You are on page 1of 15

MULTIOBJECTIVE OPTIMIZATION OF CUTTING PARAMETERS FOR

DIMENSIONAL, SURFACE QUALITY AND MATERIAL REMOVAL RATE IN HARD


TURNING USING RSM
Ahmed A. Elsadek*, Ahmed M. Gaafer and Sayed A. Zayan
Mechanical Engineering Department
Shoubra Faculty of Engineering
Benha University
Cairo, Egypt
ahmed.alsadek@feng.bu.edu.eg

Abstract — Surface roughness, and roundness error two key performance parameters for
evaluating the quality of a product in turning processes. Thus, the aim of this work is to examine
the consequence of cutting conditions, namely the cutting speed, the feed rate and the depth of
cut as well as the workpiece hardness on the above-mentioned performance parameters during
turning operation of AISI H13 tool steel. Response surface methodology (RSM) is employed for
experimental-design and for correlating the relationship between different cutting parameters and
the various performance parameters. The efficiency of the proposed models was adequated
through analysis of variance (ANOVA). Desirability optimization techniquie, is then applied to
find the optimal combination of process parameters with an objective of minimizing the values
of both surface roughness and roundness error using the employed range of cutting parameters
values. The optimal solutions were confirmed experimentally.
I. INTRODUCTION

Tool Steel H13 is widely used in hot work and cold work tooling applications. cores, and
cavities for die casting dies, die casting shot sleeves, hot forging dies, extrusion dies, and plastic
mold cavities and components that require high toughness and excellent polishability. Thus, it is
important to optimize the factors that affect the whole quality of this widely used material.
Turning operation is among the most important metal cutting operations utilized in producing
round and cylindrical steel components. The quality of the surface finish and dimensional
tolerance are two major important factors that must be controlled in turning operation. In
turning, cutting parameters such as the cutting depth, feed-rates, cutting-speeds as well as
utilizing of coolants will exert an influence on the machining qualities such as the surface
roughness and the roundness-error of circular works and dimensional deviances of the product
[1]. Surface-roughness of machining operations was studied intensely, mostly through
experiments. M. Rahman and V. Naranayan [2] tried to optimize the roundness error to achieve
the maximum possible metal removal rate with the wanted precision. Yang and Tarng [3]
utilized Taguchi method in investigating the machining performance of S45C steel employing a
tool from tungsten carbide. The optimum values of the cutting speed, feed rate and the depth of
cut in turning linked to performance indices such as surface roughness and tool life were
measured. The outcome of cutting speed, feed and depth of cut along with the cutting time on
turning of metal-matrix-composites was investigated by Davim [4]. Manna and Bhattacharyya
[5] utilized models of multiple-linear regressions in correlating the roughness in form of R a & Rt
to the diverse machining factors in turning of Aluminum-Silicon metal matrix composites. Aslan
et al. [6] improved parameters of machining operation in turning of AISI 4140 steel of hardness
63 HRC using a ceramic tool with a titanium carbon nitride as well as aluminum oxide using
orthogonal array and analysis of variance. Soft computing is a way to deal with figuring which
parallels the amazing capacity of the human personality to reason and learn in a domain of
vulnerability and imprecision [7]. Rangwala and Dornfeld [8] predicted the machining
