International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

Modeling of material removal and surface roughness in

abrasive flow machining process
Rajendra K. Jain, Vijay K. Jain*, P.M. Dixit
Mechanical Engineering Department, Indian Institute of Technology, Kanpur 208016, India
Received 16 July 1998; received in revised form 15 March 1999

Abstract
Abrasive flow machining process provides a high level of surface finish and close tolerances with an
economically acceptable rate of surface generation for a wide range of industrial components. This paper
deals with the theoretical investigations into the mechanism of abrasive flow machining (AFM) process.
A finite element model is developed for the flow of media during AFM and the same is used to evaluate
the stresses and forces developed during the process. Theoretical analysis to estimate the material removal
and surface roughness obtained during AFM is also proposed. The theoretical results are compared with
the experimental data available in the literature, and they are found to agree well.  1999 Elsevier Science
Ltd. All rights reserved.
Keywords: AFM; Modeling; FEM; Machining; Finishing

Nomenclature
a radius of the projected area of indentation, m
[A] matrix of A
{A} column vector of A
[B] matrix of differential coefficients of shape functions
C percentage concentration of abrasives by weight
dg diameter of abrasive grain, m
dw diameter of workpiece to be finished, m
Hw hardness of workpiece material, N/m2
Fn normal force applied to abrasive grain, N

* Corresponding author. Tel.: ⫹ 91-512-257-916; fax: ⫹ 91-512-250-260; email: [email protected]

0890-6955/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 8 9 0 - 6 9 5 5 ( 9 9 ) 0 0 0 3 8 - 3
1904 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

f average coefficient of friction

{F} global right hand side vector
[Ke] elemental coefficient matrix
Li actual contact length in ith stroke, m
ls stroke length, m
lw workpiece length, m
M mesh size of abrasive grains
N number of abrasive grains acting per unit contact area of media and workpiece,
mm2
n number of cycles
{Nb} one dimensional matrix of biquadratic shape functions
[Nv] matrix of biquardatic shape functions for the approximation of velocities
{Np} column vector of bilinear shape functions for the approximation of velocities
p hydrostatic pressure, N/m2
r reduction ratio
r, ␪, z cylindrical coordinates
Rw radius of workpiece, m
Rc radius of media cylinder, m
Roa initial surface roughness of the workpiece, ␮m
Ra final surface roughness of the workpiece, ␮m
T finishing time, s
t depth of indentation of abrasive into workpiece material, m
ti component of traction vector
V volume of total material removed, m3
Vi volume of material removed in ith stroke, m3
vf velocity of media inside the workpiece, m/s
vp velocity of piston or velocity of media flow at inlet, m/s
vi component of velocity vector
{we} vector of elemental nodal weight functions for primary variables
{wve} vector of elemental nodal weight functions for velocities
{wpe} vector of elemental nodal weight function for pressure
{w} global vector of nodal weight function
␣ angle made by normal at tool surface with r-axis
␦ij Kronecker’s delta
⌫ boundary of the domain
⑀˙ ij strain rate tensor
␳ density of media, kg/m3
␴ij stress tensor
␴o initial yield stress of the material, N/m2
␴r normal stress, N/m2
␩a apparent viscosity function of the media, Pa s
␩o constant used for non-dimensionalization of ␩a
␶ij deviatoric component of stress
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1905

1. Introduction

Abrasive flow machining (AFM) is a non-traditional finishing process that is used to deburr,
polish, radius, remove recast layer, and to produce compressive residual stresses even in difficult
to reach areas. Applications of this process include finishing of precision medical and aerospace
components, dies and high volume production of electronic and automotive parts. Materials from
soft aluminum to tough nickel alloys, ceramics and carbides have been successfully micro
machined with this process [1]. AFM process provides a high level of surface finish and close
geometric tolerances with an economically acceptable rate of surface generation for a wide range
of industrial components. The ability of media in AFM process to finish difficult to reach areas,
to follow complex contours and to simultaneously work on multiple edges and surfaces, makes
it more versatile than other finishing processes.
AFM removes small quantities of material by flowing a semi solid abrasive-laden media through
or across a workpiece. Two vertically opposed cylinders extrude media back and forth through
passages formed by the workpiece and tooling (Fig. 1). The media is composed of semi solid
carrier and abrasive grains. The media acts as a ‘self deformable stone’ having protruding abrasive
particles acting as cutting tools.
AFM process is in its early stages of development and the available published literature mainly
deals with a qualitative description of the process and its applications. The abrasion ability of
abrasive media is governed by many factors, especially by grain size, abrasive concentration,
extrusion pressure and hardness of workpiece material. In order to analyze the influence of such

Fig. 1. Schematic diagram of abrasive flow machining process.

