool path optimisation. Computer-Aided Design
20(2v − 1) cos 4 (1 − 2u) 2008;40:938–50.


S (u, v) = 37.5(2u − 1)  [19] Stute G, Storr A, Sielaff W. NC programming of ruled surfaces for five axis
20(2v − 1) sin π (1 − 2u) machining. Annals of the CIRP 1979;28(1):267–71.
 

4 [20] Rehsteiner F, Rewker HJ. Collision free five axis milling of twisted ruled
surface. Annals of the CIRP 1993;42(1):457–61.
[21] Liu XW. Five-axis NC cylindrical milling of sculptured surfaces. Computer-
(5) The issue of geometrical error has been discussed widely. But Aided Design 1995;27(12):87–94.
[22] Li SX, Jerard RB. Five-axis machining of sculptured surfaces with a flat-end
this is only calculated in the workpiece coordinate system.
cutter. Computer-Aided Design 1994;26(3):165–78.
If we transform the toolpath into the motions of each axis [23] Chappel IT. The use of vectors to simulate material removed by numerically
respectively, extra kinematics error will be introduced. In controlled milling. Computer-Aided Design 1983;15(3):156–8.
addition, dynamics issue will be raised too. Therefore, toolpath [24] Wu CY, Raw JA, Thompson RA, Townsend PR, Wilford AP. A CAD/CAM system
for flank milling an axial compressor blisk. In: 2nd Canadian CAD/CAM &
generation incorporating consideration of post-processing robotics conference proceedings; 1983.
may be paid more attention to in future research. [25] Brown L, Plucinski J. A modern method for productivity and profit — the
(6) Flank milling of planar surfaces has not been treated. Geomet- application of CAD/CAM at Pratt & Whitney aircraft of Canada Ltd. In: CSME
rically, flank milling planar surfaces does not represent any conference; 1979.
[26] Wu CY. Arbitrary surface flank milling of fan, compressor and impeller blades.
challenge as any selected tool direction is optimal (error is In: 1994 International gas turbine and aeroengine congress and exposition.
null). The complexity remains in the toolpath energy mini- ASME Journal of Engineering for Gas Turbines and Power 1995;117(3):534–9.
mization and optimal boundary manufacturing combination. [27] Wu ChungY. Arbitrary surface flank milling & flank SAM in the design &
manufacturing of jet engine fan and compressor airfoils. In: Proceedings of
the ASME IGTI turbo expo conference in Copenhagen; 2012.
6.2. Test surfaces [28] Elber G, Fish R. 5-Axis freeform surface milling using piecewise ruled surface
approximation. Journal of Manufacturing Science and Engineering 1997;119:
383–7.
Tables 9–14 constitute the test surfaces used in several [29] Redonnet J, Rubio W, Dessein G. Side milling of ruled surfaces; optimum
examples. Future studies aiming at studying flank milling should positioning of the milling cutter and calculation of interference. The
provide results on the surfaces below in comparison with other International Journal of Advanced Manufacturing Technology 1998;14(7):
459–65.
methodologies.
[30] Rubio W, Lagarrigue P, Dessein G, Pastor F. Calculation of tool paths for a
torus mill on freeform surfaces on five-axis machines with detection and
Acknowledgment elimination of interface. International Journal of Advanced Manufacturing
Technology 1998;14:13–20.
[31] Abdel-Malek K, Yeh HJ. Geometric representation of the swept volume using
This work was done with the help of a Lebanese American Jacobian rank-deficiency conditions. Computer-Aided Design 1997;29(6):
University research grant. 457–68.
 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808 807

[32] Leu MC, Wang L, Blackmore D. A verification program for 5-axis machining [63] Zheng G, Bi QZ, Zhu LM. Smooth tool path generation for five-axis flank
with general APT tools. Annals of the CIRP 1997;46(1):419–24. milling using multi-objective programming. Proceedings of the Institution
[33] Monies F, Redonnet JM, Rubio W, Lagarrigue P. Improved positioning of a of Mechanical Engineers Part B-Journal of Engineering Manufacture 2012;
conical mill for machining ruled surfaces. Journal of Engineering Manufacture 226(2):247–54.
