INTRODUCTION
Cold forging can be described as a process in which a piece of metal
is shaped to the desired form by plastic deformation of a simple starting
form such as bar, billet, bloom or ingot at room temperature (Altan et
al., 2005). Initially forging operation has various advantages compared
to other metal manufacturing including little loss of material, improve
strength, geometrical precision of components and high production rates
(www.forging.org). There
are variety of processes that can be classified as forging such as open
die, impression die, ring rolling, warm forging and cold forging. Currently
in practice forging process design (mainly die) repeatedly carries out
trial and error based on skills and experience of the designer who is
familiar with the forging problem. Currently most of the companies still
rely on formulas, standards and experience to aid their die design. Several
machining processes can be used to fabricate the forging die. The behavior
of machined components under fatigue is highly influenced by the residual
stress and surface roughness (Gurgel et al., 2006). The quality
of the surface plays a very important role in the performance of milling
as a goodquality milled surface significantly improves fatigue strength,
corrosion resistance, or creep life. Therefore, the desired finish surface
is usually specified and the appropriate processes are selected to reach
the required quality (Lou et al., 1999). Özel and Karpat (2005)
utilized the regression and neural network to study the effect of turning
parameter on surface quality and tool wear, while recently Korkut and
Donertas (2007) study the effect of feed rate and cutting speed to the
cutting force, surface roughness and toolchip contact length.
The objective of this study is to determine the effect of two machining
parameter i.e. feed rate and cutting speed analytically and experimentally.
A sensitivity analysis program is then developed to measure level of sensitivity
of the surface roughness. The research begins with brief introduction
and then followed by research related to the die design process and recent
finding in cold forging process.

Fig. 1: 
General forging dies design process (Mynors et al.,
1997) 
After that the machining parameter is explained. Then the methodology
of the research is presented. The framework of sensitivity analysis program
is then outlined and followed by the verification sections. Before ends,
the results of the works are then discussed and concluded
DIE DESIGN PROCESS
In the forging process, performance of the die is measure based
on quality of the forged parts and reliability of the die itself. Due
to demands for complex part, long life die and accurate tooling recently
becomes crucial. The success of cold forging process depends on two main
criteria i.e. the selected tool material and dies design (Vazquez et
al., 2000). Since the process involves huge pressure, selection of
material which exhibits the both criteria i.e. high strength and tough
is very critical (Destefani, 1994). Commonly material that owns these
criteria is very expensive. Compared with the tooling material, the design
of die is claimed has bigger influences. This is because choosing in different
material may not extend the die life until the die is properly design.
In practice the die design process begins upon requirement and specifications
from customer and then followed by the tooling design and forging process
stage. The process ends upon customer satisfaction in terms of die performance
and auxiliary tools required, the general die design process as practice
by the industry is as shown in Fig. 1.
MACHINING PARAMETER
The effect of cutting parameters on the final surface roughness of machined
surface has brought great challenge for engineers and researchers. Some useful
techniques of prediction the surface roughness of a product before milling is
necessary in order to evaluate the fitness of machining parameters for keeping
a desired surface roughness and increasing product quality. It is well known
that the prediction technique should be accurate, reliable, lowcost and nondestructive.
Surface roughness analytical equation proposed here as in Eq.
1 will be useful to predict of the criterion variable finish surface roughness
of work piece depend on variables such as feed rate, spindle speed, or depth
of cut as following:
Where: 
r_{n} 
= 
Tool tip radius (mm). 
f 
= 
The feed (mm tooth^{1}). 
A simple theoretical model was proposed using the cutting speed and the
feed rate as major parameters to predict the sensitivity of parameters
on surface roughness of work piece. With the help of the experimental
analysis, surface roughness of machined surface is measured and compared
with predicted values using the theoretical method. In this case, depth
of cut has no or little influence on surface roughness and thus it can
be neglected. Since the controllable machining parameters are easily measurable
input variable, the proposed model is quite efficient in providing a convenient
guideline for the estimation of the surface quality in any given coldforging
operation. This method shows the sensitivity of each surface parameter
for each input variable. The Eq. 2 defined the function
of spindle speed, N (rpm) with respect to cutting speed, Vc (m min^{1}).
Where: 
D 
= 
Tool diameter (mm). 
The functional of feed rate, fr (mm min^{1}) given in Eq.
3 with respect to feed (mm tooth^{1}), allowing the determination
of parameters as:
Where: 
r_{n} 
= 
No. of teeth. 
Thus, from Eq. 3, it can be modified as:
So from Eq. 1 and 4, the functional
Ra can be rewritten as:

