2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Surface Review and Letters, (2023) 2350082 (25 pages)

© World Scientific Publishing Company
DOI: 10.1142/S0218625X23500828

INVESTIGATION AND OPTIMIZATION OF TRIBO-MECHANICAL

BEHAVIOR OF SQUEEZE CASTED Al-Si PISTON ALLOY-BASED
METAL MATRIX COMPOSITE USING RESPONSE SURFACE
METHODOLOGY AND NEURAL NETWORK

K. PRATHEESH*,‡ and V. V. MONIKANDAN†

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

*
Mechanical Engineering Department,
Mangalam College of Engineering,
Kottayam, Kerala, India
†
Research Associate,
Materials Research and Innovation Centric Solutions,
Nagercoil, Tamil Nadu 629204, India
Surf. Rev. Lett. Downloaded from www.worldscientific.com

‡
[email protected]
Received 17 March 2023
Revised 23 June 2023
Accepted 30 June 2023
Published 4 September 2023

The piston in the automobile engine must withstand high stress and temperatures and also have
low weight. The alloy of Al-Si is the most commonly used in the piston. The reinforcement using
ceramics, fibres or nanoparticles will increase the properties of the piston alloy. In this work, the
piston alloy is fabricated as a metal matrix composite of Al-Si reinforced by TiC-MoS2 using the
squeeze casting method. The squeeze casting effect on the matrix and the reinforcement is stud-
ied using tensile testing and microstructure analysis. The strength hardness obtained from the
experimentation gives the highest tensile strength of 330 MPa and the hardness of 110HBN. The
fractography and morphology are investigated using SEM (scanning electron microscope), result-
ing in the lowest porosity of 3.21% obtained in the composite material. The tribological behavior
was also investigated at the condition of dry Sliding using pin on disc tribometer gives the lowest
coefficient friction of 0.31 and the wear rate of 0.0051 mm3/m. The experiment is numerically
designed and optimized using the response surface methodology (RSM). The obtained value is
predicted and validated using a hybrid approach of deep belief network-reptile swarm algorithm
(DBN-RSA). The regression of the parameters is about 99.97% showing that the model is very
accurate to the experimental results. The RMSE of the proposed method implies less error and
shows the accuracy level of the parameters. The result shows that the designed model is best
fitted for the tribological and mechanical properties investigation of metal matrix composite.

Keywords: Squeeze casting; metal matrix composite; tensile testing; microstructure; tribological
behavior.

‡
Corresponding author.

2350082-1

2350082.indd 1 09/02/23 6:39:19 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

1. Introduction the aluminium alloy. This metal reduces the weight

and density by bonding better between the matrix
Aluminium is one of the most available materials
and reinforcement. The titanium carbide increases the
from the earth crust. This material is used in a wide
material’s wear resistance by increasing the reinforce-
range of industrial applications due to its low den-
ment.11 The titanium carbide and molybdenum disul-
sity, high heat performance, thermal expansion, low
fide materials have more advantages for employing
weight to high strength, and mechanical properties.
them in pistons or other engine parts. Because TiC
The automotive industries use cast alloy of Al-Si,
has a high melting point and MoS2 is a dry lubricating
which contains the composition of nickel, copper,
material with a low coefficient of friction.12
and magnesium. It has the characteristics of a low
In order to achieve high wear resistance with low
coefficient of thermal expansion and better fatigue
density and low thermal expansion, the application
properties.1 The component is often used in pistons,
of Al-Si alloy with hyper eutectic phase is the best
cylinder heads, valves and other heat related parts.
choice for automotive parts like pistons, cylinder
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

While employing this component in the engine piston,

heads etc. The wear is classified into three types:
it must undergo cyclic and high temperature thermal
mild, moderate, and severe.13 The eutectic Al-Si
stresses. The engine piston undergoes a temperature
alloy metal matrix composite is achieved by squeeze
above 500°C.2 There is a limit to the use of aluminium casting. Squeeze casting is more advantageous for
and its alloys: low hardness, melting point, elasticity, the mass production of material and is also applied
and resistance to wear. It is overcome by employing
Surf. Rev. Lett. Downloaded from www.worldscientific.com

in producing pistons with high strength and high

the reinforcement of ceramic particles.3 For the appli- mechanical properties. It is also suitable for metal
cation of automobile industries, the alloy of Al-Si is matrix composite preparation because it enhances
developed into a metal matrix composite enriching the alloy matrix during solidification and has bet-
the mechanical and physical properties of the mate- ter reinforcement in a selected component area. The
rial. Metal matrix composites (MMC) are developed porosity formation during composite material man-
by the matrix containing a metal component of alu- ufacturing is neglected in squeeze casting due to the
minium alloy and particle fiber materials as reinforce- pressure applied in the solidification.14
ment.4 The reinforcement increases the base material’s Jiang et al.15 enhanced the aluminium alloy
wear resistance, hardness, strength, and density. The MMC with SiC microns using the squeeze casting
materials such as Al2O3, SiC, B4C, zirconia, TiC and method. The mechanical and microstructural prop-
carbon fibers are the reinforcement.5 erties were investigated, and the result obtained as
The reinforcement of two different particles in the increasing SiC in 6082 aluminium alloy decreases
base matrix is called a hybrid metal matrix compos- the porosity while casting. The author obtained the
ite (HMMC). This method improves the property and optimal SiC particle of 2 wt.% improved by 10.73%
performance by overcoming the drawbacks of each tensile strength, 23.5% yield strength and hardness
other. The commonly used particles for combinations of 25.2% with ductile fracture. Zhu et al.16 studied
are alumina, silicon carbide, boron carbide, tungsten the combined squeeze and stirred casting employed
carbide etc.6 The hybrid composite of Nanoparticles to fabricate 6082 aluminium alloy metal matrix
and ceramic contents in aluminium alloy metal also composites. The strength increases achieved during
added the high property outcomes of high perfor- experimentation were 8.92% in ultimate tensile
mance, machinability, lightweight, wear resistance strength, 12.16% in yield strength, and an introduc-
and improved tribological behavior, which are multi- tion of SiC nanoparticles also redefined 12.60% in
walled carbon nanotube (MWCNT) and SiC,7 Al2O3- elongation, including grain structure.
graphene,8 Al2O3-SiC.9 Attractive results are achieved Shabani et al.17 employed three different casting
from these ceramic materials by the ceramic particle of methods: gravity sand casting, compo casting and
MoS2. It has the specific characteristics of softness, high squeeze casting. The scanning electron microscope is
modulus of elasticity, specific gravity and thermal sta- used to analyze the microstructure. As a result, the
bility. It is a promising material for reinforcing alumin- squeeze casting is better than other castings because
ium alloy composite.10 The high hardness and strength some dendritic fractures and agglomerations were
were achieved during the reinforcement of TiC with observed in compo casting. Hao et al.18 investigated

