Sådhanå (2023)48:73 Ó Indian Academy of Sciences

https://doi.org/10.1007/s12046-023-02106-2
Sadhana(0123456789().,-volV)FT3](012345
6789().,-volV)

Experimental investigation and optimization of process parameters

of hybrid Al/SiC/B4C–MMCs finished by MAFM process using RSM
modeling with supervised machine learning algorithm
GAGANDEEP CHAWLA* and VINOD KUMAR MITTAL

Department of Mechanical Engineering, National Institute of Technology, Kurukshetra, Haryana 136119, India
e-mail: [email protected]; [email protected]

MS received 28 April 2022; revised 16 January 2023; accepted 23 January 2023

Abstract. Magnetic abrasive flow machining (MAFM) has an astonishing capability for improving the surface
quality of advanced materials viz. composites, ceramics and hard alloys. The surface quality and finishing have
major dependency on various process parameters of the focused surface while finishing through MAFM process.
The MAFM process procures several applications in medical fields (Knee joint implant and surgical instru-
ments), automotive, aerospace and tool manufacturing industries. The newness of current study is in the
development of an MAFM setup for machining of SiC/B4C hybrid MMCs with aluminium-6063 as a base
material and the measurement of parametric effects on the process performance. The efforts made have led
towards the modeling of two responses viz. MRR and DRa with response surface methodology. Box-Behnken
design approach has been adopted for analyzing six MAFM factors and a total of 54 trials have been conducted
for finding their influence on MRR and DRa. SEM and EDX have been applied to examine the surface
topography. The significance of various process parameters has been analyzed by using ANOVA. The outcomes
showed that Ep (extrusion pressure), M (mesh size), N (number of cycles), and Mf (magnetic flux density) are the
most essential factors. The optimal solutions have been attained by applying a multi-objective optimization
‘desirability’ function using statistical and supervised machine learning algorithms which led to the parametric
machine learning algorithms reflection for surmising the efficiency of MAFM process. A fine consonance has
been obtained among the predicted and actual values. The graphical abstract of the current research work is
shown below.

*For correspondence
73 Page 2 of 35 Sådhanå (2023)48:73

Graphical abstract

Keywords. MAFM; Al/SiC/ B4C-MMCs; MRR; DRa; SEM; EDX; ANOVA; machine learning.

1. Introduction methods for finishing. Extrude Hone Corporation of USA

developed one of the most extensively used finishing
In the modern era, finishing processes are extensively techniques i.e. Abrasive Flow Machining (AFM) in 1960s
needed in manufacturing industries to enhance the surface for both manufacturing industries and research academics.
quality of several geometric shaped peripherals. At the As a replacement to the tedious and labor-intensive con-
mating surfaces, stress concentrations are produced by ventional finishing processes like polishing, honing, buffing
asymmetrical surfaces which results in cracks on the sur- and grinding etc., AFM process has a potential as a fin-
faces under numerous machining conditions. There is a ishing solution [4, 5]. The AFM technique can be efficiently
significant role of surface quality in view of surface employed to finish burrs, edges and recast layers of parts
roughness, as it helps in determining the fatigue lifespan of with different geometric shapes and internal holes of hard
the component [1–3]. From experimental observations, it material components which are unreachable by conven-
has been reported that the partially machined profile has a tional finishing methods.
shorter fatigue lifetime than the finished profile. Therefore, Depending on the applications, the AFM machines are
to improve the component’s life span, polishing or finishing classified as: (1) One-way AFM; (2) Two-way AFM and (3)
of the component plays a very important role. Hence, from Orbital AFM. The analysis of AFM process mainly sta-
the last few decades, researchers found the finishing pro- tes two-way AFM, because of the simplicity for analyzing
cesses as an active research area and invented several the physics. The major elements of the AFM process are the
Sådhanå (2023)48:73 Page 3 of 35 73

with experimental results

parameters affecting MR
abrasive media, machine and tooling. Abrasion emerges at

flow speed of media [5]

MRR and surface finish

and surface finish were

abrasive concentration,
neural simulation were
The influence of AFM
process parameters on

found to be consistent
propagation algorithm

number of cycles and

model. The results of

trailed by mesh size,

the places where the media flows from the extremely

was investigated by
Findings

employing a back
restrictive passage. There are two hydraulic actuators

The dominant
opposed vertically, that extrude the media to and fro over
the passage generated through the tooling and the work-
piece. AFM has presented its promising performance to

[3]
achieve superior finishing quality of components over
decades and is appropriate for processing the surfaces
which are inaccessible by hand and multifaceted external

MRR and surface finish

and internal geometries [6–8]. Therefore, AFM has been

parameters
Responses
widely used in manufacturing industries like medical

MR and surface
(prosthetic knee joint implantation), automotive, precision

roughness
dies and moulds, aerospace and electronics, gears, turbine
blades, etc. Continuous developments are taking place to
overcome the limits of AFM process viz. low material
removal rate and low finishing rate. To achieve better
surface texture, integrity and performance, several new

Flow speed of media, percentage

concentration by weight, media

abrasive concentration, number

variations of AFM process like centrifugal force-assisted

Process parameters

flow speed, mesh size and

AFM, electrochemical magnetic abrasive finishing, mag-

of cycles and mesh size

netorheological honing process, ultrasonic flow polishing,

Percentage abrasive
electrochemical polishing and magnetic abrasive finishing

number of cycles
have been introduced by modifying the existent AFM
process and machine configuration as discussed in the next
section.

2. State of the art review

Abrasives/media

Silicon carbide (SiC)

Silicon carbide (SiC)

Abrasive

Abrasive
Machined surface texture
back propagation neural
technique/model/

AFM process has been

Experimental

was studied by SEM

modeled through the
method
design/

technique
networks

Brass and Aluminum

Material
Aluminium
Jain et al (1999)

Jain et al (2000)
Authors/year
 continued
73

Experimental design/
Authors/year Material technique/model/method Abrasives/media Process parameters Responses parameters Findings

Sehijpal et al (2002) Brass Response surface Abrasive Media flow rate, number of MR and percentage MR and DRa were
methodology (RSM) Brown super emery cycles, extrusion pressure, media improvement in surface affected by magnetic
having ferromagnetic viscosity, magnetic flux density, roughness (DRa) field as both increased in
Page 4 of 35

constituents (40%), grain size, workpiece material, the presence of magnetic

Al2O3 (45%), Si2O3 reduction ratio and abrasive field. To eradicate the
(15%) concentration same amount of material
from the workpiece
Media surface, the number of
Abrasive grains, cycles required was less
hydrocarbon gel and by applying the magnetic
silicon based polymer field [9]

Jha et al (2007) Stainless steel One variable at a time Abrasive Number of finishing cycles and Change in surface The Magnetorheological
approach Silicon carbide (SiC) and hydraulic extrusion pressure roughness abrasive flow finishing
carbonyl iron powder (MRAFF) process was
Media used for complicated
Abrasives spread in a geometries and
viscoplastic base of performance was
mineral oil and grease obtained in terms of
surface roughness
reduction [10]

Sehijpal et al (2008) Aluminum (Al), SEM technique A semi-solid abrasive Workpiece material, media MR and percentage It was revealed from
Brass laden medium volume flow rate, extrusion improvement in surface SEM analysis that the
pressure, abrasive concentration, finish (DRa) MR occurred mainly in
media viscosity, reduction ratio terms of microchips from
and average grain size of abrasive the surface due to
particles plowing and continued
abrasives flow [11]

Mamilla et al (2009) Al alloy One parameter at a time Abrasive Number of cycles, extrusion MR and change in A prominent impact of
Al alloy ? 10% SiC Silicon carbide (SiC) pressure and wt. percentage of oil average surface input parameters was
Al alloy ? 15% SiC Media in AFF media (viscosity of the roughness (DRa). observed on MR and DRa
Polymer carrier (styrene media) as MR was increased
butadiene rubber), with an increase in
processing oil extrusion pressure and
(hydraulic oil) and number of cycles, while
abrasive particles it was decreased by
(SiC) increasing the processing
oil content in the media
[12]
Sådhanå (2023)48:73
 continued

Experimental design/
Authors/year Material technique/model/method Abrasives/media Process parameters Responses parameters Findings

Sankar et al (2009) Al alloy Central composite Abrasive Extrusion pressure, rotational MR and average surface Introduced a rotary
Al alloy ? 10% SiC rotatable design (CCRD) Silicon carbide (SiC) speed of workpiece, viscosity of roughness (DRa) motion to finish Al alloy
Al alloy ? 15% SiC of RSM Media media and number of cycles and its composites by
Viscoelastic polymer improving the
material, abrasive performance parameters
Sådhanå (2023)48:73

particles and MR and DRa in the AFF

processing oil process. An
improvement of 44% and
81.8% was obtained in
DRa and MR
respectively with
rotational abrasive flow
finishing (R-AFF) as
compared to AFF [13]

Sankar et al (2010) Al alloy SEM micrographs Abrasive Extrusion pressure, workpiece MR, DRa, change in Employed R-AFF
Al alloy ? 10% SiC Silicon carbide (SiC, rotational speed and medium workpiece hardness and process for exploring the
Al alloy ? 15% SiC #220) compositions surface topology impact of workpiece
Media rotational speed on
Soft styrene polymer, surface topology,
hydrocarbon oil and material removal (MR),
abrasive particles change in workpiece
(SiC, #220) hardness and surface
roughness (DRa). The
surface topology was
analyzed by SEM
micrographs. The results
revealed that DRa, MR
and hardness of the
workpiece were
increased by increasing
the rotational speed of
the workpiece [14]

Brar et al (2011) Brass Taguchi experiments Abrasive Extrusion pressure, number of MR and percentage A robust AFM setup was
design (L9 OA) Aluminium oxide cycles and abrasive to media ratio improvement in surface developed and the
(Al2O3) finish (DRa) experiments were
Media conducted using Taguchi
Abrasive laden media method. From the
experiments it was
observed that the process
parameter abrasive to
media ratio had a
Page 5 of 35

maximum contribution
towards the response
73

parameters i.e. 75.75%

for MR and 73.78%
improvement in DRa [15]
 continued
73

Experimental design/
Authors/year Material technique/model/method Abrasives/media Process parameters Responses parameters Findings

