ARTIFICIAL NEURAL NETWORK MODELING AND OPTIMIZATION IN HONING PROCESS

http://www.iaeme.com/IJCET/index.asp 67 editor@iaeme.com
International Journal of Computer Engineering & Technology (IJCET)
Volume 7, Issue 3, May-June 2016, pp. 67–77, Article ID: IJCET_07_03_006
Available online at
http://www.iaeme.com/IJCET/issues.asp?JType=IJCET&VType=7&IType=3
Journal Impact Factor (2016): 9.3590 (Calculated by GISI) www.jifactor.com
ISSN Print: 0976-6367 and ISSN Online: 0976–6375
© IAEME Publication
ARTIFICIAL NEURAL NETWORK
MODELING AND OPTIMIZATION IN
HONING PROCESS
Benu Singh
Research Scholar, Department of Computer Application,
B B Das University, Lucknow, India
Sunita Bansal
Professor, Department of Computer Application,
B B Das University, Lucknow, India
Puneet Mishra
Professor, Department of Computer Science,
Lucknow University, Lucknow, India
ABSTRACT
Determination of process parameters and maximum utility of the resources
are the two main concerns while designing the manufacturing process for the
products. The process parameters affect the final product shape and aesthetic
look; whereas the utility refers to the output. The present study is devoted to
the development of ANN models for the analysis of the honing process applied
to an actual industrial component, namely the connecting rod of a motor bike.
The surface quality of the honed components is measured with the help of a
Talysurf Intra machine. Traditionally accepted measure of surface roughness
is Center Line Average (Ra) and therefore the same has been adopted for the
present study. Six process parameters, namely grit size, coolant temperature,
honing speed, honing feed, honing time and operator experience are
considered as the input variables. In general, honing experiment is time
consuming and expensive, especially when carried out in an industrial setting.
Fractional factorial design is, therefore, used to minimize the number of
experimental observations. The relationship between the inputs and outputs of
the honing process is determined with the help of ANNs. The ANN models
developed during the present investigation have a three layer structure,
namely one input layer, one output layer and one hidden layer. The developed
model is used to study the effect of the process parameter on Ra values.
Key words: Surface Roughness, Honing, Artificial Neural Network, and
Optimization

Benu Singh, Sunita Bansal and Puneet Mishra
Cite this Article: Benu Singh, Sunita Bansal and Puneet Mishra, Artificial
Neural Network Modeling and Optimization In Honing Process, International
Journal of Computer Engineering and Technology, 7(3), 2016, pp. 67–77.
http://www.iaeme.com/IJCET/issues.asp?JType=IJCET&VType=7&IType=3
1. INTRODUCTION
Surface roughness plays an important and valuable role in the practical and theoretical
applications of a manufactured component. It is widely used as an index of product
quality and is in most cases a technical requirement for mechanical products. [20]
This is an important design consideration as it impacts many part characteristics such
as fatigue strength, assembly tolerances, coefficient of friction, wear rate, corrosion
resistance and aesthetics. [7] Roughness of the machined surface is important while
planning for the machining operation such as turning, milling and honing etc.
Production process parameters selected for the machining of a part influence the
surface roughness and reliability of components. [19]
Honing is an important fine finishing operation, often used for internal cylindrical
surfaces such as gun barrels, hydraulic cylinders, bearings and engine cylinder bores.
[1] Excess material is removed by means of slow moving abrasive sticks pressed
against the surface to be machined. Two kinds of motion, namely rotational and
reciprocating, are imparted by the honing machine to the hone (or honing tool)
carrying the abrasive sticks. Although honing can be used on flat and external
cylindrical surfaces too, it is predominantly used for finishing internal cylindrical
surfaces (holes). Guide pads on the hone allow proper alignment (or float) of the
abrasive sticks with the axis of the cylindrical hole. Size of workpiece determines the
thickness of the hone. Larger the size, greater is the number of sticks. The usual
length of the abrasive sticks is about one-half of the workpiece length. The stroke
length during reciprocation is such that the abrasive sticks protrude one-fourth to one-
third of their length on either side of the hole. This is done to ensure uniform wear of
the abrasive sticks. The honing tool has an inbuilt mechanism to expand the honing
sticks radially outwards and impart to them the needed radial in-feed. When the
honing of the piece is completed, the honing sticks are retracted radially inward and
the hone is withdrawn from the workpiece.
2. LITERATURE REVIEW
The aim of a neural network is to learn from events that have happened during
experiments and to be able to apply this to future situations or for new independent
data. Experimental studies in neurophysiology show that the response of a biological
neuron is random and it is possible to obtain predictable results only by averaging
many observations. [22]. The development of neural networks began with the
pioneering work of McCulloch and Pitts. [17] The physiological learning rule for
synaptic modification was proposed for showing the connectivity of the brain with the
continually changing as an organism learns differing functional tasks. [10] The first
so-called self-organising maps using competitive learning was developed by Kohonen
in 1972 and were capable of reproducing important aspects of the structure of
biological neural networks. [13] Later, the idea of an energy function to formulate a
new way of understanding computations performed by recurrent networks with
symmetric synaptic connections was used. [11] Also, the development of a back
propagation algorithm was reported in 1986. [24]

