Electrodeposition of composite is a technique of co-deposition of fine particles
like metallic, non-metallic compounds or polymer particles in the coated layer
to enhance material properties such as wear resistance, corrosion resistance,
or lubrication. During the plating process, the insoluble particles are in suspended
in a conventional plating electrolyte and are co-deposited to the growing metal
layer. The second phase material (co-deposition material) can be powder, encapsulated
particle or fibre. Existence of the second-phase particles in the deposition
which improve properties like micro hardness, yield strength, tensile strength
and wear resistance (Ferkel et al., 1997).
Tribological coatings improve physical properties such as hardness, lubricity
and corrosion resistance, to lower valued substrates that improve overall quality
of the base substrate material (Meneve et al., 1997).
Composite coatings consist of a metal or alloy matrix comprised with a dispersion
of second phase particles. The second phase particles are generally oxide or
carbide fine particles such as SiC, Al2O3, TiO2,
diamond and SiO2 or solid lubricants like MoS2, graphite
and PTFE to reduce friction or improve wear resistance (Guglielmi,
1972; Roos et al., 1990; Hovestad
and Janssen, 1995; Benea et al., 2001).
Fine particles of oxides, carbides and nitrides having micron and sub-micron
sizes can be co-deposited with a variety of metal electroplating metals such
as Ni, Co and Cu. In addition, metal-inorganic particles composite can lead
to accomplish special properties of coatings such as good hardness and wear
corrosion resistance and metal-organic particles can establish new characteristics
such as self-lubricating coatings (Kim and Yoo, 1998;
Peng et al., 1998; Yeh and
Wan, 1994; Foster and Cameron, 1976; Pena-Munoz
et al., 1998).
Hou et al. (2002) reported that electrodeposited
Ni-SiC composites improve in the wear resistance. Stott
and Ashby (1978) investigated that Ni-Al2O3 composite
coatings have better oxidation resistance than unreinforced nickel in the oxygen
enriched atmosphere (Stott and Ashby, 1978). Oxide particles
of Al2O3 and SiO2 with fine sizes can be co-deposited
along with Ni without difficulty by means of conventional Watts and sulfamate
bath (Nwoko and Shreir, 1973).Various investigators
have successfully co-deposited second phase particles (like Al2O3,
SiC, diamond, TiO2, WC, Cr3C2, TiC, etc.) in
a variety of metal matrices such as Ni, Cr, Co, Re, etc. (Narayan
and Narayana, 1981).
Fabrication of composite coatings by electroplating method is one of the most
advanced fabrication technique of Functionally Graded Materials (FGMs). The
main advantages of this method over other fabrication techniques are the simplicity
to control of process, low initial investment capitals for equipments and possibility
to manufacture complicated parts (Guo and Zhang, 1991).
In this experimental work, the electrolytic co-deposition of micron and submicron
sized Al2O3 particles from Nickel Watts bath solution
were carried out. The effect of plating parameters such as current density,
temperature of the bath, pH value the bath, concentration of Al2O3
particles in electrolyte bath and agitation speed were adjusted for different
level of intervals and its effects were analysed. Properties of final deposit
such as volume fraction of Al2O3 particles, mass of deposit
and micro hardness has been studied. Surface morphological studies of Ni- Al2O3
coating were carried out with Scanning Electron Microscope (SEM) and metallurgical
microscope for identification of behaviour deposited layer. After the experimental
investigations a neural network model was developed for the prediction of similarities
between the results of experimental work and neural network trainings and simulations.
At the end results of neural network model and experimental outcomes such as
mass of deposit, volume fraction and micro hardness of coated specimens were
compared for similarities.
MATERIALS AND METHODS
The schematic diagram of electroplating setup shown in Fig. 1.
Electrodeposition experiments were conducted in a BOROSIL 2000 mL glass container.
