You are on page 1of 7

Materials Today: Proceedings xxx (xxxx) xxx

Contents lists available at ScienceDirect

Materials Today: Proceedings


journal homepage: www.elsevier.com/locate/matpr

Experimental investigation and optimization of surface grinding on


EN31AM steel using Taguchi and ANN
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri ⇑, Tarun Kumar Kotteda,
Rajesh Mudunuri
Department of Mechanical Engineering, SRKR Engineering College, Bhimavaram, Andhra Pradesh, India

a r t i c l e i n f o a b s t r a c t

Article history: Grinding is the machining processes; it is used to enhance the dimensional accuracy and surface quality
Received 3 January 2021 of workpiece. The surface grinding process is affected by the different process parameters such as grind-
Received in revised form 3 January 2021 ing wheel speed, material removal rate (MRR), number of passes, grinding wheel grain size, work piece
Accepted 17 January 2021
speed, material hardness and depth of cut respectively. The feed and speed are the important factors
Available online xxxx
because by increasing these factors will directly impacts on surface roughness (SR). Even though, the
SR is reduced by improving the MRR. Surface grinding is the major process in metals cutting applications;
Keywords:
it is widely used in finishing operations of cutting machining processes. SR and MRR are the significant
Surface grinding
EN31AM steel
output responses in the manufacturing relating to quality and quantity. In this article, an experimental
Taguchi investigation of surface grinding of EN31AM steel is studied to minimize the SR and maximize MRR
MRR for the optimization of grinding process parameters. In addition, artificial neural network (ANN) and
SR Taguchi also employed for optimizing the input process parameters such as depth of cut, feed, and speed.
ANN Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the Emerging Trends in
Materials Science, Technology and Engineering.

1. Introduction Finally, Power controlled grinding wheel controls the speed


mandatory speed of operation of grinding machine. The required
Grinding is important finishing process compared to the con- speed is obtained by the various dependencies such as manufac-
ventional processes utilized in production. Generally, the work turer’s rating and wheel’s diameter, respectively. If the grind head
piece surface is affected by using the grinding process and its resides in the fixed permanent position, then work piece will be
material is removed. Because, cylindrical wheel presented with moved and the grinding head will rotate around the permanent
the abrasive particles will be relatively moved on the surface. work piece, respectively. The surface grinder consisting of ‘‘head”,
The porous body is formed by the bonding of abrasive particles, it is lowered to the workpiece and it is moves around the grinding
so results in removal of material in work piece. Here, abrasive par- wheel based on back-and-forth conditions on a table. This table
ticle grits distribution, size and wheel abrasive structure plays the contains a controllable and enhanced permanent magnet for effec-
major role in improvement of grinding operation. Finally, the plane tively utilizing magnetic stock but can have other fixturing means
or cylindrical surface are using the application of grinding. In these or a vacuum chuck, respectively. The general surface grinders con-
applications, due to improper grinding the material removal rate sisting of the rotating grinding wheel and it cuts in horizontal axis
will be excessively increased. To solve this issue, various aspects throughout the area of entire grinding wheel. ‘‘Blanchard” manner
of the grinding process must be perfectly analyzed using the grinders are the major form of Rotary surface grinders. They con-
empirical and physical models. The proposed article considers tain the grinding head; it is used cuts the vertical axis throughout
the major input process parameters named as number of passes, the end face of grinding wheel. Mainly, the table rotates in opposite
depth of cut and work speed, respectively. The grinding machine direction around the workpiece underneath. But it removes the
consisting of bed with fixture to direct and it holds the work piece. bulk number of materials and also grinds smooth areas with
prominent curved grind marks. It is also utilized to flat shear
blades, parallel surfaces, fixture bases, and makes metal stamping
⇑ Corresponding author.
die sets with sharpening.
E-mail address: chekurirbr@gmail.com (R.B.R. Chekuri).

https://doi.org/10.1016/j.matpr.2021.01.495
2214-7853/Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the Emerging Trends in Materials Science, Technology and Engineering.