Artificial Neural Network Modeling and Optimization In Honing Process
Manufacturing process parameter modelling and optimization is one of the
advanced issues in the present environment. The process parameters if modeled well,
resulting in low tool wear, improved tool shelf life and better surface asperities etc.
The process parameters like feed rate, production quantity, workpiece hardness,
cutting tool point angle, spindle speed and depth of cut etc are the affecting parameter
of surface roughness. [9] The new methods to predict the surface roughness of the
workpiece material during the manufacturing process are developed time to time by
the researchers. [27] The machining forces, tool-wear relationship, effect of cutting
conditions and tool geometry were analyzed earlier during the manufacturing process
by regression analysis. [2, 14, 15, 23, 26] The series of experiments were conducted
to determine the affects of these parameters on surface roughness [18]
However, it is not easy to design the process parameter for complex shaped
products involving a number of factors that are not linearly related. Still, this method
has been widely used in manufacturing situation when there is a single output
parameter of interest, and the correlations with the input variables are not so complex.
[25]
3. PROBLEM FORMULATION
Honing involves complex physical phenomena, including shearing, ploughing, heat
transfer and lubrication. A variety of input variables such as physical and mechanical
properties of workpiece and abrasive grits, feed, speed, coolant temperature and man-
machine interaction affect the process. Exact mathematical analysis of the honing
process is not available in literature. Therefore, process planners have traditionally
resorted to hand book knowledge/experience or rules of thumb for designing the
process. As literature depicted that CLA values corresponding to various input
variables has the influence on the surface quality in the honing process. [16]
In the present environment, several computer application software are available
for modeling and optimization in real life problem. These software’s help in selecting
the appropriate significant parameters to obtain the required product quality, also they
enhance the manufacturing system performance. Artificial Neural Network Model is
versatile to design and optimize the process parameters. Further, the threefold cross
over approach is helpful to design the test datasets based on the results of the
experiment. Each network is trained, validated and tested with the help of these
datasets.
4. METHODOLOGY
The practitioners already have used regression analysis, genetic algorithms, expert
systems or artificial neural networks (ANN) for analyzing the process. [4, 8] In
particular, ANNs offer a new and intelligent alternative for relating the input variables
of the honing process to its output parameters. An ANN has the capability to learn or
to be trained about the honing process on account of its robustness, its ability to
incorporate non-linear dependencies between the input and output variables and its
power of generalization. [21] Artificial neural network seems to be more robust
because it exhibits the highest level of integration with computers. In contrast with
other methods, ANN can manage noisy or incomplete data with ease. The neural
networks possess the ability to learn from a set of data without the need for a full
specification of the decision model; they are believed to automatically provide any
needed data transformations. There is no need to explicitly formulate the problem, the
solution algorithm or to write code and information processing is distributed over the
neurons which operate in parallel. In the present work, the back propagation artificial