Sankar et al (2011) Al alloy Rotational abrasive flow Workpiece materials, Vol% of MRR and DRa Adopted R-AFF process
Al alloy ? 10% SiC finishing (R-AFF) Abrasive plasticizer in medium, Number of to characterize the
Al alloy ? 15% SiC process, Atomic force Silicon carbide (SiC, cycles, Workpiece rotational rheological properties of
Page 6 of 35

microscopic #220) speed, Extrusion pressure abrasive media by

Media finding the influence of
base polymer (soft rheological parameters
styrene butadiene co- viz. stress relaxation
polymer), plasticizer modulus, shear stress,
(hydrocarbon oil) and storage modulus and
abrasives (SiC #220) percentage viscous
component on MRR and
DRa. An increase in
MRR and DRa was
obtained with an increase
in the yield shear stress
[16]

Yamaguchi et al (2012) Titanium alloy (Ti– SEM with Energy Abrasive Pole tip feed, gap between pole Roughness Ra Employed MAF process
6Al–4V) Dispersive Spectroscopy Diamond tip and target, finishing time and for improving the wear
(EDS) Media spindle speed characteristics of the tool
Abrasive (Diamond by conditioning the
paste), steel grits, iron uncoated carbide tool
particles mixed with surfaces. It was
lubricant concluded that the MAF-
processed tools have a
life of two times longer
than the unprocessed
tools [17]

Brar et al (2013) Brass Full factorial and L8 Abrasive Extrusion pressure, helical drill MR and percentage An improvement in MR
(OA) Taguchi design Aluminium oxide bit status, media flow rate and improvement in surface was obtained by a factor
methodology (Al2O3) number of cycles finish (DRa) of 2.5 using a stationary
Media drill-bit in HLX-AFM
Abrasive laden media process when compared
with the basic AFM [18]

Mittal et al (2015) Al/SiC- MMCs Taguchi L27 OA, SEM Abrasive Mesh number, number of cycles, MRR and change in The most significant
and X-ray diffraction Silicon carbide (SiC) workpiece material, abrasive surface roughness (DRa) parameter affecting both
(XRD) techniques Media concentration, extrusion pressure MRR and DRa was
Blending of SiC and percentage of oil in media extrusion pressure as an
abrasives, liquid increase in both
silicon and hydraulic responses was obtained
oil no. 68 by increasing the
extrusion pressure [19]
Sådhanå (2023)48:73
 continued

Experimental design/
Authors/year Material technique/model/method Abrasives/media Process parameters Responses parameters Findings

Mittal et al (2016) Al/SiC- MMCs Box–Behnken design of Abrasive Extrusion pressure, media oil MRR, change in surface A desirability approach
RSM, SEM and XRD Silicon carbide (SiC) percentage, mesh number, roughness (DRa) was used to optimize the
techniques Media abrasive concentration, number response variables. The
Blending of SiC of cycles and workpiece material most prominent factors
abrasives, liquid that affected MRR and
Sådhanå (2023)48:73

silicon and hydraulic DRa were workpiece

oil no. 68 material, extrusion
pressure and number of
cycles while a negligible
effect of media oil
percentage was found
[20]

Guo et al (2017) Inconel 718 SEM and EDX analysis Abrasive Work piece material, abrasive MRR and surface The experiments were
SiC and Al2O3 size, abrasive type, gap, robot roughness conducted using a
Media arm feed speed, tool rotation magnetic field-assisted
Blending of carbonyl speed, initial roughness and finishing process and a
iron powder, SiC/ process time significant effect of
Al2O3 powder with surface roughness, type
lubricating fluid of abrasives and size on
(ecocool) and the surface morphologies
machining oil was found. It was also
observed that a change in
the volume of Al2O3 and
SiC in the magnetic
abrasives had no effect
[21]

Nagdeve et al (2018) Stainless steel Rotational Abrasive Finishing time, magnets Change in finishing rate A fixture similar to a
magnetorheological Corbonyl iron particles rotational speed, number of (FR) and percentage replica of a knee joint
abrasive flow (CIPs), boron carbide cycles and extrusion pressure change in Ra (% DRa) implant was designed
finishing (R-MRAFF) (B4C) and rotational-
process Media magnetorheological
Blending of abrasives AFM was used for
with paraffin oil finishing it [22]
(carrier) and AP3
grease (surfactant)

Kajal et al (2019) Gun barrel First stage: Central Abrasive Working gap, mesh size, abrasive Surface roughness The MAF process was
composite design (CCD), Boron carbide (B4C) particles volume percentage, used for finishing a 0.32
Second stage: Full Media rotational and vertical in revolver barrel and
factorial design approach Blending of carbonyl reciprocation speed attained greater than
iron particles (CIPs), 80% refinement in
Page 7 of 35

B4C abrasives and surface finish i.e. a low

lubricant value of 150 nm at
73

different optimum
finishing conditions [23]
 continued
73

Experimental design/
Authors/year Material technique/model/method Abrasives/media Process parameters Responses parameters Findings

Sudhakara et al (2020) Al7075/SiC SEM, optical microscope Media Extrusion pressure, mesh size and MR and surface AFM technique was
NMMCs (OM), EDS/EDXS and Fluid silicon number of cycles. Other constant roughness (DRa) adopted to improve
XRD analysis parameters are: stroke length, material strength. With
Page 8 of 35

media volume, room temperature the presence of

and volume flow rate reinforcement particles,
density of the composites
was increased and with
the addition of 4% 7075
alloys, an improvement
of 96 Mpa in tensile
strength was observed
[24]

Jindal et al (2021) Al/SiC/B4C-MMCs Taguchi’s L27 orthogonal Abrasive Number of cycles, extrusion MRR Evaluated the influence
array, SEM, XRD SiC and B4C pressure, mesh size, abrasive of machining factors on
analysis Media concentration, magnetic field and MRR during MAFM of
Carbonyl iron particles, workpiece material aluminum reinforced
abrasive and silicon carbide and boron
hydraulic oil carbide (Al/SiC/B4C)
MMCs. ANOVA
outcomes revealed that
by increasing the number
of cycles, magnetic field
and extrusion pressure,
the MRR was increased
[25]

Sharma et al (2022)
Al-6061/SiC/ Al2O3/ Box–Behnken Design Abrasives Magnetic flux density, number of Change in surface
a
The Box–Behnken
Rare earth oxides and Levenberg– Magnetic abrasive cycles and extrusion pressure roughness (DR ) design (BBD) model and
hybrid composite Marquardt Algorithm- particles (MAP) Grit Levenberg–Marquardt
based artificial neural Size: 200 lm Algorithm-based
networks artificial neural networks
model were compared
and the comparison
demonstrated that trained
artificial neural network
(ANN) models had
significantly superior
prediction abilities than
the Box-Behnken Design
models [26]
Sådhanå (2023)48:73
Sådhanå (2023)48:73 Page 9 of 35 73

Table 1. Comparison of findings of the present study and past studies.

Past studies Present study

It has been noted from the literature that a vast amount of effort Although the heterogeneous material is widely used now-a-days as it
has been put into finishing homogeneous materials offers high strength, light weight, corrosion resistance and low density
[1, 3–11, 15, 18, 21–23, 27–55] viz. aluminium, stainless steel, while the material (Al/SiC/B4C-MMCs) used in the present study has
brass, copper, mild steel, etc. with different AFM processes. not been finished yet. The accomplishment of fine finishing of such
While little work has been reported on heterogeneous materials composites is a challenging task. These factors motivated the usage of
[12–14, 16, 19, 20, 56, 57] like Al/SiC-MMCs MAFM process for finishing such metal matrix composites in the
present study
The workpiece fixture employed by various researchers The design of fixture should be such that it can enable part cleaning,
comprises materials like nylon loading, and unloading and can generate a maximum magnetic field
[5, 9, 15, 18–20, 25, 39, 44, 46, 50, 52, 54, 57], brass [51], around the workpiece surface. Therefore, the present study sheds light
stainless steel [33] etc. The nylon fixture is known to have a low on the newly fabricated aluminium fixture in which the workpiece
magnetic permeability for the magnetic line of forces. This comes exactly in between magnetic field generated by electromagnet
hinders the finishing process of the inner workpiece cavity coils in the MAFM setup. This enhances the effectiveness of the
MAFM process, thus making this process an interesting field of
research
From the literature, most of the studies have been reported on The permanent magnets have the capability to produce a fixed value of
finishing of homogeneous material by using either the magnetic flux density while the magnetic flux density can be varied in
permanent magnets [8, 10, 17, 22, 23, 29, 38, 40, 47, 49, 51] or coil-type magnets. Therefore, the novelty of the present study is in the
the coil-type magnets [9, 28, 32, 33, 37, 41, 48]. While the use of coil-type magnets for finishing the heterogeneous material (Al/
finishing of heterogeneous material (Al/SiC/B4C-MMCs) has SiC/B4C-MMCs) making the developed MAFM setup more versatile
not been reported using either permanent or coil-type magnets and flexible in operation
No relevant literature has been found relative to optimizing the The present work makes use of a supervised machine learning
parameters with machine learning algorithm after processing algorithm to attain an optimal solution that encouraged the parametric
with the MAFM process machine learning algorithms to reflect on the effectiveness of the
MAFM process
From the past studies it has been observed that a little effort has The present study employs the surface analysis of finished workpieces
been made for evaluating the surface integrity of hybrid Al/SiC- of hybrid Al/SiC/B4C-MMCs with the scanning electron microscope
MMCs [12, 14, 16, 19, 20, 26] after finishing with different (SEM) and energy dispersive X-ray analysis (EDX) techniques
AFM processes

Table 2. Chemical composition (Al-6063).

Al-6063 Si Mg Cu Fe Zn Mn Pb Ti Cr Al
% Composition 0.48 0.74 0.02 0.22 0.0010 0.0063 0.0021 0.0086 0.0018 Remainder

From the literature, it has been reported that several This method used Numpy and Pandas for data analysis.
researchers have made an effort to create the mathematical Python was used for computing and embedding all the
models for MAFM of various homogenous materials with values in the data structures numerically. After finishing
traditional optimization techniques. The newness of current with MAFM, the machined surfaces were examined
investigation was mainly focused on analyzing the influ- through SEM. A good agreement was presented by
ence of MAFM factors on MRR and DRa using heteroge- applying a supervised machine learning algorithm with
neous material (Al/SiC/B4C). RSM based desirability RSM based desirability function for optimization. The
function optimization technique was used to optimize the research gaps are also identified by comparing the pre-
MRR and DRa. Then, the machine learning algorithm was sent study with past studies to explore the novelty as
employed for comparing the obtained optimum results. depicted in table 1.
73 Page 10 of 35 Sådhanå (2023)48:73

Table 3. Properties of metal matrix and reinforcement particulates.