2000;214(7):625–34. [64] Abrari F, Elbestawi MA, Spence AD. On the dynamics of ball end milling:
[34] Monies F, Rubio W, Redonnet JM, Lagarrigue P. Comparative study of modeling of cutting forces and stability analysis. International Journal of
interference caused by different position settings of a conical milling cutter Machine Tools and Manufacture 1998;38:215–37.
on a ruled surface. Journal of Engineering Manufacture 2001;215(B9): [65] Ismail F, Elbestawi MA, Du R, Urbasik K. Generation of milled surfaces
1305–17. including tool dynamics and wear. Journal of Engineering for Industry 1993;
[35] Monies F, Felices JN, Rubio W, Redonnet JM, Lagarrigue P. Five axis NC milling 115:245–52.
of ruled surfaces: optimal geometry of a conical tool. International Journal of [66] Ismail F, Ziaei R. Chatter suppression in five-axis machining of flexible parts.
Production Research 2002;40(12):2901–22. International Journal of Machine Tools & Manufacture 2002;42:115–22.
[36] Senatore J, Monies F, Redonnet JM, Rubio W. Improved positioning for side [67] Tlusty J, Ismail F. Basic nonlinearity in machining chatter. Annals of the CIRP
milling of ruled surface: analysis of the rotation axis’s influence on machining 1981;299–304.
error. International Journal of Machine Tools and Manufacture 2007;47: [68] Elbestawi MA, Ismail F, Du R, Ullagaddi BC. Modelling machining dynamics
934–45. including damping in the tool workpiece interface. Journal of Engineering for
[37] Senatore J, Landon Y, Rubio W. Analytical estimation of error in flank milling Industry 1994;116:435–9.
of ruled surfaces. Computer-Aided Design 2008;40:595–603. [69] Chou JJ, Yan D. Command generation for three-axis CNC machining. ASME
[38] Lin RS, Koren Y. Ruled surface machining on five-axis CNC machine tools. Journal of Engineering Industry 1991;113:305–10.
Journal of Manufacturing Processes 2000;1(2):25–35. [70] Ramaraj TC, Eleftheriou E. Analysis of the mechanics of machining with
[39] Tsay DM, Her MJ. Accurate 5-axis machining of twisted ruled surfaces. ASME tapered end milling cutters. Transactions of ASME 1994;116:398–404.
Journal of Manufacturing Science and Engineering 2001;123:731–8. [71] Wang X, Cheng F, Barsky B. Energy and B-spline interproximation. Computer-
[40] Tsay DM, Chen HC, Her MJ. A study on five-axis flank machining of centrifugal Aided Design 1997;29(7):485–96.
compressor impellers. ASME Journal of Engineering for Gas Turbines and [72] Erkorkmaz K, Altintas Y. High speed CNC system design, part I: jerk limited
Power 2002;124:177–81. trajectory generation and quintic spline interpolation. International Journal
[41] Kim JH, Ryuh BS, Pennock GR. Development of a trajectory generation of Machine Tools & Manufacture 2001;41:1323–45.
method for a five-axis NC machine. Mechanism and Machine Theory 2001; [73] Chiou JCJ, Lee YS. A shape-generating approach for multi-axis machining
36:983–96. G-buffer models. Computer-Aided Design 1999;31:761–76.
[42] Tournier C, Duc E, Lartigue C, Contri A. The concept of the machining [74] Dugas A, Lee J, Hascoet JY. An enhanced machining simulator with error
surface in 5-axis milling of free form surfaces. In: Integrated design and analysis. Journal of Manufacturing Systems 2003;21(6):451–564.
manufacturing in mechanical engineering. Kluwer Academic Publishers; [75] Larue A, Anselmetti B. Deviation of a machined surface in flank milling.