1906 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

parameters and other AFM conditions upon material removal and surface roughness of the
machined surface, experimental investigations have been carried out by many researchers [1–3].
A stochastic modeling and analysis technique called data dependent system (DDS) was used by
Williams et al. [2] to study the surface roughness profiles before and after AFM. Jain and Adsul
[3] have presented the effect of various parameters on material removal and surface roughness
in terms of polynomial equations. Rajeshwar et al. [4] presented a simulation model to determine
the characteristics of media flow during machining. The finite difference method was chosen for
obtaining the solution. Although these investigations are excellent, they seem to be rather lacking
in theoretical treatment. Both theoretical and empirical studies of abrasive flow machining are
greatly hampered by the inherent random nature and multiplicity of variables.
Not much information is available in the literature, which deals with theoretical analysis of
material removal mechanism in AFM. The relationship between process parameters and perform-
ance characteristics of the process are not completely known. The exact mechanism by which the
individual abrasive particles accomplish material removal is only partially understood. There is
a need to carry out systematic theoretical analysis of viscous media flow to determine the stresses
and machining forces acting on the workpiece. This will help in modeling and proper understand-
ing of the machining action during AFM. Once machining forces acting on the workpiece surfaces,
developed due to media flow, are known, an expression can be established for material removal
and surface roughness in terms of extrusion pressure and other process variables. The numerical
simulation of viscous flow has been the topic of many researchers [5–8]. Early numerical formu-
lation of viscous materials involved finite element techniques based on the Galerkin models, and
finite difference techniques based on stream function-vorticity formulations. Zienkiewicz et al.
[5] have presented the flow-formulation approach in forming and extrusion, investigating two
techniques, viz. the pressure-velocity formulation with Lagrangian constraints and the penalty-
function approach. Dixit and Dixit [6] carried out deformation analysis of the steady-state wire
drawing process by FEM using the mixed pressure-velocity formulation. The initial developments
of finite difference and finite element methods used for solving non-Newtonian flow has been
reviewed by Crochet et al. [7].
The objective of the present paper is to establish a model for media flow, and hence to determine
the stresses developed during the flow. These computed stresses can be used to evaluate the
machining force and hence, material removal and surface roughness can be obtained in AFM for
the given machining conditions. The present paper describes the simulation of flow of media in
AFM process as a viscous fluid. The method involves the use of finite element technique to solve
the governing equations in terms of primary variables (velocity and pressure). The authors have
also presented a theoretical approach to the problems of estimation of material removal and surface
finish during abrasive flow machining process.

2. Modeling the flow of media in AFM process

The mathematical representation of the flow of media in AFM process involves the equations
of continuity and of momentum and the constitutive equations. The following assumptions are
made to simplify the analysis.
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1907

1. The media used in AFM is composed of semisolid carrier mixed with abrasives which exhibits
linear viscous flow property [9].
2. The media is isotropic and homogeneous. The media properties are independent of fluid tem-
perature, and constant with time and space.
3. Since cylindrical workpiece is considered, media flow is taken as axi-symmetric.
4. The media flow is steady.
Steady state, incompressible and axi-symmetric forms of the continuity, momentum and consti-
tutive equations are given below.

2.1. Basic equations

2.1.1. Continuity equation (conservation of mass)

For steady state, incompressible axi-symmetric media flow, neglecting the volume of material
removed from work surface, the condition of volume constancy can be expressed as,
⑀˙ rr ⫹ ⑀˙ ␪␪ ⫹ ⑀˙ zz ⫽ 0 (1)
where, ⑀˙ rr, ⑀˙ ␪␪ and ⑀˙ zz are components of strain rate tensor in r, ␪, and z directions, respectively
and can be expressed in terms of the velocity as

⑀˙ ij ⫽ 再
1 ∂vj
⫹
∂vk
2 ∂xk ∂xj 冎 (2)

2.1.2. Momentum equations (balance of momentum)

The momentum equations for a steady-state axisymmetric flow of media (neglecting the body
forces) have the following form.

冋
␳ vr
∂vr
∂r
⫹ vz
∂vr
∂z
⫹册∂p
∂r
⫺
r ∂r 冋
1 ∂(r␶rr) ␶␪␪ ∂␶rz
⫺
r
⫹
∂z
⫽0 册 (3)

and

冋
␳ vr
∂vz
∂r
⫹ vz
∂vz
∂z
⫹册∂p
∂z
⫺
r ∂r 冋
1 ∂(r␶rz) ∂␶zz
⫹
∂z
⫽0 册 (4)

where, ␳ is density of media, p is pressure (hydrostatic part of stress tensor), and ␶ij is the extra-
stress (deviatoric part of stress tensor).

2.1.3. Constitutive equations

The media behaves as a Newtonian fluid with shear stress being the linear function of shear
rate [10]. Hence, the constitutive equation for AFM media can be written as
␶ij ⫽ ␩a⑀˙ ij (5)
where, ␩a is the apparent viscosity function of media.
1908 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

For an incompressible fluid, the total stress tensor ␴ij is given by

␴ij ⫽ ⫺ p␦ij ⫹ ␶ij (6)
where, ␦ij are components of the identity tensor (or Kronecher delta).

2.2. Non-dimensionalization

Non-dimensionalization is usually required to avoid ill-conditioning and numerical difficulties.