Materials

Material properties Al-6063 SiC B4C

Density (g/cm3) 2.70 3.20 2.51

Modulus of elasticity (GPa) 68.9 410 450
Co-efficient of thermal expansion (m/m-°C) 21.8 9 10-6 4 9 10-6 5 9 10-6
Fracture toughness (MPaHm) 54.2 4.6 3.5
Melting point (°C) 654 2730 2445
Tensile strength (MPa) 241 137.9 261
Hardness (Knoop) 96 2480 2750
Thermal conductivity (W/m-K) 200 120 30-42

Table 4. Spectro analysis testing of MMCs.

Element Al Fe Mg Cu Si Zn Mn Pb B Cr Ti
Sample-1 Remainder 0.28 0.76 0.03 9.15 0.0010 0.0064 0.0025 1.01 0.0016 0.0096
Sample-2 Remainder 0.16 0.53 0.02 8.02 0.0012 0.0050 0.0026 1.96 0.0021 0.0090
Sample-3 Remainder 0.18 0.72 0.04 6.82 0.0012 0.0068 0.0027 2.86 0.0021 0.0091

Table 5. Chemical analysis testing of MMCs.

Elements percentage Aluminium-6063 (Al-6063) (%) Silicon carbide (SiC) (%) Boron carbide (B4C) (%)
Workpiece-1 89.96 9.06 0.98
Workpiece-2 90.13 7.95 1.92
Workpiece-3 90.21 6.87 2.92

3. Material and methods having different weight percentages of the reinforcement of

SiC (9, 8 and 7%) and B4C (1, 2 and 3%) are fabricated
All the investigations related to the finishing of Al/SiC/B4C using stir casting technique. The percentage composition of
hybrid MMCs are executed using MAFM setup. The various elements in workpiece materials is analyzed using
experimental setup was designed and developed in the spectro and chemical analysis testing of MMCs and is
laboratory. shown in tables 4 and 5. The parametric effects of the
fabricated MAFM setup for finishing the internal cylindri-
cal surfaces of sample-1: Al-6063/9 wt.% SiC/1 wt.% B4C-
3.1 Material evaluation and experimentation MMC, sample-2: Al-6063/8 wt.% SiC/2 wt.% B4C-MMC
and sample-3: Al-6063/7 wt.% SiC/3 wt.% B4C-MMC is
For the current investigation, the material used is Al-6063 investigated by carrying out a set of experimentations. An
(as base material) and SiC and B4C with mesh size 220 are electrical resistance furnace (capacity 1,150°C) with a
employed as reinforcement. The chemical composition of graphite stirrer and temperature controlling device was used
aluminium alloy (Al-6063) and the properties of metal for stirring and melting purposes. These samples were
matrix and reinforced particulates have been shown in prepared by liquefying the Al-6063 alloy in a graphite
tables 2 and 3, respectively. The Al-6063/SiC/B4C MMCs crucible furnace at 850°C temperature. Afterwards, both the
Sådhanå (2023)48:73 Page 11 of 35 73

Figure 1. (a) Pictorial representation, (b) Schematic arrangement of MAFM set-up.

reinforcements were preheated at 450°C temperature into obtained as 20.38 HRB, 25.02 HRB and 36.98 HRB with
another graphite crucible and blended with the Al alloy 9% SiC and 1% B4C, 8% SiC and 2% B4C and 7% SiC and
(liquefied) in the stir casting machine for vigorous stirring. 3% B4C MMCs respectively.
Finally, the prepared blended material was poured into the
moulds of required shape and solidified for fabricating the
MMCs which can be used to manufacture the metallic 3.4 Experimental procedure and arrangements
components. The other samples with different compositions
In this paper, an MAFM setup has been presented by
were also prepared using the same technique.
modifying the existing AFM setup with addition of a
magnetic field around the workpiece. In the past publica-
tions, work was focused on finishing of aluminium, stain-
3.2 Workpiece preparation less steel, Al/SiCp, brass, titanium alloy and mild steel etc.
To perform the experiments, three different samples with while not much work has been described on the hybrid
weight fractions viz. 9%, 8% and 7% SiC particles and 1%, MMCs [27–29]. The current study sheds light on investi-
2% and 3% B4C particles were taken with Al-6063 alloy. gating the MAFM of SiC/B4C hybrid MMCs with alu-
The fabricated rods of different compositions were initially minium as a base material. The past publications mentioned
faced, drilled and reamed using a lathe machine and finally the nylon fixtures usage in AFM setup. Such fixtures
samples were finished by MAFM setup. The size of hollow obstruct the machining of the inner cavities of the work-
cylindrical shaped workpiece was taken as external diam- piece which results in the low magnetic permeability for
eter (20 mm), internal diameter (12.5 mm) and length (40 lines of forces in the magnetic field. This section takes into
mm) as depicted in figure 2c, on the basis of the guidelines account, a newly fabricated setup having an aluminium
given in [9]. fixture which enhances the effectiveness of AFM process.
Figure 1a and b demonstrate the pictorial representation
and schematic arrangement of MAFM setup respectively.
To generate an abrasion effect around the workpiece sur-
3.3 Workpiece hardness
face, a magnetic field was employed in the present setup.
The Digital Rockwell Hardness Tester was employed for The permanent magnets were replaced with the coil type
indentation on five different points chosen arbitrarily for magnets having the copper windings for producing a
calculating the average hardness value. Ball Indenter (1/ maximum impact of magnetic field and the range for
16’’) at 100 Kgf was used. The Rockwell hardness was magnetic flux density (Mf) was taken from 0 to 2 Tesla.
73 Page 12 of 35 Sådhanå (2023)48:73

Table 6. Specifications of MAFM setup.

Sr. no. Specifications Description

1. Maximum extrusion pressure 10MPa
2. Media flow volume per pass 300 cm3
3. Stroke length 250 mm
4. Hydraulic cylinders material EN8
5. Hydraulic cylinders diameter 95 mm
6. Piston material Grey cast iron
7. Piston diameter 90 mm
8. Pump Hydraulic gear pump
9. Motor Flange type motor (2 H.P. and 1440 rpm)
10. Pressure gauges 2 Nos. (each 3000 psi)
11. Work piece fixture material Aluminium
12. Working gap 1–2 mm
13. Temperature 32°C (Room Temperature)

Table 7. Specifications of the electromagnet.

Sl. Constituents of
no. electromagnet Description of attributes
1. Electromagnet coils Coil weight = 3.820 kg
Number of turns = 2500 per
coil
Coil material = 23 gauge
copper wire
2. Magnetic flux density (Mf) 0–2 T
3. DC power supply 0–240 V
4. Core size Core length = 175 mm
Radius of core = 25 mm
Diameter of core rod =
35 mm
5. Core material Mild steel (M.S)
Figure 2. (a) An aluminium fixture, (b) cylindrical finished
workpieces, (c) workpiece geometry, (d) cross sectional view of
finished workpieces.

The key elements of MAFM setup are: (1) media fixture is shown in figure 2a where the cross-sectional path
cylinders (upper and lower) having pistons; (2) the sup- was reduced gradually so that the media can flow in a
porting frame; (3) the electromagnets; (4) the workpiece smooth manner with least vibrations. For pushing the
fixture; (5) a hydraulic unit. Table 6 shows the specification hydraulic oil from the tank to whole circuitry, a hydraulic
of MAFM setup. A striking ability of media cylinders was unit was designed having a hydraulic gear pump. An in-
to have an essential amount of media and to guide the house magneto-rheological polishing (MRP) fluid was
piston for extrusion of media during the reciprocating prepared by blending different compositions of constituents
motion. The copper winded (2500 turns) coil type magnets as shown in table 8. During MAFM, the finishing of the
were positioned around the cylindrical shaped samples to workpiece surface was done by flowing the MRP fluid from
provide a strong magnetic field. For the current research the internal cavity of cylindrical workpiece that causes the
work, the limit of magnetic field density (Mf) was taken material abrasion from the workpiece surface. The finished
from 0.15 to 0.45 Tesla. Table 7 depicts the specifications workpieces, their geometry and cross-sectional view are
of the electromagnets. The fabrication of an aluminium
Sådhanå (2023)48:73 Page 13 of 35 73

Table 8. Composition of MRP Fluid (Media).

Sl. no. Constituents Value of the constituents % Volume concentration in media

1. Silicon carbide (SiC) abrasive particle 150 Mesh size 45%, 50%, 55%
220 Mesh size
400 Mesh size
2. Iron (Fe) particle 300 Mesh size 10%
3. Hydraulic oil 68 Number 10%
4. Liquid silicon rubber (Carrier) White colour (RTV-2615) 35%, 30%, 25%

Figure 3. Material removal mechanism.

Table 9. Factors with their levels (actual and coded).

Parameter Symbols Units Lower range Upper range Coded low Coded high Mean SD
A Ep MPa 3.00 7 –1 $ 3 ?1 $ 7 5.00 1.35
B M Number 150 400 –1 $ 150 ?1 $ 400 244.44 88.52
C C Wt.% 45 55 –1 $ 45 ?1 $ 55 50.00 3.36
D Wp Type 1 3 –1 $ 1 ?1 $ 3 2.00 0.6729
E N – 100 200 –1 $ 100 ?1 $ 200 150.00 33.65
F Mf Tesla 0.15 0.45 –1 $ 0.15 ?1 $ 0.45 0.3000 0.1009
73 Page 14 of 35 Sådhanå (2023)48:73

Table 10. Design matrix for main experimentation.