2350082-2

2350082.indd 2 09/02/23 6:39:19 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

the hyper eutectic alloy by testing its mechanical and maximum strength of 329 MPa. Karthik et al.24
properties microstructure using SEM, transmission investigated the metal matrix composite employing
electron microscopy (TEM) and optical systems. AA2219, and the reinforcement of CeO2 was varied
It studied the squeeze casting effects in the hyper at wt.%. The microstructure showed a significant
eutectic Al-Si-Cu-Mg alloy. After the squeeze casting reduction of dendritic, and the grain structure was
is redefined, silicon and rosette grains are observed refined. The result obtained that 2% of CeO2 gives
in the alloy metal. The result showed 187.67 MPa of the best outcome of 124HV hardness and 87 MPa
tensile strength, 2.33% of elongation and 142.5 HB yield strength by using these factors in the Taguchi
of hardness. Also, less wear rate was obtained from method and gives an optimum influencing factor of
the semi-solid squeeze casting. wear and coefficient of friction (COF) by using sig-
Chandra et al.19 analyzed the micrographics of nal to noise ratio and ANOVA analysis.
aluminium alloy manufactured using squeeze cast- From this literature, the aluminium alloy with
ing. The material properties were improved at 15% metal matrix composite increases its properties and
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

strength and 51% ductility. The less dissolution of microstructure and is also suitable for many appli-
Al2Cu was observed in high pressure squeeze cast- cations. The squeeze casing provides a refined grain
ing, showing the highest wear property resistance. structure, increasing its hardness, ductility and
Sohag et al.20 fabricated the Al-Cu metal matrix reduced porosity. This work focused on studying
composite with the reinforcement material of TiC the fabrication of metal matrix composite using the
Surf. Rev. Lett. Downloaded from www.worldscientific.com

and SiC using the stir casting process. The uniform Al-Si piston alloy matrix and hybrid TiC and MoS2
distribution of reinforcement at 2% to 8% variation reinforcement by the Squeeze casting method. This
of SiC-TiC particles is observed in the XRD (X-Ray combination of hybrid materials is rarely used in
Diffraction) analysis. The investigation also achieved the existing work. Also, these materials will enhance
a low wear rate and density with high hardness. the hardness, wear, strength, self-lubrication and
Sarfraz et al.21 fabricated the Al6061-sic compos- mechanical properties of the piston alloy. The work
ite by squeeze casting and studied the mechanical dealt with improving the properties of the piston
and microstructural parameters. The Box Behnken alloy as it is undergoing a high temperature envi-
design is used for experimentation and finding out ronment and high stress loadings. The tribological
the influence of casting parameters on the strength and mechanical tests are conducted, and the pro-
and hardness of the material. The result obtained cess parameters are optimized using RSM, then val-
was optimum strength elongation and hardness, idated and predicted using a hybrid neural network.
and analysis of variance (ANOVA) shows the sig- The proposed methodology, material fabrication
nificance of the experimentation. Christy et al.22 and experimentation of this work are explained in
optimized the mechanical properties, process param- Sec. 2. The design of the experiment using RSM and
eters and alumina-reinforced aluminium alloy wheel DBN-RSA is detailed in Sec. 3. The result obtained
microstructure using squeeze casting. The optimiza- from this work is discussed in Sec. 4. Finally, the
tion performed using the L5 and L6 results of the findings of the work and future development are
Taguchi-Grey relational method experiment was detailed in the conclusion in Sec. 5.
the best out of 9 experiments and it achieves a low
porosity, hardness and high compressive strength,
respectively. 2. Materials and Methodology
Natrayan et al.23 fabricated AA6061/Al2O3/
SiC/Gr hybrid metal matrix composite using the
2.1. Research methodology
squeeze casting technique. The optimization of pro- In this work, an Al-Si piston alloy-based metal
cess parameters of squeeze casting used the L16 matrix composite with hybrid reinforcement mate-
orthogonal array method and found out the con- rial of TiC-MoS2 was developed by squeeze casting
sequences of hardness and tensile strength. The to study the mechanical and tribological behaviors.
optimum parameter from the Taguchi method is The squeeze casting enhances the eutectic phase
predicted using ANN and gives an accuracy of 95%. of MMC, which improves the strength, cast ability
The result showed a maximum hardness of 131 HV and wear resistance. The combination of TiC-MoS2

2350082-3

2350082.indd 3 09/02/23 6:39:19 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.
Surf. Rev. Lett. Downloaded from www.worldscientific.com

Fig. 1. (Color online) Layout of the proposed methodology.

increases the hardness and frictional resistance of 2.2. Materials and experimentation
the piston alloy. This is determined by experiment-
In the proposed study, the piston alloy of Al-Si is
ing with hardness, strength, and wear tests. The selected as the matrix material, and the composition
wear is analyzed using a pin on the disc tribometer of this material is given in Table 1. The prepara-
by varying the load velocity and sliding distance. tion process of the composite material is done by
The mechanical properties are analyzed by testing using squeeze casting because the automobile engine
the hardness and tensile strength of the material. parts are effectively cast in this method, with heat
The material’s microstructure was also analyzed treatable and high strength. Also, it is a suitable
using a scanning electron microscope to study grain alternative material for pistons.25 The reinforcement
size and fracture mechanism. In this proposed work, material of 3 mm particle size of Tic is used as the
the statistical design is modeled in the response sur- primary and secondary material of self-lubricating
face methodology of the Box Behnken design. The MoS2 at 30 mm with 99% purity. The properties of
tribological and mechanical properties obtained the reinforcement material are given in Table 2.
from experimentation are optimized in this method
Table 2. Properties parameters of TiC particle.
and predicted and validated using a hybrid neural
network called DBN-RSA. The layout of the pro- Parameters TiC MoS2
posed methodology is given in Fig. 1.
Density 15.63 g/cm3 5.06 g/cm3
Hardness 2400 kg/mm –
Table 1. Composition of Al-Si piston alloy.
Thermal conductivity 84.02 w/mk 3.5 w/mk
Youngs modulus 69.1 GPa 238 GPa
Alloy Si Cu Mg Ni Ti Mn Fe Al
Electrical resistivity 2*10−7 Ωm 105 Ωm
Composition 12.09 3.10 1.13 1.78 0.01 0.04 0.19 Bal Melting point 3160°C 1185°C

2350082-4

2350082.indd 4 09/02/23 6:39:20 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

For squeeze casting, these materials are melted in

the electrical furnace.26,27 The melting temperature
is 800°C, measured with a thermocouple connected
to the furnace. After adding the reinforcement, the
alloy gets a complete melt stage. The degasification
is done by introducing the nitrogen for 60 min, and
slag from the alloy is also stripped out. The addi-
tion of TiC and MoS2 increases the matrix materi-
al’s nucleation, reduces crystallization, and enhances
grain refinement. The die for casting is preheated Fig. 2 Tensile testing specimen dimension.
to a temperature of 250°C; then, the molten metal
is poured into the die at a temperature of around
800°C. The hydraulic press applies the squeeze pres-
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

sure in which the molten metal is kept at the same

pressure until the metal solidification. The compos-
ite is prepared at different levels of three casting
parameters: melt temperature, squeeze pressure,
and die temperature. The squeezing pressure varies
Surf. Rev. Lett. Downloaded from www.worldscientific.com

from 50 MPa to 150 MPa, and preheating 100*100*

100 mm of a die at 100°C to 300°C.

