Development of ANN modeling for coefficient of friction of electro deposited nickel-graphite composites

S.Dinesh1, G.N.K.RameshBapu2, K.Ramanathan3, T.LouieFrango4 P.G. scholar, A.C.College of Engineering &Tech, Karaikudi-4, dineshbeme@gmail.com ,9003645530. 2 Scientist, Central Electrochemical Research Institute, Karaikudi, 630006, bapu2657@yahoo.com , 9486563834. 3 Assistant Professor of Mechanical Engg, A.C.College of Engineering&Tech, Karaikudi-4., kalirams@yahoo.com ,9994607024. 4 Assistant Professor of automobile Engg, Shanmuganathan Engineering College, Pudukkottai 622507, LouieFrango@yahoo.co.in ,9443612523 Abstract: Nickel-Graphite composite coatings are produced by electro deposition using conventional techniques at various cathode current densities, pH and temperature. Electro deposition was carried out from a conventional Watts bath. Natural graphite powder of 20-30 m size was used in this study . 33 full factorial designs of experiments were designed by adopting the Design of Experiments (DOE) approach with three level of experiment namely Low, Medium and High. The volume percentage of graphite
1

deposition in composite coated specimens were measured gravimetrically. The coefficient of friction of coated specimen was measured using scratch tester. An Artificial Neural Network (ANN) model was developed using 27 practical data obtained to predict the co efficient of friction of Ni-Graphite metal matrix. Within the range of input variables for the present case (pH) = 3 to 5; current density (i) = 3 to 5 A/dm2; temperature (T)= 40 to 600C, the prediction capability of Artificial Neural Network(ANN) is very close to the experimental measurement of friction of Ni-Graphite metal matrix. Keywords: Coefficient of Friction, Scratch tester, ANN model, Ni-Graphite composite coatings, MATLAB 1. INTRODUCTION Particle-reinforced metal matrix composites generally exhibit wide engineering applications due to their enhanced hardness, frictional resistance, wear and corrosion resistance compared to pure metal or alloy. Composite electroplating has been identified to be a technologically feasible and economically superior technique for the preparation of such kind of

composites. Graphite is a natural material, is known to be the good lubricant material. Because of their outstanding properties, Nickel-Graphite has been used in EMI gasket and shield applications. Because of its excellent mechanical and electrical properties Nickel-Graphite is of great interest for a number of applications (e.g. bearings, engine parts, electronic gaskets, etc.). Searching the literature, the impression arises that the applications of Graphite are kept a bit secret. Most of the relevant references are patents (up to 90% depending on the topic) giving less exact data about the process. Papers published in journals giving detailed information are rare. The volume percent incorporation of Graphite powder in the Ni-Graphite composite coating measured gravimetrically was earlier demonstrated. The amount of Graphite deposited in the composite metal matrix is mainly affected by the process parameters such as current density, pH value, temperature of the path solution and concentration of Graphite dispersed in the electrolyte. Volume fraction of Graphite influences the Coefficient of Friction of Ni-Graphite composite coatings and hence it is essential to develop a prediction model for estimating the Coefficient of Friction of Nickel-Graphite composite using the above parameters.

S. No

Graphite g/lit 5 10 20 30 40

Weight of Graphite (gm) 0.0157 0.0316 0.061

Weight of nickel (gm)

Vol. of Graphite (v1)

Vol. of Nickel (V2)

Vol. % Of graphite 20.62 35.07 49.76 13.79 12.20

Vol. % Of nickel 79.38 64.93 50.23 86.20 87.79

1 2 3 4 5

0.2416 0.0271 0.0074 0.2348 0.0142 0.0263 0.2407 0.0273 0.02704 0.2348 0.0263 0.0094 0.0042 0.2466 0.0277 0.0086 0.0038 TABLE 1 Graphite incorporation in nickel