The non-dimensionalization of all physical quantities used above, has been done using character-
istic dimensional quantities as given below.
r z vr vz ⑀¯ ij p ␶ij ␩a
r̄ ⫽ ; z̄ ⫽ ; v̄r ⫽ ; v̄z ⫽ ; ⑀˙ ij ⫽ ; p̄ ⫽ ; ␶¯ ij ⫽ ; ␩¯ ⫽ (7)
Rw Rw vp vp (vp/Rw) ␴o ␴o ␩o
where, ␩o ⫽ (␴o)/(vp/Rw); ␴o is yield stress of material; Rw is radius of workpiece, and vp is
velocity of media flow at inlet.
Substituting non-dimensional parameter in continuity, momentum and constitutive Eqs. (1),
(3)–(5), we obtain
⑀˙ rr ⫹ ⑀˙ ␪␪ ⫹ ⑀˙ zz ⫽ 0 (8)
(␳vpRw)
␩o
v̄r
∂r冋
∂v̄r
⫹ v̄z
∂v̄r
∂z
⫹册 冋
∂p̄ 1 ∂(r̄ ␶¯ rr) ␶¯ ␪␪ ∂␶¯ rz
∂r̄
⫺
r̄ ∂r̄
⫹
r̄
⫺
∂z̄
⫽0 册 (9)

(␳vpRw)
␩o
v̄r
∂r冋
∂v̄z
⫹ v̄z
∂v̄z
∂z
⫹册 冋
∂p̄ 1 ∂r̄ ␶¯ rz ∂␶¯ zz
∂z̄
⫺
r̄ ∂r̄
⫺
∂z̄
⫽0 册 (10)

␶¯ ij ⫽ ␩¯ ⑀˙ ij (11)

For convenience in writing, the overbar (-) from non-dimensional parameters is dropped in the
succeeding description. Since Reynolds number (␳vpRw)/(␩o) is found to be typically of the order
of 10−8 for all the cases considered, hence inertial terms of Eqs. (9) and (10) are ignored in the
next section.

2.3. Finite element formulation

The finite element formulation of flow of the rigid plastic material in terms of primary variables
(pressure and velocity) has been presented by Dixit and Dixit [6]. Similar formulations with some
modifications are possible for the flow of AFM media which is a viscous incompressible material.
In the present analysis, the Galerkin weighted residual method [11] is employed to reduce the
governing differential equations by using appropriate weighting functions for the momentum and
continuity equations. Let vr, vz, and p constitute the functions which satisfy all the essential bound-
ary conditions. Then vr, vz, and p constitute a weak solution if the following integral equation
is satisfied.
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1909

冕冋
⍀
{⑀˙ rr ⫹ ⑀˙ ␪␪ ⫹ ⑀˙ zz}wp ⫹ 再 ∂p 1 ∂(r␶rr) ␶␪␪ ∂␶rz
∂r
⫺
r ∂r
⫹
r
⫺
∂z r
w 冎
⫹ 再 ∂p 1 ∂(r␶rz) ∂␶zz
⫺
∂z r ∂r
⫺
∂z z 冎册
w 2␲r dr dz ⫽ 0 (12)

where, wp, wr and wz are the weight functions which satisfy the homogeneous versions of the
boundary conditions and ⍀ represents the area of the domain.
Performing the integration by parts on second and third segments of the above equation, the
following weak form is obtained

冕⍀
[I1 ⫹ I2]2␲r dr dz ⫺ 冕
⌫1
I3d⌫1 ⫺ 冕
⌫2
I4d⌫2 ⫽ 0 (13)

where, I1 ⫽ {⑀˙ rr ⫹ ⑀˙ ␪␪ ⫹ ⑀˙ zz}wp

I2 ⫽ ⫺ p{⑀˙ rr(w) ⫹ ⑀˙ ␪␪(w) ⫹ ⑀˙ zz(w)} ⫹ ␩a{⑀˙ rr⑀˙ rr(w) ⫹ ⑀˙ ␪␪⑀˙ ␪␪(w) ⫹ ⑀˙ zz⑀˙ zz(w) ⫹ 2⑀˙ rz⑀˙ rz(w)}
I3 ⫽ tzwz
I4 ⫽ trwr
and ⌫1 and ⌫2 are the boundaries where the traction components tr and tz are specified (Fig. 2).
The term ⑀˙ rr(w), ⑀˙ ␪␪(w), ⑀˙ zz(w) and ⑀˙ rz(w) are defined as

⑀˙ rr(w) ⫽
∂wr
∂r
; ⑀˙ zz(w) ⫽
∂wz
∂z
; ⑀˙ ␪␪(w) ⫽
wr
r
; and ⑀˙ rz(w) ⫽
2 ∂r冉
1 ∂wr ∂wz
⫹
∂z 冊
In the present analysis, Langrangian biquardatic shape function for vr and vz, and bilinear shape
function for p over a 9 noded quadrilateral element are used. The final finite element equation is
obtained by assembling the elemental coefficient matrices and right side vectors into global coef-
ficient matrix and right side vector respectively. The elemental coefficient matrices [Ke] are calcu-