Please cite this article as: K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al., Experimental investigation and optimization of surface
grinding on EN31AM steel using Taguchi and ANN, Materials Today: Proceedings, https://doi.org/10.1016/j.matpr.2021.01.495
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al. Materials Today: Proceedings xxx (xxxx) xxx

Vishal Francis et. al. [1], authors proposed that if depth and feed  To analyze the effect of surface grinding process parameters like
of cut will change continuously by making the spindle speed as depth of cut conditions, table feed, work piece speed and grind-
constant. Thus, the effect on SR will be perfectly observed and ing wheel speed. The SR and MRR are enhanced with ANN opti-
the feed rate also perfectly obtained as important factors in the mization for EN31AM steel.
for-iron cast. Moreover, other factors were obtained as important  The grinding response is modeled by grinding variables; thus,
for stainless and mild steel, respectively. H. Adibi et. al. [2], authors the predictive quantitative relationships will generate in practi-
proposed that the by rising the depth of cut will increases the stur- cal and applicable manner. These models will be effectively
dily with respect to the amount of loading on surface of the wheel. used for the optimization purposed in the EN31AM steel surface
But it is not affected by changing the table speed. Prakash et. al. [3], grinding.
authors proposed that the SR is manipulated by work speedcon-  In addition, optimization of process parameters also done using
stantly in the AISI 1040. It is also influenced by the number of ANN. Finally, confirmation test results are performed using
passes, grinding wheel speed and depth of cut respectively. Shih obtained experimental values of MRR and SR with optimized
et. al. [4] proposed that theaverage grinding average chip thickness values computed using ANN approach.
will eventually be reduced by increasing in grinding wheel speed.
And hardness of the wheel is increased by the in-grinding wheel
2. Materials and methods
speed, thus it will result in the enhancement in MRRs. Hassui et.
al. [5], proposed that theworkpiece quality and vibration will be
Table 1 describes the chemical compositions of EN31AM steel
directly affected by the wear of a grinding wheel. Nathan et al
workpiece with the element type and the percentage of weights.
[6], stated that by estimating the accurate life of grinding wheel
The values of input process parameters with their levels are
is useful in the process of grinding. Becauseit is needed to be
demonstrated in Table 2. Fig. 1 discloses the experimental set up
redressing after the wheel life expires. In this cylindrical grinding
of surface grinding process. Following are the parameters utilized
machine, an A463–K5–V10 wheel is developed by the Hardened
for the experimental investigation (Fig. 2).
C60 steel (Rc 40) type of specimens. The results show that the by
Depth of cut –0.03 mm, 0.04 mm, 0.02 mm
occurring the grinding burn will directly affect theground compo-
Cutting speed – 1000 rpm, 700 rpm, 500 rpm
nentservice behavior and surface quality respectively.
Feed – 0.120 mm/min,0.095 mm/min, 0.075 mm/min,
Li Xue et al [7], stated that the reproducible SR was achieved by
the process monitoring of dressing, and the grinding wheel quality Work piece material – EN 31 AM Tool Steel
also increases continuously. Here, SR was monitored by the acous-
tic emission sensors. By using these sensors, the unwanted grind- 2.1. Taguchi parameter design for grinding process
ing wheel circumstances also detected effectively, thus the wheel
contour errors and mismatches in roundness wheel will be directly The machine quality characteristic features of turning are
detected. The results show that the faults occurred during the affected by the process parameters. Thus, in the proposed work
dressing process were identified by efficiently using the AE signal. considered process parameters are as follows: depth of cut (C), feed
And more prominent SR and grinding operation was achieved by rate (B) and cutting speed (A). Here, the selection of accurate
continuously monitoring the dressing parameters. J. Kundrak parameters is based on the interest and they are chosen by study-
et al [8], proposed that thecalculations of not direct decisive factor ing the literature reviews. Finally, the preliminary experiments
forexact power of grinding will cause assessment of rough capabil- were conducted.
ity of diamond. The rough analysis of diamond dressers was easy to
measure by calculating it. At lower speeds of wheel, the rough
dressing shocks will be effectively used and high rotating speeds 2.1.1. Selection of orthogonal array
the finishing dressing was performed. Table 2 presents the different input process parameters with
A. Sudiavso et al [9] proposed that the limitations presented in their range with the three levels. In this work, two – factor interac-
the traditional dressing approaches will be overcome by using the tion effects also studied by analyzing the process parameters, so
electrical dressing approaches. Because these approaches will turning based selected characteristics are extracted. The interac-
enhance the wheel abrasiveness along with durability of bonding tions are majorly considered among the cutting speed and depth
material gains. It also results in better surface finish by suitable of cut (AXC), speed and feed rate (AXB) andfeed rate and depth
maintaining the process conditions. A. Sudiavso et al [9] proposed of cut (BXC) respectively. The three-dimension dynamometer is
that electrolytic in process dressing (ELID) grinding is best suitable used for measuring the cutting forces and turned specimen with
for structural ceramics precision grinding process. In the ELID the help of randomization approach. Table 3 presents the detailed
method was mainly uses the metal bonded grinding wheel because outcomes of experimental output data of feed rate (mm/min),
this wheel is dressed in the electrolytical manner at the situation of depth of cut (mm),and spindle speed (rpm).The best machining
grinding process. Thus, the better abrasive wheels generate the quality characteristics are generated by ‘‘lowering and improving”
uninterrupted protrudent abrasive. By utilizing micro grit cast iron
fiber in ELID approach, the mirror finish surface grinding also effec-
Table 1
tively possible. This technology has wide applications in case of EN31 AM steel chemical compositions.
optical and semiconductor industry. Manojkumar Sinha et al.
Element % weight
[10] studied that the optimal dressing parameters are used to
grinding wheel grit size. And they studied about the dressing Iron (Fe) Rest
Manganese (Mn) 0.53
parameters selection during grinding process. The dressing depth
Molybdenum (Mo) 0.06
was effectively calculated as 60-grit size for alumina material- Nickel (Ni) 0.33
based grinding wheels. By increasing in the dressing lead will Phosphorus (P) 0.022
manipulates the specific grinding forces and SR significantly. Sulphur (S) 0.015
Finally, to overcome the drawbacks of existing approaches, the Carbon (C) 1.08
Chromium (Cr) 1.46
proposed method is contributed as follows.