2.3.  xperimentation of composite

E
material Fig. 3. (Color online) Pin on disc tribometer.
After casting, the composite ingots are cut into the
cross section of 20*20*20 mm. It was given to the with a spinning disc. The material having the dimen-
heat treatment at the temperature of 400°C at 5 h. sion of 30 mm*6 mm*6 mm is used for testing. The
The sample specimens are then polished for metal- steel disc of EN31 is used as a counterpart for the test
lurgical analysis done by Tescan Vega II Scanning specimen, which is cleansed before the experiment.
Electron Microscope. With the help of a computerized The pin is also cleansed and polished using acetone.
image processing tool called ImageJ, the %porosity The volumetric wear loss obtains the wear rate of the
of obtained SEM images is calculated. Stress-strain, study to the sliding distance in which the losses and
strength, elongation and elasticity of the material are frictional forces are recorded during the experimenta-
studied by tensile testing, which is performed on a tion. The friction coefficient was also obtained using
universal testing machine using the Instron Model the frictional force of sliding action in the tribometer.
1195-5500R. These experimentations are carried out Figure 3 represents the pin on the disc tribometer.
at room temperature. An Indentec Brinell hardness
machine tests the hardness of this material with an
indenter ball diameter of 2.5 mm and 62.5 kg of load. 3. Design of Experiment
The sample selected for hardness testing has a dimen-
3.1. Response surface methodology
sion of 15 mm*20 mm. The dimension of the tensile
testing material is given in Fig. 2. The statistical design of this work is mathemati-
The tribological behavior of the material was also cally modeled in the response surface methodology.
evaluated using a pin on the disc tribometer using It evaluates the effects of factors in the variables
a DUCOM TR-20LE apparatus. The pin and disc and finds the significance of these factors over the
diameters selected for testing are 6 mm and 30 mm, interest variables.28 The main aim of the response
respectively. The material is tested under dry condi- surface methodology is to optimize the expected
tions and at room temperature with an applied load responses. This is achieved by implementing second

2350082-5

2350082.indd 5 09/02/23 6:39:21 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

Table 3. Mechanical factors and levels of mechanical k k k k

parameters. y = b0 + ∑ bi xi + ∑ bii xi2 + ∑ ∑ bij xij . (1)
i =1 i =1 i =`1 j =i +1
Level
From the equation, responses are denoted as Y,
Code Variables Unit Low High intercepts of the model are b0, the first order, linear and
A Squeeze pressure MPa 50 150 quadratic coefficients are bi, bii, and bij. The analysis of
B Melt temperature °C 600 800 variance (ANOVA) is used to evaluate the significant
C Die temperature °C 100 300 level of the model. The RSM analysis is performed to
identify the optimal parameters of mechanical proper-
Table 4. Factors and levels of tribological parameter. ties and tribological properties. The factors and level
of input parameters are given in Tables 3 and 4.
Level
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

Code Variables Unit Low High

3.2. Hybrid DBN-RSA
A Load N 10 30
B Sliding velocity m/s 1 2 This work is predicted and validated using a hybrid
C Sliding distance m 250 750 deep belief network. The DBN is a neural network
based on the number of RBM layers and neurons.29
Surf. Rev. Lett. Downloaded from www.worldscientific.com

order polynomial based Box Benhken design (BBD) The multiple layers of restricted Boltzmann machines
method which is available in RSM tool. The regres- have no direct connection between one layer to another.
sion obtained from the polynomial solution gives the The major layers of RBM are visible and hid-
fitness percentage of the model. The second-order den layers, and a convergence algorithm trains each
equation for BBD is module of this RBM. Also, the output from the

Fig. 4. (Color online) Layout of DBN.

2350082-6

2350082.indd 6 09/02/23 6:39:24 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

visible layer is given to the output of the hidden using the given parameters, and hence the optimized
layers. The tribological and mechanical properties weights are observed. The contrastive divergence
are predicted and validated using this hybrid net- is utilized to simplify the kth iteration probability
work. The input considered in this work is squeeze function because of the slower divergence rate in
casting process parameters and tribological testing RBM. The training set data from the previous unit
parameters. The inputs are given to the visible layer is initialized to the proceeding visible unit of RBM.
of the RBM, which are Squeeze pressure, melt and
die temperature, velocity, load, and sliding distance. ∆wij = h〈pi q j 〉data − 〈pi q j 〉recon. (9)
The output desired for prediction is tensile strength, The weights are updated by using the above formula
hardness, COF, and wear rate. The layout of the based on the CD, where recon is the reconstruction of
proposed model is given in Fig. 4. the visible unit and ‘h’ is the learning rate. In RBM, the
It consists of no interconnection between visible training data of the first hidden layer is fed into the vis-
and hidden layers incorporated with energy based
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

ible layer of the second RBM layer. The training data

networks. The energy and joint distribution func- of the second hidden layer is fed into the third visible
tions for visible and hidden layers are as follows: layer of RBM. The reptile search optimization algo-
1 rithm is employed to fine-tune all completed pre-train-
p(p, q) = exp(−E (p, q)), (2) ing weights from all layers of the RBM. The optimal
z
units are obtained by using this swarm algorithm.
Surf. Rev. Lett. Downloaded from www.worldscientific.com

Z = ∑ exp(−E (q, p)), (3)

p,q This optimization begins with population X, in
n m n m which the optimum solution is obtained by generat-
E (p, q) = − ∑ ai pi − ∑ bj q j − ∑ ∑ pi wijq j , (4) ing this best solution by iterations.30 This algorithm
i =1 j =1 i =1 j =1
is based on the hunting behavior of crocodiles. Its
encircling and hunting behavior of prey is depicted as
where p and q are the binary state units, and ai
the exploration and exploitation of searching optimal
and bi are the biases of hidden and visible layers,
solutions. The algorithm initializes with the set of
respectively. The probability of visible and hidden
populations, which are the weights obtained from the
layers p(p) and p(q) is calculated using their respec-
neural network. The first phase of the algorithm is an
tive expression below.
The probability of visible layer p exploration of search space it discovers the wide-area
after the endeavors. The space searching is based
1 on two conditions similar to crocodile movement of
p(p) = ∑ p(p, q) = ∑ exp(−E (p, q)). (5)
q q z high and belly walking. To obtain a significant search
area, half the iterations are satisfied by high walk-
The probability of a hidden layer q ing and half by belly walking conditions. The given
1 equation updates the position of the search space.
p(q) = ∑ p(p, q) = ∑ exp(−E (p, q)). (6)
p p z
Best (t ) ×−η (t )
 j (i , j )

Since there is no interlinking between the visible  × β − R (t ) × rand, t ≤ T
(i , j )
layer and the hidden layer, the activation function  4
Xij(t +1) = 
probabilities of ith visible layer and jth hidden layer Best (t ) × x
j (r , j ) × Es(t )

are given as  T T
 × rand, t ≤ 2 and t >
p   4 4
P(q j = 1 p) = s  ∑ Wij pi + bj  , (7)
i =1  (10)
q  Bestj(t) is obtained best solution of position jth, a
P(p j = 1 q) = s  ∑ Wij qi + a j  , (8)
i =1  random number between 0&1 is denoted as randand
r1 is the random number between [1 N] with the
where s(p) = 1 + exp(p)−1, is the sigmoid function. random position of x(r,j). The current iteration is
The probability function of RBM was maximized depicted as t and a total number of iterations is T,

2350082-7

2350082.indd 7 09/02/23 6:39:38 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

the accuracy of the exploration is controlled by . The The best obtained solution for the position is given as
hunting operator is given in the following equation:
x(r , j ) − M (xi )
p(i , j ) = α + , (14)
h(i, j ) = Best j (t ) × P(i , j ). (11) Best j (t ) − UB( j ) − LB( j ) + ε

The search area is reduced by using the value of the where the average position is denoted as M(i, j)
reduction function
1 n
M(xi ) = ∑ x(i, j ). (15)
Best j (t ) − x(r , j ) n j =1
R(i, j ) = , (12)
Best j (t ) + e After this exploration phase, the optimal solution is
obtained by exploitation which conducts intensified
where e is the smallest value search near the optimal solution thus, the communi-
The probability ratio of evolutionary sense is cation between optimal solutions is increased. This
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

obtained by phase uses coordination and cooperation strategies

in the search space to obtain the target. The equa-
Es (t ) = 2 × r3 × (1 − T1 ). (13) tion for exploitation is given as follows:
Surf. Rev. Lett. Downloaded from www.worldscientific.com