Fig. 2. The domain and finite element mesh for axisymmetric flow of media in AFM.
1910 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

lated by evaluating the integrals by Gauss quadrature intregation technique. The global finite
element equation after application of all the boundary conditions is as follows:

冋 [0] [kpv]
[kvp] [kvv]
册再 冎 再 冎
p
v
⫽
0
F
(14)

where,

[kpv] ⫽ ⫺ 冕 {Np}{m}T[B]2␲r dr dz (15)

Ae

{m}T ⫽ {1 1 0 1} (16)

[kvv] ⫽ 冕 ␩[B]T[B]2␲r dr dz (17)

Ae

[kvp] ⫽ ⫺ 冕 [B]T{m}{Np}T2 ␲ r dr dz (18)

Ae

{F} ⫽ 冕 {Nb}{Nb}T{tb}2␲r d⌫ (19)

⌫b

where, [B] is matrix of differential coefficients of shape functions for the velocities, {Np} is
column vector of bilinear shape functions for the approximation of pressure, {Nb} is one dimen-
sional matrix of biquardatic shape functions, and {tb} is traction vector at boundaries.
The above Eq. (14) is solved after substitution of the boundary conditions.

2.4. Boundary conditions

In the present work, axi-symmetric abrasive media flow is considered for the analysis taking
into consideration the schematic diagram of the AFM process shown in Fig. 1. The domain along
with the boundary conditions is shown in Fig. 2 and boundary conditions are described as
given below:
1. Entry (at AB)
vr ⫽ 0; vz ⫽ vp
where, vp is the velocity of piston or velocity of flow at inlet.

2. Exit (at HG)

 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1911

vr ⫽ 0; tz ⫽ 0

3. Surface of media cylinder (at BC and FG)

vr ⫽ 0; tz ⫽ 0

4. Surface of workpiece (DE)

vr ⫽ 0; tz ⫽ 0

5. Media tool interface (CD and EF). The normal component of velocity (vn) and the tangential
component of the traction (ts) must be zero. Hence,
vn ⫽ 0; ts ⫽ 0 at CD and EF.

6. Axis of symmetry (AH)

vr ⫽ 0; tz ⫽ 0

The geometry of flowing media is discretized using a mesh of 5 ⫻ 20, 9-noded elements, in
the r and z directions (Fig. 2). The mesh emphasizes radial and axial refinement near the singular-
ity. The experience with inelastic flows [8] has shown that radial and axial mesh refinement is
very important to resolve adequately steep gradients. Also it has been known in viscous calcu-
lations that elements near the singularity with large aspect ratios produce numerical instabilities.
The fine mesh offers a compromise between these two concerns.

3. Mechanism of material removal in AFM

The abrasion in AFM referred to in this paper is defined as the removal of solid material from
a surface by the unidirectional sliding action of discrete particles of another material. The basic
mechanism of abrasion in general has been the subject of many investigators. Khrushchov and
Bavichov [12] identified two processes taking place when abrasive grains make contact with the
wearing surface. (1) The formation of plastically impressed grooves which did not involve material
removal, and (2) the separation of material particles in the form of micro chips. It is found [13,14]
that the chip cutting and rubbing depend upon the shape of indenting particles. In particular,
spherical indentors have been observed to show a change over from rubbing to at least partial
chip formation when the indentation strain (defined as the depth of indentation divided by the
diameter of the indentor) exceeds a certain value. In abrasive flow machining, the interaction
between the abrasive particle and machined surface can be divided into microploughing, and
microcutting.
In the ideal case, microploughing due to a single pass of one abrasive particle does not result
into any detachment of material from the workpiece surface. A prow is formed ahead of the
1912 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

abrading particle and material is continuously displaced sideways to form ridges adjacent to
groove produced. Volume loss can however occur owing to the action of many abrasive particles
or the repeated action of a single particle. Material may be ploughed aside repeatedly by passing
particles and may break off by low cycle fatigue (microfatigue). Pure microcutting results in a
volume loss in the form of chips and equal to the volume of the grooves produced. Microploughing
and microcutting are the dominant processes on ductile materials while microcutting becomes
important on brittle materials.

4. Modeling for material removal and surface finish

Abrasive flow machining is considered to be a scratching action performed by abrasive grains

in the AFM media. The normal force applied to a spherical grain will cause it to penetrate in the
surface (Fig. 3(a)) [15]. The grain produces a groove on the workpiece surface whose section
corresponds to the profile of the grain. As the grain is translated horizontally, it removes material
from the workpiece surface. The amount of stock removal is then equal to the total volume of
the grooves produced on the workpiece surface by each grain in the media. The idealized classical
model of abrasive wear [16] provides a theoretical basis for the determination of material removal
by AFM process. If the number of active grains, their shape and depth of the groove produced
are known, the volume of stock removal can be calculated.