neural network models is used to simulate the honing process and establish cause and
affect relationships between various input factors and output characteristics.
For the selected models the difference between observed and predicted values of
the response variable is checked to be within the specified tolerance limit. The R2
(a
common statistic in ANN applications) is equal to one minus ratio of the sum of
squares of estimated errors (i.e. the deviation between predicted and actual value of
the dependent variable) to the sum of squared deviations about the mean of the
dependent variable. It is a measure of the total variation of the dependent variable.
Where yi indicates the observed response,
and ti indicates target or predicted response.
4.1. Experimental Study
Connecting rod (Shown in Figure 1) of an internal combustion engine is subjected to
high rate cyclic loading. High accuracy and tight tolerances of cylindrical holes in the
connecting rod of an automobile are ever more demanding requirements. Close
tolerances and fits to mating components, namely the crankshaft on one end and
piston head on the other, are specified for the sake of component reliability.
Figure 1 Geometry of connecting rod
The connecting rod is manufactured by the powder metallurgy technology,
forging, and sometimes even casting. The powder forged (PF) rod is fabricated by
consolidating metal powders into a form, sintering the form and machining to final
dimensions. The forged connecting rod is fabricated by starting with a wrought steel
billet, followed by forging and machining to the required dimensions. The quality of
the bore surface of a connecting rod influences oil consumption, noxious emissions,
and running performance. Inner surfaces of the big end bore and small end bore of a
connecting rod are finished by rough honing followed by finish honing. In the present
investigation, the honing experiments were conducted on connecting rods used in the
Yamaha motorcycles. SCM 420H steel is used as the material of construction of the

connecting rods. The chemical composition of the connecting rod material is given in
Table 1, standard values in Table 2 and the level of the process variable is given in
Table 3.
Table 1 Chemical composition of material of the connecting rod
Ingredient C Si Mn S P Cr Mo
Composition .17-.23 .15-.35 .55-.90 .00-.03 .00-.03 .85-1.25 .15-.35
Table 2 Standard values for contact pressure
Abrasive material Rough honing (N/cm2) Finish honing (N/cm2)
Ceramic honing stones 50-250 20-100
Plastic bonded honing stones 200-400 40-250
Diamond honing edges 300-700 100-300
Boron nitride honing ledges 200-400 100-200
Table 3 Levels of the variables
Grit size Temperature Speed Feed Time Operator
Skill
Number 0C rpm µm Second Years
400 20 800 5 20 1
600 30 1000 10 30 7
800 40 1200 40
1000
4.2. Model Development
In the present study, six input factors relevant to the chosen honing process have been
considered. It was decided to adopt a 2-level approach for varying each factor in order
during a given set of observations. The most intuitive scheme to study these factors
and how they affect the responses would be to vary the factors of interest in a 2k
full
factorial design, i.e. to try all possible combinations. [5] To generate this design, 24
design in the factors P1, P2, P3, P4, and then added two columns for P5 and P6. In this,
P1, P2, P3, P4, P5 and P6 represent the six input variables taken for experiments. To
find the two new columns select the two design generators I = P2 P3 P4 P6 and I=
P1P2P3P5. Thus column P5 would be found from P5= P1P2P3 and column P6 would be

P6= P2P3P4 (Table 4). Therefore columns P1P2P3P5 and P2P3P4P6 are equal to
identity column.
Table 4 2
6-2
fractional factorial design of Experiment
Ex. No. Grit Size Temperature Speed Feed Time Operator
P1 P2 P3 P4 P5 Skills P6
1 -1 -1 -1 -1 -1 -1
2 -1 -1 -1 +1 -1 +1
3 -1 -1 +1 -1 +1 +1
4 -1 -1 +1 +1 +1 -1
5 -1 +1 -1 -1 +1 +1
6 -1 +1 -1 +1 +1 -1
7 -1 +1 +1 -1 -1 -1
8 -1 +1 +1 +1 -1 +1
9 +1 -1 -1 -1 +1 -1
10 +1 -1 -1 +1 +1 +1
11 +1 -1 +1 -1 -1 +1
12 +1 -1 +1 +1 -1 -1
13 +1 +1 -1 -1 -1 +1
14 +1 +1 -1 +1 -1 -1
15 +1 +1 +1 -1 +1 -1
16 +1 +1 +1 +1 +1 +1
Two levels of each parameter were taken for the model construction and model
validation experiments. Five (05) different set of experiments (readings) was
conducted, with each set comprising of 16 experiments.
Table 5 Factors and levels for various set of readings
Set Levels
Factors
P1 P2 P3 P4 P5 P6
No. ºC Rpm µm Sec. Years
1st
High Level (+) 600 30 1000 10 40 7
Low Level (-) 400 20 800 5 30 1
2nd
High Level (+) 600 40 1000 10 30 7
Low Level (-) 400 30 800 5 20 1
3rd
High Level (+) 800 40 1000 10 30 7
Low Level (-) 600 30 800 5 20 1
4th
High Level (+) 800 30 1200 10 40 7
Low Level (-) 600 20 1000 5 30 1
5th
High Level (+) 1000 40 1200 10 40 7
Low Level (-) 800 30 1000 5 30 1

