Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 97 (2014) 365 – 371

12th GLOBAL CONGRESS ON MANUFACTURING AND MANAGEMENT, GCMM 2014

Optimization and Prediction of Parameters in Face Milling of

Al-6061 Using Taguchi and ANN Approach
M. S. Sukumar1*, P. Venkata Ramaiah2, A. Nagarjuna3
1,2&3
Department of Mechanical Engineering; Sri Venkateswara University College of Engineering; Tirupati – 517502; Andhra Pradesh; INDIA

Abstract

In this paper, Taguchi Method has been used to identify the optimal combination of influential factors in the milling process.
Milling experiment has been performed on Al 6061 material, according to Taguchi orthogonal array (L16) for various
combinations of controllable parameters viz. speed, feed and depth of cut. The surface roughness (R a) is measured and recorded
for each experimental run and analyzed using Taguchi S/N ratios and the optimum controllable parameter combination is
identified. An Artificial neural network (ANN) model has been developed and trained with full factorial design experimental data
and a combination of control parameters have been found from ANN for the surface roughness (Ra) value, obtained from
confirmation test, for the optimum control parameters which are obtained from Taguchi S/N ratios analysis. Taguchi method and
ANN found different sets of optimal combinations but the confirmation test revealed that both got almost same Ra values.
© 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2014 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of the Organizing Committee of GCMM 2014.
Selection and peer-review under responsibility of the Organizing Committee of GCMM 2014

Keywords: Controllable parameters; Al-6061; milling; Taguchi method; S/N ratio analysis; Artificial Neural Networks.

1. INTRODUCTION

Roughness is often a good predictor of the performance of a mechanical component since irregularities in
the surface may form nucleation sites for cracks or corrosion. Although roughness is usually undesirable, it is
difficult and expensive to control during manufacturing. Decreasing roughness of a surface will usually
exponentially increase its manufacturing costs. This often results in a trade-off between the manufacturing cost of a
component and its performance in application. Increasing the productivity and the quality of the machined parts
are the main challenges of metal-based industry. There has been increased interest in monitoring all aspects of
the machining process. Quality of machining can be judged by surface roughness. Higher the surface finish higher
will be the quality. Surface finish mainly depends on cutting speed, Depth of cut, Feed. Most of the operators use
trial and error method to find the appropriate cutting condition.
2
Corresponding author. Tel.:+91-9010222176
E-mail address: [email protected]

1877-7058 © 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of the Organizing Committee of GCMM 2014
doi:10.1016/j.proeng.2014.12.260
366 M.S. Sukumar et al. / Procedia Engineering 97 (2014) 365 – 371

It is not the effective way to find out optimal cutting parameters. So the main objective of the study is to find the
optimum parameters (speed, feed, depth of cut) so that surface roughness is optimized. Aluminium has much
application in industries. Also automotive aircraft and train companies need to replace steel and cast iron with
lighter metal like aluminum. So, it is important to know the machining behaviour of aluminium. There are various
optimization techniques like Genetic Algorithm, Artificial Neural Network, Grey Analysis, Utility Concept,
Response Surface Methods, Taguchi technique, Fuzzy Logic, etc. to find out optimum cutting conditions.

2. LITERATURE REVIEW

Şeref Aykut [1] developed an ANN model to predict the surface roughness of Castamide material after
machining process. In the study, experiments on Castamide were done in CNC milling using high speed steel and
hard metal carbide tools. Dr. Mike S. Lou et al. [2] developed a multi-regression model that can predict the surface
roughness on the surface of the specimen (Al-6061) on which end milling operation has been carried out using a
CNC machine. Surasit Rawangwong et al. [3] investigated the effects of cutting parameters on the surface
roughness in semi-solid AA 7075 face milling. The results of the research could be applied in the manufacture of
automotive components. Mathew A. Kuttolamadom et al. [4] studied the effects of machining feed on surface
roughness in milling Al-6061. A controlled milling experiment on 6061 aluminium depicted the relationship
between feed and surface quality. Yang Yang et al. [5] proposed a method based on gene expression programming
(GEP) to construct the prediction model of surface roughness. GEP combines the advantages of the genetic
algorithm (GA) and genetic programming (GP). Bharat chandra routara et al. [6] studied a multi-objective
optimization problem by applying utility concept coupled with Taguchi method through a case study in CNC end
milling of UNS C34000 medium leaded brass. The study aimed at evaluating the best process environment which
could simultaneously satisfy multiple requirements of surface quality. B. Vijaya Krishna Teja et al. [7] conducted an
experimental study on performance characteristics of AISI 304 stainless steel during CNC milling process. The
work represents multi-objective optimization of milling process parameters using Grey-Taguchi method in
machining of AISI 304 stainless steel. Sanjit Moshat et al. [8] studied the highlights of optimization of CNC end
milling process parameters to provide good surface finish as well as high material removal rate (MRR).