4.1. Assumptions

The following assumptions are adopted for the analysis of material removal by AFM process.
1. Most of the abrasive grains are blocky crystals that may be approximated to be spherical in
shape [15]. This is also reasonable considering the large negative rake angles presented by the
grains. Observation of abrasive particles also shows that for the most part, they are generally
rounded, though irregular, in shape and are not composed of acute cutting edges.
2. It is assumed that each grain consists of a single active cutting edge. If there are more than
one cutting edges on one grain, there is no space to store the chip between the cutting edges.
So that cutting edge can not be active any more. As a result, the grain acts as one cutting
edge [17].
3. To render the mathematics tractable, it is necessary to assume that the load on each particle
is constant and equal to the average load.

Fig. 3. (a) Schematic diagram of a spherical abrasive grain removing material from the workpiece surface. (b) Simpli-
fied surface geometry
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1913

4. Every abrasive grain is assumed to achieve the same penetration depth depending upon the
applied force.
5. All active grains are of the same size and there is metal available to be cut by every abras-
ive grain.

4.2. Material removal

The normal force acting on a spherical grain will cause it to penetrate the surface just as a
Brinell ball makes a dent during a hardness test. When this grain is translated horizontally, the
plastically deformed zone beneath the surface will be inclined and give rise to upward flow thus
forming a chip which is subsequently sheared from the surface.
The indenting force Fn (normal force) on a spherical grain of diameter dg is given as
␲d 2g
Fn ⫽ ␴r (20)
4
where, ␴r is normal stress acting on the grain.
If a is the radius of the projected area of indentation ⵜA, t is depth of indentation, and Hw is
hardness of workpiece material, then
Fn ⫽ HwⵜA ⫽ Hw␲a2 (21)

From the geometry of Fig. 3(a), radius of the projected area ‘a’ of the indent made and depth
of indentation (t) can be obtained as given below.
a ⫽ √t(dg ⫺ t) (22)

冪冉 冊
dg d 2g
t⫽ ⫺ ⫺ a2 (23)
2 4

Substituting the value of ‘a’ from Eq. (21), into Eq. (23), we obtain,

冪
dg d 2g Fn
t⫽ ⫺ ⫺ (24)
2 4 Hw␲

The cross sectional area of groove generated A⬘ (shaded portion of the grain, as shown in Fig.
3(a)) can be derived from the geometry of the figure.

A⬘ ⫽
d 2g −1 2√t(dg ⫺ t)
4
sin
dg
⫺ √t(dg ⫺ t)
dg
2
⫺t冉 冊 (25)

Therefore, volume of the material removed (Va) by an abrasive grain is equal to the product
of area A⬘ and length of contact Li of grain with workpiece surface.
1914 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

Va ⫽ 冋 d 2g −1 2√t(dg ⫺ t)
4
sin
dg
⫺ √t(dg ⫺ t)
2冉 冊册
dg
⫺t Li (26)

In practice, the total material removal is made up of a number of similar processes. Let N be
the number of abrasive grains simultaneously acting per unit area of contact. Then, total no. of
abrasive grains (Ns) indenting in the workpiece surface per stroke are given by
Ns ⫽ 2␲RwNvfls/vp (27)
where, ls is stroke length, vp is velocity of piston, vf is velocity of flow across workpiece surface,
and Rw is radius of cylindrical workpiece (Fig. 1).
From continuity equation
vf R2c
⫽ 2 (28)
vp Rw
where, Rc is radius of media cylinder.
By substituting the value of (vf)/(vp) from Eq. (28) into Eq. (27), Ns can be calculated as
given below.
R2c
Ns ⫽ 2␲RwNls (29)
R2w
Hence, volumetric material removal in ith stroke (Vi) in AFM is given as

Vi ⫽ 2␲Nls 冋
R2c d 2g −1 2√t(dg ⫺ t)
Rw 4
sin
dg
⫺ √t(dg ⫺ t)
dg
2
⫺t冉 冊册 Li (30)

This simplified model suggests that material removal per unit length of sliding depends on
depth of indentation (Eq. (24)), size of abrasive grains, and total number of abrasive grains (Eq.
(27)) for the given size of workpiece and media cylinder.

4.3. Surface roughness

It is assumed that the surfaces of the workpiece have uniform profile without statistical distri-
bution, that they have an initial surface roughness Roa, and that the abrasives move in the length
direction of the scratches [18]. There is a constant ratio between peak to valley rougness Rt and
arithmetic average roughness Ra. Let Ria be the surface roughness obtained after ith stroke during
AFM. From Fig. 3(b), the actual contact length (Li) between the workpiece and a spherical abras-
ive is given as

Li ⫽ 冉 冊 冋
ⵜlc
ⵜlg
lw ⫽
(Roa ⫺ Ria)tan␪
Roatan␪
lw 册 冋
⫽ 1 ⫺
Ria
l
Roa w 册 (31)

where, ⵜlc and ⵜlg are shown in Fig. 3(b).

