Int J Adv Manuf Technol (2005) 26: 477–490

DOI 10.1007/s00170-004-2180-x

ORIGINAL ARTICLE

S.C. Jayswal · V.K. Jain · P.M. Dixit

Modeling and simulation of magnetic abrasive finishing process

Received: 27 November 2003 / Accepted: 13 March 2004 / Published online: 2 February 2005
© Springer-Verlag London Limited 2005

Abstract Magnetic abrasive finishing (MAF) is one of the ad- tively new fine finishing method, magnetic abrasive finishing
vanced finishing processes, which produces a high level of sur- (MAF) is an advanced finishing process in which the cutting
face quality and is primarily controlled by a magnetic field. In force is primarily controlled by the magnetic field. It minimizes
MAF, the workpiece is kept between the two poles of a mag- the possibility of microcracks on the surface of the workpiece,
net. The working gap between the workpiece and the magnet particularly in hard brittle material, due to controlled low forces
is filled with magnetic abrasive particles. A magnetic abrasive acting on abrasive particles [1]. This process is able to produce
flexible brush (MAFB) is formed, acting as a multipoint cutting surface roughness of nanometer range on flat surfaces as well as
tool, due to the effect of the magnetic field in the working gap. internal and external cylindrical surfaces [2]. It can also be used
This paper deals with the theoretical investigations of the MAF for internal finishing of non-rotatable workpieces such as elbows
process. A finite element model of the process is developed to and bent tubes [3]. The MAF process offers many advantages,
evaluate the distribution of magnetic forces on the workpiece such as self-sharpening, self-adaptability, controllability, and the
surface. The MAF process removes a very small amount of ma- finishing tools require neither compensation nor dressing [4].
terial by indentation and rotation of magnetic abrasive particles In MAF, the workpiece is kept between the two poles of a mag-
in the circular tracks. A theoretical model for material removal net (Fig. 1a). The working gap between the workpiece and the
and surface roughness is also proposed accounting for microcut- magnet is filled with magnetic abrasive particles (MAPs) (Fig. 2),
ting by considering a uniform surface profile without statistical composed of ferromagnetic particles and abrasive powder. MAPs
distribution. Numerical experiments are carried out by providing can be used as bonded or unbonded. Bonded MAPs are pre-
different routes of intermittent motion to the tool. The simula- pared by sintering of ferromagnetic particles and abrasive powder
tion results are verified by comparing them with the experimental whereas unbonded MAPs are a mechanical mixture of ferromag-
results available in the literature. netic particles and abrasive powder with a small amount of lu-
bricant [4–6]. The purpose of the lubricant is to provide some
Keywords FEM · Magnetic abrasive finishing (MAF) · holding strength between the constituents of MAPs. The bonded
Modeling · Nanometer (nm) finish · type of MAPs are considered in the present work due to their ex-
Non-conventional finishing · Simulation cellent finishing effects [4]. The magnetic abrasive particles join
each other along the lines of magnetic force and form a magnetic
abrasive flexible brush (MAFB) between the workpiece and the
1 Introduction magnetic pole (Fig. 1a). This brush behaves like a multi-point cut-
ting tool for the finishing operation. When the magnetic N-pole
Traditional fine finishing operations such as grinding, lapping, or is rotating, the MAFB also rotates like a flexible grinding wheel
honing employ a rigid tool that subjects the workpiece to sub- and finishing is done according to the forces acting on the abrasive
stantial normal stresses which may cause microcracks resulting particles. It is usually assumed that there is no slip between the N-
in reduced strength and reliability of the machined part. A rela- pole and MAFB. The arrangement of magnetic abrasive particles
is shown only in the last track (Fig. 1b), however, in reality many
S.C. Jayswal∗ · V.K. Jain (u) · P.M. Dixit such tracks are formed on the workpiece.
Department of Mechanical Engineering, Shinmura et al. [7, 8] found that magnetic flux density and
Indian Institute of Technology Kanpur, working gap greatly affect the surface roughness and stock re-
Kanpur-208016, India
E-mail: [email protected] moval. A vibrational motion given to cylindrical workpieces en-
∗ On leave from Department of Mechanical Engineering, M.M.M. Engineer- hances the finishing efficiency. Shinmura et al. [6, 9] have also
ing college, Gorakhpur-273010, India analyzed the effect of the size of magnetic abrasive particles, and
478

Fig. 1a,b. Magnetic abrasive finishing setup:

a Front view, b Top view

found that stock removal and surface roughness value increase

as the magnetic abrasive particle diameter “Dmap” increases.
Surface roughness is improved by decreasing the abrasive par-
ticle diameter “ds”. Deburring of grinding burrs can easily be
performed with this process. Fox et al. [10] have found that
unbonded MAPs yield a higher material removal rate (MRR)
and bonded MAPs give a better surface roughness. The sur-
face roughness value (Ra) of a ground rod after MAF has been
achieved as low as 10 nm. It was observed during finishing the
external surface of a cylindrical workpiece by Jain et al. [2], that
the working gap and circumferential speed of the workpiece are
Fig. 2. Schematic view of a bonded magnetic abrasive particle; Dmap = the parameters which significantly influence material removal
Diameter of magnetic abrasive particle; ds = diameter of abresive particle and the change in surface roughness value.
 479

