Ceramic Coatings Using GP Approach

7/30/2019 Ceramic Coatings Using GP Approach

1/13

International Journal of Mechanical Engineering and Technology (IJMET), ISSN 0976

6340(Print), ISSN 0976 6359(Online) Volume 3, Issue 1, January- April (2012), IAEME

77

PREDICTION OF MECHANICAL AND TRIBOLOGICAL

CHARACTERISTICS OF INDUSTRIAL CERAMIC COATINGS

USING A GENETIC PROGRAMMING APPROACH

Mohammed Yunus1

, Dr. J. Fazlur Rahman2

and S.Ferozkhan3

1. Research scholar, Anna University of Technology Coimbatore

Professor, Department of Mechanical Engineering H.K.B.K.C.E.,

Bangalore, India.

[email protected]

Mobile: +919141369124

2. Supervisor, Anna University of Technology Coimbatore

Professor Emeritus, Department of MechanicalEngineering

H.K.B.K.C.E., Bangalore, India.

[email protected]

3. Lecturer, Department of Mechanical Engineering,

H.K.B.K.C.E., Bangalore, India.

[email protected]

ABSTRACT

The state of the art methods used to manufacture the coating materials in

atmospheric plasma spray process and the level of process control employed in

todays coating equipment provides an excellent coating over a broad range of

application requirements. The various characteristics of coatings depend oncoating material, spray parameters, spray equipment and componentconfigurations. Amongst the many characteristics, the controlled porosity,

optimized hardness which is the demanding requirements of wear-resistantapplication, specific coating thickness and resistance to wear plays an

important role in deciding the quality of coating material.

In the technical paper, wear tests and Rockwell hardness tests were

conducted on different types of industrial coatings, namely, Alumina, Alumina-

INTERNATIONAL JOURNAL OF MECHANICALENGINEERING AND TECHNOLOGY (IJMET)

ISSN 0976 6340 (Print)ISSN 0976 6359 (Online)

Volume 3, Issue 1, January- April (2012), pp. 77-89

IAEME: www.iaeme.com/ijmet.html

Journal Impact Factor (2011) -1.2083 (Calculated by GISI)

www.jifactor.com

IJMET

I A E M E


2/13



78

Titania (AT) and Partially Stabilized Zirconia (PSZ), Super- Z alloy, Zirconia

Toughened Alumina, AT, PSZ under different parameters.

Genetic programming (GP) is an automated method for creating a working

computer program from a high-level problem statement of a problem. Theprediction of mechanical and tribological characteristics of ceramic oxide

coatings, depending on input parameters (Power input, standoff distance, typeof coatings, normal pressure, sliding velocity etc.), was performed by means of

genetic programming and data on the outputs ( hardness, weight loss, percent

porosity, coefficient of friction etc.) of Mechanical and tribological properties

already made. This technical paper highlights how we use GP technique in the

prediction output parameters. Commercial Genetic Programming (GP)

software-Discipulus is used to derive a mathematical modelling of relations for

various input and output parameters used in characterisation. Genetic approach

has been used for the modelling the properties in coated components is

proposed on the basis of a validation, training and applied data set. Various

different genetic models for prediction of different Mechanical and tribological

properties with greater accuracy (less than 2%) were also proposed bysimulated evolution.

Keywords: Evolutionary computation, Genetic Programming, Hardness, Bond

strength, wear and coefficient of friction, PSZ, ZTA, Plasma, Super-Z alloy..

1.INTRODUCTION

Thermal Sprayed Surface Coatings are used extensively for a wide range of

industrial applications [1-2]. The selection of a technology to engineer the

surface is an integral part of the component design. Accordingly, the first step

in selecting surface modification techniques is to determine the surface and the

substrate engineering functional requirements [11]. These involve therequirements of one or more of the properties like wear resistance, corrosion

resistance, erosion resistance, thermal resistance, fatigue strength, creep

strength and pitting resistance.