Pseudo-code of algorithm
Initialization phase
Initialize the sensitive parameters of RSA
Population initialization randomly, X
While (t<T)do
Calculate the X fitness function
Best solution
Es equation updation
Start the RSA
For (i=1 toN) do
For(j=1 to n)do
h, R, P values are updated using equation
if (t ≤ T4 ) then
X(i , j ) = Best j (t ) ×−η(i , j )(t ) × β − R(i , j )(t ) × rand (high walking)
else if (t ≤ 2 T4 and t > T4 ) then
X(i , j ) = Best j (t ) × x(r , j ) × Es(t ) × rand, belly walking
else if (t ≤ 3 T4 & T > 2 T4 ) then
Xij (t +1) = Best j (t ) × R(i, j )(t ) × rand (hunting coordination)
else if (t ≤ T and t > 3 T4 )
Xij (t +1) = Best j (t ) − η(i, j )(T ) × ε − R(i , j )(t ) × rand (hunting
cooperation)
end if
end for
t=t+1
end while
return best (X)

2350082-8

2350082.indd 8 09/02/23 6:39:56 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

Best j (t ) × R(i , j )(t ) Tensile strength: The tensile strength is carried

 out for all experimental runs obtained from RSM,
 T T and the values are tabulated in Table 5. Three exper-
 × rand, t ≤ 3 & T > 2 ,
Xij(t +1) =   4 4 iments were conducted for each experimental run,
Best (t ) − η (T ) × ε − R (t )
 j (i , j ) (i , j ) and the average was calculated for further analysis
 T using RSM. The statistical model is developed to
 × rand, t ≤ T and t > 3 .
 4 find the accuracy of a design. The significant result
(16) obtained from the analysis is shown in Table 6. The
ANOVA implies the significance of the experimented
This mechanism traps the local optima to form the
tensile strength of squeeze-cast composite material.
optimal solution and maintain the diversity over the
The polynomial equation of the statistical design of
candidate solution. The Pseudo-code for this algo-
tensile strength is given as
rithm is given in Sec. 3.
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

155.125 + 0.518A-0.07225B

 + 0.07525C − 0.00045AB
4. Result and Discussion 
Tensile strength = −0.00265AC-5e-05BC ,

4.1. Mechanical properties 2
 + 0.00566A + 0.00019B2

 + 0.00079C2
Surf. Rev. Lett. Downloaded from www.worldscientific.com

The material characteristics, namely mechanical

strength and hardness of the prepared composite by (17)
squeeze casting, are explained in this section. The
where A, B, and C denote squeeze pressure, melt
best outcomes of mechanical properties are obtained temperature and die temperature, respectively.
with squeeze casting. In this work, the composite Equation (17) implies the influence of varying
materials showed excellent hardness, high strength squeeze casting parameters on tensile strength. The
and reduced porosity in most casting processes. significance of the model is given in ANOVA, as
The variation of pressure, melt temperature and die shown in Table 6. The p-value in the analysis of vari-
temperature influences on mechanical properties are ance shows the significant term in which the value
given in Table 5. should be less than 0.0500. The f-value obtained for
Table 5. Experimental factors and the response of mechanical properties.

Factor 1: A Factor 2: Factor 3: Response: 1 Response: 2 Response:

squeeze pressure B melt die temperature hardness tensile strength 3 porosity
Run (MPa) temperature (°C) (°C) (HBN) (MPa) (%)
1 50 600 200 75 220 6.20
2 100 700 200 87 260 5.30
3 100 700 200 86 261 5.20
4 150 800 200 104 330 3.95
5 100 700 200 87 259 5.40
6 50 800 200 81 252 5.70
7 150 600 200 91 307 4.95
8 50 700 100 74 218 6.05
9 100 700 200 87 264 5.20
10 100 700 200 86 262 5.30
11 150 700 100 92 323 4.70
12 100 600 100 78 251 5.70
13 100 600 300 93 263 4.40
14 100 800 100 87 280 5.30
15 50 700 300 84 270 5.10
16 150 700 300 111 322 3.21
17 100 800 300 102 290 3.81

2350082-9

2350082.indd 9 09/02/23 6:39:58 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

Table 6. ANOVA table for tensile strength.

Source Sum of squares df Mean square F-value p-value Remark

Model 17090.19 9 1898.91 103.81 <0.0001 Significant
A-Squeeze pressure 12960.50 1 12960.50 708.50 <0.0001 Significant
B-Melt temperature 1540.13 1 1540.13 84.19 <0.0001 Significant
C-die temperature 666.13 1 666.13 36.41 0.0005 Significant
AB 20.25 1 20.25 1.11 0.3277 Not significant
AC 702.25 1 702.25 38.39 0.0004 Significant
BC 1.0000 1 1.0000 0.0547 0.8218 Not significant
A² 843.04 1 843.04 46.09 0.0003 Significant
B² 15.20 1 15.20 0.8309 0.3923 Not significant
C² 262.78 1 262.78 14.37 0.0068 Significant
Residual 128.05 7 18.29 – – –
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

Lack of Fit 113.25 3 37.75 10.20 0.0241 Significant

Pure Error 14.80 4 3.70 – – –
Cor Total 17218.24 16 – – – –

tensile strength is 103.81, and the chance of noise polynomial equation for the statistical design of
Surf. Rev. Lett. Downloaded from www.worldscientific.com

is only 0.01%. Also, the obtained lack fit and the hardness response is given as
noise chance are only 10.20 and 2.41%. This value
88.625 − 0.181A − 0.05175B
shows that the model terms are significant. Besides, 
the regression analysis showed the best accuracy  − 0.08925C + 0.00035AB

with the experimental results. The regression coeffi- Hardness =  + 0.00045A AC + 1.05303e ,

cient (R2) achieved for the tensile strength is 0.9926;  − 19BC + 0.00028A2

the adjusted R2 is 0.9830, and the predicted R2 is  + 4.5e-05B2 + 0.000295C2
0.8934. It shows the agreeable fitness result with a (18)
difference of less than 0.2. The signal-to-noise ratio
measured for the tensile strength is 33.076, showing where A, B, and C denote squeeze pressure, melt
the experimental values’ high precision. temperature and die temperature, respectively.
Figure 5 shows the comparison of predicted and Equation (18) shows the impact of squeeze con-
actual results, the contour plot and the perturbation trol variables on the hardness of the composite mate-
plot of tensile strength. The maximum tensile strength rial. It is used to predict the experimental response
obtained from the experimentation is 330 MPa at the for the given input factors. The significant level of
squeeze pressure of 150 MPa, melt and die tempera- the experimental and modeled results is given in the
ture of 800°C, and 200°C, respectively. Due to the high ANOVA Table 7. For hardness, the F-value obtained
temperature of the melt, nuclei formation is increased from the statistical result is 204.22, which shows the
and accelerates nucleation rates in the melt’s solidifi- significance of the model with 0.015 of noise possi-
cation. On the other hand, the die temperature also bility. A, B, C, AB, AC, and C² are the terms used
enhances during the solidification process. Therefore, to predict significant values. A p-value less than
the equal dispensation of reinforcement particles 0.050 defines the significant terms from the designed
increases with increasing of all parameters. At the model, whereas more than 0.050 makes the model
same time, the reinforcement dissolves in the matrix insignificant. The lack of fit obtained is 5.28, and the
chance of noise is also 7.09%.
to strengthen the alloy-based composite’s structure.
The graphs obtained from the statistical design
Hardness: The statistical equation obtained for are given in Fig. 6. The highest hardness of 111HBN
hardness response from the model is explained as is obtained for the experimental result of the com-
follows. An analysis of variance explains the accu- posite material. The pressure and temperature used
racy of the model. It implies the proposed model to obtain such hardness values are 150 MPa, 700°C
is significant for the experimental results. The (melt) and 300°C (die), respectively. The graphs

2350082-10

2350082.indd 10 09/02/23 6:39:59 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a) (b)
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(c)
Fig. 5. (Color online) Statistical result of tensile strength, (a) prediction plot, (b) perturbation plot, (c) contour plot.