2
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al. Materials Today: Proceedings xxx (xxxx) xxx

Table 2 Table 3
Input process parameters. Taguchi orthogonal array.

Process parameters Level 1 Level 2 Level 3 Job no. Depth of cut (mm) Feed rate (mm/min) Spindle speed (rpm)
Depth of cut (mm) 0.4 0.5 0.6 1 0.02 0.075 1000
Feed rate (mm/rev) 200 250 300 2 0.03 0.095 1000
Cutting speed(rpm) 600 1200 1800 3 0.04 0.120 1000
4 0.02 0.075 700
the radial and feed forces. This type of responses S/N ratiois given 5 0.03 0.095 700
6 0.04 0.120 700
in the equation (1) as follows: 7 0.02 0.075 500
  8 0.03 0.095 500
1 2  9 0.04 0.120 500
S=Nratio ¼ 10 log y1 þ y22 þ ::: þ y2n ð1Þ
n
Here y1 ; y2 ;    ; yn indicates the various responses, they are used
kDN/1000. Then feed rate was changed into machining feed using
to control the machining characteristicsin various levels for each
formula (F  N). Table 4 presents the various experimental outputs
parameter.
obtained from the experiments. Thevalues of cutting parameters
were adjusted in the surface grinding machine (Table 5).
3. Results and discussions The main effect S/N ratio plot presented in Fig. 3, mean values
(Table 6) are measured with respect to the feed rate, depth of cut
Initially the technical parameters were converted into machin- and cutting speed, respectively. The minimum SR values with opti-
ing parameters. Such as cutting speed in the mm/min was con- mal condition are obtained by observing the main effect plot is as
verted into the spindle speed in RPM by using the Formula V= follows:

Fig. 1. Experimental setup.

Fig. 2. Final work pieces.

3
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al. Materials Today: Proceedings xxx (xxxx) xxx

Table 4
Obtained experimental values of SRwith Taguchi orthogonal array.

Job no. SR (lm) Spindle speed (rpm) Depth of cut (mm) Feed rate (mm/min)
1 0.235 1000 0.02 0.075
2 0.248 1000 0.03 0.095
3 0.354 1000 0.04 0.120
4 0.312 700 0.02 0.075
5 0.325 700 0.03 0.095
6 0.297 700 0.04 0.120
7 0.212 500 0.02 0.075
8 0.225 500 0.03 0.095
9 0.230 500 0.04 0.120

Table 5
Obtained MRR values with experimentation of surface grinding on EN31AM steel.

Speed (Rpm) Depth of cut(mm) Feed (mm/min) Weight (after grinding) Weight (before grinding) Time (in sec) MRR(mm3/min)
1000 0.02 0.075 137.456 150.516 38 0.343684
1000 0.03 0.095 141.956 150.983 38 0.23755
1000 0.04 0.120 145.538 155.610 38 0.26505
700 0.02 0.075 151.279 153.706 60 0.0652
700 0.03 0.095 147.329 151.257 60 0.10336
700 0.04 0.120 145.630 150.992 60 0.141105
500 0.02 0.075 143.934 150.185 70 0.1645
500 0.03 0.095 148.598 150.870 70 0.05978
500 0.04 0.120 146.809 149.438 70 0.0691842

Fig. 3. Effect of surface grinding parameters for S/N ratio (SR).