The present paper describes the development and application of a artificial neural
network model for the prediction of surface roughness in a honing process. The
proposed ANNs have a 3-layer architecture, comprising of 6 neurons in the input layer,
optimized number of neurons h in the hidden layer and a single neuron in the output layer
shown in Fig. 2.
Figure 2 Schematic diagram for the Single Output Type Neural Network
Graphs are plotted to illustrate the comparisons of ANN predictions with the
experimental observations for both trained and validated data corresponding to the
above models. Fig.3 and 4 show the graphical plot for the R2
value and the slope of
ANN predictions of Ra training and validation datasets respectively for the Model-1S.
Similarly, figure 5 & 6 shows the graphical plot for Model-2S, and figure 7 & 8
shows the graphical plot for Models 3S.
Figure 3 Predictions against Observations of Ra for Model-1S (Training dataset)

Figure 4 Predictions against Observations of Ra for Model-1S (Validation dataset)
Figure 5 Predictions against Observations of Ra for Models-2S (Training dataset)
Figure 6 Predictions against Observations of Ra for Models-2S (Validation dataset)

Figure 7 Predictions against Observations of Ra for Models-3S (Training dataset)
Figure 8 Predictions against Observations of Ra for Models-3S (Validation dataset)
The correlation coefficients of the training and validation data of Ra component of
surface roughness is measured for all the three models the values of R2
show a good
correlation for both training and validation data.
5. RESULTS AND CONCLUSIONS
The performance of a connecting rod can be expressed in terms of Ra surface
roughness parameter. The present work is based on modeling of the honing process
carried out on a precision automobile component, namely the big end bore of a
connecting rod. The component is designed to be a load bearing and wear resistant, to
offer low friction and possess long life. The proposed models relate the input
parameters of the honing process to Ra values. The control of the surface roughness
parameter leads to achieve functional requirements like reduction of wear,
improvement of oil efficiency and improved life of the component.
Further, the graph between observed and predicted values shows the R2
and the
linear relationship. The R2
value plotted in graph reflects the amount of noise in the