performance through turning employing a feedforward neural network in which they utilized a
set of input arrays. Natarajan et al. [9] optimizied a neural network utilizing PSO for tool life
prediction, which diminished the time of training by half. Li et al. [10] employed neuro-fuzzy
techniques in estimating force due to feed by measureming the current motor using a sensor
introduced to a computer numerical controlled lathe. Palanisamy et al. [11] optimized cutting
speed, depth of cut, the machining time and feed rate employing genetic algorithm optimization
technique in milling operation. Saravanan and Janakiraman [12] optimized turning operation
using genetic agorithm to diminish the machining time by optimizing the magnititudes of both
the machining speed and feed rate. R V Rao and P J Pawar [13] optimized the limits of wire
electrical discharge cutting employing the aritifical bee colony algorithm (ABC) the target work
was boosting the speed of machining operation with improving the surface roughness as the
constrained function. Zheng and Ponnambalam [14] utilized PSO to advance the multi-pass
turning Process the target work was limiting the of unit generation cost. Ali R. Yildiz [15]
utilized cuckoo search (CS) algorithm for improving the machining parameters in milling and
contrasted the outcomes and the aftereffects of other evolutionary algorithms such as PSO and
GA algorithm. The outcomes got showed that the CS is a compelling and hearty methodology
for the improvement of machining optimization issues. Girish Kant et al. [16] combined the
artificial neural network (ANN) with genetic algorithm (GA) to for predicting and optimizing the
surface roughness. The model lead to minimizing the surface roughness value to 0.099 μm at
optimized parameters of 4.650 m/sec of cutting speed, 0.1420 mm/tooth of feed, 0.670 mm depth
of cut and 0.080 mm of flank-wear. Kuldip Singh Sangwan and Girish Kant [17] combined
response surface methodology (RSM) and genetic algorithm (GA) for forecasting the energy
consumption along with the corresponding cutting parameters on turning of AISI 1045 steel. Rao
et al. [18] employed a novel algorithm called Jaya algorithm for optimizing the machining
parameters in wire electrical discharge machining (WEDM) and electrochemical machining as
well the obtained results showed a good performance for Jaya algorithm in comparison with
other algorithms. M. Mia et al. [19] optimized flank wear, cutting force and surface roughness
whicle turning AISI 1060 steel by combining Taguchi and Grey relation. Pardeep Kumar et al.
[20] employed RSM in optimizing and predicting radial and main cutting forces as well as
surface roughness when turning of AISI H13 steel. As it clear from literature, optimizing the
cutting processes no matter what their type is i.e. traditional or non-traditional, is an important
issue as it affects the whole manufacturing process from different aspects. Thus, the target of the
current work is to investigate the usage of a RSM expremintal design as well as desirability
optimization to solve the machining optimization problem . The study utilized tool Steel H13 due
to its wide usage in industry as stated before and the optimization parameters are surface
roughness as a measure of surface quality and roundness error as a measure of dimensional
quality. The study will provide a reliable investigation on optimizing the machining process for
tool steel in industry, based only on the available and/or required cutting ranges of the available
machine tools. Moreover, it provides a full preview on employing such technique in different
manufacturing optimization problems.

