Proceedings of the National Conference on Recent Innovations in Design and Analysis (RIDA) 18 March 2009

Department of Mechanical Engineering, Government College of Engineering, Tirunelveli-627007. Tamil Nadu, India

Modeling and optimization of CNC drilling operations in Natural Fiber

reinforced composites using Factorial, Nelder Mead, Genetic Algorithm
methods
R.Kumarasubramanian1 and S.Jayabal2
1
P.G Scholar, Department of Mechanical Engineering, A. C. College of Engineering and Technology,
Karaikudi. E-mail id: rkscad@gmail.com
2
Senior lecturer, Department of Mechanical Engineering, A. C. College of Engineering and Technology,
Karaikudi. E-mail id: jayabalsubbaian@rediffmail.com

ABSTRACT
The useful life of a cutting tool and its operating conditions largely control the economics of the
machining operations. Hence, it is imperative that the condition of the cutting tool, particularly some
indication as to when it requires changing, to be monitored. The drilling operation is frequently used as a
preliminary step for many operations like boring, reaming and tapping; however, the operation itself is
complex and demanding. This paper discusses a mathematical model for correlating the interactions of
some drilling parameters such as drill bit diameter, speed and feed rate their effects on some responses
such as thrust force, torque and tool wear during drilling of roselle fiber reinforced composites by means
of factorial method. A three level full Factorial design was chosen for experimentation using CNC
milling machine and drill tool dynamometer. The measures of Central Tendency, Dispersion and
Interaction between variables were estimated using Minitab 15 and Design Expert 7.1.6 statistical
packages. The significance of the mathematical model developed was ascertained using statistical
packages Design Expert 7.1.6. The optimum value of responses was calculated using desirability
considerations. The equations are not only for predicting optimum operating conditions and also for
minimizing power and maximizing metal removal rate, hole quality etc during drilling operations.

Keywords: Factorial Design; Genetic Algorithm; Nelder Mead method; CNC milling machine; Machine
vision system; Tool wear
.

1. INTRODUCTION processing equipment, renewability and

biodegradability. It has been observed that natural
Over the last decades, composites of
fiber reinforced composites have properties similar
polymers reinforced by natural fibers received
to traditional synthetic fiber reinforced
increased attention. Natural fibers such as coir,
composites. Natural fiber composites have been
sisal, roselle and jute possess good reinforcing
studied and reviewed by a number of researchers.
capability when properly compounded with
During the past decade, a number of significant
polymers. Natural fibers are now considered as
industries such as the automotive, construction or
serious alternative to synthetic fibers for use in
packaging industries have shown massive interest
various fields. The use of natural fibers as
in the progress of new biocomposites materials.
reinforcing materials in both thermoplastic and
One of the most appropriate examples of this is
thermo set matrix composites provides positive
the substitution of inorganic fibers such as glass
environmental benefits with respect to ultimate
oraramid fibers by natural fibers. All these
disposability and best utilization of raw materials.
properties have made natural fibers very attractive
Currently, studies on use of
for various industries currently engaged in
lignocelluloses bio fibers in place of synthetic
searching for new and alternate products to
fibers as reinforcing materials are being pursued
synthetic fiber reinforced composites. Tagliaferri
vigorously. These bio fibers are being extensively
et al. reported that the machining of composite
used for the production of cost effective eco
materials requires a better understanding of cutting
friendly bio The advantages of natural fibers over
processes to achieve accuracy and efficiency.
traditional reinforcing materials such as glass
reported that drilling operations using twist drill
fiber, carbon fiber etc are their specific strength
were the most commonly applied process amongst
properties, easy availability, lightweight, ease of
all machining operations for generating holes for
separation, enhanced energy recovery, high
riveting and fastening structural assemblies. The
toughness, non-corrosive nature, low density, low
achieving the dimensional accuracy, good surface
cost, good respiratory irritation, less abrasion to
finish and deep holes etc are impracticable during