2002. p. 279–86. International Journal of Machine Tools and Manufacture 2003;43(2):
[43] Duc E, Lartigue C, Tournier C, Bourdet P. A new concept for the design and 129–39.
the manufacturing of free-form surfaces: the machining surface. Annals of [76] Larue A, Altintas Y. Simulation of flank milling process. International Journal
the CIRP 1999;48(1):103–6. of Machine Tools and Manufacture 2005;45:549–59.
[44] Pechard PY, Tournier C, Lartigue C, Lugarini JP. Geometrical deviations versus
[77] Landon Y, Segonds S, Lascoumes P, Lagarrigue P. Tool positioning error
smoothness in 5-axis high-speed flank milling. International Journal of
(TPE) characterisation in milling. International Journal of Machine Tools and
Machine Tools and Manufacture 2009;49:453–61.
Manufacture 2004;44(5):457–64.
[45] Beudaert X, Pechard PY, Tournier C. 5-Axis tool path smoothing based on
[78] Ferry WB, Yip-Hoi D. Cutter-workpiece engagement calculations by parallel
drive constraints. International Journal of Machine Tools & Manufacture
slicing for five-axis flank milling of jet engine impellers. ASME Journal of
2011;51:958–65.
Manufacturing Science and Engineering 2008;130.
[46] Lartigue C, Duc E, Affouard A. Tool path deformation in 5-axis flank milling
[79] Ferry WB, Altintas Y. Virtual five-axis flank milling of jet engine impellers—
using envelope surface. Computer-Aided Design 2003;35:375–82.
part I: mechanics of five-axis flank milling. Journal of Manufacturing Science
[47] Bedi S, Mann S, Menzel C. Flank milling with flat end milling cutters.
and Engineering 2008;130.
Computer-Aided Design 2003;35:293–300.
[80] Ferry WB, Altintas Y. Virtual five-axis flank milling of jet engine impellers—
[48] Li C, Mann S, Bedi S. Error measurements for flank milling. Computer-Aided
part II: feed rate optimization of five-axis flank milling. Journal of
Design 2005;37(14):1459–68.
Manufacturing Science and Engineering 2008;130.
[49] Menzel C, Bedi S, Mann S. Triple tangent flank milling of ruled surfaces.
[81] Lavernhe S, Tournier C, Lartigue C. Kinematical performance prediction
Computer-Aided Design 2004;36(3):296–389.
in multi-axis machining for process planning optimization. International
[50] Chiou JCJ. Accurate tool position for five-axis ruled surface machining by
Journal of Advanced Manufacturing Technology 2008;37(5–6):534–44.
swept envelope approach. Computer-Aided Design 2004;36(10):967–74.
[82] Quinsat Y, Lavernhe S, Lartigue C. Characterization of 3D surface topography
[51] Weinert K, Du S, Damm P, Stautner M. Swept volume generation for the
in 5-axis milling. Wear 2011;271(3–4):590–5.
simulation of machining processes. International Journal of Machine Tools
& Manufacture 2004;44:617–28. [83] Chaves-Jacob J, Poulachon G, Emmanuel D. New approach to 5-axis flank
[52] Lee SW, Nestler A. Complete swept volume generation, part I: swept volume milling of free-form surfaces: computation of adapted tool shape. Computer-
of a piecewise C1-continuous cutter at five-axis milling via Gauss map. Aided Design 2009;41:918–29.
Computer-Aided Design 2011;43(4):427–41. [84] Loewen EG, Shaw MC. On the analysis of cutting-tool temperatures.
[53] Gong H, Cao LX, Liu J. Improved positioning of cylindrical cutter for flank Transactions of the ASME 1954;76:217–31.
milling ruled surfaces. Computer-Aided Design 2005;37(12):1205–13. [85] Yellowley I, Barrow G. The influence of thermal cycling on tool life in
[54] Chu CH, Chen JT. Tool path planning for five-axis flank milling with peripheral milling. International Journal of Machine Tools and Manufacture
developable surface approximation. International Journal of Advanced 1976;16:1–12.