Table 7. ANOVA of hardness.

Source Sum of squares df Mean square F-value p-value Remark

Model 1562.29 9 173.59 204.22 <0.0001 Significant
A-Squeeze pressure 882.00 1 882.00 1037.65 <0.0001 Significant
B-Melt temperature 171.13 1 171.13 201.32 <0.0001 Significant
C-die temperature 435.13 1 435.13 511.91 <0.0001 Significant
AB 12.25 1 12.25 14.41 0.0067 Significant
AC 20.25 1 20.25 23.82 0.0018 Significant
BC 0.0000 1 0.0000 0.0000 1.0000 Not significant
A² 2.06 1 2.06 2.43 0.1632 Not significant
B² 0.8526 1 0.8526 1.00 0.3499 Not significant
C² 36.64 1 36.64 43.11 0.0003 Significant
Residual 5.95 7 0.8500 – – –
Lack of Fit 4.75 3 1.58 5.28 0.0709 Not significant
Pure Error 1.20 4 0.3000 – – –
Cor Total 1568.24 16 – – – –

2350082-11

2350082.indd 11 09/02/23 6:39:59 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a) (b)
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(c)
Fig. 6. (Color online) Statistical results of hardness (a) prediction, (b) perturbation, (c) contour plot.

imply the variation of these parameters in the mate- 4.27 + 0.012A + 0.0033125B
rial hardness. It is achieved by the high density 
 + 0.0177375C − 2.5e-05AB
obtained during the casting process, which restricts 
Porosity =  − 2.7e-05AC − 4.75e-06BC , (19)
the particles’ dislocation and increases the hardness 
 − 2.35e-05A2 − 2.125e-06B2
due to high temperature and pressure. 
 − 4.5625e-05C2
Porosity: Porosity is the important parameter
where ‘A’ represents squeeze pressure, ‘B’ is for
that depicts the strength of the composite during
melt temperature and ‘C’ denotes die temperature
the squeeze casting. The statistical result obtained
respectively.
from the response surface methodology is explained From Eq. (19), the influence of squeeze param-
in this section. The fitness of the experimental result eters on the porosity of composite material is
to the statistical design is analyzed using ANOVA described. It is used to predict the experimental
in Table 8. In contrast, the polynomial equation for response for the given input factors. The significant
porosity is expressed in the following equation: level of the experimental and modeled results are

2350082-12

2350082.indd 12 09/02/23 6:40:00 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

Table 8. ANOVA of porosity.

Source Sum of squares df Mean square F-value p-value Remark

Model 10.12 9 1.12 94.82 <0.0001 Significant
A-Squeeze pressure 4.87 1 4.87 410.36 <0.0001 Significant
B-Melt temperature 0.7750 1 0.7750 65.34 <0.0001 Significant
C-die temperature 3.42 1 3.42 288.27 <0.0001 Significant
AB 0.0625 1 0.0625 5.27 0.0554 Not significant
AC 0.0729 1 0.0729 6.15 0.0423 Significant
BC 0.0090 1 0.0090 0.7609 0.4120 Not significant
A² 0.0145 1 0.0145 1.23 0.3049 Not significant
B² 0.0019 1 0.0019 0.1603 0.7008 Not significant
C² 0.8765 1 0.8765 73.90 < 0.0001 Significant
Residual 0.0830 7 0.0119 – – –
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

Lack of Fit 0.0550 3 0.0183 2.62 0.1875 Not significant

Pure Error 0.0280 4 0.0070 – – –
Cor Total 10.20 16 – – – –
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(a) (b)

(c)
Fig. 7. (Color online) Statistical result of porosity, (a) prediction plot, (b) perturbation plot, (c) contour plot.

2350082-13

2350082.indd 13 09/02/23 6:40:01 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

given in the ANOVA (Table 8). For porosity, the temperature of 300°C, the percentage of porosity is
F-value obtained from the statistical result is 94.82, 3.21%, respectively.
showing the model’s significance with 0.015 of the
possibility of noise. A, B, C, AB, AC, and C² are
the terms used to predict significant values. The lack 4.2.  icrostructure and fractography
M
of fit obtained is 2.62, and the chance of noise is analysis
18.09%. The alloy matrix and reinforcement undergo sev-
The statistical results are depicted in the graph eral phase transformations like eutectic Si, a-Al,
given in Fig. 7. The lowest porosity is achieved in intermetallic Al, Ti, and AlCuNi. Therefore, the
the squeeze casting of composite material. Usually, microstructure of processed samples is obtained
the composite material shows more porosity. But the with the help of the SEM method. As expected, the
prepared Al-Si alloy with TiC-MoS2 shows the low- MoS2 and TiC formed stable thermodynamic prop-
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

est level at high temperature and pressure of squeeze erties with the matrix alloy metal. According to a
casting. This is due to the enhancement of nucle- microstructural study, unreinforced material in the
ation, which increases grain growth in the pores. composite makes a coarse grain structure than the
The improved bonding between Al-Si and reinforced reinforced composite. The squeeze pressure appli-
materials also helps reduce porosity. The maximum cation in the casting process during solidification
Surf. Rev. Lett. Downloaded from www.worldscientific.com

reduction in porosity is achieved at a pressure of makes the uniform dispersion of reinforcement over
150 MPa. At a melt temperature of 700°C and a die the matrix. It also causes an increase in Al alloy

(a) (b)

(c)
Fig. 8. Microstructure and fractography of composite material (a) Microstructure of composite material, (b) Worn
­surface, (c) Tensile fracture surface.