67 RIDA
 Proceedings of the National Conference on Recent Innovations in Design and Analysis (RIDA) 18 March 2009
Department of Mechanical Engineering, Government College of Engineering, Tirunelveli-627007. Tamil Nadu, India

machining of FRP composite materials. Tsao et al. x3 = f = Feed rate in mm/rev

used multivariable linear regression analysis to
Table 1 Assignment of the levels to the factors
correlate machining parameters (feed rate, spindle
speed, and drill diameter) and delamination in a
CFRP laminate. Davim et al. and Davim et al. Drill-bit
Speed Feed rate
used multivariable linear regression analysis to Level diameter
(rpm) (mm/rev)
correlate cutting parameters (cutting velocity, feed (mm)
rate) and cutting power, specific cutting pressure
1 6 600 0.1
and delamination factor in CFRP composite
laminate Manufacturing requires reliable models 2 8 1200 0.2
and methods for the prediction of output
performance of machining process due to the wide
spread use of highly automated machine tools in 3 10 1800 0.3
the industry.
In this investigation, natural fiber sisal Figure 1 Schematic of experimental set up.
composite was fabricated using molding method.
A series of experiments were conducted using
CNC milling machine to drill the coir composites.
A drill tool dynamometer was used to record the
cutting force, torque and tool wear. A machine
vision system was used to measure the tool wear.
In this work, thrust force, torque and tool wear are
taken as response variables (output variables) and
cutting speed, feed rate and drill-bit diameter were
taken as input variables.
2. EXPERIMENTAL SETUP
The three axis CNC Vertical Machining
Center was used to drill the natural fiber
reinforced composites using HSS twist drills for
the machining of roselle fiber reinforced polyester
The response variables in this response surface
composites. A two-component drill tool
modeling as follows.
dynamometer was used to record the thrust force
y =th or tq or tw
and torque. Conventional high-speed steel twist
where th = thrust force in N
drills were used as much as cemented tungsten
tq = torque in Nm
carbide drills. Tool geometry is a relevant aspect
tw = tool wear in mm
to be considered in drilling of fiber-reinforced
Coefficient (βi) is a constant multiplicative factor
plastics, particularly when the quality of the
of the mathematical model. The coefficients
machined hole is critical.
calculated using SPSS package for thrust force,
Direct measurements by dynamometers
torque and tool wear are shown in the Table 2.
have won general acceptance and since then, many
Mean and ratio value calculated for each response
research workers in the field of metal cutting have
was given as Table 3.
tackled the problem of constructing dynamometers
to measure forces. The common feature in all
3.1. Mathematical model of thrust force
types of dynamometers is the measuring springs
whose deflections are proportional to cutting
The mathematical relationship for correlating the
forces. The schematic of experimental set up is
thrust force and the considered process variables
shown as Figure 1.
was obtained from the coefficients resulting from
The assignment levels of factors listed as
the SPSS software output. Coefficient of
Table 1.
determination, R2 is used in the context of
3. REGRESSION ANALYSIS statistical models whose main purpose is the
prediction of future outcomes on the basis of other
Polynomial curve fitting equations
related information.
normally exist in first degree.
It is the proportion of variability in a data
y = β0 + β1x1 + β2x2 + β3x3+ β12 x1x2+ β23
set that is accounted for by the statistical model. It
x2x3+ β31 x3x1+ β123 x1x2 x3
provides a measure of how well future outcomes
The process variables in this response
are likely to be predicted by the model. R squared
surface modeling as follows.
value was 0.94154.The following equation shows
x1 = d = Drill diameter in mm
the mathematical model for thrust force.
x2 = s = Speed in rpm