Shinmura et al. [11, 12] have also conducted an experimental material properties of the MAFB. Use of analytical methods to
study on plane workpieces using the MAF process. They ob- obtain the solution of the governing equation under such hetero-
served that the surface roughness value decreases with increasing geneous conditions would be very difficult. Therefore, the finite
finishing time upto a certain limit of time beyond which no element method is used to obtain the solution. Here, the govern-
further improvement was noticed. Addition of machining fluid ing equation of the process is expressed in terms of the magnetic
(such as stearic acid, straight oil type of grinding fluid) to un- potential, which is the primary variable.
bonded MAPs has shown remarkable effects on stock removal
and surface roughness [5]. 2.1 Governing equation
Hou et al. [13] presented the thermal aspect of magnetic
abrasive finishing of a ceramic roller using the bonded type of The following assumptions are made to derive the governing
MAPs. Waigaonkar et al. [14] presented the optimization of the equation which is based on Maxwell’s equations [19]:
magnetic abrasive process. Yamguchi et al. [15] developed an in- 1. The intensity of the magnetic field is not varying with time
ternal magnetic abrasive finishing process using a pole rotation as the operating parameters are held constant during MAF
system to produce a highly finished inner surface of a workpiece. process.
They also studied the effect of a magnetic field on MAFB config- 2. Leakage of the magnetic field, if any, is negligible due to
uration and force acting on abrasive particles. a small working gap.
Kremen et al. [16] developed an empirical expression to es- 3. When the magnetic field is applied, the magnetic abrasive
timate machining time to produce a workpiece with specified particles are closely packed in the form of a brush between
out-of-roundness. Kim and Choi [17] modeled and simulated the the tool (north pole) and workpiece. It means that there is no
MAF process for finishing cylindrical workpieces and concluded air pocket in the brush, and relative permeability of MAPs is
that the magnetic flux density in the air-gap is affected greatly by calculated by considering relative fractions of ferromagnetic
the length of the air-gap; magnetic flux density increases as the particles and abrasive particles.
air-gap length decreases. They have also found that simulation 4. The chips removed during the MAF process have an insignif-
results for surface roughness agree better with the experimental icant effect on the properties of the MAFB due to very low
data for the low magnetic flux density than they do for high mag- material removal during the process.
netic flux density. Kremen et al. [18] proposed a theory to explain 5. The magnetic potential drop in the core is neglected due to
the “out-of-roundness” phenomenon based on force analysis and very low reluctance of the core.
the material removal mechanism. 6. The solution domain is considered as axisymmetric because
From the above literature survey, it can be concluded that of the axisymmetic nature of the geometry of the MAFB as
very little effort has been made toward the modeling and simu- shown in Fig. 1b and magnetic field.
lation of the plane magnetic abrasive finishing process. Most of
Intensity of the magnetic field H is a gradient of magnetic scalar
the work done on this process is experimental and consists of
potential φ and is expressed as:
the study of the process principle and effects of various pro-
cess parameters on the material removal and surface roughness. H = −∇φ . (1)
This paper presents a mathematical model for mechanics of ma-
terial removal during the MAF process. A theoretical model of Based on the above assumptions, in the axisymmetric form, the
surface roughness has also been proposed to predict surface qual- governing equation of the process becomes:
ity. To determine surface quality, distribution of magnetic forces


needs to be estimated first. A finite element based code has been 1 ∂ ∂φ ∂ ∂φ
rµr + µr =0, (2)
developed to evaluate the distribution of magnetic forces, con- r ∂r ∂r ∂z ∂z
sidering magnetic flux density, type and size of MAPs, and the
where, φ is a magnetic scalar potential and µr is the relative
working gap as the main parameters.
permeability of MAPs. The permeability of most of the ferro-
magnetic materials is not constant, and varies with the magnetic
field. It implies that Eq. 2 is non-linear in nature.
2 Mathematical modeling
The MAF process is not completely understood as there is a lack 2.2 Boundary conditions
of quantitative relationships between process parameters and
Since the problem is axisymmetric, only the right half of the
process performance. Mathematical modeling and analysis of the
cross section (region ABCD of Fig. 1a) is used as a solution do-
process would help in understanding the forces acting on the
main for the analysis. The solution domain is shown in Fig. 3.
workpiece and mechanism of material removal.
The total boundary consists of four parts: C1 , C2 , C3 , and C4 .
In MAF, magnetic abrasive particles are placed under the in-
The boundary conditions of the problem are as follows.
fluence of a magnetic field. MAPs acquire magnetic polarization
and join each other along the lines of magnetic force, forming (1) Essential boundary conditions:
a magnetic abrasive flexible brush. The MAF process involves The magnetic potential φ is specified on the boundaries C1
irregular geometry in the working gap and nonlinear composite and C3 (Fig. 3). The magnetic potential is assumed to have a zero
480

where the vector {φ}, called the global magnetic potential vector,
contains the nodal value of φ at all the nodes of the domain and
the matrix [K], called the global coefficient matrix. The value
of [K] is obtained after assembling the elemental coefficient ma-
trices over all the elements. The typical elemental coefficient
matrix [k]e is given by