2.0 EXPERIMENTAL PROCEDURE 2.1 The electrolyte: The conventional Watts bath of the following composition was used: Nickel sulphate - 225 g/l; Nickel chloride- 30 g/l; Boric acid- 40 g/l. The electrolyte was purified

in the conventional manner for removal of organic and inorganic impurities [5]. The pH value of the electrolyte was adjusted electrometrically using dilute H2SO4 or NH3. 0.01-g/l sodium lauryl sulphate was added to the electrolyte as anti-pitting agent before plating. The temperature of the electrolyte was maintained using a thermostat. 2.2. Plating procedure: Deposition was carried out on a 500 ml capacity using conventional technique. Nickel anodes and mild steel cathodes were used. The cathodes of 7.52.5 cm area were mechanically polished, degreased, bent to 90, suitably masked to expose an effective plating area of 12.5 cm2, electro cleaned, first cathodically and then anodically, washed rinsed and then introduced into the plating electrolyte with the area to be plated in the vertical plane closer to the bottom of the cell facing the anode. A bagged nickel anode bent similarly was placed above the area to be coated. Graphite powder (20 to 20 m) was added to the electrolyte in the form of slurry. The solution was stirred using a magnetic stirrer. Stirring was given initially for 30 s to bring all the Graphite powder into the suspension and then stopped. The deposition was continued for 40 minutes to allow the particles to settle on the substrate while the deposition proceeded. The same process was repeated to obtain various thicknesses.

Figure 1

Electro deposition experimental setup

2.3. Nickel-Graphite deposition: Natural grade graphite powder of 2030 m sizes was used. Prior to the codeposition, the graphite particles were ultrasonically dispersed in the bath for 10 min.

Experiments were conducted at a fixed Graphite concentration of 20 g/l, varying the plating parameters like temperature, pH, and current density. Ranges of coating parameters in the coating process are as follows: Current density, I = 3 5 A/dm2 ; pH value = 3 5; Temperature = 40 to 60
0

C .For the prediction of Coefficient of Friction of Graphite under a variation of coating

conditions, a training database with regard to different coating parameters needs to be established. For the above combination of parameters, twenty seven numbers of Nickel-Graphite composite coatings were obtained and their Coefficient of Friction was measured from scratch tester. 2.4 Design of experiment: Levels Process parameter Units Level 1 pH Current density Temperature A/dm2
0

Level 2 4 4 50

Level 3 5 5 60

3 3 40

TABLE 2

Process parameters with different levels

S.NO 1 2 3 4 5 6 7 8

CURRENT DENSITY (A/dm2) 2 4 6 2 4 6 2 4

TEMPERATURE (OC) 50 50 50 60 60 60 40 40

pH VALUE 4 4 4 4 4 4 4 4

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

6 2 4 6 2 4 6 2 4 6 2 4 6 2 4 6 2 4 6

40 50 50 50 60 60 60 40 40 40 50 50 50 60 60 60 40 40 40

4 3 3 3 3 3 3 3 3 3 5 5 5 5 5 5 5 5 5

TABLE 3 2.5 Coefficient of Friction:

27 Parameters-Nickel Graphite coatings

Nickel- graphite composites have been tested through the Scratch tester with constant loadcondition at starting load should be 10 Newtons, loading rate should be zero, stroke length is 10mm, scratch speed should be 0.20 mm/sec and scratch offset is 0.25mm .Then we have transfer loading condition enter the file name. Finally we have seen the view file and how much of Coefficient of friction is obtained.