68 RIDA
 Proceedings of the National Conference on Recent Innovations in Design and Analysis (RIDA) 18 March 2009
Department of Mechanical Engineering, Government College of Engineering, Tirunelveli-627007. Tamil Nadu, India

th = -23.1483+5.8334d-0.0175s+63.8894f- squared value is 0.92677.The following equation

shows the mathematical model for thrust force.
0.0001ds+0.0861sf+4.1666fd+0.0104dsf
tw = -0.4689+0.0863d+0.0002s+1.2778f-
(1) 0.00002ds -0.0011sf-0.1458fd+0.0026dsf
(3)
Table 2 Calculated values of coefficients of
4. FACTORIAL OPTIMIZATION
second order polynomial
A 33 full factorial design with a experimental
Coefficients th tq tw runs were carried out. The thrust force, torque and
flank wear were the response variables recorded
β0 -23.1483 -3.9729 -0.4689 for each run. The effect of the machining
parameters is another important aspect to be
considered. The optimum value of responses was
β1 5.8334 0.6664 0.0863 given as Table 4.
Table 4 Optimum parameter in factorial method
β2 -0.0175 0.0034 0.0002
Run d s f th tq tw
β3 63.8894 17.6278 1.2778
10 6 600 0.1 15 1.47 0.1
β12 -0.0001 -0.0005 -0.0002
The main effects plot for responses were
shown as Figure 2.The effects of individual
β23 0.0861 -0.0204 -0.0011 variable on responses were clearly studied using
main effects plot. The Interaction plot for
responses was shown as Figure 3.The interaction
Β31 4.1666 -0.3955 -0.1458
effects of two variables on responses were studied
using Interaction plot. The Cube plot, 3D surface
β123 0.0104 0.0026 0.0002 plot, Contour plot and surface plot for the
responses were shown in the figure 4, 5, 6 and 7.

Table 3 Calculated values of mean and ratio

Figure 2 Main Effects plot for responses
Signal to
Std.
Variables Mean Noise
Dev.
ratio
Thrust
62 31.54 1.97
force(N)
Torque (N) 3.76 1.69 2.22

Tool wear 0.27 0.12 2.25

3.2. Mathematical model for torque

The mathematical relationship for
correlating the torque and the considered process
variables was obtained from the coefficients
resulting from the SPSS software output. R
squared value was 0.91165.The following
equation shows the mathematical model for thrust
force.
tq = -3.9729+0.6664d+0.0034s+17.6278f-
0.0005ds-0.0204sf-0.3955fd+.0026dsf
(2)
3.3. Mathematical model of tool wear
The mathematical relationship for
correlating the torque and the considered process
variables is obtained from the coefficients
resulting from the SPSS software output. R

69 RIDA
 Proceedings of the National Conference on Recent Innovations in Design and Analysis (RIDA) 18 March 2009
Department of Mechanical Engineering, Government College of Engineering, Tirunelveli-627007. Tamil Nadu, India

Figure 5 3D surface for responses

Figure 3 Interaction plot for responses

Figure 6 Contour plot for responses

Figure 4 Cube plot for responses

70 RIDA
 Proceedings of the National Conference on Recent Innovations in Design and Analysis (RIDA) 18 March 2009
Department of Mechanical Engineering, Government College of Engineering, Tirunelveli-627007. Tamil Nadu, India

5. OPTIMIZATION USING NELDER-MEAD

METHOD
A pattern search algorithm computes a
sequence of points that get closer and closer to the
optimal point. A pattern is a collection of vectors
that the algorithm uses to determine which points
to search at each iteration. At each step, the
algorithm searches a set of points, called a mesh,
around the current point -- the point computed at
the previous step of the algorithm. The algorithm
forms the mesh by adding the current point to a
scalar multiple of a fixed set of vectors called a
pattern. If the algorithm finds a point in the mesh
that improves the objective function at the current
point, the new point becomes the current point at
the next step of the algorithm. The pattern search
algorithm finds a sequence of points, x0, x1, x2...
that approaches the optimal point. The value of the
objective function decreases from each point in the
sequence to the next. The pattern search algorithm
stops the current iteration as soon as it finds a
mesh point whose fitness value is smaller than that
of the current point. Consequently, the algorithm
might not poll all the mesh points. You can make
the algorithm poll all the mesh points by setting
Figure 7 Surface responses complete poll to on. The design criteria and
optimum parameters calculated using MatLab 7.0
Pattern Search Tool was given in the Table 5.
Table 5 Design criteria and Optimum
parameter in Nelder mead method
Design Parameters value
Lower bound [6 600 0.1]
Upper bound [10 1800 .3]
Stopping criteria 1E-6
Iteration 20