[k] = µr [B]eT [B]e 2πr drdz ,
e
(8)
Ae
Fig. 3. Finite element mesh
where the matrix [B]e contains derivatives of the shape functions
and Ae is the domain of a typical area element.
value at the workpiece (C1 -south pole). The value of magnetic
The integral in expression 8 is computed numerically using
potential on the north pole (C3 ) can be calculated as a product of
the Gauss–Legendre quadrature with 3 Gauss points in each di-
input current (I) and the number of turns (N) in the electromag-
rection. The value of µr of MAPs depends on the field strength
net (due to assumptions 2 and 5). Thus,
which, in turn, depends on the magnetic scalar potential. There-
φ = 0 on C1 , (3) fore, the values of µr at the Gauss points are found iteratively.
φ = NI on C3 . (4) The relations between µr and H (graphical as well as in the form
of equations) are given in Appendix-I. The variation (ε) in the
(2) Natural boundary conditions: solution in the consecutive iterations is computed as
On the line of symmetry (AD) (Fig. 3), the normal derivative 
  (k)
tnn 
(k−1) 2
of the magnetic potential is zero. Thus, φi − φi 
i=1
∂φ ε=  , (9)
= 0 on C4 . (5) 
tnn  2
∂n φi(k−1)
The boundary BC is a line of magnetic force. Then the equipo- i=1
tential lines are perpendicular to this boundary. Therefore, the where tnn is the total number of nodes and k is iteration number.
normal derivative of the magnetic potential is zero on the bound- Iterations are continued till the value of ε is equal to or less than
ary BC. Thus, the prescribed tolerance (say, equal to 10−3 ).
∂φ The finite element equations (Eq. 7) are a set of linear alge-
= 0 on C2 (6) braic equations. These equations are solved by the Gauss elimi-
∂n
nation method after imposing essential boundary conditions.
The shape of this line BC and the location of point C are
unknown. In this work, the boundary BC is approximated as 3.2 Evaluation of secondary variables
a parabola and the location of point C is found by the trial and
error method. For this purpose numerical experiments were per- The values of secondary variables are most accurate at the Gauss
formed by varying the length of the fringing zone from one to points. Therefore intensity of the magnetic field, derivative of in-
six times the working gap. When EC is equal to two times the tensity of magnetic field, and magnetic force are calculated at the
working gap, magnetic potential is just 0.2% in comparison to Gauss points. For the calculation of magnetic forces, H and its
its maximum value. Hence, the distance EC (Fig. 3) is taken as derivatives are needed. These are calculated from Eq. 1 using the
being equal to twice the working gap (BE). finite difference method [21].
The magnetic force on a particle depends on the magnetic
field strength and the magnetic property of the particle. By virtue
3 Finite element analysis of its position in the magnetic field, the magnetic potential en-
ergy, Em , of the particle is given by Eq. 10 [22]:
Galerkin’s finite element method [20] is used to evaluate the dis- 
µ0
tribution of magnetic potential (φ) within the solution domain. Em = χr H · H dv , (10)
2
The domain is discretised using an isoparametric eight noded v
quadrilateral elements (Fig. 3). where, v is the volume of the particle and χr is its susceptibil-
ity. The magnetic force F acting on the particle is defined as the
3.1 Finite element equations and solution procedure gradient of the magnetic potential energy. The radial and normal
components of F can be expressed as
Application of the Galerkin finite element method to the prob-
lem consisting of differential Eq. 3 and boundary conditions µ0 ∂
Fr = v (χr H · H) , (11)
(Eqs. 3–6) leads to the following algebraic equations: 2 ∂r
µ0 ∂
[K] {φ} = {0} , (7) Fz = v (χr H · H) . (12)
2 ∂z
 481

Fig. 4. A typical element with MAPs

Since the magnetic abrasive particle is composed of ferro-

magnetic particles and abrasive particles, according to Wiede- 4 Material removal and surface roughness
mann’s law [23], its susceptibility can be expressed as
The mechanism of material removal is based on microcut-
ting [17]. The MAF process removes material in the form of
χr = αχr ferr + (1 − α)χrabr , (13) tiny chips. The volume of these chips is equal to the volume of
grooves produced on the workpiece surface during the MAF pro-
where α is the volume fraction of ferromagnetic particles, and cess. Magnetic and mechanical energies are utilized in the MAF
χr ferr and χrabr are the susceptibilities of ferromagnetic material process. A magnetic abrasive flexible brush is formed due to the
and abrasive particles, respectively. effect of a magnetic field in the working gap. The rotation of the
From the known values of the intensity of the magnetic field north pole makes the MAFB rotate. It generates the tangential
and its derivatives at the Gauss points, the magnetic forces are force (Ft ) on the cutting edges of MAPs. The magnetic energy
calculated first at the Gauss points of the elements in contact develops the normal (Fz ) and radial (Fr ) magnetic forces on the
with the workpiece surface. Then, they are extrapolated on the MAPs. The normal magnetic force creates compressive reaction
points indicated by
in Fig. 4. Further, extrapolation or interpo- on the surface of the workpiece. This compressive force is re-
lation at the contact surface is necessary to determine magnetic sponsible for the penetration of cutting edges into the workpiece.
forces at the actual contact points between the MAPs and the The resultant of radial magnetic force and mechanical tangential
work surface. For this, the locations of the actual contact points force on the cutting edges of MAPs removes (shears out) the ma-
of MAPs need to be determined. This is discussed in the follow- terial from the workpiece along the circular paths. It is assumed
ing paragraph. that the total mechanical power available at the end of the spin-
In the present work, the tool (N-pole) is cylindrical and the dle (N pole) is utilized in finishing the workpiece. The tangential
workpiece surface is flat. Therefore, the cross sectional area of force, which removes material from the workpiece, is assumed to
the brush is circular. It is assumed that MAPs are closely packed be uniformly distributed. In the present case, it is assumed that
in the gap and rotate in a particular track as shown in Fig. 1b. The all the cutting edges work simultaneously. Therefore, finishing
location of the contact point of a MAP depends on the radius of power is the summation of the product of the tangential force and
the track, in which the particle rotates. The location of a contact cutting speed on each cutting edge.
point is, therefore, given by Total number of cutting edges (nct(tr) ) in the tr-th track is
given by
Dmap 2πR(tr) na
R(tr) = (tr − 1) Dmap + , (14) nct(tr) = , (16)
2 Dmap

where R(tr) is radius of the tr-th track and Dmap is diameter of where na is the number of active cutting edges of the MAPs (as-
the magnetic abrasive particle (assumed to be the same for each sumed to be 1 in this particular case). Cutting speed of a particle
MAP). The total number of tracks (nt ) in a brush is given by in the tr-th track (Vc(tr) ) is given by
Vc(tr) = 2πR(tr) Nrs , (17)
radius of magnetic abrasive brush
nt = , (15) where Nrs is the rotational speed of the magnetic pole. Thus, the
Dmap
finishing power is given by
The radius of the magnetic abrasive brush at the workpiece in- nt
terface is equal to the length DC of Fig. 1 and Dmap is shown in P= Ftedge nct(tr) Vc(tr) , (18)
Fig. 2 (or width of a track in Fig. 1b). tr=1
482

where Ftedge is the tangential force on the cutting edge. Substi- (1) Freq = Fcedge
tuting the value of nct(tr) from Eq. 16 and Vc(tr) from Eq. 17 in It is the equilibrium condition. It indicates starting of the
Eq. 18, the tangential force is given by finishing operation.
(2) Freq < Fcedge
P
Ftedge = . (19) Under this condition, material is removed.