Figure 2

Coefficient of friction experimental setup

2.5.1 Initial Loading Conditions

Figure 3 initial loading conditions for Coefficient of Friction

2.5.2 Graphs shown Coefficient of Friction:

Figure 4

Coefficient of Friction

2.5.3 Scratch Depth:

Figure 5 scratch is depth measured

3. ARTIFICIAL NEURAL NETWORK (ANN) ANN is a neural system of imitative biology, and the principle of human brain operation. Using a large amount of data out of which they build knowledge bases, ANN establishes analytical model to solve the problem in the estimation, prediction, decision making and diagnosis. Neural network consist of simple processors, which are called neurons, linked by weighted connections. Each neuron has inputs and generates an output that can be seen as the reflection of local information that is stored in connections. The output signal of a neuron is fed to other neurons as input signals via interconnections. Since the capability of a single neuron is limited, complex functions can be realized by connecting many neurons. It is widely reported that structure of neural network, representation of data, normalization of inputs outputs and appropriate selection of activation functions have strong influence on the effectiveness and performance of the trained neural network . A Neural network consists of at least three layers i.e., input layer, hidden layer, and output layer, where inputs are applied at the input layer and outputs are obtained at the output layer and learning is achieved when the associations between a specified set of input output pairs as established in Figure.1 .Here Feed forward back propagation (FFBP) algorithm is used for the prediction of Coefficient of Friction of NickelGraphite in the deposit for the given condition. Figure.vi shows the architecture of a standard supervised training FFBP ANN and Figure vii shows the perception.

Figure 6 ANN Model with two hidden layers Figure7 a typical processing element (Perception).

Figure. 8 Standard supervised training feed forward neural network 4.0 TRAINING THE ARTIFICIAL NEURAL NETWORK The neural network has to be first trained and then tested to use for application. The training was done with MATLAB software using a computer. MATLAB is a software package used for high performance numerical computations and visualization. It provides an interactive environment with hundreds of built in functions for technical computations, graphic and animations. MATLAB stands for matrix lab. Built in functions provides excellent tools for linear algebra computation data analysis, signal processing, optimization and others scientific computations. In this work ANN module is utilized for predicting the Coefficient of Friction of Graphite deposition in Nickel-Graphite composite matrix. The features current density, pH and temperature are the inputs and the Coefficient of Friction of Nickel-Graphite is the output for training the neural networks. Weights between input layer & hidden layer and weights between hidden layer & the output layer are generated randomly for the selected topology of the network. The number of patterns used for the training of Artificial Neural Network using Feed forward back propagation algorithm is 27. Training of the ANN was performed without any allowable error. The patterns are selected for training and testing the ANN. These selected patterns were normalized so that they lie between 0 and 1. Twenty Seven patterns were selected for training the ANN. The inputs and outputs are normalized by,
Xi Xi Where Xi is the value of a X max

feature and Xmax is the maximum value of the feature. A 3-6-1 Feed forward back propagation network was trained and the structure of the network is shown in figure-4.

Figure 9 3-6-1 Feed forward back propagation network.

Figure 10 Performance curve of ANN model for 100 epochs 5. VALIDATING THE ARTIFICIAL NEURAL NETWORK Once the network was trained such that the maximum error for any of the training data was less than allowable error, the weights and the threshold values were automatically saved by the program. As the input values from the validation experiments were given to the ANN program, the program predicts the required output. To validate the results of the Artificial Neural Network analysis eight data as shown in Table I were used. Once the pH, current density and temperature are fed into the trained networks, the Coefficient of Friction

Graphite that can be obtained in the Nickel-Graphite composites could be calculated quickly using ANN model.

10

i 4 2 4 4 2 6

T 50 50 60 40 60 40

pH 4 3 3 3 5 5

Coefficient of Friction 0.22 0.26 0.29 0.26 0.24 0.28

Table 4 Experimental coating parameters and Coefficient of Friction value. 6. RESULTS AND DISCUSSION A comparison of measured Coefficient of Friction value and predicted hardness value using the prediction technique Artificial Neural Network (ANN) is presented in the Table II

Test No 1
2 3 4 5 6

Measured Friction 0.709677 0.83871 0.935484 0.83871 0.774194 0.863

Predicted Friction (ANN model) 0.71007 0.8501 0.93538 0.83271 0.77247 0.85234

Percentage Error (ANN model) -0.05532 -1.35808 0.011103 0.715346 0.222625 1.235226

Table 5 Comparison of measured & predicted Friction of Nickel-Graphite deposition. The Average Absolute Error of the trained ANN model is found to be only 0.128484 % and it is shown apically in Figure.6. The closeness of ANN model in predicting the hardness of CBN deposition with actual value is shown in Figure 7. It is shown clearly that the prediction technique ANN is better in predicting the Volume fraction of CBN deposition in the Nickel-CBN composite coated specimens.