Among the various surface modification methods, the thermal spray

processes are widely recognized. Two-wire electric arc and atmospheric plasma

spraying process are most commonly used in industries. The typical

applications [1-2] & [6] of ceramic coatings are listed as under

1. General manufacturing industry:- Extrusion dies, threaded guides,

forging tools, wire drawing Capstans, cam followers, roller bearings, hot

forming dies.

2. Gas turbine industry:- Turbine Nozzles , Jet engine, Jet engine manifold

rings, Gas turbines fan seals, Aircraft flap tracks, expansion joints, fan

blades.

3. Petroleum Industry: - Pump plungers, compressor rods.

4. Chemical Process Industry: - Gate Valve, pump components.


3/13



79

5. Paper / Pulp Industry: - Printing rolls, liquor tanks.

6. Automotive Industry: - Piston rings, cylinder liners.

1.2 Genetic Programming (GP) Method

Genetic Programming is a form of machine learning that automatically

writes computer programs. It uses the principle of Darwinian Natural Selectionto select and reproduce fitter programs. GP applies that principle to apopulation of computer programs and evolves a program that predicts the target

output from a data file of inputs and outputs [10]. The programs evolved by GP

software Discipulus [13], in this case Java, C/C++ and assembly interpreter

programs represents a mapping of input to output data. This is done by

Machine Learning that maps a set of input data to known output data. The aims

of using the machine learning technique on engineering problems are to

determine data mining and knowledge discovery. GP provides a significant

benefit in many areas of science and industry [ 10-12]. The Discipulus GP [13 ]

system uses AIM Learning Technology. AIM stands for Automatic

Induction of Machine Code. AIM Learning and Discipulus deal with themachine learning speed problem. This speed allows the analyst to able to make

many more runs to investigate relationships between data and output, assess

information content of data streams, uncover bad data or outliers, assess time

lag relationships between inputs and outputs, and the like. The evolved models

have been used to develop process prediction or control algorithms. Hence GP

technology has been selected for the present work.

GP solutions are computer programs that can be easily inspected,

documented, evaluated, and tested. The GP solutions are easy to understand the

nature of the derived relationship between input and output data and to examinethe uncover relationships that were unknown before. Genetic Programming

evolves both the structure and the constants to the solution simultaneously.Discipulus GP strongly discriminates between relevant input data and inputs

that have no bearing on a solution [10-13]. In other words, Discipulus performs

input variable selection as a by-product of its learning algorithm.

The following step by step procedure will be implemented for a steady state

GP algorithm [10]:

1. Initialization of population: Generate an initial population of random

compositions of the functions and terminals of the problem (computer

programs).

2.

Fitness evaluation: Execute each program in the population, randomly itselects some programs and assign it a fitness value according to how

well it solves the problem by mapping input data to output data. Some

programs are selected as winners (best programs), and the others as

losers.

3. Create a new population of computer programs by exchanging parts of

the best programs with each other (crossover).

4. Copy the best existing programs


4/13



80

5. Create new computer programs by randomly changing each of the

tournament winners to create two new programs mutation.

6. Iterate Until Convergence. Repeating steps two through four until a

program is developed which predicts the behaviour sufficiently.

GP has been successfully used to solve problems in a wide range of broad

categories [15-23]:

1. Systems Modelling, Curve Fitting, Data Modelling, and Symbolic

Regression

2. Industrial Process Control

3. Financial Trading, Time Series Prediction and Economic Modelling

4. Optimisation and scheduling

5. Medicine, Biology and Bioinformatics

6. Design

7. Image and Signal processing

8. Entertainment and Computer games

2. EXPERIMENTAL PROCEDURE

Five different commercially available ceramic coatings powders namely,Alumina (Al2O3), Alumina-Titania (Al2O3-TiO2), Partially StabilizedZirconia (PSZ), Zirconia toughened alumina (ZTA consist of 80% alumina and20% PSZ) and Super-Z alloy (20% alumina and 80% PSZ) were used for thepreparation of coatings [1-2 ]. A 40 KW Sulzer, Metco plasma spray systemwith 7MB gun is used for this plasma spraying of coatings. Mild steel plates of50x50x6 mm and cylindrical pins of 6 mm diameter and 21mm length were

used as substrate to spray the ceramic oxides. They were grit blasted, degreasedand spray coated with a 50 to 100 microns Ni Cr Al bond coat. The ceramicTBC were then plasma sprayed using optimum spray parameters. In this study,two response parameters such as wear and hardness tests of the coating wereconsidered.