Run no. Ep M C Wp N Mf MRR (lg/s) DRa (lm)

1 5 220 45 1 150 0.15 6.31 1.52
2 7 400 50 1 150 0.3 6.78 2.24
3 5 150 55 2 200 0.3 7.78 2.68
4 3 400 50 3 150 0.3 2.32 0.79
5 3 150 50 1 150 0.3 5.01 1.42
6 5 220 50 2 150 0.3 6.28 1.73
7 7 220 50 1 100 0.3 6.79 1.91
8 5 400 55 2 200 0.3 6.94 2.21
9 3 400 50 1 150 0.3 4.56 1.31
10 7 220 50 3 100 0.3 4.87 1.72
11 5 150 45 2 200 0.3 7.66 2.51
12 3 220 55 2 150 0.45 4.97 1.39
13 5 400 45 2 200 0.3 6.82 1.98
14 5 220 45 3 150 0.45 4.94 1.73
15 5 220 45 1 150 0.45 6.83 1.98
16 5 400 50 2 100 0.45 7.08 2.19
17 7 220 50 3 200 0.3 5.87 2.13
18 7 220 50 1 200 0.3 8.21 2.79
19 5 400 50 2 200 0.15 6.58 1.27
20 5 400 45 2 100 0.3 6.35 1.81
21 3 220 45 2 150 0.15 4.68 1.12
22 5 220 55 1 150 0.15 6.49 1.65
23 5 220 50 2 150 0.3 6.42 1.75
24 5 400 55 2 100 0.3 6.45 1.87
25 3 220 50 3 200 0.3 3.29 1.07
26 7 220 55 2 150 0.45 8.24 2.83
27 5 150 50 2 100 0.45 7.62 2.74
28 5 220 55 3 150 0.45 5.02 1.96
29 5 220 55 3 150 0.15 4.37 1.38
30 3 150 50 3 150 0.3 3.37 1.01
31 5 150 50 2 200 0.15 7.29 2.15
32 3 220 50 1 200 0.3 4.98 1.38
33 5 400 50 2 200 0.45 7.23 2.41
34 7 220 45 2 150 0.45 7.93 2.61
35 5 150 45 2 100 0.3 7.2 2.01
36 5 150 50 2 100 0.15 6.82 1.89
37 5 220 50 2 150 0.3 7.05 2.01
38 5 220 55 1 150 0.45 6.95 2.29
39 3 220 50 3 100 0.3 3.21 0.73
40 3 220 45 2 150 0.45 4.89 1.21
41 5 400 50 2 100 0.15 5.74 1.03
42 5 150 55 2 100 0.3 7.32 2.17
43 3 220 50 1 100 0.3 4.91 1.19
44 5 220 50 2 150 0.3 6.93 1.76
45 5 150 50 2 200 0.45 8.41 2.88
46 7 150 50 1 150 0.3 7.69 2.68
47 7 400 50 3 150 0.3 3.93 1.91
48 5 220 50 2 150 0.3 6.86 1.89
49 7 150 50 3 150 0.3 5.28 2.1
50 3 220 55 2 150 0.15 3.95 1.13
51 5 220 45 3 150 0.15 4.12 1.19
52 7 220 45 2 150 0.15 7.41 2.16
53 5 220 50 2 150 0.3 7.11 2.15
54 7 220 55 2 150 0.15 6.68 2.23
Sådhanå (2023)48:73 Page 15 of 35 73

Table 11. Responses model summary.

Responses Units Lower output Higher output Mean SD Ratio Model

MRR lg/s 2.32 8.41 6.09 1.47 3.63 Reduced quadratic (polynomial)
DRa lm 0.73 2.88 1.85 0.5519 3.95 Reduced quadratic (polynomial)

Table 12. ANOVA for MRR (reduced quadratic after backward elimination).

Source SS DF MS F-value p Value Remarks

Model 111.20 11 10.11 142.55 \ 0.0001 Significant
Ep 27.47 1 27.47 387.29 \ 0.0001
M 4.74 1 4.74 66.89 \ 0.0001
Wp 25.88 1 25.88 364.87 \ 0.0001
N 1.87 1 1.87 26.37 \ 0.0001
Mf 3.90 1 3.90 54.94 \ 0.0001
Ep*M 0.2132 1 0.2132 3.01 0.0903
Ep*N 0.6441 1 0.6441 9.08 0.0044
Ep2 5.33 1 5.33 75.18 \ 0.0001
M2 0.2032 1 0.2032 2.87 0.0979
Wp2 14.78 1 14.78 208.45 \ 0.0001
N2 1.30 1 1.30 18.36 0.0001
Residual 2.98 42 0.0709
LOF 2.39 37 0.0646 0.5469 0.8688 Not significant
SD 0.2663 R2 0.9739
Mean 6.09 AR2 0.9671
C.V. % 4.37 PR2 0.9539
AP 46.6031

Table 13. ANOVA for DRa (reduced quadratic after backward elimination).

Source SS DF MS F-value p Value Remarks

Model 15.07 11 1.37 53.68 \ 0.0001 Significant
Ep 7.66 1 7.66 300.22 \ 0.0001
M 1.14 1 1.14 44.49 \ 0.0001
C 0.1601 1 0.1601 6.27 0.0162
Wp 0.8971 1 0.8971 35.15 \ 0.0001
N 0.7350 1 0.7350 28.80 \ 0.0001
Mf 2.47 1 2.47 96.79 \ 0.0001
Ep*N 0.0722 1 0.0722 2.83 0.1000
M* Mf 0.1950 1 0.1950 7.64 0.0084
Ep2 0.1647 1 0.1647 6.45 0.0149
M2 0.6489 1 0.6489 25.43 \ 0.0001
Wp2 0.6595 1 0.6595 25.84 \ 0.0001
Residual 1.07 42 0.0255
LOF 0.9281 37 0.0251 0.8729 0.6460 Not significant
SD 0.1597 R2 0.9336
Mean 1.85 R2 0.9162
C.V. % 8.64 AR2 0.8866
PR2 26.8346
73 Page 16 of 35 Sådhanå (2023)48:73

Figure 4. Parametric effects for MRR.

depicted in figure 2b, c and d, respectively. After finishing a between the electromagnet poles. This provides the bond-
workpiece for the desired number of cycles counted ing strength to the surrounded abrasive particles and the
through a digital counter meter, another workpiece was flow of magnetic abrasive particles changes their movement
positioned in the slots by detaching the fixture along with path. Moreover, it impinges at a small angle on the work-
the finished workpiece. The finished workpiece was washed piece surface resulting in micro-chipping of the surface.
with the acetone (CH3)2CO. This magnetic field also affects the abrasive distribution
pattern at the workpiece surface which requires machining.
Due to this, the maximum abrasive particles strike the
3.5 Mechanism of material removal for Al/SiC/B4C surface and actively participate in the abrasion process. As
hybrid MMCs a result, the amount of cutting wear is increased that causes
an overall improvement in the material removal rate.
In conventional AFM process, a semi-solid media con-
taining the polymer-based carrier and abrasives in a specific
ratio is extruded through the workpiece surface for finishing 4. Experimental design and methodology
purpose. This abrasive media works as a flexible cutting
tool. A magnetic field has been added and a solidified paste The actual and coded form of various factors and their
(a mixture of liquid silicon rubber (carrier), silicon carbide levels are shown in table 9. The starting value of the coded
(SiC) abrasive particles, iron particles (Fe), and hydraulic form is center value (0) and extended between upper and
oil 68 nos.) is used as a medium during MAFM process as lower values (?1 to –1). The preliminary experimentation
shown in figure 3. The media is expelled backward and
was conducted with one-factor-at-a-time (OFAT) approach.
forward through the workpiece passage and fixture. By
Based on the past survey and preliminary experiments, six
applying the magnetic field, a chain-like structure is formed
Sådhanå (2023)48:73 Page 17 of 35 73

prominent factors i.e. Ep, M, N, Wp, C, and Mf were chosen and final sample. For measuring the weight of each
for finding their influence on MRR and DRa. To analyze the sample, BL-220H, an electronic balance with 10-3 g
surface morphology, SEM (Model: JEOL, USA, JSM- accuracy was employed.
6510LV) integrated with EDX apparatus was employed. An
experimentation design matrix with a total 54 experimental
runs having six parameters at three levels was planned on 4.1 Mathematical expressions
the basis of Box–Behnken’s design as shown in
table 10. The value of parametric effect on MRR and The mathematical relations 1, 2 and 3 were used for cal-
DRa was computed by Design ExpertÒ software (ver- culating the abrasive concentration (C), MRR and DRa
respectively.
sion 12.0). The weight and surface roughness of each
sample was determined before and after each trial. The
roughness of the workpiece surface was calculated by 4.1.1 The abrasives concentration (C) The abrasives
E35-B, a digital handysurf machine for every initial concentration (C) in medium was determined by:

Weight of the abrasives in medium

Abrasive concentration in medium ¼  100 ð1Þ
ðWeight of the abrasives in medium þ Weight of the carrier mediumÞ

Figure 5. Parametric effects for DRa.

73 Page 18 of 35 Sådhanå (2023)48:73

Figure 6. 3D Response surface interaction and perturbation plots for MRR.

4.1.2 Material removal rate (MRR) MRR denotes 5. Results and discussion
the amount of material removed from the workpiece
surface while finishing with MAFM. The time duration 5.1 ANOVA for predictive modeling
of every process was noted and workpiece weight
was calculated initially and at the end of each The adequacy of the models has been evaluated by per-
trial. Mathematically, MRR (lg/sec) was calculated forming three different tests (model summary statistics, SS
as: and LOF). The BBD matrix can be developed effectively for
an experimentation plan and its obtained MRR and DRa
values by employing the RSM. The significance level (a =
MRR ¼ 0.05) at 95% CI and ANOVA has been computed for the
ðInitial weight of the sample  Final weight of the sampleÞ reduced quadratic model after backward elimination method.
Time The reduced quadratic (polynomial) has been attained for
ð2Þ each response parameter and model summary statistics for
every response is depicted in table 11. A quadratic model of
second-order mathematical polynomial regression equation
4.1.3 Change in surface roughness (DRa) The surface for MRR and DRa is presented below:
roughness of every sample was calculated before and after
finishing with MAFM. Mathematically, DRa (lm) was MRR ¼ 1:22559 þ 2:08397  Ep þ 0:00669834
calculated as:  M þ 3:618  Wp þ 0:0500642  N þ 2:68611
DRa ¼ Ra ðinitialÞ  Ra ðfinalÞ ð3Þ  Mf þ 0:000614549  Ep  M þ 0:0028375
 Ep  N þ 0:174771
Ra (final) = Surface roughness after finishing
Ra (initial) = Surface roughness before finishing
 E2p þ 1:12989e  05  M2
þ 1:16408  W2p þ 0:0001382  N2 ð4Þ
Sådhanå (2023)48:73 Page 19 of 35 73

Figure 7. 3D Response surface interaction and perturbation plots for DRa.

Figure 8. SEM micrographs (a) before, (b) after finishing of Al/9 wt.% SiC/1 wt.% B4C-MMCs.
73 Page 20 of 35 Sådhanå (2023)48:73

Figure 9. SEM micrographs (a) before, (b) after finishing of Al/8 wt.% SiC/2 wt.% B4C-MMCs.

Figure 10. SEM micrographs (a) before, (b) after finishing of Al/7 wt.% SiC/3 wt.% B4C-MMCs.