The plating electrolyte was a watts type nickel bath. Mild steel plate of sized
50.8x25.4x1.8 mm3 thick was used as a cathode substrate and area
of deposition taken as 25.4x25.4 mm2 (1 inch2). The left
behind portions of plating area were masked. A Nickel plate of sized 102x43x5
mm3 thick was employed as anode in the circuit. Fine sized and decontaminated
Al2O3 particles (supplied by NICE chemicals (p) Ltd.,
Cochin, India) with an average size of 5Fm and lesser than sized were used as
the reinforcement element. A regulated power supply unit made by Spark Tek (0-2
A, 0-30 V, DC supply) was used for the electrodeposition. Temperature of the
bath was accurately controlled using Osham temperature controller unit (230
volt, 50 Hz, 30-110°C, AC type). A motorised mechanical stirrer was used
to agitate and keep the Al2O3 particles in suspension
in the electrolyte and whose rotations were monitored by a digital tachometer.
|| Process parameters and their levels
|| Experimental setup
The pH of the solution was monitored by digital pH meter (Made by HANNA Instruments,
Mauritius). The pH of bath was adjusted with diluted H2SO4
and NaOH solutions depending upon the plating conditions. The plating conditions
taken for this experimental test are given in Table 1.
Before plating, each mild steel plate was smoothly polished in dry cloth buffing
wheel, degreased with acetone. Initial weight of the each specimen was observed
before deposition using electronic weighing machine. Subsequently the specimen
was masked and cleaned by anodic and cathodic preparations. In plating bath,
the distance between the anode (Ni) and cathode (Mild steel substrate) was maintained
constantly for all experiments. Plating time was taken as 60 min in all cases.
After co-deposition, the final weights of samples were absorbed. The mass of
deposition was measured from the difference of initial and final weights of
Surface morphology and distribution of Al2O3 particles
in the deposit was examined by scanning electron microscope with various magnifications
and the volume percentage of embedded alumina particles were determined from
micrograph images using METZER-METZ 56 model (Magnification range: 10-2500X)
metallurgical microscope. Assessments of micro hardness of the deposits were
carried out in VAISESHIKA-Vickers microhardness tester TYPE 7005 with the pay
load of 200 g for 15 sec indentation period. The indentation dimensions were
examined at 400X magnification and the hardness values were formulated.
After the experimental investigations, a feed forward back propagation neural
network model was developed. The neural network model was framed and trained
up with experimental parameters and the outcomes of the same. The input layer
of network composed with five parameters such as current density, temperature
of the bath, pH value the bath, concentration of Al2O3
particles in electrolyte bath and agitation speed and target composed with mass
of deposit, volume percentage of Al2O3 and micro hardness
were considered in network training and modelling. In the end, the consequences
of experimental and neural networks were compared for the proportion of resemblances.
RESULTS AND DISCUSSION
The experimental results mainly discussed with effect of current density, pH,
temperature of the bath, alumina concentration in bath and agitation speed which
took part in an important role in the determination of mass of deposit, hardness
and volume fraction.
Effect of plating parameters: The effects of selected plating parameters
on Ni-Al2O3 coating were determined. Figure
2 shows the results of volume percentage of alumina at different current
density and pH levels. It was observed that increasing current density produced
higher deposition. When current density amplified from 2-3 A dm-2,
volume percentage of alumina increased in the deposit. It appeared that the
amount of alumina in the nickel matrix increased with current density of 2 A
dm-2 and reached a maximum about 46.33% with the pH level of 2.5.
From the experimental outcome, the volume fraction of alumina increased with
pH value shown in Fig. 3.
||The relation of current density A dm-2 with volume
of alumina in deposit
The greater effect of pH value of bath for better deposition of alumina was
positioned between 2.5-3.5. The maximum amount of alumina in deposit, when the
pH value was around 2.5 and the least value of alumina deposition (14.31%) at
pH 4. It is clarified that the deposition of alumina was reduced with increase
The volume percentage of alumina incorporated into the nickel matrix was affected
by bath temperature as shown in Fig. 4. It appeared that the
amount of alumina in the nickel matrix increased with increase in current density
and temperature level of 45°C.