2.1 Wear Test

The pin-on-disc testing machine was used to measure the wear of material

weight loss by conducting dry sliding wear tests [ 4-5]. This instrument

consists of a pin is mounted on a stiff lever, designed as a frictionless force

transducer and pressed against a rotating disk. Generally pin surface is coated

with ceramic oxide to different thicknesses using plasma spray process, fixed to

an arm and pressed with a known force. The measurement includes RPM, Wearand Frictional force to measure effect of sliding speed, applied pressure, and

weight loss on the wear characteristics of different types of coatings. As the

disc is rotated, resulting frictional forces acting between the pin and the disc are

measured by using strain gage sensors.

The main object of this study is to evaluate the behavior of A, AT and PSZ

ceramic coatings subjected to different grinding conditions. The performance

[14] was evaluated by measuring


5/13



81

1. Grinding force ( both normal and tangential forces)

2. Surface finish produced which also includes the bearing area

characteristics

2.2 Hardness Test

The Rockwell hardness number was determined by pressing a hardenedsteel ball indenter or diamond cone penetrator against a test specimen and

resulting indentation depth was measured as a gauge of the specimen hardness

using c-scale.

3. Genetic Programming MethodologyG Genetic programming can be the most general approach among

evolutionary computation methods in which the coatings subject to thermal

changes adaptation are those hierarchically organized computer programs

whose size and form dynamically change during simulated evolution. The

initial population in GP is obtained by the creation of random computer

programs consisting of available function genes from set F and available

terminal genes from set T. The next step is the calculation of individualsadaptation to the environment. Fitness is a guideline for modifying those

structures undergoing adaptation. After finishing the first cycle, which includes

creation of the initial population, calculation of fitness for each individual of

the population, and genetic modification of the contents of the computer

programs, an iterative repetition of fitness calculation and genetic modification

follows. The evolution is terminated when the termination criterion is fulfilled.

This can be a prescribed number of generations or sufficient quality of the

solution. Instruction set for the present program used are +,-, * and /. Eachindividual GP run started with the training phase by the training data set, the

testing data set was not included within the training range.

Process Inputs:Normal Pressure (MPa),Velocity of Sliding (m/sec),Sliding Distance (m),Power Input (KW),Standoff Distance (mm)Thickness of Coating (m)Toughness (MPa m)Thermal conductivity (W/m K)Thermal Diffusivity (10

-7m

2/sec)

Measured Process OutputsPercentage Porosity

Weight loss (mg)Hardness (RHC)Coefficient of frictionBond strength (MPa)


6/13



82

TABLE 1. Experimental results of evaluating percentage porosity during under different process parameters for different coatings

Table2. Experimental results of evaluating hardness during under different process parameters for different coatings.

Power in KW

(Vo)

Standoff distance

in mm (V1)

% Porosity

for PSZ

% Porosity for

Alumina-Titania

% Porosity

for Alumina

16 100 7 8 9.4

16 110 6.7 7.2 8.8

16 120 6.25 6.8 8.5

16 135 6.7 6.7 8.2

16 150 6.8 7 8.35

25 100 5.5 6.5 825 110 5.2 5.85 7.3

25 120 4.8 5.4 7

25 135 5.2 5.3 6.7

25 150 5.3 5.6 6.85

35 100 4.9 5.9 6.7

35 110 4.62 5.25 6.1

35 120 4.2 4.8 5.75

35 135 4.55 4.75 6

35 150 4.65 5 6.1

40 100 4.5 5.75 6.7

40 110 4.22 5.1 6.1

40 120 3.8 4.65 5.6

40 135 4.15 4.6 5.8

40 150 4.2 4.75 5.9

Powerin KW (V3)

Thickness in m (V5)

Standoff distancein mm (V2)