2350082-14

2350082.indd 14 09/02/23 6:40:02 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

nucleation; thereby, reduced dendritic formation and MPa squeeze pressure, 700°C melt temperature
fine grain structure were noticed. Thus, squeeze cast- and 300°C die temperature, had more favorable
ing improves the interfacial bonding and enhances mechanical properties, the tribological properties
the grain refinement, whereas enhanced mechanical of squeeze-cast samples are only examined for these
properties and thermal stability are achieved from casting recipes. The major tribological parame-
casting. Based on the mechanical strength evalua- ters obtained from this analysis are friction and
tion results, the maximum strength of the TiC-MoS2 wear rate coefficients. The experimental results
-reinforced Al-Si composite can be achieved for 150 are designed in the RSM using the Box Behnken
MPa squeeze pressure, 700°C melt temperature and design. Similar to the mechanical property study,
300°C die temperature. The sample’s microstructure three samples cast at a 150 MPa squeeze pressure,
prepared at such parameter is analyzed using SEM, 700°C melt temperature and 300°C die temperature
as shown in Fig. 8. are used to calculate the average values of each
Figure 8(c) shows the tensile fracture surface of tribological property for all specified experimental
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

a squeeze casted composite and the plastic defor- runs, given in Table 9.
mation of the broken Si in the matrix. Whereas
Coefficient of friction: The coefficient of fric-
the worn surface and the microstructure of the
tion is the important tribological property related
prepared composite are analyzed using Figs. 8(b)
to the ratio of frictional force to the normal force of
and 8(a). A mixed fracture mode is obtained in the
Surf. Rev. Lett. Downloaded from www.worldscientific.com

the component. The experimental result is obtained

composite, less than the alloy matrix evaluation.
from the pin on disc testing at dry sliding condi-
Also, the dimples and porosity are reduced due to
tions under sliding velocity, sliding distance, and
the squeezing pressure. The higher tensile strength,
load variation. The statistical design is modeled in
yield strength and elongation were also observed
REM using the Box Behnken design. The polyno-
in the reinforced composite. According to Fig. 8,
mial equation obtained from the design is also given
no debonding is observed in the fractured surface
in the following equation:
because there is good bonding between TiC-MoS2
and Al-Si material. The MoS2 in the reinforce- 0.64425 + 0.002695A
ment forms a thin lubricating layer in the sliding 
 − 0.2379B − 0.0001772C
surface, thus increasing the wear rate of the com- Coefficient 
=  − 0.001AB-1e-06AC , (20)
posite during tribological testing. This indicates of friction 
 + 2e-05BC + 1.7e-05A 2
that the wear rate of the proposed composites is 
highly dependent on the sliding surface. However,  + 0.0468B2 + 6.72e-08C2
the coefficient of friction gets reduced due to the
where A, B and C are the load, sliding velocity and
solid lubrication formed by MoS2 on the compos-
sliding distance, respectively.
ite surface. The thickness of the lubricating layer
The analysis of variance is used to find the
increases with decreasing coefficient of friction
significant level of the statistical result, which is
thereby reduced shear stress and surface contact
given in Table 10. For the coefficient of friction,
noticed from the composites. Based on the results, the F value obtained for the model is 85.2, which
it was confirmed that the addition of TiC-MoS2 shows the model best fits the experimental results.
increases the mechanical and tribological proper- The significant terms AB, AC, BC, A², B², and
ties of the Al-Si composite. C² are used for prediction. The significant term is
obtained by the P-value, which should be less than
0.05. The regression of this analysis shows the con-
4.3. Tribological properties fidence level that the regression achieved for COF
The tribological test is conducted on the pin on is 0.9910, showing that the design is best fitted to
the disc tribometer by varying the load, sliding the experiment.
velocity and sliding distance. Since TiC-MoS2- The statistical results of prediction, pertur-
reinforced Al-Si composite samples, which were bation and contour plots are given in Fig. 9. The
produced according to the processing recipe of 150 graph depicts how the increase in load, velocity and

2350082-15

2350082.indd 15 09/02/23 6:40:03 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

Table 9. Factors and responses of tribological parameters.

Factor B Factor C Response 2

Factor A sliding sliding Response 1 wear rate
Run load (N) velocity (m/s) distance (m) COF (mm3/m)
1 20 1 750 0.4 0.0039
2 20 2 750 0.31 0.0051
3 20 1 250 0.46 0.0035
4 10 1.5 250 0.37 0.0037
5 20 1.5 500 0.36 0.0042
6 30 1 500 0.44 0.004
7 10 1 500 0.41 0.0032
8 10 2 500 0.31 0.0046
9 20 1.5 500 0.36 0.0042
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

10 30 2 500 0.31 0.005

11 20 1.5 500 0.36 0.0042
12 10 1.5 750 0.32 0.0041
13 20 2 250 0.34 0.0048
14 30 1.5 250 0.41 0.0043
15 30 1.5 750 0.34 0.0045
Surf. Rev. Lett. Downloaded from www.worldscientific.com

16 20 1.5 500 0.36 0.0042

17 20 1.5 500 0.36 0.0043

Table 10. ANOVA of COF.

Source Sum of squares df Mean square F-value p-value Remark

Model 0.0305 9 0.0034 85.82 < 0.0001 Significant
A-load 0.0015 1 0.0015 38.33 0.0004 Significant
B-Velocity 0.0231 1 0.0231 585.76 < 0.0001 Significant
C-sliding distance 0.0050 1 0.0050 126.72 < 0.0001 Significant
AB 0.0001 1 0.0001 2.53 0.1554 Not significant
AC 0.0000 1 0.0000 0.6336 0.4522 Not significant
BC 0.0000 1 0.0000 0.6336 0.4522 Not significant
A² 0.0000 1 0.0000 0.3084 0.5960 Not significant
B² 0.0006 1 0.0006 14.61 0.0065 Significant
C² 0.0001 1 0.0001 1.88 0.2124 Not significant
Residual 0.0003 7 0.0000 – – –
Lack of Fit 0.0002 3 0.0001 2.31 0.2185 Not significant
Pure Error 0.0001 4 0.0000 – – –
Cor Total 0.0308 16 – – – –

distance greatly impacts the coefficient of friction. from the experimentation is 0.31 at the load of 20N
The increasing coefficient of friction is obtained in and the velocity of 2 m/s.
the high load. Still, it is reduced in the large veloc-
ity because the self-lubricating behavior in the rein- Wear rate: The wear rate is defined as the volume
forcement improves the properties of the composite of losses observed in the material at a given time. It
material. The lubricating layer on the surface of is caused during the application of a high load in the
the composite material reduces the COF values.
component. The statistical analysis of the wear rate
The thin film layer formation may happen to break
down, which causes increases in COF during the is performed in the Box Behnken design of RSM. The
mating of metal surfaces. The low COF obtained results obtained from the RSM are explained below

2350082-16

2350082.indd 16 09/02/23 6:40:03 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a) (b)
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(c)
Fig. 9. (Color online) Statistical result of COF, (a) prediction plot, (b) perturbation plot, (c) contour plot.

in this section. The polynomial equation obtained is 197.46, showing that the model is best fitted to
from the model is given in the following equation: the experimental results. AB, AC, BC, A, B, and C
are the model terms used for prediction. The signif-
0.000975 + 0.0001065A icant term is obtained by the P-value, which should

 + 0.000795B be less than 0.05. The confidence level of the wear
 rate regression model is 0.9961, indicating that the
Wear rate =  + 9.1e-07C-2e-05AB , (21)
 design best fits the experiment. A better correla-
−2e-08AC-2e-07BC-9.75e-07A 2
 tion between experimental and predicted values is
 + 0.00031B2 + 4.4e-10C2 observed at 0.9665 of R2 and 0.9910 adjusted R2,
which is less than 0.025 of variation. The lack of fit
where A, B and C are the load, sliding velocity and
obtained is insignificant, showing that the error is
sliding distance, respectively. much less. The model is best fitted for the experi-
From Table 11, the varying factors of the load mental wear rate.
are A, sliding distance is B, and sliding velocity is The statistical result obtained from the RSM is
C. For wear rate, the F value obtained for the model given in Fig. 10. The influence of factors over the

2350082-17

2350082.indd 17 09/02/23 6:40:04 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

Table 11. ANOVA of wear rate.