Manufacturing Technology 2006;29(7–8):707–13. [86] Giardini C, Ceretti E, Maccarini G. A neural network architecture for tool wear
[55] Gong H, Cao LX, Liu J. Second order approximation of tool envelope surface for detection through digital camera observations. The International Journal of
5-axis machining with single point contact. Computer-Aided Design 2008; Advanced Manufacturing Technology 1996;372:137–44.
40(5):604–15. [87] Jeon JU, Kim SW. Optical flank wear monitoring of cutting tools by image
[56] Sprott K, Ravani B. Cylindrical milling of ruled surfaces. International Journal processing. Wear 1988;127:207–17.
of Advanced Manufacturing Technology 2008;38:649–56. [88] Park JJ, Ulsoy AG. On-line flank wear estimation using an adaptive observer
[57] Wu PH, Li YW, Chu CH. Optimized tool path generation based on dynamic and computer vision, part 2: experiment. Transactions of the ASME Journal
programming for five-axis flank milling of rule surface. International Journal of Engineering for Industry 1993;115:30–43.
of Machine Tools and Manufacture 2008;48:1224–33. [89] Bhattacharyya A, Ham I. Analysis of tool wear-part 1: theoretical model of
[58] Ding H, Zhu LM. Global optimization of tool path for five-axis flank milling flank wear. ASME Journal of Engineering for Industry 1969;790–8.
with a cylindrical cutter. Science in China Series E-Technological Sciences [90] Karthik A, Chandra S, Ramamoorthy B, Das S. 3D tool wear measurement and
2009;52(8):2449–59. visualization using stereo imaging. International Journal of Machine Tools
[59] Gong H, Wang N. Analytical calculation of envelope surface for generic and Manufacture 1997;37(11):1573–81.
milling tool directly from CL-data based on moving frame. Computer-Aided [91] Loney G, Ozsoy T. NC machining of free form surfaces. Computer-Aided
Design 2009;41(11):848–55. Design 1987;19:85–90.
[60] Zhu LM, Zhang XM, Zheng G, Ding H. Analytical expression of the swept [92] Shpitalni M, Koren Y, Lo CC. Real time curve interpolators. Computer-Aided
surface of a rotary cutter using the envelope theory of sphere congruence. Design 1994;26(11):832–8.
Journal of Manufacturing Science and Engineering-Transactions of the ASME [93] Koren Y, Lin RS. Five-axis surface interpolators. Annals of the CIRP 1995;44:
2009;131(4):041017. 379–82.
[61] Zhu LM, Zheng G, Ding H, Xiong YL. Global optimization of tool path for five- [94] Bedi S, Ali I, Quan N. Advanced interpolation techniques for CNC machines.
axis flank milling with conical cutter. Computer-Aided Design 2010;42(10): ASME Journal of Engineering Industry 1993;115:329–36.
903–10. [95] Vickers GB, Guan KW. Ball-mills versus end-mills for curved surface
[62] Zhang XM, Zhu LM, Zheng G, Ding H. Tool path optimisation for flank machining. ASME Journal of Engineering for Industry 1989;111:22–6.
milling ruled surface based on the distance function. International Journal [96] Choi BK, Park JW, Jun CS. Cutter-location data optimization in 5-axis surface
of Production Research 2010;48(14):4233–51. machining. Computer-Aided Design 1993;25(6):377–86.
808 R.F. Harik et al. / Computer-Aided Design 45 (2013) 796–808

Dr. Ramy Harik is an Assistant Professor at the Lebanese

[97] You CF, Chu CH. Tool-path verification in five-axis machining of sculptured
American University (LAU) and holder of a Ph.D. special-
surfaces. International Journal of Advanced Manufacturing Technology 1997;
ized in CAD/CAM in Industrial Engineering from Henri
13(4):248–55.
Poincaré University, France (2007). He holds a joint ap-
[98] Blackmore D, Leu MC, Wang LP. The sweep-envelope differential equation
pointment in Mechanical and Industrial engineering and
algorithm and its application to NC machining verification. Computer-Aided
teaches courses in the areas of design and manufacturing.