Hardnessfor PSZ

Hardness forAlumina-Titania

Hardnessfor Alumina

Hardnessfor ZTA

Hardness for

Super-Z

16 100 100 73 107 112 104 110

16 100 110 80 115 120 112 115

16 100 120 84 122 127 118 120

16 100 140 90 118 123 116 118

16 150 100 83 117 122 114 116

16 150 110 87 122 126 118 120

16 150 120 94 130 135 126 126

16 150 140 90 124 129 121 123

16 200 100 78 112 117 109 114

16 200 110 82 117 122 114 117

16 200 120 90 125 131 123 122

16 200 140 86 120 125 117 119

16 300 100 75 108 114 106 108

16 300 110 78 112 117 109 112

16 300 120 87 120 125 117 116

16 300 140 82 115 120 112 104

25 100 100 75 113 118 113 110

25 100 110 83 118 123 116 119

25 100 120 88 123 128 117 125

25 100 140 93 119 124 118 121

25 150 100 86 115 120 112 119

25 150 110 90 120 125 118 125

25 150 120 97 125 130 123 130

25 150 140 94 121 126 119 126

25 200 100 82 102 107 112 116

25 200 110 86 117 122 115 121

25 200 120 94 122 127 119 124

25 200 140 90 119 124 118 11625 300 100 76 97 102 97 110

25 300 110 81 114 120 112 117

25 300 120 88 118 123 115 120

25 300 140 85 115 120 123 108

30 100 100 78 115 121 112 115

30 100 110 87 120 125 117 124

30 100 120 91 125 131 122 130

30 100 140 96 119 121 115 126

30 150 100 89 118 123 118 124

30 150 110 94 123 128 119 130

30 150 120 102 129 134 125 136

30 150 140 97 124 129 123 131

30 200 100 86 105 110 104 120

30 200 110 89 121 126 118 126

30 200 120 98 126 131 125 130


7/13



83

Table 3. Experimental results of evaluating weight loss during under different process parameters for different coatings.

30 200 140 94 123 127 119 118

30 300 100 80 103 108 103 115

30 300 110 84 118 123 116 123

30 300 120 91 122 127 119 125

30 300 140 89 118 122 117 113

40 100 100 82 118 123 115 121

40 100 110 90 123 128 121 127

40 100 120 95 128 133 115 134

40 100 140 100 124 129 124 129

40 150 100 92 121 126 118 12840 150 110 97 126 131 123 134

40 150 120 106 134 139 134 141

40 150 140 100 127 132 124 136

40 200 100 89 109 114 106 125

40 200 110 91 125 131 122 131

40 200 120 111 130 136 127 133

40 200 140 97 127 132 124 123

40 300 100 84 107 112 104 120

40 300 110 88 121 126 117 128

40 300 120 94 125 130 123 130

40 300 140 93 121 127 119 118

when D = 4Km (V2) Alumina

Normal

pressure (V0)

sliding,z velocity v

=2.5 m/s(V1)

v=5 m/s

(V1)

v=7.5 m/s

(V1)

v=10 m/s(

V1)