nt 4π 2 R2(tr) Nrs na (3) Freq > Fcedge
Dmap
tr=1
It shows no cutting condition. Under this condition, the depth
The cutting force (Fc ) is resultant of tangential force (Ft ) and of penetration of the cutting edge gets adjusted by rotation of
radial force (Fr ): the magnetic abrasive particle until the cutting force required be-
 comes equal to the cutting force available Fcedge , keeping the
Fc = F2t + F2r . (20) cross sectional area of indentation (Ai ) the same as shown in
Fig. 6:
Total tangential force and radial force on the MAFB for the
specified machining conditions, are found to be 97.6368 N and F̄req = Fcedge , (22)
2.5285 N, respectively. Therefore the resultant cutting force (Fc )
is given by where F̄req is the modified required cutting force:
 
Fc = 97.63682 + 2.53852 N = 97.6695 N . F̄req = τs Ap , (23)

Since the overall effect of the radial force is negligible, the cut- where Ap is the modified projected area of penetration. Substitut-
ting force on the edge (Fcedge ) can be taken as equal to the ing the value of F̄req from Eq. 22 in Eq. 23, we get
tangential force (Ftedge ).
The force required (Freq ) to remove material from the work- Fcedge
Ap = . (24)
piece depends on the shear strength (τs ) of the workpiece mate- τs
rial and projected area of the penetration (Ap in Fig. 5c). Thus,
Assumptions
For simplification of computational work, the following assump-
Freq = τs Ap . (21)
tions are made in developing the model for material removal and
In removing the material in the MAF process, one of the three surface roughness:
situations may occur: 1. The surface of the workpiece has a uniform triangular profile.

Fig. 5a–c. Schematic diagram of

abrasive grain showing: a Depth
of penetration b Projected area
of indentation c Projected area of
penetration

Fig. 6. Rotation of a magnetic abrasive particle.

hs = depth of penetration before rotation, hs =
depth of penetration after rotation, Ap = pro-
jected area of penetration before rotation, Ap =
projected area of penetration after rotation
 483

2. There is no relative motion between the MAPs because these

are arranged closely along the lines of the magnetic force in
the form of a semi-rigid brush. This results in insignificant
porosity in the brush between the particles.
3. The shape of MAPs is approximated to be spherical and of
the same size. Only one cutting edge of the MAP is assumed
to be in contact with the the workpiece. This cutting edge of
a MAP in contact with the workpiece removes material in its
own track. Each cutting edge of a particular track removes
the same amount of material. Fig. 7. Schematic diagram of workpiece surface profile
4. To render the problem mathematically tractable, it is neces-
sary to assume that the working gap remains constant during
finishing.
5. The size of the magnetic abrasive particle is larger than the
spacing between the two consecutive peaks of surface un-
evenness because the MAF process is applied for finishing
the semi finished components, having the spacing of the
order of a few micrometer.
6. There is no slip between the rotating spindle (N pole) and
the MAFB due to the force of magnetic attraction between Fig. 8. Shape of the surface profile
them.

where, ds is the diameter of the abrasive particle. Substituting

4.1 Material removal model the value of “r” from Eq. 27, we get
MAF process removes a very small amount of material by pene- 
tration and rotation of MAPs on the work surface to be finished. ds ds 2 Fzedge
hs = − − . (29)
The volume of material removed by an abrasive grain is equal 2 2 Hm π
to the product of sheared area and length of finishing of the
workpiece (ie., the length of contact of the abrasive grain on the The projected area (Ap ) of penetration subjected to the shearing
workpiece surface). force during the MAF (shaded area of Fig. 5c) as obtained from
The normal magnetic force on the abrasive particles causes the geometry of Fig. 5a is
penetration into the workpiece. The normal force acting on a cut-
 
(ds)2 −1 2hs ds
ting edge (Fzedge ) of the magnetic abrasive particle is given as Ap = cos 1− − − hs hs(ds − hs) .
4 ds 2
Fz (30)
Fzedge = , (25)
na
Figure 7 shows a schematic diagram of the assumed work-
where Fz is the normal magnetic force acting on the MAP. The piece surface profile along with the path (or track) of rotation of
depth of penetration is calculated by equating the applied force a MAP, which is circular. The cutting edges of MAPs in contact
to the reaction of the workpiece [24]: with the workpiece remove the material along the circular path.
However, the brush is provided intermittent feed after comple-
Fzedge = Hm , ∆A (26) tion of each revolution of the brush/pole. Therefore, the location
of each MAP varies with respect to the surface of the workpiece
where Hm is hardness of the workpiece material and ∆A is the after each revolution. To simplify the computation of material re-
projected area of indentation (shaded area of Fig. 5b). The pro- moval and surface roughness, the workpiece surface is divided
jected area of indentation is circular. Let, r be the radius of this in small square cells. The cell is specified by the Cartesian co-
circular area. Then, r is given by ordinates of its center. It is assumed that material in the cell
 is removed by the MAP whose track passes through the cell.
Fzedge The procedure to calculate the material removal is as follows.
r= . (27)
Hm π The flow chart for computation of material removal and surface
roughness is given in Appendix-II.
From the geometry of Fig. 5a, the depth of penetration (hs) is Let the initial surface roughness of the workpiece be R0max .
obtained as The volume of material removed by a MAP passing through the
 cell (i,j) in the nth revolution is
ds ds 2
hs = − − r2 , (28) (n) (n)
2 2 ∆Ved(i,j)) = Ap(tr) lt(i,j) , (31)
484

To compute the total material removed in each cycle, the ma-

terial removed in each cell can be summed up. The total material
removed in the nth rotation is given by
∆V(n) = ∆V(n)
(i,j) . (39)