 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1915

Substituting the value of Li from Eq. (31) into Eq. (30), the volumetric material removal in ith
stroke Vi is given as

Vi ⫽ 2␲Nls
Rw 4 冋
R2c d 2g −1 2√t(dg ⫺ t)
sin
dg
⫺ √t(dg ⫺ t)
dg
2 冉 冊册冋 册
⫺t 1⫺
Ria
l
Roa w
(32)

Let D ⫽ 2␲Nlslw
Rw 4 冋
R2c d 2g −1 2√t(dg ⫺ t)
sin
dg 冉 冊册
⫺ √t(dg ⫺ t)
dg
2
⫺t (33)

Hence, Vi ⫽ D 1 ⫺ 冋 Rai
Roa 册 (34)

Total material removal in n number of cycles is evaluated as

冘
2n

V⫽ Vi
i⫽1

冋
R2c d 2g −1 2√t(dg ⫺ t)
冉 冊册 冘 冋 册
2n
dg Ria
V ⫽ 2␲Nls sin ⫺ √t(dg ⫺ t) ⫺t 1⫺ l (35)
Rw 4 dg 2 i⫽1
Roa w
Weight of material removed Wm in n number of cycles is given by

Wm ⫽ 2␲␳wNls
Rw 4 冋
R2c d 2g −1 2√t(dg ⫺ t)
sin
dg
⫺ √t(dg ⫺ t)
dg
2 冉 冊册 冘 冋
⫺t
2n

i⫽1
1⫺ 册
Ria
l
Roa w
(36)

where, ␳w is density of workpiece material.

For a machined surface, the ratio of peak-to-valley roughness (Rt) and arithmetic average rough-
ness (Ra) is closer to 5 for ground surfaces and is about 10 for honed surfaces [19]. This ratio is
assumed as 7 for the surface used for AFM in the present case. The same ratio is also obtained
during experimentation by Jain et al [21].
According to continuity equation, the sum of all the grooves produced constitute a volume
equal to the volumetric material removed from the workpiece surface.
Material removed in ith stroke ⫽ actual contact length ⫻ width of workpiece ⫻ total height
of material removed

冋
Vi ⫽ 1 ⫺ 册
Ria
l ⬵2␲Rw ⫻ 7(Rai ⫺ 1 ⫺ Ria)
Roa w
(37)

Substituting the value of Vi from Eq. (34) into Eq. (37), and simplifying, we find,
D
Ria ⫽ Ria ⫺ 1 ⫺ (38)
14␲Rwlw
Substituting the value of D from Eq. (33) into Eq. (38),

Ria ⫽ Ria ⫺ 1 ⫺
1
7
Nls 2
Rw 4
sin 冋
R2c d 2g −1 2√t(dg ⫺ t)
dg
⫺ √t(dg ⫺ t)
dg
2
⫺t冉 冊册 (39)
1916 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

A critical surface roughness (Rc r) may exist at the given machining force, because of the
indentation of the cutting edge into the workpiece, such that the surface roughness will no longer
improve beyond that. The critical surface roughness is equal to the depth of indentation of spheri-
cal abrasive grain (given by Eq. (24)) in the machined surface.

冪
dg d 2g Fn
Rcr ⫽ ⫺ ⫺ (40)
2 4 Hw ␲

5. Results and discussion

The simulation procedure involves the determination of stresses on the workpiece surface by
the solution of flow model by finite element technique as described in Section 2 of this paper.
The normal stress, so obtained from the flow model, is used for the estimation of material removal
and surface finish according to model mentioned in Section 4 of the paper. The numerical values
used in the simulation are given in Appendix A.
In Fig. 4, the piston pressures are plotted against the piston velocity for various angles of tool
(tool directs the media, a polymer based abrasive laden compound, to the areas of the workpiece
to be abraded, Fig. 1). It is observed that the present FEM analysis gives a linear relationship
between the piston pressure and the velocity of the piston. The piston pressure increases linearly
with piston velocity and its value is more for higher angle of tool (Fig. 4). Further to achieve the
same flow velocity of media, tool angle should be low at low pressure. Thus for a given piston
velocity, it requires highest pressure at 90° tool angle which is a commonly used configuration.
Hence, the tool should be designed rather than using 90° angle configuration. Normal stress also
varies linearly with piston velocity (Fig. 5) and its value is higher for higher value of angle of
tool. At lower radial pressure, the penetration depth expected is lower. If radial pressure is very
high and the abrasive grain size is also high, then it may lead to the final surface finish worse

Fig. 4. Variation in piston pressure with piston velocity for various tool angles.
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1917

Fig. 5. Variation in normal stress on workpiece surface with piston velocity for various tool angles.