v=12.5

m/s(V1)0.05 0.17 0.19 0.23 0.3 0.48

0.1 1.6 1.9 2.1 2.3 5.2

0.15 2.1 2.2 2.7 4.7 5.8

0.2 3.4 3.4 3.4 6.5 10

0.25 4.9 5.4 5.8 8.3 16

0.3 6 6.3 6.9 13 18

when D=6Km (V2) Alumina

0.05 0.8 0.7 0.8 1.2 1.5

0.1 4.3 4.3 4.6 6.8 13

0.15 9.3 7.3 7.3 7.3 14

0.2 12 10.1 10.5 14 17

0.25 14 11 11.6 16 20

0.3 16 13 13.3 18 22

when D = 8Km (V2) Alumina

0.05 1.4 1.6 1.3 2.2 2.5

0.1 5.6 5 6 7 100.15 12 10.5 12.5 14 18

0.2 13 13 12 13 25

0.25 14 14 14 18 28

0.3 15 15 15 23 33

when D = 4Km (V2) Alumina- Titania

0.05 0.39 0.23 0.18 0.26 0.52

0.1 2.8 2.3 1.8 2.8 7.5

0.15 3 2.8 2.5 6 8.1

0.2 3.8 4.8 5 8.1 15

0.25 4.4 6.3 6.3 9.2 17

0.3 6.5 8 10 14 19

when D=6Km (V2) Alumina- Titania

0.05 0.7 0.59 0.28 0.32 0.8

0.1 3.7 3.5 2.8 3.7 11

0.15 9.2 9.2 7.8 8.5 14

0.2 10.1 9.8 10.1 11 200.25 10.2 10 12 14 22

0.3 11 11 16 16 24

when D = 8Km (V2) Alumina- Titania

0.05 0.4 0.6 0.375 0.75 0.9

0.1 3.8 4 4 4.2 8

0.15 9 9 8.5 10 15

0.2 10 11 11 12 20

0.25 11 12 13 16 26

0.3 12 13 15 18 29

when D = 4Km (V2) PSZ

0.05 2.8 2.7 1.8 3 4.5


8/13



84

Table 4. Experimenta l results of evaluating coefficient of friction during under different process parameters for differentcoatings.

0.1 3.2 3 2.3 3.5 9.6

0.15 3.5 3.3 3 5.5 11

0.2 4 4.3 6.8 9.2 13

0.25 4.6 4.9 8.8 10 15

0.3 6.2 6.8 10 11 17


0.05 3.2 3.2 2.6 4.2 8.6

0.1 7.2 4.7 3.7 4.9 12

0.15 8 7.3 8 8.8 14

0.2 10 8.2 10 12 18

0.25 9.5 9.1 12 13 20

0.3 9 9.5 16 15 22


0.05 5 4.75 4.5 4.75 8

0.1 8 6.5 7 8 14

0.15 9 9 9 11 17

0.2 10 9.5 9 11 19

0.25 9 8.5 13 13.5 23

0.3 8 11 16 17 25

Sliding Distance

D=4km (V2)

PSZ

Coating

(V1)

Normal;

pressure P (V0)

0.05 0.1 0.15 0.2 0.25 0.3 Sliding

Velocity V in m/s

0.82 0.82 0.83 0.84 0.85 0.86 V = 2.5

0.83 0.829 0.838 0.849 0.858 0.868 V = 5

0.8 0.79 0.77 0.77 0.81 0.82 V = 7.5

D=6km PSZ

0.05 0.1 0.15 0.2 0.25 0.3

0.03 0.07 0.071 0.1 0.1 0.09

0.028 0.045 0.07 0.075 0.081 0.095

0.042 0.04 0.08 0.09 0.102 0.13

D=4km Alumina

0.05 0.1 0.15 0.2 0.25 0.3

0.075 0.072 0.07 0.071 0.067 0.065

0.07 0.068 0.068 0.067 0.065 0.063

0.078 0.074 0.074 0.073 0.071 0.07

D=6km Alumina0.05 0.1 0.15 0.2 0.25 0.3

0.072 0.07 0.069 0.07 0.068 0.067

0.071 0.071 0.072 0.07 0.067 0.66

0.068 0.067 0.072 0.062 0.061 0.058

D=8km Alumina

0.05 0.1 0.15 0.2 0.25 0.3

0.073 0.072 0.071 0.07 0.068 0.067

0.072 0.072 0.072 0.07 0.067 0.66

0.067 0.066 0.064 0.062 0.061 0.058

D=4km Alumina-

Titania (AT)

0.05 0.1 0.15 0.2 0.25 0.3

0.085 0.086 0.088 0.088 0.089 0.089

0.085 0.84 0.088 0.088 0.082 0.082

0.081 0.079 0.081 0.078 0.08 0.081

D=6km AT

0.05 0.1 0.15 0.2 0.25 0.30.081 0.078 0.078 0.082 0.083 0.084

0.082 0.083 0.084 0.085 0.086 0.085

0.08 0.078 0.078 0.075 0.079 0.082

D=8km AT

0.05 0.1 0.15 0.2 0.25 0.3

0.083 0.082 0.085 0.083 0.085 0.087

0.082 0.083 0.086 0.088 0.089 0.089

0.081 0.082 0.084 0.08 0.081 0.085


9/13



85

4. GENETIC MODELS RESULTS AND DISCUSSION

Using GP simulation, percent porosity can be determined from the following

mathematical model (refer Table1.),

Where V3 =Thermal conductivity, V2 =Thermal diffusivity and V4=Toughness (refer table 1.)