Source Sum of squares df Mean square F-value p-value Remark

Model 3.935E-06 9 4.372E-07 197.46 < 0.0001 Significant
A-load 6.050E-07 1 6.050E-07 273.23 < 0.0001 Significant
B-Velocity 3.001E-06 1 3.001E-06 1355.40 < 0.0001 Significant
C-sliding distance 2.113E-07 1 2.113E-07 95.40 < 0.0001 Significant
AB 4.000E-08 1 4.000E-08 18.06 0.0038 Significant
AC 1.000E-08 1 1.000E-08 4.52 0.0712 Not significant
BC 2.500E-09 1 2.500E-09 1.13 0.3233 Not significant
A² 4.003E-08 1 4.003E-08 18.08 0.0038 Significant
B² 2.529E-08 1 2.529E-08 11.42 0.0118 Significant
C² 3.184E-09 1 3.184E-09 1.44 0.2695 Not significant
Residual 1.550E-08 7 2.214E-09 – – –
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

Lack of Fit 7.500E-09 3 2.500E-09 1.25 0.4028 Not significant

Pure Error 8.000E-09 4 2.000E-09 – – –
Cor Total 3.951E-06 16 – – – –
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(a) (b)

(c)
Fig. 10. (Color online) Statistical result of wear rate, (a) prediction plot, (b) perturbation plot, (c) contour plot.

2350082-18

2350082.indd 18 09/02/23 6:40:04 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

response is given in the contour plot. The predicted 4.4.  alidation of experiment using
V
output from the RSM design is given in the pre- hybrid DBN-RSA
dicted graph. The low wear rate is observed in the
low load condition, and the sliding distance pro- This work’s tribological and mechanical properties
vides a lowering wear rate. The maximum wear rate are predicted and validated using a hybrid neural
of 0.005 mm3/m is obtained in the load of 30N at ­network of DBN-RSA. It is achieved by the correla-
2 m/s of sliding velocity, respectively. tions derived using sigmoid and probability functions.
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(a) (b)

(c)
Fig. 11. (Color online) Comparison of predicted mechanical properties (a) tensile strength, (b) hardness, (c) porosity.

(a) (b)
Fig. 12. (Color online) Comparison of tribological parameters (a) COF, (b) wear rate.

2350082-19

2350082.indd 19 09/02/23 6:40:07 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a)
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(b)

(c)
Fig. 13. (Color online) Regression plots of mechanical properties (a) Tensile strength, (b) Hardness, (c) Porosity.

The comparisons of the predicted mechanical from the network layers are higher, making a large
properties from different approaches are shown in deviation in the output values. The weight obtained
Fig. 11. The tensile strength is higher in the 4th run, from the neural network layers is optimized using
the high hardness is achieved in the 16th run, and
the hybrid approach’s reptile search optimization
porosity is minimum in the 16th run, respectively.
algorithm. The accuracy also obtained from the
The hybrid DBN-RSA also predicted the values
nearer to the experimental, and RSM predicted val- network is higher. The algorithm increases the net-
ues. A large impact is observed in non-hybrid DBN work’s learning rate and tunes the output responses
predicted values because the weight units obtained with less computing time.

2350082-20

2350082.indd 20 09/02/23 6:40:08 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a)
Surf. Rev. Lett. Downloaded from www.worldscientific.com

(b)
Fig. 14. (Color online) Regression plots for tribological properties. (a) COF (b) wear rate.

The comparison of the tribological properties is

given in Fig. 12. The graph shows the high wear
rate and low coefficient of friction achieved in the
2nd run of the model. The result obtained from the
hybrid network is closer to the experimental and
the RSM predicted values. A large deviation is
observed in the non-hybrid approach of DBN due
to the large weight obtained from the network. The
tribological properties are best fitted to the designed
model with high accuracy.
The regression plots of the mechanical properties
are shown in Fig. 13. The obtained regression val-
ues are closer to 1, depicting the statistical result
from the hybrid network fitted to the experimen-
Fig. 15. (Color online) The RMSE of mechanical and tal and RSM results. The tensile strength of the
tribological results. composite material gives fitness values of 0.99098

2350082-21

2350082.indd 21 09/02/23 6:40:10 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.
Surf. Rev. Lett. Downloaded from www.worldscientific.com

Fig. 16. (Color online) Criteria settings for obtaining optimum solutions.

Fig. 17. (Color online) Optimization of tribological parameters.

2350082-22

2350082.indd 22 09/02/23 6:40:10 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.
Surf. Rev. Lett. Downloaded from www.worldscientific.com

Fig. 18. (Color online) Optimization parameter of mechanical properties.

at training and 0.99771 at validation. The hardness 4.5. Optimization of parameters

of the material shows a fitness value of 0.99988 and
The parameters are optimized using the response
0.99977 and porosity values of 0.99871 and 0.99843.
surface methodology based on the Box-Behnken
It implies that the predicted mechanical properties
design. In order to obtain optimal solutions, the
best fit the experimental and RSM results.
RSM requires some input criteria for all parameters
The regression plots for tribological behavior are
involved in the design. The default criteria setting
shown in Fig. 14. The training and validation of
of RSM is the in-range condition. According to the
the hybrid network give the best fitted result than
study requirements, only the input parameters for
the non-hybrid network. The graph implies that
mechanical and tribological RSM models are main-
the DBN-RSA approach’s result is more accurate
than the non-hybrid DBN approach. The regression tained at in-range criteria, as shown in Fig. 16.
values obtained from the hybrid network are closer This is because the study used to find the optimal
to unity, that is, 0.99297 for friction coefficient and input combinations suitable for given output con-
0.99324 for wear rate, respectively. This regression ditions. In order to obtain better tribological prop-
gives the fitness determination of the experimental erties, the initial parameters COF and wear rate
results. The proposed method gives an agreeable criteria are given as maximum and minimum condi-
regression outcome. tions. Likewise, the mechanical output parameters
Figure 15 represents the RMSE of predicted val- hardness and tensile strength criteria are set to the
ues from the different mechanical and tribological maximum value. In contrast, the porosity criteria
properties approach. The error analysis of the pro- must be set to the minimum value, as shown in
posed DBN-RSA is compared with the predicted val- Fig. 16.
ues of the RSM and DBN approaches; it shows that Figure 17 represents the optimum tribological
high error is obtained from the DBN. The obtained output obtained from the RSM method. The tribo-
error in RSM is 0.027 for COF, 0.035 for wear rate, logical result shows the optimum load of 13.2897N,
0.022 for hardness, 0.044 for porosity, and 0.059 for a velocity of 1 m/s, and a sliding distance of
tensile strength. For the DBN-RSA approach, the 256.117 m. It achieves the optimum output of fric-
error values are 0.031, 0.039, 0.025, 0.045, and 0.07 tion coefficient, which is 0.439421 and the wear
for COF, wear rate, hardness, porosity, and tensile rate of 0.003199 mm3/m at the desirability factor
strength. of 0.964.