After this temperature level alumina deposition progressively decreased for
all current conditions.
The Fig. 5 corresponds to the effect of concentration of
alumina in the bath. The bath concentration was varied from 10-30 g L-1
with an interval of 5 g L-1. The highest of alumina deposit was obtained
between the bath concentrations of 10-30 g L-1.
||The relation of pH value of bath with volume of alumina in
||The relation of temperature with volume of alumina in deposit
||The relation of bath concentration with volume of alumina
||The relation of agitation speed with volume of alumina in
The mass of deposit was gradually increased with respect to increase in bath
concentration. The deposition rate of alumina was decreased gradually after
current density level of 3 A dm-2 for all concentration levels.
The purpose of agitation to kept the alumina particles in suspension in the
electrolyte. In this experiment the mechanical stirrer was used to keep the
particles in suspension in the electrolyte solution by continuous stirring.
Figure 6 shows the effect of agitation speed for enhanced
deposition of alumina in the deposit. It was observed that highest amount of
alumina deposit was obtained between agitation speed levels of 200-300 rpm with
current density level of 2 A dm-2.
Effect of current density on mass of deposit: The effect of current
density on mass of deposit was shown in Fig. 7.
|| The relation of current density with mass of deposit
It is observed that increase in current density slightly increase in mass of
deposit and similar trends observed at all pH levels. At the stage of 3.5 pH
and current density 3 A dm-2 utmost mass of deposition was found.
Effect of vol.% of alumina on micro hardness: Micro hardness test of
the composite coating was carried out in VAISESHIKA-Vickers micro hardness tester
TYPE 7005 with the pay load of 200 g for 15 sec indentation period. The indentation
in the coated layer was examined at 400X magnification and it was positioned
for measurement of indentation diagonal length D. The micro hardness was calculated
by the system based on:
||Load in kg
||Diagonal length in mm
The effects of volume fraction of alumina and current density in micro hardness
of deposit were shown in Fig. 8. Micro hardness of the specimen
was varied with volume of alumina in the matrix. It was examined that the micro
hardness of Ni-Al2O3 deposit increased with increase in
volume percentage of reinforcement particles in the nickel matrix and accomplished
highest micro hardness of 405.48 Hv and good hardness values were attained around
30-40% of alumina in the matrix with current density span of 2-4 A dm-2.
Surface morphology of Ni-Al2O3 composite coating:
The photographic image of specimens after deposition is exposed in Fig.
|| Effect of vol. fraction of alumina in micro hardness at different
|| Snapshot image of specimens after deposition
The surface morphologies of Ni- Al2O3 composite coating
were observed by Scanning Electron Microscope (SEM) revealed in Fig.
10 and 11 with different volume fractions of alumina
with 46.33 and 32.76%, respectively. It can be seen that incorporation of Al2O3
particles in nickel matrix with the mentioned volume percentages. Also
the Fig. 12 illustrates the cross-section microstructure
of the deposit containing 46.33% of alumina.
Exploration of experimental works with artificial neural network for the
analysis of resemblances: Artificial Neural Network (ANN) techniques are
vastly supple modelling equipment with a facility to learn the mapping between
input variables and output characteristics (Thillaivanan
et al., 2010). In this study a feed forward back propagation neural
network model was trained and developed using experimental statistics and results.
||Microstructure obtained at Current: 2 A dm-2, pH:
3.5, Temperature: 30°C, with 10 g L-1 of Al2O3
effected with 46.33% of alumina in the nickel matrix
||Microstructure obtained at Current: 2 A dm-2, pH:
3.5, Temperature: 45°C, with 10 g L-1 of Al2O3
effected with 32.76% of alumina in the nickel matrix
||Cross sectional view of the deposited specimen with 46.33%
|| Neural network structure used for training with 1 input layer,
1 hidden layer with 5 neurons and 1 output layer
The input parameters current density, temperature, concentration of bath, volume
percentage of Al2O3, pH of the solution and the output
constrains such as percentage of Al2O3, mass of deposit
and micro hardness were taken for network training. The architecture of network
model is given in Fig. 13. The network consists of one input
layer, one hidden layer with 5 neurons and one output layer which was utilized
for training of network.