Using GP simulation, Weight loss in can be determined from the following

mathematical model (refer Table3),

Where V3 =Hardness of coatings, V5 =Thermal conductivity, V4 =Thermal

diffusivity and V6= Toughness

where

Using GP simulation, Coefficient of Friction can be determined from the

following mathematical model ( refer Table 4.),

Where V4 =Thermal conductivity, V3 =Thermal diffusivity and V5= Toughness

Where V2

Using GP simulation, Rockwell Hardness number on C- scale can be

determined from the following mathematical model (refer Table2.),

Where V4 =Thermal conductivity, V1 =Thermal diffusivity and V0= Toughness

The percentage deviation of GP (expected) and experimental results for Normal

grinding forces simulation of the Grinding Machining process are shown on Figure 1


10/13



86

and the Tangential grinding forces are presented in Figure 2. Whereas results of

surface roughness Ra during grinding and lapping operations are shown in figure 3

and figure 4 respectively. The Discipulus GP technique was able to simulate these

output variables within an average of 1.9% of their measured value, with no value

exceeding a 5% deviation.

Figure 1. Percentage deviation curve between the best models regarding individual generation and experimental results ofpercentage porosity.

Figure 2. Percentage deviation curve between the best models regarding individual generation and experimental results of wearrate.

Figure 3. Percentage deviation curve between the best models regarding individual generation and experimental results of

coefficient of friction.


11/13



87

Figure 4. Percentage deviation curve between the best models regarding individual generation and

experimental results of hardness

4 CONCLUSIONGenetic programming (GP) is a highly versatile and useful tool for identifying

relationships in data for which a more precise theoretical construct is unavailable. The

experimental data in this research were in fact the environment to which the

population of models had to be adapted as much as possible. The models presented

are a result of the self-organization and stochastic processes taking place during

simulated evolution. In the research the genetic programming was used for predicting

the mechanical and tribological characteristics. In the proposed concept the

mathematical models for verifying the machinability are subject to adaptation. After

many trials, with the help of validation and testing data, the fittest model reliability is

98%. Thus, in this case the reliability was almost nearly 100% since some of

genetically developed models of mechanical and tribological parameters of ceramic

oxide coatings, out of many successful solutions are presented here. The accuracies of

solutions obtained by GP depend on applied evolutionary parameters and also on the

number of measurements and the accuracy of measurement. In general, more

measurements supply more information to evolution which improves the structures of

models and we have provided enough data.

In this paper, the genetic programming was used for predicting the mechanical and

tribological characteristics for verifying the experimental results of Controlling

parameters subject to adaptation. Its reliability is 98% in different parameters

prediction. In the testing phase, the genetically produced model gives the same result

as actually found out during the experiment, thereby with the reliability of cent

percent. It is inferred from our research findings that the genetic programming

approach could be well used for the prediction of Mechanical and tribological

characteristics of ceramic coatings without conducting the experiments. This helps to

establish efficient planning and optimizing of process for the quality production of

ceramic coatings depending upon the functional requirements by developing a

mathematical model.


12/13



88

REFERENCES

[1]Dr.J.Fazlur Rahman and Mohammed Yunus, Benefits of TBC Coatings onEngine applications, International conference, INCAM 2009 at Kalsalingam

University, Tamil Nadu, India, March 2009.

[2]Dr.J.Fazlur Rahman and Mohammed Yunus, Mechanical and Tribologicalcharacteristics of Tungsten Carbide Cobalt HVOF coatings International

conference held at Anjuman college of Engineering, Bhatkal October 2008.