DRa ¼ 0:241676 þ 0:445871  Ep þ 0:0144985  M recognized by F-value and p value. The statistical signifi-
cance of a parameter to the response parameters can be
þ 0:0163333  C þ 0:786061  Wp
implied by the higher F-value and lower p values \ 0.05. It
þ 0:00125  N þ 0:21562  Mf þ 0:00095  Ep has been predicted from the values of ANOVA (tables 12,
 N þ 0:00783656  M  Mf þ 0:0305871  E2p 13) that parameters Ep, M, Wp, N, Mf and their interactions
(Ep 9 M, Ep 9 N) have been most significant to MRR and
þ 1:89228e  05  M2 þ 0:244848  W2p
Ep, M, C, Wp, N, Mf and their interaction (Ep 9 N, M 9
ð5Þ Mf) have been observed as significant factors to DRa. The
LOF p value for MRR is 0.8688 and for DRa is 0.6460
The adequacy and statistical implication of the devel-
which specifies that the model fits the data effectively.
oped regression model has been governed by observing the
Moreover, for MRR ‘‘PR2’’ of 95.39% is in reasonable
p value and F-value. The parametric significance can be
Sådhanå (2023)48:73 Page 21 of 35 73

Figure 11. EDX analysis (a, b, c) machined sample-1: Al/9%SiC/1% B4C-MMCs, (d, e, f) machined sample-2: Al/8%SiC/2% B4C-
MMCs, (g, h, i) machined sample-3: Al/7%SiC/3% B4C-MMCs.

Table 14. Elements migration for different Al-MMCs samples deposited on MAFM surface.

Sample no. Run no. C O Mg Al Si Fe

Sample-1 18 39.33 9.95 0.20 48.93 0.88 0.70
(Al/9%SiC/1% B4C) 46 39.75 14.06 0.26 45.69 0.02 0.22
38 42.38 11.91 0.18 44.74 0.80 –
Sample-2 27 7.21 2.38 0.01 89.89 0.43 0.07
(Al/8%SiC/2% B4C) 8 7.64 6.36 0.10 85.25 0.45 0.21
31 5.20 2.46 0.16 91.45 0.61 0.12
Sample-3 4 24.77 18.15 0.10 55.69 0.94 0.35
(Al/7%SiC/3% B4C) 39 21.64 19.65 0.46 56.64 1.13 0.47
25 18.13 15.89 0.63 63.70 1.13 0.52

agreement with the ‘‘AR2’’ of 96.71% and for DRa ‘‘PR2’’ 5.2 Parametric effect on MRR and DRa
of 88.66% is in good agreement with ‘‘AR2’’ of 91.62%.
Hence, LOF is found to be insignificant. Additionally, for Figures 4 and 5 demonstrate the one factor plot showing the
MRR (R2 = 97.39%) and for DRa (R2 = 93.36%) show that parametric effect on MRR and DRa respectively. An
changing the MRR and DRa value is attributed towards increase of 51.83% has been observed in MRR by
controlling different factors. Therefore, the polynomial increasing the extrusion pressure (Ep) from 3 to 7 Mpa. It
model’s accuracy and general ability is declared to be good. has been reported that with increased Ep, more abrasion
The signal to noise ratio is evaluated by adequate precision takes place as the abrasive particles strike with a strong
(AP) and a ratio of [4 is essential. Therefore, the ratios for impact against the surface to be machined [10, 11, 30–32].
MRR and DRa are obtained 46.60 and 26.83 that reveal the On the other hand, the effect of Ep on DRa has been
adequate signal. increased by 101% resulting in the movement of abrasive
73 Page 22 of 35 Sådhanå (2023)48:73

Table 15. Optimal solution after desirability function.

Optimal solution Ep M C Wp N Mf MRR (lg/s) DRa (lm) Desirability

1 4 352 45 1 100 0.15 5.40 0.73 0.917
2 4 333 45 1 100 0.15 5.38 0.73 0.917
3 4 365 45 2 100 0.15 5.38 0.73 0.917
4 4 346 45 1 101 0.15 5.37 0.73 0.917
5 4 353 45 1 100 0.16 5.36 0.73 0.917
6 4 340 45 2 100 0.15 5.38 0.74 0.916
7 4 341 46 2 100 0.15 5.34 0.73 0.916
8 4 312 45 1 100 0.15 5.33 0.73 0.916
9 4 367 46 2 100 0.15 5.33 0.73 0.916
10 4 306 45 1 100 0.15 5.30 0.73 0.914

Figure 12. Optimization plot for MRR and DRa using desirability function.
Sådhanå (2023)48:73 Page 23 of 35 73

particles against the peaks and the surface getting machined Table 16. Instructions for importing the desirability values for
[19, 20, 43–46]. The outcomes declared that by increasing MRR.
the number of cycles (N), a non-linear increment of 8.15%
has been observed in MRR and 20% in DRa. In the early
stages, the surface has several peaks and valleys which can
be removed by increasing the number of cycles
[12–16, 33–39]. In MAFM process, a magnetic field is
generated by applying the current to the electromagnet. The
generation of magnetic flux density (Mf) helps in creation
of a chain like configuration that increases bonding strength
of abrasive particles. Hence, the participation of abrasive
particles increases and more peaks are sheared off with
deep penetration [9, 32, 33, 37, 41, 48]. Therefore, by
increasing the magnetic field the MRR and DRa have been
increased by 13.5% and 48.61% respectively. The work-
piece material (Wp) used in the current work is Al/SiC/
B4C-MMCs. The workpiece hardness increases with the
presence of reinforcement B4C in the MMCs [58]. The
experimental results revealed that the MRR has been
enhanced by 51.52% due to an influence of change in
reinforcement percentage of B4C and SiC in Al-6063. The
sample-3 is harder than samples-1 and 2 because of the high
percentage of reinforcement B4C (3%), due to which it
requires more extrusion pressure for removing the material.
Therefore, in sample-3, MRR and DRa have been observed
as 4.27lg/s and 1.34 lm respectively. While in sample-2
the B4C is taken as 2% resulting in an improvement of
MRR and DRa by 6.48 lg/s and 1.74lm respectively as
compared to sample-3. On the other hand, in sample-1, a
maximum value of MRR and DRa has been observed as
6.52 lg/s and 1.80 lm respectively because of less per-
centage of reinforcement B4C (1%) than both the samples.
An insignificant effect of abrasive concentration (C) has
been observed for MRR while DRa has been improved by
8.18%. As by increasing the abrasive concentration, more
abrasives interact with the surface of the workpiece due to
which more cutting force is retained by the media causing
more abrasion [19, 20, 30, 31]. Further, the MRR and DRa
have been increased by 14.46% and 22.58% respectively by
decreasing the mesh number (M) [3–9, 27–31]. The main increased significantly with 55.70% as displayed by the
reason is the increase in abrasive size that causes further immediate interaction among number of cycles (N) and
increase in penetration depth and width resulting in high extrusion pressure (Ep) in figure 6. DRa has been observed
values of MRR and DRa [19, 20, 43–48]. as 152% with the simultaneous interaction among number
A fitting response surface graph is structured initially for of cycles (N) and extrusion pressure (Ep) as depicted in
every response and after that a set of operating conditions figure 7. Moreover, 22.16% DRa has been observed through
are found for the instantaneous interaction of multiple the simultaneous interaction of magnetic flux density (Mf)
factors. Figures 6 and 7 depict the 3D interaction response and mesh number (M). The parametric effect of each
surface graphs for MRR and DRa. The obtained value of parameter is equated at a specific reference point by the
MRR has an insignificant effect as it shows a constant line perturbation plots as depicted in figures 6 and 7. The
in an interaction effect. An increasing trend has been shown response can be plotted by varying only one variable over
by the parameter resulting in a significant effect. The its limit; however the remaining variables are kept constant.
immediate interaction among mesh number (M) and The sensitivity of a response to the parameter can be shown
extrusion pressure (Ep) resulted in an increased MRR per- by a steep slope or curvature while insensitivity can be
centage that is 45.17%. On the other hand, MRR has been shown by a flat line.
73 Page 24 of 35 Sådhanå (2023)48:73

Table 17. Correlation matrix and heat map for MRR.

5.3 Analysis of surface morphology SEM analysis shown in figures 8b, 9b, and 10b. It has also
been observed that several abrasive particles have been
A distinguished property of a hybrid material for showing embedded on the specimen’s surface. Energy dispersive
the essential responses in the engineering industries is X-ray (EDX) analysis is an analytical method, mainly
surface morphology. The specimens have been analyzed adopted for characterization and elemental analysis of the
using a scanning electron microscope (JSM-6510LV, chemicals present in the hybrid materials [24]. The pres-
JEOL, USA) after machining with MAFM. Each specimen ence of a more concentrated element in the workpiece is
is cleaned with the solution of acetone (CH3)2CO and dried represented by a higher peak in the EDX spectrum. The
in the air before analyzing with SEM. Magnification is energy dispersive X-ray (EDX) of finished surfaces
taken as 9200. The casted products (cylindrical rods) have obtained at 3KV accelerating voltage. The EDX spectrum
been initially faced, drilled and reamed by lathe operation. of machined sample-1 (Al/9%SiC/1%B4C-MMCs), sam-
It has been analyzed by SEM micrographs that some ple-2 (Al/8%SiC/2%B4C-MMCs) and sample-3 (Al/
irregularities and casting defects viz. crater, cracks, voids, 7%SiC/3% B4C-MMCs) demonstrates Al, C, Mg, Si, Fe
cutting tool marks, multi layers, micro and large pits were and O peaks as displayed in figure 11a–i. The EDX spec-
present on the specimen surface before machining the trum of Al-6063 with reinforced materials SiC and B4C
surfaces using MAFM as depicted in figures 8a, 9a, and having different ratios exhibits aluminium (Al) as higher
10a. The surface topography has been improved signifi- peaks and are depicted in figure 11a–i. The percentage of
cantly and all the irregularities present on the surface are element migration for various Al-MMCs samples posi-
either dissolved or eroded from the surface as analyzed by tioned on MAFM surface is displayed in table 14. A total of
Sådhanå (2023)48:73 Page 25 of 35 73

Table 18. Instructions for preprocessing the MRR dataset and splitting into training (80%) and test data (20%).