In this paper, Taguchi method is adopted experimentally to investigate surface roughness influenced by the
control parameters such as speed, feed and depth of cut. And also it presents an ANN approach for prediction of
control parameters for surface roughness in face milling.

3. CONTROL PARAMETERS AND THEIR LEVELS

The parameters which influence the surface roughness of machined surface called control parameters such as
speed, feed and depth of cut. In this work, three controllable parameters are considered and each parameter is set at
four levels. The parameters and its levels are shown in Table 1.
Table 1: control parameters and their levels
Control parameters
Levels
Speed (RPM) Feed (mm/rev) DOC (mm)
1 900 125 0.1
2 1120 160 0.15
3 1400 200 0.2
4 1800 250 0.25

4. EXPERIMENTAL DESIGN AND MILLING OF WORK MATERIAL

In this work, Taguchi L16 design is used for conducting milling experiments (see Fig.1) on Al6061 work material by
considering different speed, feed and depth of cut combinations and the values of surface roughness are measured
using Talysurf surface tester (Fig 2) recorded in Table 2. And also experiments are conducted for full factorial
design (Table.6) to train the developed ANN.
 M.S. Sukumar et al. / Procedia Engineering 97 (2014) 365 – 371 367

Fig.1Milling Operation Fig. 2 Talysurf surface tester

Table 2: Experimental design L16 and data
Exp. run Speed (rpm) Feed (mm/rev) DOC (mm) Surface roughness (μm)
1 900 125 0.1 0.695
2 900 160 0.15 0.82
3 900 200 0.2 1.04
4 900 250 0.25 1.13
5 1120 125 0.15 1.205
6 1120 160 0.1 0.7
7 1120 200 0.25 0.33
8 1120 250 0.2 0.535
9 1400 125 0.2 0.745
10 1400 160 0.25 0.62
11 1400 200 0.1 0.53
12 1400 250 0.15 0.485
13 1800 125 0.25 0.59
14 1800 160 0.2 0.605
15 1800 200 0.15 0.355
16 1800 250 0.1 0.115

5. OPTIMIZATION OF MACHINING PARAMETERS USING TAGUCHI S/N RATIO ANALYSIS

The Experimental data of surface finish (Ra) is analyzed using Taguchi design in Minitab software and signal to
noise (S/N) ratio values are determined. The optimum levels of influential parameters are determined based on the
obtained S/N ratios.
Main Effects Plot for SN ratios
Data Means
Speed Feed

8
6
4
Mean of SN ratios

2
0
900 1120 1400 1800 125 160 200 250
DOC

8
6
4
2
0
0.10 0.15 0.20 0.25

Signal-to-noise: Smaller is better

Figure 3: Main Effects Plot for S/N ratios

368 M.S. Sukumar et al. / Procedia Engineering 97 (2014) 365 – 371

After determining the S/N ratio values (Table 3), the effect of each Machining parameter is separated based
on S/N ratio at different levels and the values of S/N ratio for each level of the controllable parameters and the effect
of parameter on response (Ra) in rank wise are summarized in Table-4. Basically, large S/N ratio means it is close to
good quality, thus, a higher value of the S/N ratio is desirable. From the Table-3 and Fig.3 the cutting parameters
with the best level are spindle speed at level-4, feed at level-4 and DOC at level-1. The optimal levels for the
controllable parameters obtained from this methodology are verified by the conformation test, the surface roughness
(Ra) value obtained for these optimum control parameters is 0.115 μm and as shown in Table-5.
Table 3: S/N ratios
Exp. run Surface roughness (Ra, μm) S/N for Ra
1 0.695 3.160
2 0.82 1.723
3 1.04 -0.340
4 1.13 -1.061
5 1.205 -1.619
6 0.7 3.098
7 0.33 9.629
8 0.535 5.432
9 0.745 2.556
10 0.62 4.152
11 0.53 5.514
12 0.485 6.285
13 0.59 4.582
14 0.605 4.364
15 0.355 8.995
16 0.115 18.786

Table 4: S/N ratio for each level of control parameters

S/N Response table for Ra

Levels Speed Feed DOC
1 0.870 2.170 7.639
2 4.135 3.101 3.846
3 4.627 5.949 3.003
4 9.182 7.360 4.186
Delta 8.311 5.190 4.636
Rank 1 2 3

Table 5: optimum control parameters values for S/N ratio analysis

Speed (rpm) 1800

Feed (mm/ min) 250
Depth of Cut (mm) 0.10

6. PREDICTION OF VALUES USING ANN

A three-layer feed-forward network with sigmoid hidden neurons and linear output neurons can fit multi-
dimensional mapping problems arbitrarily well, given consistent data and enough neurons in its hidden layer. The
network is trained with Levenberg-Marquardt back propagation algorithm (trainlm). The structure of artificial neural
network with input and output parameters is shown in Fig. 4(a). Cutting speed, Feed rate and depth of cut are taken
as the input parameters whereas surface roughness (Ra) is taken as output parameter.
 M.S. Sukumar et al. / Procedia Engineering 97 (2014) 365 – 371 369