Optimum parameters
d s f tw
6 600 0.1 0.02

71 RIDA
 Proceedings of the National Conference on Recent Innovations in Design and Analysis (RIDA) 18 March 2009
Department of Mechanical Engineering, Government College of Engineering, Tirunelveli-627007. Tamil Nadu, India

Figure 8 Pattern search graph Population size 60

No. of generation 100
Crossover probability 0.7
Mutation probability 0.01
String length 30

Optimum parameters
d s f tw
6 600 0.1 0.14

7. CONCLUSION
In thrust force analysis, if the drill bit
diameter increases the thrust force
correspondingly get increase with respect to the
feed rate levels. The interaction between drill bit
6. OPTIMIZATION USING GENETIC
diameter and feed rate was maximum.
ALGORITHM
In torque analysis the most significant
Genetic algorithms are search algorithms variable was also feed rate. The interaction
for optimization, based on the mechanics of between drill bit diameter and feed rate was
natural selection and genetics. The power of these maximum. The interaction between drill bit
algorithms is derived from a very simple heuristic diameter and speed and all the three parameters
assumption that the best solution will be found in were insignificant.
the regions of solution space containing high Speed was not playing major role in tool
proposition of good solution and that these regions wear in drilling of sisal composites. The variable
can be identified by judicious and robust sampling interaction was also minimum in tool wear
of the solution space. The mechanics of genetic estimation.
algorithms is simple, involving copying of binary REFERENCES
strings and the swapping of the binary strings. The 1. A.M. Abrao, P.E. Faria, J.C. Campos Rubio,
simplicity of operation and computational P. Reis and J. Paulo Davim (2006) ‘Drilling
efficiency are the main attractions of genetic of fiber reinforced plastics: A review’
algorithm approach. The computations are carried Elsevier - Journal of Materials Processing
out in three stages to get a result in one generation Technology 186 (2007) 1–7
or iteration. The three stages are (a) Reproduction 2. U.A. Khashaba, M.A. Seif and M.A. Elhamid
(b) Cross-over (c) Mutation (2006) ‘Drilling analysis of chopped
Algorithm composites’ Elsevier - Composites: Part A 38
Step 1 chooses a coding to represent problem (2007) 61–70
parameters, a selection operator, o cross over 3. S.C. Lin and IK. Chen (2005) ‘Drilling carbon
operator and a mutation operator. Choose fiber-reinforced composite material at high
population size, n, crossover probability, pc and speed’ Composite Structures 71 (2005) 407–
mutation probability, pm. Initialize a random 413
population of strings of size l choose a maximum 4. N.S. Mohan, A. Ramachandra and S.M.
allowable generation number t max. Set t = 0 Kulkarni (2005) ‘Influence of process
Step 2 Evaluate each string in the population parameters on cutting force and torque during
Step 3 If t > tmax or other termination criteria is drilling of glass–fiber polyester reinforced
satisfied, Terminate. composites’ Elsevier - Composite Structures
Step 4 Perform reproduction on the population. 71 (2005) 407–413
Step 5 Perform crossover on random pairs of 5. K. Palanikumar and J. Paulo Davim (2006)
strings. ‘Mathematical model to predict tool wear on
Step 6 Perform mutation on every string. the machining of glass fibre reinforced plastic
Step 7 Evaluate strings in the new population. Set composites’ Elsevier - Materials and Design
t = t + 1 and go to step 3 28 (2007) 2008–2014
The design criteria and optimum value of 6. C.C. Tsao and H. Hocheng (2007)
parameters calculated using MatLab 7.0 Genetic ‘Evaluation of Thrust Force and Surface
Algorithm Tool was given in the Table 6. Roughness in Drilling Composite Material
Using Taguchi Analysis and Neural Network’
Table 6 Design criteria and Optimum parameters Elsevier - Journal of Materials Processing
in Genetic Algorithm Technology,

Design Parameters value

72 RIDA