Design 1997;29(9):629–37.
In 2011, he was a Fulbright visiting scholar at the Compu-
[99] Warkentin A, Ismail F, Bedi S. Intersection approach to multi-point ma-
tational Design and Innovation Lab at Purdue University.
chining of sculptured surfaces. Computer Aided Geometric Design 1998;15:
His research is in the areas of CAD geometric modeling,
567–84.
feature recognition, manufacturing identification and er-
[100] Rao A, Sarma R. On local gouging in five-axis sculptured surface machining
gonomics. He is a member of ASME, IIE & CAD’xx. Since
using flat-end tools. Computer-Aided Design 2000;32:409–20.
2008, he has been a member of the program advisory board of the annual CAD con-
[101] Roth D, Bedi S, Ismail F, Mann S. Surface swept by a toroidal cutter during
ference and exhibition and was appointed as the conference co-chair of the CAD’10
5-axis machining. Computer-Aided Design 2001;33(1):57–63.
edition.
[102] Roth D, Ismail F, Bedi S. Mechanistic modeling of the milling process using an
adaptive depth buffer. Computer-Aided Design 2003;35:1287–303.
[103] Chiou JCJ, Lee YS. Swept surface determination for 5-axis numerical control Dr. Hu Gong is an Associate Professor in the State Key Lab-
machining. International Journal of Machine Tools and Manufacture 2002; oratory of Precision Measuring Technology & Instruments,
42:1497–507. Centre of MicroNano Manufacturing Technology at Tian-
[104] Chiou JCJ, Lee YS. A machining potential field approach to tool path jin University. He received his Ph.D. degree from Dalian
generation for multi-axis sculptured surface machining. Computer-Aided University of Technology (2006) in P.R. China. His research
Design 2002;34:357–71. interests include CAD/CAM, multi-axis NC machining, ul-
[105] Fussell BK, Jerard RB, Hemmett JG. Modeling of cutting geometry and forces trasonic machining, precision machining and measuring.
for 5-axis sculptured surface machining. Computer-Aided Design 2003;35:
333–46.
[106] Yoon JH, Pottmann H, Lee YS. Locally optimal cutting positions for 5-axis
sculptured surface machining. Computer-Aided Design 2003;35:69–81.
[107] Ho MC, Hwang YR, Hu CH. Five-axis tool orientation smoothing using Prof. Alain Bernard, 51, graduated in 1982, gained his
quaternion interpolation algorithm. International Journal of Machine Tools Ph.D. in 1989, was Associate-Professor from 1990 to
and Manufacture 2003;43(12):1259–67. 1996 at Centrale Paris. From Sept. 1996 to Oct. 2001,
[108] Affouard A, Duc E, Lartigue C, Langeron JM, Bourdet P. Avoiding 5-axis he was Professor at CRAN, Nancy I, in the ‘‘Integrated
singularities using tool path deformation. International Journal of Machine Design and Manufacturing’’ team. Since 0ct. 2001, he
Tools and Manufacture 2004;44(4):415–25. has been Professor at Centrale Nantes and (is now?)
[109] Langeron JM, Duc E, Lartigue C, Bourdet P. A new format for 5-axis tool Dean for Research. He is at IRCCyN in the ‘‘System
path computation, using bspline curves. Computer-Aided Design 2004;36: Engineering-Product-Performance-Perception’’ team. His
1219–29. research topics are KM, system modeling, interoperability,
[110] Lamikiz A, López de Lacalle LN, Ocerin O, Díez D, Maidagan E. The Denavit performance evaluation, virtual engineering, and rapid
and Hartenberg approach applied to evaluate the consequences in the tool product development. He has published more than 60
tip position of geometrical errors in five-axis milling centres. Journal of papers in refereed international journals. He is chairman of WG5.1 of IFIP and vice-
Advanced Manufacturing Technology 2008;37:122–39. chairman of CIRP STC Design.