Figure 4(a): ANN model

Figure 4(b): ANN Training

As shown in Fig.4, The ANN has three layers, input layer, hidden layer and output layer. The hidden layer
consists of 10 neurons and the output layer consists of only one neuron. The experimental values of full factorial
design (Table 6) are used for training (ref. Fig.4) the network. A well trained ANN is well generalized which
gives proper output for those input also which has never been encountered with the network while training.
Training a network is nothing but to set optimum weights of the links of two neurons. These weights,
activation function, number of layers and neurons in a layer decide how well nonlinearity can be defined. The
performance plot and regression plots are shown in Figure 5 and 6 respectively. The best performance is obtained at
epoch 6.

Figure 5: Performance plot

370 M.S. Sukumar et al. / Procedia Engineering 97 (2014) 365 – 371

Figure 6: Regression plot

The confirmation test results for the optimal parametric combination obtained in Taguchi S/N ratio analysis is
given as input to the trained ANN, and the network predicted the control parameter values as speed: 1800 rpm, feed:
250 mm/rev and depth of cut: 0.5 mm. Again confirmation test has been conducted for this parameter combination.
The surface roughness (Ra) value obtained for these optimum control parameters is 0.113 μm (see Table.7).

Table 6: Experiments for Full factorial design

Run Speed Feed DOC Ra
1 900 125 0.1 0.695
2 900 125 0.15 0.708
3 900 125 0.2 0.852
4 900 125 0.25 1.042
- - - - -
- - - - -
- - - - -
- - - - -
60 1800 200 0.25 0.098
61 1800 250 0.1 0.115
62 1800 250 0.15
63 1800 250 0.2
64 1800 250 0.25 0.155

Table 7: Confirmation test results

Combination of controllable parameters Ra (μm)
Speed (rpm) Feed(mm/rev) DOC(mm) Predicted value Exp. Value
ANN 1800 250 0.2 --- 0.113
Taguchi 1800 250 0.1 N.A 0.115

7. CONCLUSION

Among the consider parameters, speed has the most influence on the surface finish of the work-piece. The
trained ANN is able to predict the Ra values with reasonable accuracy. Taguchi S/N ratio analysis and ANN are
useful to find the optimum combination of parameters for getting a good surface finish.
 M.S. Sukumar et al. / Procedia Engineering 97 (2014) 365 – 371 371

REFERENCES

[1] Şeref Aykut, “Surface Roughness Prediction in Machining Castamide Material Using ANN” (2011), Acta
Polytechnica Hungarica, Vol. 8, No. 2, PP 21-32.
[2] Dr. Mike S. Lou, Dr. Joseph C. Chen & Dr. Caleb M. Li, (1999), “Surface roughness prediction technique
for CNC End-milling”, Journal of industrial technology, Vol-15.
[3] Surasit Rawangwong, Jaknarin Chatthong, Romadorn Burapa, and Worapong Boonchouytan, “An
investigation of optimum cutting conditions in face milling semi-solid AA7075 using carbide tool” (2012),
International journal of innovation, Vol-3, No. 6, PP 692-696.
[4] Mathew A. Kuttolamadom, Sina Hamzehlouia, M. Laine Mears, “Effect of machining feed on surface
roughness in cutting 6061 aluminum” (2010), International center for automotive research, Clemson
university, 2010-01-0218 .
[5] Yang Yang, Xinyu Li, Ping Jiang and Liping Zhang, “Prediction of surface roughness in end milling with
gene expression programming”, Proceedings of the 41st International Conference on Computers &
Industrial Engineering, PP 441-446.
[6] Bharat chandra routara, Saumya darsan mohanty, saurav datta, Asish bandyopadhyay and siba sankar
mahapatra, “Optimization in CNC end milling of UNS C34000 medium leaded brass with multiple surface
roughness characteristics”, Sadhana, Indian Academy of Sciences, Vol. 35, Part 5, October 2010, pp. 619–
629.
[7] B. Vijaya Krishna Teja, N. Naresh and K. Rajasekhar, “Multiple response optimization of Milling
parameters on AISI 304 Stainless steel using grey-taguchi method” (2013), International Journal of
Engineering Research and Technology, Vol-2, issue 8, PP 2335-2341.
[8] Sanjit Moshat, Saurav Datta, Asish Bandyopadhyay and Pradip Kumar Pal, “Optimization of CNC end
milling process parameters using PCA-based taguchi method” (2010), International Journal of Engineering,
Science and Technology, Vol-2, No-1, PP 92-102.