II. Materials and Methods


Round bars of diameter were employed as workpiece. A conventional lathe with center distance of 1.5 meter, Spindle
speed range is 45 up to 2445, Feed range is 0.08 up to 0.38 mm/rev.

Cemented carbide tool inserts with specs shown in Table 3. are utilized in experimentation.

Table 3. Specs of Cemented carbide inserts

Rake Angle Cutting edge angle Clearance angle Nose Radius


α (º) β (º) γ (º) r (mm)
5 80 7 0.4
Surftest SJ-310 is employed for measuring surface roughness parameter (Ra). A Talyround (250) with a computer
based relating table is used for measuring the roundness error, its specifications are max. measuring diameter = 300
mm, max. measuring high = 300 mm, work table speed = 6 r.p.m c.w and work table diameter = 19 mm. The
accuracy is Concentric (0.04 +0.0003H) µm, Eccentric (0.04 +0.0008H) µm, where H: is the high above work table
in mm.

III. Experimental Design


The experimental design is performed for realizing the effect of different parameter levels and a
response as well as the interactions of the several factors. The experimental design for
investigating the consequence of different turning operation parameters (e.g. cutting-speed, feed-
rate and depth of cut) on the machining characteristics (e.g. surface roughness and roundness
error), was done. In the current study, the investigates were designed using response surface
method employing central composite-second-order consists of 2y-factorial, where “y” represents
the integer of variables, to expand the trustworthiness of outcomes and to reduce the size of
experimentation without loss of accuracy. Therefore, the numeral of experiments (E) can be
found from the next equations;
E = yc + ya + y…, (1)
yc = 2k…, (2)
ya = 2k…, (3)
Where yc is the digit of corner-points while ya is the digit of axial points of the unit cube that is
constitute of a central composite design. The factors and factor levels are summarized in Table 4.
Table 4. The factors and factor levels
Factors Levels
Cutting Speed (m/min) 100 125 150
Feed Rate (mm/rev) 0.05 0.10 0.15
Depth of Cut (mm) 0.05 0.09 0.13
Hardness (HRC) 45 50 55

IV. Response Surface Methodology


Response surface methodology (RSM) is statistical technique utilized for investigating the
connections between a few informative factors and the required response factors. RSM thought is
to utilize a lot of planned tests to enhance the response. In this investigation, response surface
methodology (RSM) is utilized for deciding the connection between the distinctive turning process
parameters, for example, speed, feed rate, cutting depth and workpiece hardness with the diverse
machining criteria and investigating the impact of these parameters on the responses, for example
the surface roughness and the roundness error. To investigate the impacts of the turning
parameters on the recently expressed machining criteria, response surface 2 nd order polynomial
mathematical models can be created. In the general case, the response surface can be represented
by the below equation:

z z z
Ru = βo + ∑ βixi + ∑ βiixi2 +∑ βijxixj + ….. (4)
i=1 i=1 i,j

Where Ru is the matching responses e.g. surface roughness and roundness error, while x(1,2,…z)
are coded levels of z computable method variables, the terms βo, βi, βii and βij are the 2nd order
regression coefficients. For the present study, Eq. 4 can be written in the below structure:
Ru=βo +β1A +β2B +β3C + β4D + β11x12 +β22x22 +β33x32 + β44x42… (5)
Where A, B, C and D are cutting speed, feed rate, depth of cut and workpiece hardness
respectively.

A. Mathematical Modelling for Surface Roughness

The effects of the different cutting parameters on the surface roughness have been evaluated by computing the
values of the different constants based on of Eq. (5) using a Design Expert computer software and utilizing the
relevant data from Table 5. The mathematical relation correlating surface roughness and the considered cutting
parameters was obtained as follows:

Surface Roughness = + 2.29587 - 0.016361* Cutting Speed + 5.94470 * Feed - 1.75990 * Depth of Cut -0.034286 *
Cutting Speed * Feed + (4.76190E-004) * Cutting Speed * Depth of Cut +1.66667 * Feed * Depth of Cut +
(5.75411E-005) * Cutting Speed2 - 1.61095 * Feed2 + 2.80631 * Depth of Cut2 (6)

B. Mathematical Modelling for Roundness

The effects of the different cutting parameters on the roundness have been evaluated by computing the values of the
different constants based on of Eq. (5) using a Design Expert computer software and utilizing the relevant data from
Table 5. The mathematical relation correlating roundness and the considered cutting parameters was obtained as
follows:

Roundness = + 4.12174 - 0.045474 * Cutting Speed + 4.62907* Feed + 0.47070* Depth of Cut + 0.015238* Cutting
Speed * Feed + (9.76190E-003) * Cutting Speed * Depth of Cut - 6.55556* Feed * Depth of Cut + (9.62224E-005)
* Cutting Speed2 - 0.72950* Feed2 + 0.19011* Depth of Cut2 (7)
Table 5. Central composite design: different controlling parameters and results