2350082-23

2350082.indd 23 09/02/23 6:40:10 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

K. Pratheesh & V. V. Monikandan

Similarly, from Fig. 18, optimized squeeze casting temperature. The optimum mechanical properties
parameters for obtaining optimum mechanical prop- obtained from the analysis were the highest tensile
erties are 149.425 MPa of squeeze pressure, 799.3 of strength of 330.665 MPa, the highest hardness of
melt temperature, and 290.68 of die temperature. 115.528HBN and the low porosity of 2.885, respec-
Thus, the corresponding tensile strength, hardness tively. The tribological result shows the optimum
and porosity are 330.665 MPa, 115.528HBN and load of 13.2897N, a velocity of 1 m/s, and a sliding
2.885, respectively. It implies the experimentation distance of 256.117 m. It achieves the optimum out-
and statistical outcome gives agreeable results of put of friction coefficient is 0.439421 and the wear
tribological and mechanical properties of the fabri- rate of 0.003199 mm3/m. From the result, the pro-
cated Al-Si piston alloy matrix-reinforced TiC-MoS2 posed material is highly suitable for high strength
composite material. and high temperature applications. Due to its better
tribological behavior, it is highly suitable for the pis-
ton of diesel engines. This work may improve find-
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

5. Conclusion ing the fabricated material’s thermal stability and

This work prepared the casting of the Al-Si alloy-­ fatigue behavior because the piston undergoes high
reinforced TiC-MoS2 composite using squeeze cyclic loads.
casting. The casting parameter’s effect on the
mechanical properties of the fabricated composite
Surf. Rev. Lett. Downloaded from www.worldscientific.com

has been experimented with, and its tribological References

behavior was also analyzed. The wear test of the 1. P. K. Yadav, S. K. Patel, V. P. Singh, M. K. Verma,
composite material is performed using a pin on the R. K. Singh, B. Kuriachen and G. Dixit, J. Bio-and
disc tribometer under dry sliding wear conditions. Tribo-Corrosion 7 (2021) 1.
2. F. Xia, X. S. Gao, M. X. Liang, Y. C. Guo, J. P. Li,
The tensile strength is tested, and the fracture
Z. Yang and L. L. Zhang, J. Mat. Res. Technol. 9
mechanism is also obtained using the SEM micro- (2020) 12926.
structural analysis. Also, the surface morphology 3. G. Anbuchezhiyan, B. Mohan, N. Senthilkumar and
was studied. The experimentation resulted in a high R. Pugazhenthi, Metals Mat. Int. 27 (2021) 3058.
strength of 330 MPa at the highest squeeze pres- 4. N. Mahaviradhan, S. Sivaganesan and N. P. Sravya,
sure and melt temperature, similar to the hardness A. Parthiban, Mat. Today Proc. 39 (2021) 743.
5. S. Senthil, M. Raguraman and D. T. Manalan, Mat.
obtained at 111HBN. The microstructural analysis Today Proc. 45 (2021) 5934.
depicts the lowest porosity of 3.21% at the melt 6. M. A. Xavior and J. A. Kumar, Proc. Eng. 174
temperature and pressure of 150 MPa and 700°C. (2017) 1110.
The wear rate obtained from the testing by varying 7. K. R. Padmavathi and R. Ramakrishnan, Proc. Eng.
load, sliding distance and sliding velocity are 0.0051 97 (2014) 660.
8. S. B. Boppana, S. Dayanand, B. V. Murthy, M.
mm3/s, and the coefficient of friction is 0.31, respec-
Nagaral, A. Telagu, V. Kumar and V. Auradi, J.
tively. The design of the experiment is performed Comp. Sci. 5 (2021) 155.
in the response surface methodology using the Box 9. S. Hossain, M. M. Rahman, D. Chawla, A. Kumar, P.
Behnken design. P. Seth, P. Gupta and A. Jamwal, Mat. Today Proc.
The RSM predicted and experimented values are 21 (2020) 1458.
compared and validated using a hybrid approach 10. C. S. Manda, B. S. Babu and N. Ramaniah, Adv.
Mat. Proc. Technol. (2021) 1.
of DBN-RSA and DBN. The result shows the pro-
11. A. Kumar, M. Y. Arafath, P. Gupta, D. Kumar, C.
posed approach gives a highly accurate value closer M. Hussain and A. Jamwal, Mat. Today Proc. 21
to the experimented result with the RMSE of 2% to (2020) 1417.
7% deviation. The regression value obtained from 12. A. Dhulipalla, B. U. Kumar, V. Akhil, J. Zhang, Z.
the network was 99.9%, which depicts the proposed Lu, H. Y. Park and J. Zhang, Manuf. Lett. 24 (2020)
method is best fitted to the experiment. The opti- 82.
13. R. Palanivel, I. Dinaharan and R. F. Laubscher,
mization result obtained from the response surface
11882 Casting Routes for Production of Metallic
methodology gives the optimum squeeze casting Based Composite Parts. (2021)
parameters of 149.425 MPa of squeeze pressure, 14. S. A. Hassasi, M. Abbasi and S. J. Hosseinipour, Int.
799.3°C of melt temperature, and 290.68°C of die J. Metalcast. 15 (2021) 852.

2350082-24

2350082.indd 24 09/02/23 6:40:10 AM

 2nd Reading WSPC/149-SRL 2350082 ISSN: 0218-625X

Character Optimization of Al-Si Reinforced Composite

15. W. Jiang, J. Zhu, G. Li, F. Guan, Y. Yu and Z. Fan, 23. L. Natrayan and M. S. Kumar, Mat. Today
J. Mat. Sci. Technol. 8 (2021) 8119. Communicat. 25 (2020) 101586.
16. J. Zhu, W. Jiang, G. Li, F. Guan, Y. Yu and Z. Fan, 24. A. Karthik, S. A. Srinivasan, R. Karunanithi, S. K.
J. Mat. Proc. Technol. 283 (2020) 116699. Babu and V. K. S. Jain, J. Mat. Res. Technol. 14
17. M. O. Shabani, A. Baghani, A. Khorram and F. (2021) 797.
Heydari, Silicon 12 (2020) 2977. 25. N. Srivastava and M. Anas, Mat. Today Proc. 26
18. J. Hao, H. Luo, J. Bian, Y. Shi, B. Yu and R. Li, Int. (2020) 1914.
J. Metalcast. 16 (2021) 1. 26. A. Karthik, R. Karunanithi, S. A. Srinivasan and M.
19. V. S. Chandra, B. Blessto, S. Divya, S. Dhanasekaran, Prashanth, Mat. Today Proc. 27 (2020) 2574.
M. Ravi and K. Sivaprasad, Trans. Indian Inst. 27. P. Krishnankutty, A. Kanjirathinkal, M. A. Joseph
Metals 73 (2020) 1409. and M. Ravi, Mat. Today Proc. 5 (2018) 8406.
20. M. A. Z. Sohag, P. Gupta, N. Kondal, D. Kumar, N. 28. H. Barabadi, S. Honary, P. Ebrahimi, A. Alizadeh, F.
Singh and A. Jamwal, Mat. Today Proc. 21 (2020) Naghibi and M. Saravanan, Inorg. Nano-Metal Chem.
1407. 49 (2019) 33.
21. M. H. Sarfraz, M. Jahanzaib, W. Ahmed and S. 29. H. B. Huang, X. R. Huang, R. X. Li, T. C. Lim and
by WSPC on 11/22/23. Re-use and distribution is strictly not permitted, except for Open Access articles.

Hussain, Int. J. Adv. Manuf. Technol. 102 (2019) W. P. Ding, Appl. Acoust. 113 (2016) 149.
759. 30. L. Abualigah, M. Abd Elaziz, P. Sumari, Z. W. Geem
22. J. V. Christy, R. Arunachalam, A. H. I. Mourad, P. and A. H. Gandomi, Exp. Syst. Appl. 191 (2022)
K. Krishnan, S. Piya and M. Al-Maharbi, J. Manuf. 116158.
Proc. 59 (2020) 287.
Surf. Rev. Lett. Downloaded from www.worldscientific.com

2350082-25

2350082.indd 25 09/02/23 6:40:10 AM