than the initial surface finish. Hence, grain size selection and design of tool to achieve the appro-
priate radial pressure is important so that the desired surface finish can be achieved.
Fig. 6 shows the effect of reduction ratio on normal stress on the workpiece surface for various
angles of tool. Reduction ratio expresses the reduction in cross sectional area of flow. It is defined
as the difference between cross sectional area of media cylinder and workpiece, divided by the
cross sectional area of media cylinder [1 ⫺ (Rw/Rc)2]. It is seen that normal stress at the workpiece
surface increases with increase in reduction ratio. With higher value of reduction ratio, the rate
of increase in normal stress is also higher. It has also been observed experimentally [21] that
high abrading action occurs at the cross section of high restriction and less abrading action in the
region of less restriction (as abrasion is directly proportional to radial stresses). The computed
stresses have been used to evaluate the machining forces acting in abrasive flow machining. The
machining force so evaluated from the analysis has been employed to carry out theoretical analysis
of material removal and surface roughness in AFM. Thus, analysis of media flow is an important
step in the direction of modeling of AFM process.
The values of the active number of abrasive grains at various concentrations and mesh sizes

Fig. 6. Variation of normal stress with reduction ratio for various tool angles.
1918 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

Fig. 7. Variation of material removal with piston velocity.

have been reported by Jain et al [20]. The same values have been used in the present computations
of material removal and surface roughness. Fig. 7 compares the experimental and theoretical
relationships between the piston velocity and material removal. It is found that material removal
increases with increase in piston velocity. The simulation results for material removal agree well
with the experimental data. But at higher velocity, number of abrasive grains taking part in actual
cutting may be less due to rolling of grains at higher velocity. Hence, actual material removal
will be less than theoretical which can be seen from Fig. 7 also. Material removal also increases
with increase in piston pressure, normal stresses and reduction ratio (Figs. 8 and 9), due to increase
in depth of indentation. At higher reduction ratio, the forces acting on the workpiece surface are

Fig. 8. Variation of material removal with piston pressure and normal stress.
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1919

Fig. 9. Variation of material removal with reduction ratio.

higher, leading to higher material removal. The same is true with higher velocity, pressure and
radial stresses.
Fig. 10 shows the effect of mesh size on material removal for brass workpiece. From the figure
it is evident that as the mesh size increases the material removal decreases. The reason for decrease
of material removal is that, with increase in mesh size, area of indented groove and depth of
indentation decrease. Experimental results of Jain and Adsul [3] also show the similar trend.
Fig. 11 shows the effect of percentage concentration of abrasives by weight on material
removal. As the percentage concentration of abrasives in media increases, material removal
increases because higher concentration results in higher viscosity of media and higher normal and
tangential forces on abrasive grains. With higher percentage concentration of abrasives, more
numbers of abrasive grains will come into contact with the surface to be machined. All these
factors lead to the increase in material removal.
Fig. 12 exhibits the role of piston velocity on surface roughness value obtained after the speci-
fied number of machining cycles. It is observed that change in surface roughness value increases
with increase in piston velocity and it tends to become constant after a certain value of velocity.

Fig. 10. Variation of material removal with abrasive mesh size.

1920 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

Fig. 11. Variation of material removal with percentage concentration of abrasives.

Fig. 12. Effect of velocity on change of surface roughness value.

The surface roughness value as a function of pressure is shown in Fig. 13. The trend is similar
to that of the effect of piston velocity, because piston pressure is proportional to the piston velo-
city. The surface roughness is seen to first decrease with increasing pressure, then it becomes
constant which is equal to the critical surface roughness value. Any further increase in pressure
will deteriorate the surface finish due to increase in depth of indentation. Change in surface
roughness increases with increase in percentage concentration of abrasives (Fig. 14). With higher
percentage concentration of abrasives, the number of abrasives taking part in machining will be
more, and hence surface roughness obtained for a specified number of cycles will be better.
Change in surface roughness value is higher with increase in grain mesh size (Fig. 15). If the
grain size is large enough, then it may lead to the surface roughness worse than the initial one
due to increase in depth of indentation. Such behavior has also been observed by Jain et al [21].
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1921

Fig. 13. Variation of surface roughness with piston pressure.

Fig. 14. Effect of percentage concentration of abrasives on change in surface roughness.

6. Conclusions

In the present paper, a model for the flow of AFM media through cylindrical workpiece is
developed and solved by finite element method. The model is shown to predict the radial stresses
at the workpiece surface with reasonable accuracy. A model for the material removal by the
abrasive media is also presented. Classical abrasion theory has been applied in the simulation
model. The model is based on accumulated plastic flow, by repeated indentation by moving abras-
ive particles. The model includes the operating conditions of the process, e.g. abrasive size, num-
ber of contacting abrasive particles, extrusion pressure, workpiece material hardness and size of
workpiece to be machined. Good agreement was found between the experimental results and
theoretical predictions. The following conclusions have been derived from the present analysis.
1. The model, developed for the media flow in AFM predicts the radial stresses at the workpiece
surface with reasonable accuracy.
1922 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923

Fig. 15. Effect of mesh size on change in surface roughness.