Ru Speed Feed Depth of Cut Hardness Roughness Roundness


n m/min mm/rev mm HRC Ra O
1 100.00 0.10 0.09 50.00 2.58 12.36
2 150.00 0.10 0.09 50.00 2.65 13.73
3 125.00 0.10 0.09 50.00 3.058 18.4
4 125.00 0.10 0.13 50.00 3.185 12.17
5 150.00 0.05 0.05 45.00 2.23 10.64
6 125.00 0.10 0.09 50.00 2.51 17.72
7 125.00 0.05 0.09 50.00 2.63 12.47
8 125.00 0.10 0.09 50.00 1.905 12.67
9 150.00 0.15 0.05 55.00 2.879 26.84
10 150.00 0.15 0.13 45.00 6.308 37.4
11 125.00 0.10 0.05 50.00 2.703 12.21
12 125.00 0.10 0.09 50.00 3.068 27.48
13 100.00 0.05 0.05 55.00 1.229 7.31
14 125.00 0.10 0.09 55.00 2.98 17.69
15 150.00 0.15 0.05 45.00 5.323 27.88
16 125.00 0.10 0.09 45.00 4.227 15.62
17 150.00 0.05 0.13 55.00 3.081 14.05
18 150.00 0.05 0.05 55.00 1.501 11.21
19 125.00 0.10 0.09 50.00 3.369 15.93
20 125.00 0.15 0.09 50.00 5.218 26.26
21 150.00 0.05 0.13 45.00 2.5 23.49
22 100.00 0.15 0.05 45.00 5.693 18.95
23 100.00 0.05 0.13 55.00 1.82 15.13
24 100.00 0.15 0.05 55.00 4.609 26.76
25 125.00 0.10 0.09 50.00 3.187 13.42
26 100.00 0.05 0.05 45.00 2.316 8.23
27 150.00 0.15 0.13 55.00 5.384 29.81
28 100.00 0.15 0.13 45.00 4.928 26.36
29 100.00 0.15 0.13 55.00 5.987 22.67
30 100.00 0.05 0.13 45.00 2.245 12.12

V. Analysis of Variance (ANOVA)

The efficiency of the three models stated in the previous section are tested through the analysis
of variance (ANOVA). The results appear in tables 6-8. Variance can be defined as the mean of
squared deviations around the mean or the sum of the squared deviations around the mean value
divided by degrees of freedom. Testing the efficiency of a model is usually done by computing
the F-ratio of the lack of fit to the pure error and compare it with the standard value. If the
standard values exceed the value of the F-ratio then the proposed model is adequate. As it clear
from the results of analysis, the lack of fit F-value is not significant which proves the adequacy
of the mathematical models and demonstrate the different effects of the different machining
parameters on the roughness, roundness and material removal rate.