Fig. 2: 
Mitutoyo surftest SV400 measuring instrument 
MATERIALS AND METHODS
The case study used in this project is a universal joint, which is model
using SolidWork 2006. The model is then converted into CAM file to develop
machining code for fabrication purpose using CNC milling machine. Before
that a simple CNC program has been generated to run the CNC milling machine
on Aluminum plate at various machining parameter. The objective of the
experiment analysis is to investigate the affect of machining parameter
on the surface roughness. The measurement of surface roughness was done
using Mitutoyo Surftest SV400 Measuring Instrument as shown in Fig.
2. Before that, the machined surface must be examined using Mitutoyo
Measuring Flatness Testing Instrument to ensure the flatness of work piece
is achieved.
The next step is implementing a parameter sensitivity analysis method
to obtain the optimal machining conditions with respect to surface quality.
There are two main machining parameters (spindle speed and feed rate)
used in the program.
SENSITIVITY ANALYSIS
The framework of machining parameter sensitivity analysis proposed
is based on four criteria. The sensitivity program is built using C++
programming. The purpose of program is to help creating a new approach
for surface finish prediction in endmilling operations.
Criteria 1
Machining Parameter Sensitivity Surface Roughness Analysis
• 
Key in No. of flutes (η_{t}). 
• 
Key in tool tip radius (r_{n}). 
• 
Key in N value (N = spindle speed/rpm). 
• 
Then key in fr value (fr = feed rate, mm min^{1}). 
Formula used:
• 
Compare A and B. Program will choose the answer, which has higher
value. 
• 
Finally, if
has higher value, then display: feed rate has the highest influence
on surface roughness. 
• 
Otherwise, if
has higher value, then display: ‘Cutting speed has the highest influence
on surface roughness. 
Criteria 2
Analytical Calculation of Surface Roughness
Where: 
r_{n} 
= 
Tool tip radius (constant in this case) and f is the feed (mm tooth^{1}) 
Criteria 3
Judgment for Accuracy of Parameter Analysis
Percentage deviation is defined as:
Where: 
φ 
= 
Percentage deviation of single sample data 
Ra_{i}` 
= 
Actual Ra measured by a surface texture measuring machine →value
obtained from experiment 

= 
Predicted Ra generated by a analytical equation →equation from
choice 2 
Criteria 4
Average percentage deviation
Where: 

= 
Average percentage deviation of all samples and m represent the
size of sample data 
The result to date has claimed that the cutting tools are identical in
properties and the material of the work piece is similar. The result of
sensitivity analysis showed that feed rate was the most significant machining
parameter used to predict surface roughness.

Fig. 3: 
(a) Machined surface for experiment No. 1 and (b) machined
surface for experiment No. 2 
RESULT VERIFICATION
The main objective of this experiment is to verify the surface roughness
obtained from the developed program. If the comparative result on aluminum
is positive (deviation percentage is not more than 10%), that`s mean the
analytical method can be used to act as benchmark to predict surface roughness.
The experiment can be divided into two;
• 
The experiment 1: To determine which machining parameter
that gives the highest influence on surface roughness when all parameters
are varies. Figure 3a show the machined plate. 
• 
The experiment 2: To investigate the effect on the surface
roughness when one of the parameter is varies and the others are constant.
Figure 3b shows the machined plate. 
DISCUSSION
Experiment 1
The predicted result and experimental output indicated similar pattern
that the feed rate is the most important parameter to provide smooth surface
because the final answer of Ra for both methods has smallest value mean
the smoothest surface produced if smaller feed rate value is used. The
positive result of experiment shows that the Ra prediction equation can
be used to determine parameter sensitivity on surface roughness. It is
well recommended that smaller feed rate and higher cutting speed (spindle
speed) can help to produce higher quality of surface. Thus, these two
parameters are adjusted to the desired value and create two critical parameter
sets. First set result (n_{1}) contains the smallest cutting speed
but slowest feed rate value while third result (n_{3}) has the
largest value of cutting speed and highest feed rate. Second set (n_{2})
is a default value. This method is done to determine which one parameter
has the greatest significant influence on surface finish when both of
main parameters are set to desired value (Table 1, 2).
The average percentage deviation is less than 10% (= 9.963%), the result
of experiment analysis in Table 3 is considered to be
accepted.
Figure 4 plots the result of surface roughness (Ra) with
cutting speed (Vc) using fitted variation data as input while the plots of average
surface roughness (Ra) with the feed (fz) is shown in Fig. 5.
The continuous line represents experimental data of surface roughness while
the dotted line means computational surface roughness. From gradient slope of
both lines, there are small difference values between analytical result and
experimental data. That`s mean the percentage deviation is small and the computational
results showed by sensitivity program are accepted. The value of gradient line
represents the sensitivity rate value that depicted influence of machining parameter
(cutting speed and feed rate) to the surface roughness.
Table 1: 
Analytical result for aluminum plate 