For uniformed training and analysis, the experimental data sets should be normalized
between 0 and 1. The input parameters and the outcomes of the experimental works
were normalized. The general method applied for normalization is:
where, Xn-normalized valve, Xmin and Xmax are the least and highest value in
the parameter dataset and the X stands for the value to be normalized. The experimental
data sets were normalized and utilized for network training. The Table
2 shows the normalized input parameters used for training and modeling of
ANN training algorithm for development of feed forward back propagation
network: The normalized experimental input and output parameters were trained
with the use of feed forward back propagation module in ANN tool. In the network
model, inputs were considered as P and targets were T. The network was created
with one hidden layer of five neurons. ANN training response result was mainly
based on number of neurons and the epoch values. Initially the network was trained
for 5 neurons and 300 epochs. But the test values were not in correlation with
validation values for the above neuron and epoch setup and testing curve is
||Training response chart for 38 neurons and 300 epochs which
shows closer relation between test values and validation values
So the value of neuron and epoch setup needs to be increased. After the network
was trained for 10 neurons and 300 epochs which clearly visualized that the
test values were not in mate with the validation values and testing curve was
also not linear. So still the neuron size needs to be increased to reach the
goal of 0. After a number of iterations with different neuron and epoch setups
the network was trained for 38 neurons and 300 epochs. This particular structure
offers the output values which are closer to the target set values with very
little error. It was clearly visualized that the test values are in near correlated
with the validation values. The predicted ANN results were well associated with
experimental values shown in training response chart of Fig.
|| Normalized data sets of input parameters for network training
Further, increase of neuron size will yield performance value near to the goal
of 0 but it requires more number of neurons.
Comparison chart for the resemblance of experimental model and ANN model:
After training the target results of the trained network were tabulated with
experimental works. The Table 3 shows the end results of experimental
work and training results of the network. The linearity between experimental
values and neural network results are shown in the statistical plots with its
straight line equations and R2 values. Figure 15
shows the comparison of volume fractions of alumina with experimental and ANN
||Evaluation of volume fraction of alumina with experimental
and ANN results with the correlation R2 = 0.9231
||Evaluation of micro hardness with experimental and ANN results
with the correlation R2 = 0.9979
||Evaluation of mass of deposit with Experimental and ANN results
with the correlation R2 = 0.998
The correlation coefficient of this relation obtained was 0.9231. Similarly,
Fig. 16 and 17 show the resemblance of
experimental model and ANN values for micro hardness and mass of deposits with
the correlation coefficients 0.9979 and 0.998, respectively. This statistics
clearly states that the experimental values were in good association with the
predicted ANN results.
|| Comparison table for experimental and ANN Training results
In this research work experimental trials were conducted for different intervals
of current densities, pH, concentration of Al2O3 in bath,
agitation speed, bath temperature. Volume fraction of alumina in the deposit
was predicted and fell down between the ranges of 10.45- 46.32%. In order to
ascertain the validity of the above experimental, microscopic image analysis
was carried out. These results also confirmed the above microscope observation
of the Al2O3 particles.
The determination of hardness value of deposits was measured from Vickers micro
hardness testing machine with the pay load of 200 g and the micro hardness values
were placed between 138.92-405.49 Hv. The mass of deposit was determined accurately
with the help of electronic balance with the utmost deposition of 420.78 mg.
The experimental input parameters and its effects such as mass of deposit, volume
fraction and the micro hardness were employed for neural network training and
simulation for the achievement of goal.
The experimental effects and the ANN results were compared by the graphical
representation for its resemblances. The above neural network simulations were
useful for the prediction of mass of deposit, volume fraction of Al2O3
and the micro hardness at any set of input parameter setups within the prescribed
level. Such predictions will avoid unnecessary conveyance of experiments in