11

% of Error ANN model

10 8 6 4 % of Error 2 0 -2 -4 -6 -8 -10 No.of Experiments 1 2 3 4 5 6 ANN Model

Figure 9 Percentage error of ANN model in the prediction of Friction

Closeness ANN Model prediction

F riction for Ni-Graphite Deposition

0.95
0.9 0.85 0.8 0.75 0.7 0.65 Actual Friction Measured Friction

2 3 4 5 No. of experiments

Figure 10 Closeness of ANN model in the prediction of friction 7. CONCLUSION A 3-9-1 Feed forward back propagation Artificial Neural Network (ANN) model was developed for predicting Coefficient of Friction in Nickel-Graphite composite coated metal matrix using 27 test data .The developed neural net work was validated with eight data. Values obtained by the above ANN model were compared with the experimental values of the response variables to decide about the nearness of the predictions with the experimental values.

12

Within the range of input variables for the present case (pH = 3 to 5; current density (I) = 3 to 5 A/dm2; temperature (T) = 40 to 600C), the results showed that Artificial Neural Network comes in nearness of the predictions to the experimental values of Coefficient of Friction as the average errors in case of ANN is very less i.e. 0.128484 % only. 8.REFERENCES 1 Jack lapinski, derekpletcher, frank c. walsh, The electro deposition of Ni-Graphite composite layers(2011), Surface and coating technology. Vol. 42, pp.70-73. 2 Haijun Zhao, Lei Liu, Wenbin Hu, Bin ShenFriction and wear behaviour of Nigraphite composites prepared by electroforming,(2007) Materials& design, Vol. 18, pp.13741378. 3 Didier Floner, Florence Geneste Homogeneous coating of graphite felt by nickel electro deposition to achieve light nickel felts with high surface area (2007), Electrochemistry communication, Vol. 9, pp.2271-2275. 4 L.Q. Zhoua,b, J.G. Tanga, Y.P. Lia, Y.C. Zhoub Predicting forming limit of electrodeposited nickel coating based on practical stressstrain relationship(2008), Journal of materials processing technology , Vol. 206, pp.431-437. 5 Haijun Zhao , Lei Liu, Jianhua Zhu, Yiping Tang, Wenbin Hu Microstructure and corrosion behavior of electrodeposited nickel prepared from a sulphamate bath(2007) Materials Letters ,Vol. 61, pp.1605-1608. 6 Y.S. Jeona, J.Y. Byunb, T.S. Oha Electrodeposition and mechanical properties of Ni carbon nanotube nanocomposite coatings(2008) Journal of Physics and Chemistry of Solids ,Vol. 69, pp.1391-1394. 7 K. Subramanian , V.M. Periasamy , Malathy Pushpavanam , K. Ramasamy Predictive modeling of deposition rate in electro-deposition of coppertin using regression and artificial neural network (2009) Journal of Electro analytical Chemistry, Vol. 636, pp.30-35. 8 M. Surender, B. Basu , R. Balasubramaniam Wear characterization of electrodeposited NiWC composite coatings (2004) Tribology International Vol. 37, pp.743-749.

13

9 D.P.Weston, P.H. Shipway, S.J. Harris, M.K. Cheng Friction and sliding wear behaviour of electrodeposited cobalt and cobalttungsten alloy coatings for replacement of electrodeposited chromium (2009) Wear ,Vol. 267, pp.934-943. 10 J. Lamovec , V. Jovi , D. Randjelovi , R. Aleksi , V. Radojevi Analysis of the composite and film hardness of electrodeposited nickel coatings on different substrates (2008)Thin Solid Films, Vol. 516, pp.8646-8654. 11 M. Ghorbani, M. Mazaheri, A. Afshar Wear and friction characteristics of electrodeposited graphitebronze composite coatings (2005) Surface & Coatings Technology, Vol. 190, pp.32-38.

14