nine random experiments for EDX spectra of three dif- 6. Multi-objective optimization through
ferent samples have been taken. The MAFMed surface desirability function
predominantly comprises Al, C, Mg, Si, Fe and O ele-
ments. From the EDX analysis, it is evident that the In the current investigation, the optimal solution is obtained
major element present in the MMCs is the highest com- by employing the desirability function. The optimization is
position of Al (44.74–91.45%). In addition, an amount of done both graphically and numerically. Further, the com-
C (5.20–42.38%) and Si (0.02–1.13%) content is present parison of the results obtained after optimization is done
after MAFM that indicates the precipitation of organic with the computational based machine learning algorithm.
molecules from SiC [59, 60]. Hence, it validates the
occurrence of reinforcement particles. The low peak of
oxygen (2.38–19.65%) reveals the protection of com- 6.1 Graphical and numerical optimization
posites during production [61]. Moreover, a lesser
amount of Mg (0.01–0.63%) and Fe (0.07–0.70%) ele- To achieve the optimum objectives for MRR and DRa, the
ments is observed. models for searching the factor space are used by numerical
73 Page 26 of 35 Sådhanå (2023)48:73

Table 19. Instructions for prediction of data for MRR using for every goal is marked from 3 pluses (???) to 5 pluses
trained model. (?????). For attaining the optimum solutions, the
‘Desirability function’ has been applied and the value di
has been varied between 0 B di B 1 [20]. The implemen-
tation of desirability function resulted in a total of 10
optimum solutions as described in table 15. The range of
overall desirability lies from 0.917 to 1. To achieve the
optimal overall desirability, figure 12 displayed the ramp,
bar graph and overlay plots. The optimum solutions are
indicated by a point on the ramp graph. The results obtained
from graphical optimization, overlay contour plots where
all responses meet the critical properties simultaneously,
the ‘optimal spot’. The critical response area has been
visualized in the parameter space by the overlay plot. The
dark grey shaded area indicates that this area did not fit the
optimization criteria. The optimal parameter settings are
satisfied by the yellow area region.

6.2 Analyzing parameters with machine learning

algorithm (PYTHON)
An exhaustive analysis of various MAFM process
parameters has been described in this section and the
results of different performance measures have been pre-
dicted through machine learning. As the data structures
will be used frequently, importing them in the local
namespace. The following steps are undertaken for para-
metric analysis:
Step 1: (Data Analysis through NumPy and Pandas)
Initially, the data is imported from the excel format and
is given to the local namespace. Then, NumPy and Pandas
are used for analyzing the data. To understand Pandas, the
fundamental requirements for data structures are data frame
and series [62, 63] as shown in tables 16 and 20.
Step 2: (Visualization of Data using Seaborn and
Matplotlib)
The second step focuses on finding the correlation among
various performance measures and MAFM factors. This
can be done using the visualization and evaluation of the
datasets. The seaborn library is used to observe the output
values for MRR and DRa. The correlation matrix and heat
map for each factor is shown in tables 17 and 21.
optimization for best trade-offs. The weights of objectives Step 3: (Scikitlearn for Data Preprocessing)
have been achieved by assigning the criteria with impor- The data obtained through step-2 is processed initially
tance (upper or lower bounds) or to mark a target value. in machine learning libraries using sci-kit-learn in the
The present work aims at setting the goal on a target for third step. The model is trained at a higher degree with a
achieving the highest MRR=8.41 lg/s and lowest DRa polynomial regression algorithm. The model is trained by
=0.73. The weights have been allocated as (upper weight= partitioning the data in two parts [training data (80%) and
0.8, and lower weight=0.2). Thereafter, the ‘‘importance’’ testing data (20%)] as depicted in tables 18 and 22.
Sådhanå (2023)48:73 Page 27 of 35 73

Table 20. Instructions for importing the desirability values for DRa.
73 Page 28 of 35 Sådhanå (2023)48:73

Table 21. Correlation matrix and heat map for DRa.

Step 4: (Scikitlearn for Prediction Analysis) relatively low and the model output is also in good agree-
Finally, the model is trained for predicting the data for ment with experimental values. The error plot of RSM
MRR and DRa as depicted in tables 19 and 23. The actual predicted values and machine learning predicted values is
and predicted values (tables 19 and 23) are in good plotted for a total of 54 values. These predicted values
agreement and the plots coincide with each other as which are obtained for MRR coincide with each other and
depicted in figure 13 (tables 19 and 23). this is clear from figure 14. Similarly, the error plot for DRa
of the RSM predicted values against the machine learning
values shows a good agreement among each other, as
7. Error plots for RSM and machine learning depicted in figure 15. Hence the machine learning values
predicted values are a true representation of the RSM predicted values. The
error obtained after the difference of the RSM and machine
Table 24 considers the error obtained in simulation through learning predicted values is quite low corresponding to
machine learning. This implies that it tabulates the errors figures 14 and 15. Hence, the results confirm the ability of
computed between the RSM and machine learning pre- machine learning to optimize the response parameters and
dicted values for MRR and DRa during the process of simulate the design experimentation results in a coherent
MAFM. From table 24, the modeling error is observed to be manner that improved the efficiency of the MAFM process.
Sådhanå (2023)48:73 Page 29 of 35 73

Table 22. Instructions for preprocessing the DRa dataset and splitting into training (80%) and test data (20%).

8. Conclusions DRa. For LOF, the p-value of MRR and DRa is (0.8688,
0.6460) which indicates that this model efficiently fits
In brief, a Behnken0 s design of RSM based ‘Desirability the data. Moreover, ‘‘PR2’’ of (95.39%, 88.66%) for
Function’ and a ‘Machine Learning Algorithm (PYTHON)’ MRR and DRa is in reasonable agreement with the
was employed in MAFM process for simultaneously opti- ‘‘AR2’’ of (96.71%, 91.62%). Therefore, the observed
mizing the MRR and DRa of hybrid Al/SiC/B4C-MMCs. value of LOF was not significant. Further, the variation
The input parameters viz. Ep, M, N, and Mf had a major of MRR and DRa shown by the R2 value of (97.39%,
role for attaining the optimum settings for both the 93.36%) was attributed to control factors. The backward
responses. The following outcomes are outlined from the elimination process was used for removing the non-
present study: significant terms.
2. Results show that the MRR and DRa were mainly
1. ANOVA was adapted at 95% CI for obtaining the improved by (51.83%, 101%) with an increased value of
impact of every MAFM factor relating to MRR and
73 Page 30 of 35 Sådhanå (2023)48:73

Table 23. Instructions for prediction of data for DRa using trained
model.

(a)

(b)
Figure 13. Comparison between RSM based experimental and
Machine learning results (a) MRR, (b) DRa.

4. The interaction plots of immediate interaction among

extrusion pressure (Ep). On the other hand, by decreas- extrusion pressure (Ep) and number of cycles
ing the mesh number (M) the MRR and DRa were (N) showed that MRR and DRa were considerably
increased by (14.46%, 22.58%). An improvement of improved by (55.70%, 152%).
8.18% was obtained in DRa by increasing the abrasives 5. The interaction plots for simultaneous interaction among
concentration (C) and an improvement of 51.52% was extrusion pressure (Ep) and mesh number (M) was
obtained in MRR due to the effect of changing the obtained to be increased by (45.17%) for MRR and the
reinforcement percentage (SiC and B4C) in the base simultaneous interaction of magnetic flux density (Mf)
material (Al-6063). with mesh number (M) showed an improvement of
3. During MAFM process, the magnetic field is produced (22.16%) for DRa.
in the abrasive particles with the electromagnet by 6. By analyzing the results, the extrusion pressure (Ep =
passing the current through the coils. A greater force is 24.05%) and workpiece material (Wp = 22.66%) were
provided by these magnetic field assisted abrasive found to be the most significant factors having a major
particles on the workpiece surface. Hence the magnetic influence on MRR while extrusion pressure (Ep =
flux density (Mf) increases which results in an increased 47.45%) and magnetic flux density (Mf = 15.30%)
value of MRR and DRa with (13.5%, 48.61%). affected DRa.
Sådhanå (2023)48:73 Page 31 of 35 73

Table 24. RSM versus machine learning results for MRR and DRa during MAFM.

Material removal rate (MRR; lg/s) Change in surface roughness (DRa; lm)

Sr. no. RSM Machine learning Error RSM Machine learning Error
1 6.31 6.318091 -0.008091 1.52 1.567753 -0.047753
2 6.78 7.248925 -0.468925 2.24 2.296115 -0.056115
3 7.78 8.141464 -0.361464 2.68 2.81318 -0.13318
4 2.32 1.662493 0.657507 0.79 0.736771 0.053229
5 5.01 5.04633 -0.03633 1.42 1.50441 -0.08441
6 6.28 6.730389 -0.450389 1.73 1.875013 -0.145013
7 6.79 6.207996 0.582004 1.91 1.812895 0.097105
8 6.94 7.078835 -0.138835 2.21 2.267238 -0.057238
9 4.56 4.026048 0.533952 1.31 1.139451 0.170549
10 4.87 3.781767 1.088233 1.72 1.571751 0.148249
11 7.66 7.619536 0.040464 2.51 2.503965 0.006035
12 4.97 6.215922 -1.245922 1.39 1.791975 -0.401975
13 6.82 6.951627 -0.131627 1.98 2.063158 -0.083158
14 4.94 4.902174 0.037826 1.73 1.684076 0.045924
15 6.83 6.774099 0.055901 1.98 2.017768 -0.037768
16 7.08 7.542964 -0.462964 2.19 2.287718 -0.097718
17 5.87 6.548421 -0.678421 2.13 2.322095 -0.192095
18 8.21 9.049113 -0.839113 2.79 2.842342 -0.052342
19 6.58 6.397247 0.182753 1.27 1.214072 0.055928
20 6.35 6.691349 -0.341349 1.81 1.835695 -0.025695
21 4.68 4.530069 0.149931 1.12 0.996319 0.123681
22 6.49 6.394723 0.095277 1.65 1.666852 -0.016852
23 6.42 6.730389 -0.310389 1.75 1.875013 -0.125013
24 6.45 6.365038 0.084962 1.87 1.843993 0.026007
25 3.29 2.835239 0.454761 1.07 0.921566 0.148434
26 8.24 7.33326 0.90674 2.83 2.786868 0.043132
27 7.62 7.711719 -0.091719 2.74 1.927341 0.812659
28 5.02 5.194836 -0.174836 1.96 1.948749 0.011251
29 4.37 4.124692 0.245308 1.38 1.288088 0.091912
30 3.37 3.634246 -0.264246 1.01 1.20023 -0.19023
31 7.29 7.86646 -0.57646 2.15 2.293323 -0.143323
32 4.98 4.550966 0.429034 1.38 1.392879 -0.012879
33 7.23 7.554139 -0.324139 2.41 2.798078 -0.388078
34 7.93 7.561972 0.368028 2.61 2.741695 -0.131695
35 7.2 7.266687 -0.066687 2.01 2.008627 0.001373
36 6.82 6.810989 0.009011 1.89 1.885101 0.004899
37 7.05 6.730389 0.319611 2.01 1.875013 0.134987
38 6.95 7.220261 -0.270261 2.29 2.317572 -0.027572
39 3.21 4.56898 -1.35898 0.73 0.914566 -0.184566
40 4.89 5.248385 -0.358385 1.21 1.27267 -0.06267
41 5.74 5.434348 0.305652 1.03 1.073725 -0.043725
42 7.32 7.335097 -0.015097 2.17 2.12206 0.04794
43 4.91 6.210244 -1.300244 1.19 1.106775 0.083225
44 6.93 6.730389 0.199611 1.76 1.875013 -0.115013
45 8.41 7.815466 0.594534 2.88 2.705577 0.174423
46 7.69 6.91514 0.77486 2.68 2.55153 0.12847
47 3.93 4.100405 -0.170405 1.91 1.844499 0.065501
48 6.86 6.730389 0.129611 1.89 1.875013 0.014987
49 5.28 4.718092 0.561908 2.1 2.198414 -0.098414
50 3.95 5.128077 -1.178077 1.13 1.314919 -0.184919
51 4.12 4.201559 -0.081559 1.19 1.22412 -0.03412
52 7.41 7.123666 0.286334 2.16 2.108075 0.051925
53 7.11 6.730389 0.379611 2.15 1.875013 0.274987
54 6.68 6.525424 0.154576 2.23 1.952543 0.277457
 73 Page 32 of 35 Sådhanå (2023)48:73