Table 6 Depth of cut at (0.04 mm)


Experimental results vs ANN values. Cutting speed at (1000Rpm)
Optimization with Taguchi Optimized using ANN Feed rate at (0.12 mm/min)
MRR SR MRR SR
Fig. 4 shows the main effect Means plot shows respectively for-
cutting speed, depth of cut and feed rate. The minimum SR values
0.343684 0.235 0.375 0.241
0.23755 0.248 0.275 0.246
with optimal condition are obtained by observing the main effect
0.26505 0.354 0.260 0.351 plot is as follows:
0.0652 0.312 0.0612 0.0.314 Depth of cut at (0.04 mm)
0.10336 0.325 0.106 0.319 Feed rate at (0.12 mm/min)
0.141105 0.297 0.140 0.301
Cutting speed at (1000Rpm)
0.1645 0.212 0.165 0.210
0.05978 0.225 0.0591 0.228 MRR: The volume of material removed to the machining time is
0.0691842 0.230 0.0802 0.234 defined as the MRR.

4
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al. Materials Today: Proceedings xxx (xxxx) xxx

ðW b  W a Þ Depth of cut at (0.02 mm)


MRR ¼ ð2Þ
Tm Cutting speed at (1000Rpm)
Fig: 5 shows the main effect Means plot shows respectively for-
T m = machining times (min/sec).
feed rate, depth of cutand cutting speed. The maximum MRR val-
W a = weight of work piece material after grinding.
ues with optimal condition are obtained by observing the main
W b = weight of work piece material before grinding.
effect plot is as follows:
Fig. 5 shows the main effect S/N ratio plot shows respectively
Feed rate at (0.075 mm/min)
for feed rate, depth of cut and cutting speed. The maximum MRR
Depth of cut at (0.02 mm)
values with optimal condition are obtained by observing the main
Cutting speed at (1000Rpm)
effect plot is as follows: (Fig. 6)
Feed rate at (0.075 mm/min)

Fig. 4. Effect of grinding parameters for means (SR).

Fig. 5. Effect of grinding parameters for S/N ratio (MRR).

5
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al. Materials Today: Proceedings xxx (xxxx) xxx

Fig. 6. Effect of grinding parameters for means (MRR).

4. Optimization of SR and MRR USINGANN An done optimized output parameter is generated at a time
through one neuron of output layer by utilizing the only one hid-
4.1. Artificial neural network den layer, respectively.
Finally, purelin and tansig transfer activation functions are used
ANN is the artificial intelligence-based data processing system, in the output and hidden layers for effective optimization. Here,
which is functioned based on the biological neural networks with linear transfer function is tansig and hyperbolic tangent sigmoid
the dedicated operations based on performance characteristics. transfer function is purelin respectively. Following equations repre-
Even though, it is an advanced technique; due to mathematical sents the operation of each transfer function with its output:
operations, it needs to maintain the following assumptions for
increasing the efficiency of system. purelinðnÞ ¼ n ð3Þ