RESULTS AND DISCUSSION

Table 6, shows the analysis of variance “ANOVA” for surface roughness response for the linear
model. The model has got an F-Value of 21.87 indicates that the model is significant, the casual
that the obtained large F-Value due to noise is only value 0.01%. Lack of Fit F-value of 1.91
suggests the Lack of Fit isn't huge with respect to the pure error. There is a 24.51% causal that a
Lack of Fit F-Value this huge could happen because of noise.
Factor 1 Factor Factor 3 Factor 4 Resp Response
2 onse 1
2
Std Run A:Speed B:Feed C:Depth D:Hardness Roun Roughness
of Cut dnes
s
m/min mm/rev mm HRC
17 1 100.00 0.10 0.09 50.00 12.36 12.06 0 12.36 2.98
18 2 150.00 0.10 0.09 50.00 14.73 13.09 13.73 13.73 2.65
27 3 125.00 0.10 0.09 50.00 13.4 12.68 18.4 18.4 3.058
22 4 125.00 0.10 0.13 50.00 15.97 14.17 12.17 12.17 4.25
2 5 150.00 0.05 0.05 45.00 13.32 11.91 10.64 8.64 2.05
28 6 125.00 0.10 0.09 50.00 12.72 12.95 17.72 17.72 2.63
19 7 125.00 0.05 0.09 50.00 12.87 15.4 12.47 23.47 2.14
29 8 125.00 0.10 0.09 50.00 12.98 15.1 12.67 6.67 2.905
12 9 150.00 0.15 0.05 55.00 14.56 26.84 26.84 26.84 3.879
8 10 150.00 0.15 0.13 45.00 20.4 20.5 37.4 57.4 5.508
21 11 125.00 0.10 0.05 50.00 12.21 12.21 12.21 12.21 2.568
25 12 125.00 0.10 0.09 50.00 13.48 17.48 27.48 17.48 2.658
9 13 100.00 0.05 0.05 55.00 11.76 13.31 7.31 13.31 1.929
24 14 125.00 0.10 0.09 55.00 12.69 17.69 17.69 17.69 2.298
4 15 150.00 0.15 0.05 45.00 16.03 27.88 27.88 27.88 4.323
23 16 125.00 0.10 0.09 45.00 17.62 15.62 15.62 15.62 3.227
14 17 150.00 0.05 0.13 55.00 13.05 14.05 14.05 14.05 1.781
10 18 150.00 0.05 0.05 55.00 12.21 11.21 11.21 11.21 1.901
26 19 125.00 0.10 0.09 50.00 13.93 15.93 15.93 15.93 3.369
20 20 125.00 0.15 0.09 50.00 16.26 26.26 26.26 26.26 4.078
6 21 150.00 0.05 0.13 45.00 12.49 13.49 23.49 13.49 2.267
3 22 100.00 0.15 0.05 45.00 16.95 18.5 18.95 36.95 5.193
13 23 100.00 0.05 0.13 55.00 12.89 15.13 15.13 15.13 3.82
11 24 100.00 0.15 0.05 55.00 14.76 17.9 26.76 26.76 4.573
30 25 125.00 0.10 0.09 50.00 13.42 13.42 13.42 13.42 3.187
1 26 100.00 0.05 0.05 45.00 10.97 11.6 8.23 8.23 3.316
16 27 150.00 0.15 0.13 55.00 17.09 22.81 29.81 29.81 5.004
7 28 100.00 0.15 0.13 45.00 18.96 19.67 26.36 32.36 5.640
15 29 100.00 0.15 0.13 55.00 19.67 16.9 22.67 22.67 3.928
5 30 100.00 0.05 0.13 45.00 13.8 22.12 12.12 22.12 2.987

Materials and Methods


Round bars of diameter 35 mm and 100 mm length of AISI H13 tool steel supplied by Deutsche Edelstahlwerke
GmbH of chemical composition shown in Table 1. and primary hardness of 207 HB were employed as workpiece

Table 1. Chemical Structure of AISI H13 tool steel

element C Mn Cr Mo Si P S Va
0.39
% 0.48 4.88 1.250 1.09 0.012 0.002 0.920
0

To achieve the required hardness values for the specimens which are 45, 50 and 55 HRC the below heat
treatment procedure was followed (heat treatment was done at Hot cycle company):

 Stress Relief:
This treatment (stress removal) was performed for all pieces at a temperature of 600 ° C for four hours
and then slow cooling in the furnace (the pieces were left in the furnace after extinguishing until
completely cooled and then removed)
 Hardening:
The pieces were heated to a temperature of 1050 ° C in phases as follows:
1. Pre-heating for 500 ° C for 2 hours.
2. Second Pre-heating for 850 ° C for half an hour.
3. Heat up to 1050 ° C for 15 minutes and then quench in oil.

After the complete cooling of the pieces were washed from oil and salts and then measure the hardness
of pieces to ensure the validity of the process of hardening and hardness were measured utilizing
Rockwell Hardness Tester (Make: INNOVATEST Europe BV Model: NEXUS 610) shown in Fig. 1.
 The Tempering Review
Due to the difference in the required hardness, the temperature was reviewed according to the required
hardness as shown in Table 2:

Group Hardness HRC 45 50 55

Tempering Temperature ° C 600 580 450

The pieces were kept in a uniform temperature of two hours. All pieces were removed from the ovens and
suspended in the air until they were completely cooled and then washed and measured to ensure that the
desired salad was achieved.