2. The piston velocity increases linearly with piston pressure and its value is higher for larger
angle of tool.
3. At higher value of reduction ratio, the rate of increase in piston pressure is higher. The normal
stresses on the workpiece surface increase with increase in reduction ratio.
4. Material removal increases with piston pressure and percentage concentration of abrasives in
media, but its value decreases with mesh size of abrasive grains.
5. Surface roughness value decreases with increase in piston velocity, piston pressure, percentage
concentration of abrasives and grain mesh size for a specified number of cycles. Beyond a
certain value of velocity and pressure, it deteriorates due to increase in depth of indentation.
The minimum surface roughness value is equal to the critical surface roughness.
The extension of the present model to three dimensional shape and consideration of change in
machining conditions of AFM process with time are some of the issues that need to be investi-
gated. Although the method has been defined for, and applied to, axisymmetric problems, the
technique is easily extendable to three dimensional cases.

Acknowledgements

The authors greatly appreciate the help and suggestions of Prof. K. Murlidhar of Mechanical
Engineering Department, Indian Institute of Technology, Kanpur during the finite element formu-
lation of the process. The financial support by the Department of Science and Technology, Govt.
of India, New Delhi for the project entitled “Abrasive flow machining process” is acknowledged.

Appendix A: Numerical values of conditions used in the simulation

Radius of cylinder, Rc ⫽ 0.0275 m

Radius of workpiece, Rw ⫽ 0.015 m
 R.K. Jain et al. / International Journal of Machine Tools & Manufacture 39 (1999) 1903–1923 1923

Length of stroke, ls ⫽ 0.05m

Length of workpiece, lw ⫽ 0.012 m
Workpiece material ⫽ brass
Hardness of workpiece material, Hw ⫽ 1500 MPa
Apparant viscosity of media, ␩a (C ⫽ 45% and M ⫽ 100) ⫽ 543.38 MPa
Concentration of abrasives, C ⫽ 45%
Abrasive grain diameter, dg ⫽ 0.152 mm. (Mesh size, M ⫽ 100)
Number of cycles, n ⫽ 50

References

[1] L.J. Rhoades, Abrasive flow machining with not-so-silly putty, Metal Finishing (1987) 27–29.
[2] R.E. Williams, K.P. Rajurkar, Stochastic modeling and analysis of abrasive flow machining, ASME Journal of
Engineering for Industry 114 (1992) 74–81.
[3] V.K. Jain, S.G. Adsul, Experimental investigations into abrasive flow machining process (to be published).
[4] G. Rajeshwar, J. Kozak, K.P. Rajurkar, Modeling and computer simulation of media flow in abrasive flow machin-
ing process, Proceedings of the International Mechanical Engineering Congress and Exposition, Chicago, PED
68 (1994) 965–971.
[5] O.C. Zienkiewicz, P.C. Jain, E. Onate, Flow of solids during forming and extrusion: some aspects of numerical
solutions, International Journal of Solids Structures 14 (1978) 15–38.
[6] U.S. Dixit, P.M. Dixit, An analysis of the steady-state wire drawing of strain-hardening material, Journal of
Materials Processing Technology 47 (1995) 201–229.
[7] M.J. Crochet, A.R. Davies, K. Walters, Numerical Simulation of Non-Newtonian Flow, Elsevier, New York, 1984.
[8] M.E. Kim-E, R.A. Brown, R.C. Armstrong, The role of inertia and shear thinning in flow of an inelastic liquid
through an axisymmetric sudden contraction, Journal of Non-Newtonian Fluid Mechanics 13 (1983) 341–363.
[9] V.K. Jain, C. Ranganatha, Evaluation of properties of media used in abrasive flow machining (to be published).
[10] R. Darby, Viscoelastic Fluids, Marcel Dekker Inc., 1976.
[11] O.C. Zienkiewicz, K. Morgan, Finite Elements and Approximation, Wiley, New York, 1983.
[12] M.M. Khrushchov, M.A. Bavichev, Research on wear of metals, Ch.8, NEL Translation No. 893, National Engin-
eering Laboratory, East Kilbride (1960).
[13] D. Graham, R.M. Baul, An investigation into the mode of material removal in the grinding process, Wear 19
(1972) 301–314.
[14] R.L. Aghan, L.E. Samuels, Mechanism of abrasive polishing, Wear 16 (1970) 293–301.
[15] M.C. Shaw, A new theory of grinding, in: Proc. of Institutions Conference on Production Science in Industry,
Melbourne, 1971, pp. 73–78.
[16] E. Rabinowicz, L.A. Dunn, P.G. Russell, A study of abrasive wear under three body conditions, Wear 4 (1961)
345–355.
[17] J. Verkerk, Final report concerning CIRP co-operative work on the characterization of grinding wheel topography,
Annals of the CIRP 26 (1977) 385–394.
[18] J.-D. Kim, M.-S. Choi, Simulation for the prediction of surface accuracy in magnetic abrasive machining, Journal
of Material Processing Technology 53 (1995) 630–642.
[19] M.C. Shaw, Principles of Abrasive Processing, Clarendon Press, Oxford, 1996.
[20] R.K. Jain, V.K. Jain, Simulation of abrasive flow machining process (to be published).
[21] V.K. Jain, S.K. Choudhary, S. Jha, On the performance of abrasive flow machining process (to be published).