4.2 Surface roughness model

The MAF process improves surface quality by reducing the

unevenness of the surface profile. The normal magnetic force
applies machining pressure on the workpiece surface through
Fig. 9. a Schematic diagram of workpiece with cells, b Exaggerated view of
a typical cell
MAPs, resulting in the penetration of cutting edges of abrasive
grains in the workpiece. Due to rotation of MAPs, grooves are
formed on the workpiece surface which decide the surface pro-
where Ap(tr) is the sheared area in track “tr” as discussed earlier file after the MAF. Surface roughness is determined on the basis
(n)
(Eq. 30) and lt(i,j) is the length of contact of the MAP with the of the surface profile achieved by equating the volume of the ma-
surface of the cell (i,j) in the nth revolution. The length of con- terial removed to the volume of groove produced. It is assumed
tact of the MAP in cell (i,j) is the product of the contact length of that the surface of the workpiece has uniform triangular profiles
(n)
a single profile l(i,j) and the number of profiles (nf ) in the cell (Fig. 8) without statistical distribution. Initial surface roughness
(i,j). is R0max and after completion of “n” revolutions of the MAFB,
surface roughness in cell (i,j) becomes R(n) max(i,j) . Volume of the
l(n) (n)
t(i,j) = l(i,j) nf . (32) groove produced in a profile is equal to the product of the cross
From Fig. 8, the contact length (BC) of a single profile can be sectional area of the removed portion of the surface profile and
obtained as average width of cut. The average width of cut is equal to the
radius of the projected area of penetration of the abrasive grain
l(n) (n)
(i,j) = 2hs(tr) tan θw , (33) (Sect. 4.1), in that profile. The cross sectional area of the re-
moved portion of surface profile varies in each revolution as
(n)
where hs(tr) is the depth of penetration of the MAP in the tr-th shown in Fig. 10.
track and 2θw is the mean angle of surface asperity. The mean From the geometry of the profile of a workpiece surface as
angle of surface asperity depends upon the initial surface rough- shown in Fig. 10, AD, BC and the volume of the groove pro-
(1)
ness and manufacturing process of the work surface. Figure 9a duced in one profile in a cell (i,j) by a cutting edge ∆Vged(i,j) pr
shows the work surface divided into various cells. An exagger- are given by
ated view of a typical cell is shown in Fig. 9b. Here, it is also
assumed that each grain cuts the peaks normal to its lay direction. AD = R0max − R(1) max(i,j) (40)
 
From Fig. 9b, we get (1)
BC = 2 R0max − Rmax(i,j) tan θw (41)
lc   1
nf = , (34) ∆Vg(1)
lb ed(i,j) pr = 2 × AD × BC × r(tr) (42)

where lc is the length of cell and lb is the base length of a single where r(tr) is average width of cut produced by a cutting edge in
profile. From Fig. 8, lb is given by track “tr”. Substituting Eqs. 40 and 41 into Eq. 42, we get
   
lb = 2R0max tan θw . (35) ∆Vg(1) 0 (i)
ed(i,j) pr = Rmax − Rmax(i,j)
2
tan θw r(tr) . (43)

Substituting Eqs. 32–35 into Eq. 31, we get Total volume of the material removed in a cell (i,j) by a cutting
(1)
edge ∆Vged(i,j) is
(n) Ap(tr) hs(n)
(tr) lc  
∆Ved(i,j) = . (36)
R0max ∆Vg(1) (1)
ed(i,j) = ∆Vged(i,j) pr nf . (44)
The total volume of material removed in a cell (i,j) in the nth The value of nf is obtained from Eqs. 34 and 35:
revolution is
lc
nf = . (45)
∆V(n)
(i,j) = ∆V(n)
ed(i,j) nct(tr) . (37) 2R0max tan θw
(n)
Substituting the value of ∆Ved(i,j) from Eq. 36 and nct(tr) from Substituting the value of ∆Vg(1) ed(i,j) pr from Eq. 43 and nf from
Eq. 16 in Eq. 37 Eq. 45 into Eq. 44,
 2
(0) (1)
(n)
2Ap(tr) hs(tr) lc πR(tr) na Rmax − Rmax(i,j) r(tr) lc
(n)
∆V(i,j) = . (38) ∆Vg(1)
ed(i,j) = (46)
R0max Dmap 2R(0)
max
 485

Fig. 10. Surface profile after 3rd revolution

The total volume of material removed in cell (i,j) by “n(tr) ” par- (FEMS-MAFP) has been developed in the Fortran 90 program-
ticles is ming language. Results obtained from the code have been pre-
sented as follows.
∆Vg(1) (1)
(i,j) = ∆Vged(i,j) nct(tr) . (47)

(1)
Substituting the value of ∆Vged(i,j) from Eq. 46 and nct(tr) from 6 Results and discussion
Eq. 16 into Eq. 47:
The magnetic forces obtained from the numerical simulation
 2 as discussed above, are used for the computation of surface
(0) (1)
Rmax − Rmax(i,j) r(tr) lc πR(tr) na roughness according to the surface roughness model discussed in
∆Vg(1)
(i,j) = . (48)
R(0)
max Dmap
Sect. 4.2. Then this computed surface roughness is compared with
the experimental results of Shinmura et al. [12] to check the valid-
The volume of material removed in cell (i,j) by “n(tr) ” particles ity of the mathematical model and surface roughness model. The
(as discussed in Sect. 4.1) is equated to the volume of groove machining parameters and material properties employed in the
produced in that cell in the first revolution: simulation are taken from reference [12] and are given in Table 1.
Before the calculation of magnetic forces, the convergence
∆V(1) (1)
(i,j) = ∆Vg(i,j) . (49) of mesh is studied. This study has revealed that mesh of 1152
isoparametric eight nodded quadrilateral elements with 3577
(1)
Putting the value of ∆Vg(i,j) from Eq. 48 into Eq. 49, we get nodes are adequate for this problem. The distribution of the
normal magnetic force on the workpiece surface along radial di-
⎛ ⎞1 rection is shown in Fig. 11. It is found that the magnitude of
∆V(1) 0 2
(1) ⎝ (i,j) Rmax Dmap ⎠ ,
Rmax(i,j) = R(0)
max − (50)
πr(tr) lc R(tr) na
Table 1. Machining conditions and material properties
where R(1)
max(i,j) is the surface roughness obtained after one revo- Magnetic flux density (B) 0.8 T
lution of the N-pole. Following the same procedure (Appendix- Working gap (lg ) 4 mm
III), it can be shown that after the nth revolution, surface rough- Diameter of N pole (Dnp ) 34 mm
Number of turns of coil to electromagnet (N) 3000
ness in cell (i,j) can be expressed as
Size of workpiece 150 mm × 150 mm
×1.2 mm
⎛ ⎛ ⎞⎞ 1 Workpiece material SUS304 stainless steel
 2 ∆V
(n) 0
R D
2
(n−1) max map Hardness of workpiece (Hm ) 5.5 GPa
= R0max − ⎝ R0max − Rmax(i,j) +⎝ ⎠⎠ .
(n) (i,j)
Rmax(i,j)
πr(tr) lc R(tr) na Shear strength of workpiece (τs ) 13.28 GPa
Type of abrasive grain Al2 O3
(51) Relative permeability of abrasive grain (µrabr ) 0.999996
Permeability of free space (µ0 ) 4π × 10−7 H/m
Volume fraction of iron particle (α) 0.70
RPM of rotating pole (Nrs ) 196
Power input (P) 1 kW
5 Computer implementation Diameter of abrasive grain (ds) 5 µm
Diameter of magnetic abrasive particle (Dmap ) 100 µm
Using the expressions given in Sects. 2–4, a code namely the fi- Mean semi angle of surface asperity (θw ) 80◦
Initial surface roughness (R0max ) 1.5 µm
nite element simulation for magnetic abrasive finishing process
486