Abbreviations
AR2 Adjusted R2
AP Adequate precision
CI Confidence interval
ANOVA Analysis of variance
LOF Lack of fit
BBD Box–Behnken design
Ep Extrusion pressure
N Number of cycles
Al/SiC/ Aluminium-6063/Silicon carbide/Boron
B4C carbide
C Concentration of abrasives
MMCs Metal matrix composites
Figure 14. Error plot for MRR. EDX Energy dispersive X-ray analysis
MAFM Magnetic abrasive flow machining
MS Mean square
MRR Material removal rate
PR2 Predicted R2
DRa Change in surface roughness
RSM Response surface methodology
Wp Workpiece material
M Mesh size
Mf Magnetic flux density
SEM Scanning electron microscope

Acknowledgements

The authors would like to thank the National Institute of

Figure 15. Error plot for DRa. Technology (NIT), Kurukshetra, Haryana, India, to provide
the essential facilities for current research work. The
authors are also thankful to Thapar University Patiala,
Punjab, India for granting the permission of using the
7. The optimal solutions were obtained by applying a resources in their laboratory (SEM and EDX). No financial
multi-objective optimization based ‘Desirability Func- assistance has been received from any institution or agency.
tion’ and overall desirability range was obtained from
0.917 to 1. Declarations
8. The SEM morphology of Al/SiC/ B4C examination of
the machined surfaces declared that the surface finish of Conflict of interest The authors declare that they have no conflict of
interest.
the workpiece was improved significantly. Various
irregularities and casting defects viz. cracks, multi
layers, micro and large pits, voids and cutting tool
References
marks have been deteriorated or integrated with the
irregular debris or craters after finishing with the MAFM [1] Williams R E and Rajurkar K P 1992 Stochastic modeling
process. and analysis of abrasive flow machining. J. Eng. Ind. 114(1):
9. The EDX analysis of machined sample-1 (Al/9%SiC/ 74–81. https://doi.org/10.1115/1.2899761
1%B4C-MMCs), sample-2 (Al/8%SiC/2%B4C-MMCs) [2] Williams R E and Melton V L 1998 Abrasive flow finishing
and sample-3(Al/7%SiC/3%B4C-MMCs) revealed that of stereolithography prototypes. Rapid Prototyp. J. 4(2):
the MAFMed surface is predominantly consisted of Al, 56–67. https://doi.org/10.1108/13552549810207279
C, Mg, Si, Fe and O elements. [3] Jain R K, Jain V K and Kalra P K 1999 Modelling of
10. A good agreement was presented by applying a super- abrasive flow machining process: a neural network approach.
vised ‘Machine Learning Algorithm’ with ‘Desirability Wear 231: 242–248. https://doi.org/10.1016/S0043-
Function’ for optimization. 1648(99)00129-5
Sådhanå (2023)48:73 Page 33 of 35 73

[4] Jain R K and Jain V K 1999 Simulation of surface generated J. Inst. Eng. (India) Ser. C. 94(1): 21–29. https://doi.org/10.
in abrasive flow machining process. Robot. Comput. Integr. 1007/s40032-012-0054-9
Manuf. 15(5): 403–412. https://doi.org/10.1016/S0736- [19] Mittal S, Kumar V and Kumar H 2015 Experimental
5845(99)00046-0 investigation and optimization of process parameters of Al/
[5] Jain V K and Adsul S G 2000 Experimental investigations SiC MMCs finished by abrasive flow machining. Mater.
into abrasive flow machining (AFM). Int. J. Mach. Tool Manuf. Process. 30(7): 902–911. https://doi.org/10.1080/
Manuf. 40(7): 1003–1021. https://doi.org/10.1016/S0890- 10426914.2015.1004704
6955(99)00114-5 [20] Mittal S, Kumar V and Kumar H 2016 Multi-objective
[6] Jain R K and Jain V K 2000 Optimum selection of optimization of process parameters involved in micro-
machining conditions in abrasive flow machining using finishing of Al/SiC MMCs by abrasive flow machining
neural network. J. Mater. Process. Technol. 108(1): 62–67. process. Proc. Inst. Mech. Eng. Part L: J. Mater.: Des. Appl.
https://doi.org/10.1016/S0924-0136(00)00621-X 232(4): 1–14. https://doi.org/10.1177/1464420715627292
[7] Jain V K, Ranganatha C and Muralidhar K 2001 Evaluation [21] Guo J, Tan Z E, Au K H and Liu K 2017 Experimental
of rheological properties of medium for AFM process. Mach. investigation into the effect of abrasive and force conditions
Sci. Technol. 5(2): 151–170. https://doi.org/10.1081/MST- in magnetic field-assisted finishing. Int. J. Adv. Manuf.
100107841 Technol. 90: 1881–1888. https://doi.org/10.1007/s00170-
[8] Sehijpal S, Shan H S and Kumar P 2002 Wear behavior of 016-9491-6
materials in magnetically assisted abrasive flow machining. [22] Nagdeve L, Jain V K and Ramkumar J 2018 Differential
J. Mater. Process. Technol. 128(1–3): 155–161. https://doi. finishing of freeform surfaces (knee joint) using R-MRAFF
org/10.1016/S0924-0136(02)00442-9 process and negative replica of workpiece as a fixture. Mach.
[9] Shan H S and Singh S 2002 Development of magneto abrasive Sci. Technol. 22(4): 671–695. https://doi.org/10.1080/
flow machining process. Int. J. Mach. Tools Manuf. 42(8): 10910344.2017.1402929
953–959. https://doi.org/10.1016/S0890-6955(02)00021-4 [23] Kajal S, Jain V K, Ramkumar J and Nagdeve L 2019
[10] Jha S, Jain V K and Komanduri R 2007 Effect of extrusion Experimental and theoretical investigations into internal
pressure and number of finishing cycles on surface roughness magnetic abrasive finishing of a revolver barrel. Int. J. Adv.
in magnetorheological abrasive flow finishing (MRAFF) Manuf. Technol. 100: 1105–1122. https://doi.org/10.1007/
process. Int. J. Adv. Manuf. Technol. 33: 725–729. https:// s00170-017-1220-2
doi.org/10.1007/s00170-006-0502-x [24] Sudhakara D, Suresh S and Vinod B 2020 Experimental
[11] Sehijpal S, Shan H S and Kumar P 2008 Experimental study on abrasive flow machining (AFM): new approach for
studies on mechanism of material removal in abrasive flow investigation on Nano-SiC in the improvement of material
machining process. Mater. Manuf. Process. 23(7): 714–718. removal and surface finishing. J. Bio- and Tribo-Corros. 6:
https://doi.org/10.1080/10426910802317110 24. https://doi.org/10.1007/s40735-019-0321-x
[12] Mamilla R S, Ramkumar J and Jain V K 2009 Experimental [25] Jindal A, Kumar P and Mittal S 2021 Effect of process
investigation and mechanism of material removal in nano parameters on material removal rate in magnetic abrasive
finishing of MMCs using abrasive flow finishing (AFF) flow machining of Al/SiC/B4C metal matrix composites. IOP
process. Wear 266(7–8): 688–698. https://doi.org/10.1016/j. Conf. Ser.: Mater. Sci. Eng. 1091: 012053. https://doi.org/10.
wear.2008.08.017 1088/1757-899X/1091/1/012053
[13] Sankar M R, Jain V K and Rankumar J 2009 Experimental [26] Sharma M, Janardhan G, Sharma V K, Kumar V and Joshi R
investigations into rotating workpiece abrasive flow finish- 2022 Comparative prediction of surface roughness for
ing. Wear 267(1–4): 43–51. https://doi.org/10.1016/j.wear. MAFM finished aluminium/silicon carbide/aluminium triox-
2008.11.007 ide/rare earth oxides (Al/SiC/Al2O3)/REOs) composites
[14] Sankar M R, Jain V K and Ramkumar J 2010 Rotational using a Levenberg–Marquardt Algorithm and a Box–
abrasive flow finishing (R-AFF) process and its effects on Behnken Design. J. Process Mech. Eng. 236(3): 790–804.
finished surface topography. Int. J. Mach. Tools Manuf. https://doi.org/10.1177/09544089211049012
50(7): 637–650. https://doi.org/10.1016/j.ijmachtools.2010. [27] Gorana V K, Jain V K and Lal G K 2004 Experimental
03.007 investigation into cutting forces and active grain density
[15] Brar B S, Walia R S, Singh V P and Sharma M 2011 during abrasive flow machining. Int. J. Mach. Tools Manuf.
Development of a robust abrasive flow machining process 44(2–3): 201–211. https://doi.org/10.1016/j.ijmachtools.
setup. Int. J. Surf. Eng. Mater. Technol. 1(1): 17–23 2003.10.004
[16] Sankar M R, Jain V K, Ramkumar J and Joshi Y M 2011 [28] Singh D K, Jain V K and Raghuram V 2004 Parametric study
Rheological characterization of styrene-butadiene based of magnetic abrasive finishing process. J. Mater. Process.
medium and its finishing performance using rotational Technol. 149(1–3): 22–29. https://doi.org/10.1016/j.jmatpro
abrasive flow finishing process. Int. J. Mach. Tools Manuf. tec.2003.10.030
51(12): 947–957. https://doi.org/10.1016/j.ijmachtools.2011. [29] Jayswal S C, Jain V K and Dixit P M 2005 Modeling and
08.012 simulation of magnetic abrasive finishing process. Int.
[17] Yamaguchi H, Srivastava A K, Tan M A, Riveros R E and J. Adv. Manuf. Technol. 26: 477–490. https://doi.org/10.
Hashimoto F 2012 Magnetic abrasive finishing of cutting 1007/s00170-004-2180-x
tools for machining of titanium alloys. CIRP Ann. 61(1): [30] Gorana V K, Jain V K and Lal G K 2006 Forces prediction
311–314. https://doi.org/10.1016/j.cirp.2012.03.066 during material deformation in abrasive flow machining.
[18] Brar B S, Walia R S, Singh V P and Sharma M 2013 A Wear 260: 128–139. https://doi.org/10.1016/j.wear.2004.12.
robust helical abrasive flow machining (HLX-AFM) process. 038
73 Page 34 of 35 Sådhanå (2023)48:73