A ceramic insert of Al2O3 coated with TiCN which possess superb wear resistance and basically utilized
for finish cutting of hardened steel (CNGA 120408 E040, Rhombic shape with angle 80o , Clearance
angle 0 ) was used so as to boost production at low cost by avoiding expensive PCD and CBN
o

cutting inserts. The Shank Tool holder was PCLNR 2525M 12 with negative rake angle of 6◦.
Taking in consideration the upcoming prospects and the concerns of the environment, dry cutting
operations were done [23 nano 2 paper]. The machining tests were performed on a conventional
lathe with center distance of 1.5 meter, Spindle speed is up to 1400 r.p.m (Make: SJR Machinery,
Tradename: SMAC ,Origin: China, Model: L6241, Power: 4kW) shown in Fig. 1.
The values of average surface roughness were obtained using a portable surface roughness tester
“Surftest SJ-310”, Make: MITUTOYO shown in Fig.

The measurements of roundness error was done using Talyrond 73 Make: Taylor Hobson, S/no.
112/2802-0123, L.R = 0.01 m
A four level central composite design, cutting speed (A), Feed rate (B) Depth of cut (C) and
Hardness (D) as the input independent entities variables while the Surface roughness (Ra) and
Roundness error (O) as the response ones were chosen. Continuous machining was done on the
composite samples, wherein the workpiece is subjected to multi sided compression. As a result
of this shear stress is developed and keeps on rising till it crosses the shear strength of the
workpiece and leads to shear deformation in the region of maximum shear stress.

The details of the variables with their levels in Table 1 and the details of the experimental
plan with observations in Table 2 are illustrated.
The experimental plans were developed for the establishment of linear models for cutting
surface roughness and roundness. Statistical tests performed, were for finding out the
significance of model, significance of individual variables and ascertaining the ‘Lack of
Fit’. These are important steps before moving further to predict the relationships.
Analysis of variance (ANOVA) is applied for summarizing and summing up the tests
performed in this study.

A model ratio greater than 10 usually indicates that a


transformation is required; a value of less than 3.0 would mean that transformations are
not significantly required. This model for cutting force has a maximum to minimum ratio
of 2.43478 for cutting force and a maximum to minimum ratio of 1.62637 for surface
roughness, which means that any transformation is meaningless.

The next task is to ascertain whether the models developed pass the lack of fit tests. Table 3
shows the lack
of fit test for cutting force. The linear model developed for cutting force has an F-Value
of 4.33, which means that there is 5.62% chance that this large “lack of fit value” could
occur due to noise. Lack of fit is Insignificant which is good for the model, as we desire it
to fit. The “Adequate Precision” measures signal/noise ratio. Ideally this value should be
above 4. A ratio of 12.465 generated by the model indicates an ample signal. Thus,
these values as discussed indicate that the generated model is good enough to navigate
the design space.
Similarly, analyzing “lack of fit value” i.e. F-Value for surface roughness in Table-4, we
observe an F-Value of 4.19, which means that there is a 6.01% chance that this large
“Lack of Fit” value could occur. This model also has an insignificant lack of fit, which is
good. Adequate precision ratio for this model comes out to be 14.009, which signifies
adequate signal. The model is good enough for navigating the design space.
This linear model generated by the design can be represented by the following equation:
n
Y = ao + ∑ ai xi + z
i=1

Where Y is the response a , a ,.. a are invariables, ‘x’ is the varying factor and ‘z’ is the unwanted
o 1 i

signal or error. The purpose of considering such a model comprises of three distinct aims, which
are: predicting relationships between variables so as to predict response, determining
significance of the factors involved and determining the optimum settings of the variables
with an aim to optimize response(s) [24].

You might also like