Fig. 11. Distribution of normal magnetic force for machining conditions Fig. 13. Variation of surface roughness at various y-levels in case of the tool
given in Table 1 movement along x-axis after 126 revolutions

Fig. 12. Schematic diagram of movement of the tool in x direction

Fig. 14. Variation of the product of normal magnetic force and contact time
of the tool at various y-levels in case of the tool movement along x-axis. No.
of revolutions of the tool = 126
the normal magnetic force is relatively higher near the edge of
a magnetic pole in comparison to other parts of the tool. It is due
to a larger variation in gradients of the intensity of magnetic field roughness is uniform in a certain region of the workpiece. Note
near the edge. Such behavior near the edges is observed experi- that the normal magnetic force is non-uniform as shown in
mentally as well (Singh personal communication, 2004). Fig. 11. However, it is the combined effect of the magnitude of
The size of a cell is taken as 5 mm × 5 mm. Numerical ex- the normal magnetic force and the contact time that is responsi-
periments were carried out by providing intermittent motion to ble for surface roughness. Figure 14 shows the variation of the
the tool in the following two fashions. product of the magnitude of the normal magnetic force and the
contact time along the x-axis at different y levels. It shows that,
(i) Tool movement in the x direction and shift in y direction
at various y levels, the product is constant in a certain region of
(ii) Tool movement in the y direction and shift in x direction
the workpiece surface. Further, one is concave in nature while
another is convex in nature. This explains the uniform nature of
(i) Tool movement in the x direction and shift in y direction surface roughness in certain regions of the workpiece surface.
The tool moves from position A to B, then it shifts to location The improvement in the surface roughness is less, when “x” is
C before it starts moving towards D (Fig. 12). The distance BC either less than 25 mm or more than 125 mm. The same thing
is taken to be equal to the cell length. This movement continues is observed, when “y” is either 2.5 mm or 72.5 mm. These parts
till the tool reaches the final position (position 2 of Fig. 12 – one lie in the fringing zones of the brush at the ends of the passes.
forward pass). The path of the tool center is indicated by a dotted In the end of fringing zones, the magnetic force almost reaches
line in Fig. 12. From position 2, the tool returns back to position the zero value, hence this can be called the fringing zone effect.
1 following the reverse path (– one backward pass). This effect can be reduced by using a dummy workpiece in the
Figure 13 shows the change in the surface roughness values fringing zone region. The surface quality of the workpiece at any
along the x-axis at different y levels, after completing one for- time is represented as the average of the surface roughness of
ward pass. It is observed that at various y levels, the surface the cells excluding the cells lying in the region of the fringing
 487

Table 2. Comparison of surface roughness with the experimental results [12]

S. Time Surface roughness (µm) % Error

no. (min)
Experi- Theoretical
mental
Tool move- Tool move- Tool move- Tool move-
ment in ment in ment in ment in
x direction y direction x direction y direction

1. 1 0.55 0.87 0.87 36.78 36.78

2. 2 0.32 0.60 0.61 46.66 47.54
3. 3 0.23 0.40 0.40 42.50 42.50
4. 4 0.22 0.23 0.24 4.34 8.33
Fig. 16. Schematic diagram of movement of the tool in y-direction. Move-
ment along Y -axis

Figure 15 and Table 2 show that the simulated values are

higher than experimental results, but the trend is the same as that
of the experimental results. Initially the discrepancy is higher but
it reduces as finishing time increases. This discrepancy may be
due to either computational limitations and/or some of the as-
sumptions of the model.