[31] Gorana V K, Jain V K and Lal G K 2006 Prediction of [45] Seifu Y, Santhosh K S and Hiremath S S 2016 Modeling and
surface roughness during abrasive flow machining. Int. simulation: machining of mild steel using indigenously
J. Adv. Manuf. Technol. 31: 258–267. https://doi.org/10. developed abrasive flow machine. Procedia Technol. 25:
1007/s00170-005-0197-4 1312–1319. https://doi.org/10.1016/j.protcy.2016.08.230
[32] Wani A M, Yadava V and Khatri A 2007 Simulation for the [46] Ali P, Walia R S, Rastogi V and Tyagi M 2016 Modelling of
prediction of surface roughness in magnetic abrasive flow CNT particles based abrasive laden media used for abrasive
finishing (MAFF). J. Mater. Process. Technol. 190(1–3): flow machining. In: Proceedings of 6th International & 27th
282–290. https://doi.org/10.1016/j.jmatprotec.2007.02.036 All India Manufacturing Technology, Design and Research
[33] Das M, Jain V K and Ghoshdastidar P S 2008 Fluid flow Conference. ISBN: 978-93-86256-27-0, pp. 178–183
analysis of magnetorheological abrasive flow finishing [47] Ahmad S, Gangwar S, Yadav P C and Singh D K 2017
(MRAFF) process. Int. J. Mach. Tools Manuf. 48(3–4): Optimization of process parameters affecting surface rough-
415–426. https://doi.org/10.1016/j.ijmachtools.2007.09.004 ness in magnetic abrasive finishing process. Mater. Manuf.
[34] Kar K K, Ravikumar N L, Tailor P B, Ramkumar J and Process. 32: 1723–1729. https://doi.org/10.1080/10426914.
Sathiyamoorthy D 2009 Performance evaluation and rheo- 2017.1279307
logical characterization of newly developed butyl rubber [48] Verma G C, Kala P and Pandey P M 2017 Experimental
based media for abrasive flow machining process. J. Mater. investigations into internal magnetic abrasive finishing of
Process. Technol. 209(4): 2212–2221. https://doi.org/10. pipes. Int. J. Adv. Manuf. Technol. 88: 1657–1668. https://
1016/j.jmatprotec.2008.05.012 doi.org/10.1007/s00170-016-8881-0
[35] Fang L, Zhao J, Sun K, Zheng D and Ma D 2009 [49] Kathiresan S and Mohan B 2017 Experimental analysis of
Temperature as sensitive monitor for efficiency of work in magneto rheological abrasive flow finishing process on AISI
abrasive flow machining. Wear 266(7–8): 678–687. https:// stainless steel 316L. Mater. Manuf. Process. 33(4): 422–432.
doi.org/10.1016/j.wear.2008.08.014 https://doi.org/10.1080/10426914.2017.1279317
[36] Siddiqui S S and Hameedullah M 2010 Abrasive flow [50] Sambharia J K, Mali H S and Garg V 2017 Experimental
machining performance measures on work-piece surfaces investigation on unidirectional abrasive flow machining of
having different vent/passage considerations for media- trim die work piece. Mater. Manuf. Process. 33(6): 651–660.
outflow. Int. J. Comput. Commun. Inf. Syst. 2(1): 194–199 https://doi.org/10.1080/10426914.2017.1364847
[37] Sadiq A and Shunmugam M S 2010 A novel method to [51] Nagdeve L, Sidpara A, Jain V K and Ramkumar J 2018 On
improve finish on non-magnetic surfaces in magneto-rheo- the effect of relative size of magnetic particles and abrasive
logical abrasive honing process. Tribol. Int. 43(5–6): particles in MR fluid-based finishing process. Mach. Sci.
1122–1126. https://doi.org/10.1016/j.triboint.2009.12.038 Technol. 22(3): 493–506. https://doi.org/10.1080/10910344.
[38] Das M, Jain V K and Ghoshdastidar P S 2011 The out-of- 2017.1365899
roundness of the internal surfaces of stainless steel tubes [52] Petare A C and Jain N K 2018 On simultaneous improve-
finished by the rotational–magnetorheological abrasive flow ment of wear characteristics, surface finish and microgeom-
finishing process. Mater. Manuf. Process. 26(8): 1073–1084. etry of straight bevel gears by abrasive flow finishing
https://doi.org/10.1080/10426914.2010.537141 process. Wear 404–405: 38–49. https://doi.org/10.1016/j.
[39] Brar B S, Walia R S, Singh V P and Sharma M 2012 Helical wear.2018.03.002
abrasive flow machining (HLX-AFM) process. Int. J. Surf. [53] Singh R K, Gangwar S and Singh D K 2019 Experimental
Eng. Mater. Technol. 2(2): 48–52 investigation on temperature affected magnetic abrasive
[40] Mulik R S and Pandey P M 2012 Experimental investiga- finishing of aluminum 6060. Mater. Manuf. Process. 34(11):
tions and modeling of finishing force and torque in ultrasonic 1274–1285. https://doi.org/10.1080/10426914.2019.1628263
assisted magnetic abrasive finishing. J. Manuf. Sci. Eng. [54] Baraiya R, Babbar A, Jain V and Gupta D 2020 In-situ
134(5): 051008. https://doi.org/10.1115/1.4007131 simultaneous surface finishing using abrasive flow machin-
[41] Judal K B, Yadava V and Pathak D 2013 Experimental ing via novel fixture. J. Manuf. Process. 50: 266–278. https://
investigation of vibration assisted cylindrical–magnetic doi.org/10.1016/j.jmapro.2019.12.051
abrasive finishing of aluminum workpiece. Mater. Manuf. [55] Guo J, Feng W, Jong H J H, Suzukic H and Kanga R 2020
Process. 28: 1196–1202. https://doi.org/10.1080/10426914. Finishing of rectangular microfeatures by localized vibra-
2013.811725 tion-assisted magnetic abrasive polishing method. J. Manuf.
[42] Patil V B, Bhanage A S and Patil R S 2014 Analysis and Process. 49: 204–213. https://doi.org/10.1016/j.jmapro.2019.
optimization of process parameters of abrasive flow machin- 11.026
ing process for super finishing of non-ferrous material [56] Harlal S M and Manna A 2012 Simulation of surface
nozzle. Appl. Mech. Mater. 612: 97–104. https://doi.org/10. generated during abrasive flow finishing of Al/SiCp-MMC
4028/www.scientific.net/AMM.612.97 using neural networks. Int. J. Adv. Manuf. Technol. 61:
[43] Kathiresan S, Hariharan K and Mohan B 2015 Prediction of 1263–1268. https://doi.org/10.1007/s00170-012-4091-6
surface roughness in magneto rheological abrasive flow [57] Yunus M and Alsoufi M S 2020 Genetic based experimental
finishing process by artificial neural networks and regression investigation on finishing characteristics of AlSiCp-MMC by
analysis. Appl. Mech. Mater. 766–767: 1076–1084. https:// abrasive flow machining. Int. J. Eng. Technol. Innov. 10(4):
doi.org/10.4028/www.scientific.net/AMM.766-767.1076 293–305. https://doi.org/10.46604/ijeti.2020.4951
[44] Vaishya R O, Walia R S and Kalra P 2015 Parametric [58] Reddy P S, Kesavan R and Ramnath B V 2018 Investigation
investigation of gun metal for surface roughness with hybrid of mechanical properties of aluminium 6061-silicon carbide,
abrasive flow machining process. Int. J. Multidiscip. Curr. boron carbide metal matrix composite. Silicon 10: 495–502.
Res. 3: 736–741 https://doi.org/10.1007/s12633-016-9479-8
Sådhanå (2023)48:73 Page 35 of 35 73

[59] Zhang X F, Sixta M E and De Jonghe L C 2001 Secondary Mater. Res. Technol. 12: 930–946. https://doi.org/10.1016/j.
phases in hot-pressed aluminum-boron-carbon–silicon car- jmrt.2021.03.034
bide. J. Am. Ceram. Soc. 84(4): 813–820. https://doi.org/10. [62] Kumar A, Sharma R, Gupta A K and Gujral R 2021
1111/j.1151-2916.2001.tb00746.x Investigation of biocompatible implant material through
[60] Zhang X F, Yang Q, De Jonghe L C and Zhang Z 2002 Energy WEDM process using RSM modeling hybrid with the
dispersive spectroscopy analysis of aluminum segregation in machine learning algorithm. Sadhana 46: 148. https://
silicon carbide grain boundaries. J. Microsc. 207(1): 58–68. doi.org/10.1007/s12046-021-01676-3
https://doi.org/10.1046/j.1365-2818.2002.01034.x [63] Vendan A S, Kamal R, Abhinav K, Liang G, Niu X and Garg
[61] Kumar J, Singh D, Kalsi N S, Sharma S, Mia M, Singh J, A 2020 Welding and cutting case studies with supervised
Rahman M A, Khan A M and Rao K V 2021 Investigation on machine learning. Eng. Appl. Comput. Methods. https://doi.
the mechanical, tribological, morphological and machinabil- org/10.1007/978-981-13-9382-2
ity behavior of stir-casted Al/SiC/Mo reinforced MMCs. J.