[3]Noordin M. Y., Venkatesh V. C., Sharif.S, Elting. S., Abdullah. A., Applicationof response surface methodology in describing the performance of coated carbide

tools when turning AISI 1045 steel, Journal of Materials Processing Technology

46, 145 ,2004.

[4]Buckley, D.H. and Kazuhisa Miyoshi, Friction and wear of ceramics, J.of wear,Vol. 100, pp. 333-353 , 1984.

[5]Gane, N. and Beardsley, R., Measurement of wear of engineering materials usingpin on disc machine, Proc. Industrial Tribology Conference, Melbourne, pp.

187,1987.

[6]Hannink, R.H.J., Murray, M.J. and Scott, H.G., Friction and wear of Partially

Stabilized Zirconia: basic science and practical applications, J. of wear, Vol. 100,pp. 355-366,1984.

[7]Ishigaki, J., Nagata, R. and Iwasa, M., Friction and wear of Partially StabilizedZirconia, I. Mech. E., pp. 609-614, 1987.

[8]Pfender, E., Fundamental studies associated with plasma spray process, J. ofsurface coating technology, Vol. 34, No.1, 1988, pp. 1-7. 1988.

[9]Nordin, J.P. and Banzhaf, W., Controlling an Autonomous Robot with GeneticProgramming. In: Proceedings of 1996 AAAI fall symposium on Genetic

Programming, Cambridge, USA, 1996.

[10] Koza, J.R., (1992) Genetic Programming: On the Programming of Computersby Natural Selection.MIT Press, Cambridge, MA,1992.

[11] J. R. Koza, Genetic programming II, The MIT Press, Massachusetts, 1994.[12] Koza, Bennett, Andre, & Keane, (1999) GENETIC PROGRAMMING III

Darwinian Invention and Problem Solving, Morgan Kaufmann Publishers, Inc. pp.

1154, 1999.

[13] Francone, F., Discipulus Owners Manual and Discipulus Tutorials, RegisterMachine Learning Technologies, Inc, 1998-2000.

[14] Spector, L., Langdon, W., B., OReilly, U., Angeline, P.J., Advances inGenetic Programming Volume 3, MIT Press, pp. 476, 1999.

[15] Tanguy, B. Besson J., Piques R., Pineau A., Ductile to brittle transition of anA508 steel characterized by Charpy impact test Part 1: experimental results,

Engineering Fracture Mechanics 72, pp.49 72, 2005.

[16] M. Kovacic, J. Balic and M. Brezocnik, Evolutionary approach for cutting

forces prediction in milling, Journal of materials processing technology, 155/156,pp.1647-1652, 2004.

[17] H. Kurtaran, B. Ozcelik and T. Erzurumlu, Warpage optimization of a busceiling lamp base using neural network model and genetic algorithm, Journal of

materials processing technology, 169(2), pp.314-319, 2005.

[18] Sette S., Boullart L. Genetic programming: principles and applications,Engineering Applications of Artificial Intelligence 14, pp.727 736, 2001


13/13



89

[19] Pierreval H., Caux C., Paris J.L., Viguier F. Evolutionary approaches to thedesign and organization of manufacturing system, Computers & Industrial

Engineering 44, pp.339-364, 2003.

[20] Gusel L., Brezocnik M. Modeling of impact toughness of cold formed materialby genetic programming, Comp. Mat. Sc. 37, pp.476 482, 2006.

[21] Chang Y.S., Kwang S.P., Kim B.Y. Nonlinear model for ECG R-R interval

variation using genetic programming approach, Future Generation ComputerSystems 21, pp.1117-1123.

[22] Brezocnik M., Gusel L, Predicting stress distribution in cold-formed materialwith genetic programming, International Journal of Advanced Manufacturing

Technology, Vol. 23, pp.467-474, 2004.

[23] M. Brezocnik, M. Kovacic and M. Ficko, Prediction of surface roughness withgenetic programming, Journal of materials processing technology, 157/158, pp.28-

36, 2004.

Documents

Ceramic Coatings Using GP Approach