7 Conclusions
In the present investigation, modeling and numerical simulation
of surface roughness in the MAF process have been performed.
This helps in understanding the mechanism of forces applied and
Fig. 15. Comparison of surface roughness with the experimental results [12] the material removal. The following conclusions are drawn on
in case of the tool movement along x-axis the basis of the results discussed.
The magnitude of the normal magnetic force is relatively
higher near the edge of the magnetic pole due to the edge effect.
zone effects. Table 2 shows a comparison of the surface quality
The surface roughness of the workpiece can be found in almost
computed as above with experimental results [12] at four discrete the same way by providing the intermittent motion to the tool ei-
times. Figure 15 shows the comparison when the time is varied
ther along the x-axis or y-axis. These simulated results compare
continuously upto 4 min. favorably well with the experimental results after finishing for
a period of 4 min.
(ii) Tool movement in the y direction and shift in x direction
In this motion, the tool first moves intermittently along the Nomenclature
positive y-axis. Starting from position 1 (Fig. 16), it moves a dis-
tance equal to the cell breadth after each revolution. Upon reach- Ap Projected area of the penetration (m2 )
ing the opposite side, the tool shifts along the negative x-axis B Magnetic flux intensity (T)
by a distance equal to the cell length and again it resumes the D Electric flux density (C/m2 )
intermittent motion along the y-axis (this time along the nega- Dmap Diameter of a magnetic abrasive particle (m)
tive y-axis). This sequence is repeated till the tool reaches the Dnp Diameter of the N-pole (m)
final position (position 2 of Fig. 16). When the tool reaches pos- ds Diameter of an abrasive particle (m)
ition 2, it completes 126 revolutions and takes 0.64 min. From E Electric field intensity (V/m)
position 2, the tool returns back to the position 1 following the Em Magnetic potential energy (J)
reverse path. In this way, intermittent motion is provided upto Fcedge Cutting force available on a cutting edge (N)
4 min. It is completed in 784 revolutions of the tool. Fzedge Normal force acting on a cutting edge (N)
The change in the surface roughness is also estimated along Ftedge Tangential force on a cutting edge (N)
the x direction after completing 126 revolutions of the tool, at Fredge Radial magnetic force on a cutting edge (N)
different y levels. It gives the same results as seen in Fig. 13. The H Magnetic field intensity (A/m)
surface quality of the workpiece is compared with the experi- Hm Hardness of workpiece material (N/m2 )
mental results [12] in Table 2 at four discrete times. hs Depth of penetration of abrasive particle (m)
488

I Input current to electromagnet (A) 6. Shinmura T, Takazawa K, Hatano E, Matsunaga T (1990) Study on
J Current density (A/m2 ) magnetic abrasive finishing. Ann CIRP 39(1):325–328
7. Shinmura T, Takazawa K, Hatano E, Aizawa T (1985) Study on mag-
[k]e Elemental coefficient matrix netic abrasive process- process principles and finishing possibility. Bull
lg Working gap (m) Japan Soc Precis Eng 19(1):54–55
lb Base length of single profile (m) 8. Shinmura T, Takazawa K, Hatano E, Aizawa T (1984) Study on mag-
lc Length of cell (m) netic abrasive process- finishing characteristics. Bull Japan Soc Precis
Eng 18(4):347–348
M Magnetization of magnetic abrasive particles (A/m) 9. Shinmura T, Takazawa K, Hatano E (1987) Study on magnetic abrasive
N No. of turns in the electromagnet – effects of various types of magnetic abrasives on finishing character-
Nrs Rotational speed of the tool (RPM) istics. Bull Japan Soc Precis Eng 21(2):139–141
nrev Total number of revolution 10. Fox M, Agrawal K, Shinmura T, Komanduri R (1994) Magnetic abra-
sive finishing of rollers. Ann CIRP 43(1):181–184
nt Total number of tracks 11. Shinmura T, Aizawa T (1989) Study on magnetic abrasive finishing
na Number of active cutting edges on a magnetic abrasive process-development of plane finishing apparatus using a stationary
particle type electromagnet. Bull Japan Soc Precis Eng 23(3):236–239
12. Shinmura T, Takazawa K, Hatano E (1985) Study on magnetic abra-
nct(tr) Total number of cutting edges on MAPs in track tr-th sive process-application to plane finishing. Bull Japan Soc Precis Eng
nf Number of profiles in a cell 19(4):289–291
R(tr) Radius of the tr-th track (m) 13. Hou ZB, Komanduri R (1998) Magnetic field assisted finishing of
R0max Initial surface roughness of the workpiece (µm) ceramics-on the thermal aspect of magnetic abrasive finishing of ce-
ramic rollers. Trans ASME J Eng Ind 20:660–667
r Radius of projected area of indentation (m) 14. Waigaonkar SD, Navale LG, Kajale SR (2000) Optimization of mag-
t Time (s) netic abrasive finishing process. Proc 19th AIMTDR Conference, Chan-
∆A Projected area of indentation (m2 ) nai, India, pp 147–152
α Volume fraction of ferromagnetic particles 15. Yamguchi H, Shinmura T (2000) Study on internal magnetic abrasive
finishing using a pole rotation system-discussion of characteristics abra-
φ Magnetic potential (AT) sive behaviour. Precis Eng 24:237–244
µ0 Permeability of free space (H/m) 16. Kremen GZ, Elsayed EA, Ribeiro JL (1994) Machining time estimation
µr Relative permeability of magnetic abrasive particle for magnetic abrasive processes. Int J Prod Res 32(12):2817–1825
θw Semi-mean angle of surface asperity 17. Kim J, Choi M (1995) Simulation for the prediction of surface-accuracy
in magnetic abrasive machining. J Mater Process Technol 53:630–642
µrabr Relative permeability of abrasive particle. 18. Kremen GZ, Elsayed EA, Rafalorich VI (1996) Mechanism of material
χr Susceptibility of magnetic abrasive particle removal in magnetic abrasive process and the accuracy of machining.
χr ferr Susceptibility of ferromagnetic material Int J Prod Res 34(9):2629–2638
19. Jefimenko OD (1966) Electricity and magnetism. Meridith, New York
χrabr Susceptibility of abrasive particle 20. Reddy JN (1993) An introduction to the finite element method, 2nd ed.
τs Shear strength of workpiece material (N/m2 ) McGraw-Hill, New Delhi
ρ Density of workpiece material (kg/m3 ) 21. Griffiths DV, Smith IM (1991) Numerical methods for engineers.
i, j Coordinates x and y Blackwell Scientific, Oxford
22. Stradling AW (1993) The physics of open-gradient dry magnetic sepa-
r, θ, z Coordinates r, θ and z ration. Int J Miner Process 39:1–18
Superscripts 23. Mulay LN (1963) Magnetic susceptibility. Wiley, New York
e Element 24. Rabinowicz E, Dunn LA, Russel PG (1961) A study of abrasive wear
Transpose under three-body conditions. Wear 4:345–355
T
25. Edminister JA (1986) Theory and problems of electromagnetics.
McGraw-Hill, New York
Acknowledgement The authors acknowledge the suggestions of Prof.
Deepak Gupta of the Materials and Metallurgical Engineering Department
and Prof. M. Sachchidanand of the Electrical Engineering Department, In-
dian Institute of Technology, Kanpur during this work. The financial support
for this work from the Department of Science and Technology, Government
of India, through project no. SR/S3/RM/25/2003, is also acknowledged.
Appendix – I
Relationship between µr ferr and H, χr ferr and H
The graphical relationship between the relative permeability
References µr ferr and field strength H of the ferromagnetic material [25] is
1. Jain VK (2002) Advanced machining processes. Allied Publishers, shown in Fig. 17. It shows a highly non-linear behavior. There-
Delhi fore, to find out the mathematical expression between µr ferr and
2. Jain VK, Kumar P, Behra PK, Jayswal SC (2001) Effect of working H with minimum possible error, the graphical representation is
gap and cicumferential speed on the performance of magnetic abrasive divided into three regions as indicated in Fig. 17. The expres-
finishing process. Wear 250:384–390
3. Yamguchi H, Shinmura T (1999) Study of surface modification result- sions for these three regions are obtained as given below. In the
ing from an internal magnetic abrasive finishing. Wear 225:246–255 following expressions, H is the magnitude of the vector H.
4. Chang GW, Yan BH, Hsu RT (2002) Study on cylindrical magnetic (a) For region I:
abrasive finishing using unbonded magnetic abrasives. Int J Mach Tools
Manuf 42:575–583
5. Shinmura T, Takazawa K, Hatano E (1986) Study on magnetic abrasive µr ferr = 743.4520 + 16.1679 H + 0.7794 H2 (52)
finishing-effects of machining fluid on finishing characteristics. Bull
Japan Soc Precis Eng 20(1):52–54 χr ferr = 742.4520 + 16.1679 H + 0.7794 H 2
(53)
 489