data. The lower value of R2
indicates that important and systematic factors have been
omitted from the model and the model is inadequate.
REFERENCES
[1] Armergo, E.J.A. and Brown, R.H. (1969), The machining of metals, Prentice-
Hall Inc., Englewood Cliffs, New Jersey, pp370.
[2] Arora, P.K. (2008), Statistical modelling of surface roughness in turning
process, M.Tech. Dissertation, Department of Mechanical Engineering, CITM,
MD University, Rohtak, Haryana, India.
[3] ASME, (1996), Surface Texture (Surface roughness, Waviness, and Lay): An
American Standard, ASME B46.1-1995 (Revision of ANSI/ASME B46.1-
1985), and ASME Press, New York.
[4] Bernados, P.G. and Vosniako, G.C. (2002), Predicting surface roughness in
CNC face milling using neural networks and Taguchi’s design of experiments,
Robotics and Computer Integrated Manufacturing, 18, pp343–354.
[5] Box, G.E.P, Hunter, W.G., and Hunter, J.S. (1978), Statistics for experimenter:
An Introduction to design, data analysis, and model building, New York, John
Wiley and Sons.
[6] Dabnun, M.A., Hashmi, M.S.J. and El-Baradie, M.A. (2005), Surface roughness
prediction model by design of experiments for turning machinable glass ceramic
(Macor), Journal of Materials Processing and Technology, 164(65), pp1289–
1293.
[7] Feng, C-X. and Wang, X. (2002), Surface roughness predictive modeling Neural
Networks vs. regression, IIE Transactions,.35, pp11–27.
[8] Feng, C-X. (2001), Experimental study of the effect of turning parameters on
surface roughness, Proceedings of 2001 Industrial Engineering Research
Conference, Institute of Industrial Engineers, Norcross, GA, Paper 2036.
[9] Hebb, D.O. (1949), The organization of behavior, Wiley, New York.
[10] Hopfield, J.J. (1982), Neural network and physical system with emergent
collective computational capabilities, In proceedings of national academy of
science (USA), 79, pp2554–2558.
[11] Kohonen, T. (1972), Correlation matrix memories, IEEE Transactions on
Computers, 21, pp353–359.
[12] Lin, S.C. and Chang, M.F. (1998), A study on the effects of vibrations on the
surface finish using a surface topography simulation model for turning,
International Journal of Machine Tools and Manufacture, 38, pp763–982.
[13] Lin, J.T, Bhattacharya, D. and Kecman, V. (2003), Multiple regression and
neural networks analyses in composite machining, Journal of Composite Science
and Technology, 63, pp539-548.
[14] Malburg, M.C., Raja, J. (1993), Characterization of surface texture generated by
plateau honing process, CIRP Annals, 42 (1), pp637–640.
[15] McCulloch, W. and Pitts, W. (1943), A logical calculus of the ideas imminent in
nervous activity, Bulletin of Mathematical Biophysics,.5, pp115–133.
[16] Munoz-Escalona, P and Cassier, Z. (1998), Influence of the critical cutting speed
on the surface finish of turned steel, Wear,.218, pp103–109.
[17] Ogodorov, V.A. (2008), Identification of honing processes, Russian Engineering
Research, 28, pp93–98.

[18] Ozcelik, B. and Bayramouglu, M. (2005), The statistical modeling of surface
roughness in high speed flat end milling, International Journal of Machine Tools
and Manufacture, 46, pp1395–1402.
[19] Ozel, T. and Nadgir, A. (2002), Prediction of flank wears by using back
propagation neural network modeling when cutting hardened H-13 steel with
chamfered and honed CBN tools, International Journals of Machine Tools and
Manufacture, 42, pp287–297.
[20] Patrikar, R.M., Ramanathan, K., and Zhuang, W. (2002), Surface roughness
modeling with Neural Networks, Proceedings of 9
th
International Conference on
Neural Information Processing, 4, pp1895–1899
[21] Rangwala, S.S. and Dornfeld D.A. (1989), Learning and optimization of
machining operations using computing abilities of neural networks, IEEE Trans
Systems, Man Cybernetics, 19, pp299–314.
[22] Rumelhart, D.E. and McClelland, J. (1986), Parallel distributed processing, MIT
Press, Cambridge, Mass.
[23] Smith, A.E. and Mason, A.K. (1997), Cost estimation predictive modelling:
regression versus neural network, The Engineering Economist, 42(2), pp137–
161.
[24] Wang, M.Y. and Chang, H.Y. (2004), Experimental study of surface roughness
in slot end milling, International Journal of Machine Tools and Manufacture,.44,
pp51–57.
[25] Sweety H Meshram, Vinaykeswani and Nileshbodne, Data Filtration and
Simulation by Artificial Neural Network, International Journal of Electronics
and Communication Engineering & Technology, 5(7), 2014, pp. 46–55.
[26] Nikhil Balkrishna Bole and Rohansing R. Kait, Experimental Verification of
Honing Process Parameters on Surface Roughness with Form Talysurf 120
Machine In Taper Bearings, International Journal of Mechanical Engineering
and Technology, 5(1), 2014, pp. 68–78.
[27] Zhou, X. and Xi, F. (2002), Modeling and predicting surface roughness of the
grinding process, International Journal of Machine Tools and Manufacture, 46,
pp969–977.

ARTIFICIAL NEURAL NETWORK MODELING AND OPTIMIZATION IN HONING PROCESS

More Related Content

What's hot (20)

Viewers also liked (15)

Similar to ARTIFICIAL NEURAL NETWORK MODELING AND OPTIMIZATION IN HONING PROCESS (20)

More from IAEME Publication (20)

Recently uploaded (20)

ARTIFICIAL NEURAL NETWORK MODELING AND OPTIMIZATION IN HONING PROCESS