Table 2: 
Experimental result for aluminum plate 

Table 3: 
Comparison between analytical and experimental result 

Table 4: 
Analytical result for aluminum and sensitivity analysis 


Fig. 4: 
Chart for average surface roughness, Ra (μm) by cutting
speed, Vc (m min^{1}) 
Table 5: 
Experimental result for aluminum 

Table 6: 
Analytical and experimental result for aluminum 

For example, the gradient value for experimental data in Fig.
4 is 0.128, which is much smaller value than gradient value in Fig.
5 is 6.667, which means that feed rate has much more influence on surface
roughness than spindle speed. The experimental result also show that feed rate
has the highest influence on surface roughness in end milling of aluminum platen
and then followed by cutting speed and depth of cut. The similar result can
be proved by theoretical calculation of surface roughness shown in sensitivity
analysis.
Experiment 2
Differ from experiment No. 1, The higher sensitivity value of each machining
parameter in term of surface roughness indicates that greater influence on surface
finish. From the sensitivity model obtained, feed rate is the most critical
parameter compared to cutting speed (Table 4, 5).
It shows that the smoothest surface was obtained by using parameter set 1, which
based on decreases in feed rate while maintaining the cutting speed value. Decline
in feed rate value also has a tendency of improving the surface finish and thus
reducing the R_{a} parameter. Positive R_{a} sensitivity value
means there is proportionality between R_{a} and feed rate. Since the
average percentage deviation is less than 10% i.e., 9.785% (Table
6), the experiment result once again is accepted and theoretical model can
be referred as benchmarking to provide a good indication of the influence of
machining parameters on surface roughness.
To prove the finding, the same experiment for different material is conducted.
Table 7 exhibited the resultant predicted surface roughness
for finishing cut of Cast Iron using HSS cutter tool in end milling. The
resultant Ra is predicted and the sensitivity value of each parameter
variable in term of Ra shows that the feed rate is still become the most
significant parameter compared to cutting speed.
Table 7: 
Analytical result for cast iron and sensitivity analysis 

The result indicates that by increasing cutting speed, the surface finish
can be improved. It is because the occurrence of BuiltUp Edge (BUE) when
machining multiphase materials at lower cutting speeds. It generates large
burr quantity on the machined surface, consequently deteriorating surface
finish. This phenomenon may be associated with increase in the cutting
forces and the consequent dynamic instability of the cutting process.
Hence, increase in the depth of cut increases surface roughness values.
CONCLUSIONS
The purpose of this study is to develop analytical based surface prediction
technique which can be more accurate, flexible, reliable and nondestructive
and then evaluate its prediction ability. The sensitivity analysis program
proposed in this study is a useful computational tool to help analysis
of the relationship between the cutting parameters and the surface roughness
of machined surfaces without embarking on laborious time consuming and
often expensive machining trials. The sensitivity results showed that
the feed rate is the most important cutting parameter for determining
the machined surface roughness, Ra when end milling aluminum platen. This
is followed by the cutting speed and the depth of cut. The sensitivity
result was verified by experimental analysis, which was showing result
that the percentage deviation between analytical method and experimental
is less than 10%. The same result also obtained for different material.
This application of this model can be used in predicting the machining
condition of die and mold design with some minor modification and upgrading.
ACKNOWLEDGMENT