Fig. 17. Relationship between relative permeability and intensity of mag-

netic field

(b) For region II:

µr ferr = − 247748.4573 + 9689.6642 H

− 123.2216 H2 + 0.5224 H3 (54)
χr ferr = − 247749.4573 + 9689.6642 H
− 123.2216 H2 + 0.5224 H3 (55)

(c) For region III:

µr ferr = 8231.1280 − 25.6849 H + 0.0283 H2 (56)

χr ferr = 8230.1280 − 25.6849 H + 0.0283 H 2
(57)

After region III, the saturation region starts. In this region, the
susceptibility and relative permeability remain constant. These
are found by substituting the value of H at the saturation point
in the expressions of region III. In the present case, saturation
occurs at the value of H = 400 A/m.

Appendix – II
Flow chart for determination of material removal and Appendix – III
surface roughness Computation of surface roughness in 2nd and 3rd
revolutions

(i) 2nd revolution:

Cross sectional area of material (BCFE) removed in the 2nd
revolution is obtained from the geometry of the profile of the
workpiece surface as shown in Fig. 10.

Cross sectional area of BCFE =

    
(2) (1)
R0max − Rmax(i,j) 2 − R0max − Rmax(i,j) 2 tan θw (58)
490

The volume of material removed in cell (i,j) by a cutting edge in workpiece surface as shown in Fig. 10.
the 2nd revolution is given by
Cross sectional area of EFIH =
(2)
 2  2 
∆Vged(i,j) = Cross sectional area of BCFE × r(tr) × nf (59) (3) (2)
R0max − Rmax(i,j) − R0max − Rmax(i,j) tan θw (63)
Substituting the value of the cross sectional area of BCFE from
Eq. 58 and nf from Eq. 45 in Eq. 59, The volume of material removed in cell (i,j) by a cutting edge in
the 3rd revolution is given by
 2  2
R0max − R(2) 0 (1)
max(i,j) − Rmax − Rmax(i,j) r(tr) lc ∆Vg(3)
ed(i,j) = Cross sectional area of EFIH × r(tr) × nf (64)
(2)
∆Vged(i,j) =
2R0max Substituting the value of the cross sectional area of EFIH from
(60) Eq. 63, nf from Eq. 45 in Eq. 64,
 2  2
The total volume of material removed in cell (i,j) by “n(tr) ” par- (3) (2)
R0max − Rmax(i,j) − R0max − Rmax(i,j) r(tr) lc
ticles in the 2nd revolution is given by (3)
∆Vged(i,j) =
2R0max
(2)
∆Vg(i,j) = (65)
 2  2 The total volume of the groove produced in cell (i,j) by “n(tr) ”
(2) (1)
R0max − Rmax(i,j) − R0max − Rmax(i,j) r(tr) lc πR(tr) na
particles in the 3rd revolution is given by
R0max Dmap
∆Vg(3)
(i,j) =
(61)  2  2
(3) (2)
R0max − Rmax(i,j) − R0max − Rmax(i,j) r(tr) lc πR(tr) na
(2)
Equating the volume of material removed ∆V(i,j) to the groove
(2)
produced ∆Vged(i,j) in cell (i,j) in 2nd revolution, we get R0max Dmap
(66)
(2)
=
Rmax (i,j) Equating the volume of material removed ∆V(3)
(i,j) to the groove
⎛ ⎛ ⎞⎞ 1 (3)
 2 ∆V
(2) 0
R D
2 produced ∆Vged(i,j) in cell (i,j) 3rd revolution, we get
(1) max map
⎝
Rmax − Rmax − Rmax (i,j) +
0 0 ⎝ (i,j) ⎠⎠ (62)
πr(tr) lc R(tr) na (3)
Rmax (i,j) =
⎛ ⎛ ⎞⎞ 1
(ii) 3rd revolution:  2 ∆V
(3) 0
R D map
2
(2) max
Rmax − ⎝ Rmax − Rmax (i,j) + ⎝ ⎠⎠
0 0 (i,j)
The cross sectional area of material (EFIH) removed in the (67)
πr(tr) lc R(tr) na
3rd revolution is obtained from geometry of the profile of the