 Each neuron holds the important data, so it must be perfectly


processed. 2
 The connection links are used to transmit the signal between tansig ðnÞ ¼ 1 ð4Þ
1 þ e2n
various neurons with the weight.
For accurate optimization, Lavenberg–Marquardt algorithm is
 Here, the weight is multiplied (convolved) with the input signal
used to train the data along with the back-propagation operation.
and transmitted into neural net.
Some initial exhaustive trial attempts are done initially, and the
 Theoutput signal is generated by the help of nonlinear activa-
training algorithm will select the transfer functions after some
tion function applied oneach neuron with respect to each input.
time. Thus, the set target error occurrence is effectively reduced
during the training process with lesser number of epochs. Finally,
Multiple authors have studied the different types of neural net-
mean squared error (MSE) and average error percentage is used
works by using various types of layers for the Grinding process.
to calculate the proposed network performance with optimal num-
This work majorly focused on generation of surface grinding pro-
ber of hidden layers to that of corresponding neurons.
cess with best optimal parameters by utilizing the ANN optimizer.
For this purpose, systematic method of back-propagation based
multilayer ANN is used for the training of network with multilayer
feed forward operations for simultaneous testing. Here, the train- 5. Conclusion
ing operation performed in four stages as follows,
The proposed work majorly focuses on optimization of various
 Initially neuron weights are initialized. cutting parameters forEN31AM tool steelsurface grinding process.
 Generation of back propagation model between the input to Here, depth of cut, feed rate and cutting speed parameters are opti-
output layers to reduce the errors. mized by using the Taguchi optimization approach effectively. So,
 After error identification, the weights will be re-updated for these parameters are majorly used for turning of work piece
error correction. The squared errors occurred in the output are EN31AM tool steel. Finally by applying the Back propagated based
effectively reduced by using the gradient-descent property of ANN is used to generate the optimal values are as depth of cut are
the back-propagation procedure. 0.0.04 mm, 0.03 mm, and 0.02 mm, feed rate are 0.120 mm/min,
 Finally, user specific activation function based on biases is 0.095 mm/min and 0.075 mm/min, and cutting speed are
applied to generate the resultant output. 500 rpm, 700 rpm and1000rpm respectively. By using these opti-
mal values, resultant MRR and SR values are generated and vali-
In this work, majorly five input parameters are needed to opti- dated using Experimental work. The following conclusions are
mize, thus five input neurons are considered in the proposed ANN. obtained by analyzing taguchi and experimental results.
6
K. Siddartha Varma, U.R. Prasad Varma, Rama Bhadri Raju Chekuri et al. Materials Today: Proceedings xxx (xxxx) xxx

 To obtain the improved SR value, the optimal parameters com- References


bination is feed rate – 0.12 mm/min,depth of cut – 0.04 mm and
speed – 1000 rpm. [1] V. Francis, A. Khalkho, J. Tirkey, R.S. Tigga, N.A. Tirkey, Experimental
investigationandprediction of surface roughness in surface grinding
 To obtain the improved maximum MRR value, the optimal operation using factorial method and regression analysis, Int. J. Mechan. Eng.
parameters combination is feed rate – 0.12 mm/min,depth of Technol. 5 (5) (2014) 108–114.
cut – 0.04 mm and speed – 1000 rpm. [2] H. Adibi, S.M. Rezaei, A.A.D. Sarhan, Analytical modeling of grinding wheel
loading phenomena, Int. J. Adv. Manufact. Technol. 68 (2013) 473–485.
[3] D.B. Prakash, G. Krishnaiah, N.V.S. Shankar, Optimization of process
parameters using taguchi techniques when turning AISI 1040 steel with
coated tools, Int. J. Mechan. Eng. Technol. 7 (6) (2016) 114–122.
CRediT authorship contribution statement [4] A.J. Shih, M.B. Grant, T.M. Yunushonis, T.O. Morris, S.B.mcspadden, Vitreous
bond CBN wheel for high-speed grinding of Zirconia and M2 tool steel, Trans.
K. Siddartha Varma: Conceptualization, Data curation, Investi- NAMRI/SME 26 (1998) 195–200.
[5] A. Hassui, A.E. Diniz, Correlating surface roughness and vibration on plunge
gation, Methodology, Writing - original draft, Writing - review & cylindrical grinding of steel, Int. J. Mach. Tools Manuf 43 (2003) 855–862.
editing. U.R. Prasad Varma: Investigation, Methodology. Rama [6] R.D. Nathan, L. Vijayaraghavan, R. Krishnamurthy, In-process monitoring of
Bhadri Raju Chekuri: Investigation, Methodology. Tarun Kumar grinding burn in the cylindrical grinding of steel, J. Mater. Process. Technol. 91
(1999) 37–42.
Kotteda: Investigation, Methodology. Rajesh Mudunuri: Writing [7] L. Xue, F. Naghdy, C. Cook, Monitoring of wheel dressing operations for
- original draft, Writing - review & editing. precision grinding, IEEE International Conference on Industrial Technology,
Bankok, Thailand, 11-14 Dec. 2002, 2:1296-1299.
[8] J. Kundrak, V. Fadorovich, A.P. Markopoulos, Improvements of the dressing
process of super abrasive diamond grinding wheel, Manufact. Technol. 14
(2014) 545–554.
Declaration of Competing Interest [9] A. Sudiarso, J. Atkinson, In-process electricaldressing of metal-bonded
diamond grinding wheels, Eng. Lett. 16 (3) (2008) 308–310.
[10] M. K. Sinha, D. Sctti, S. Ghosh, P. V. Rao. An investigation intoselection of
The authors declare that they have no known competing finan- optimum dressing parameters based on grinding wheel grit size, AIMTDR
cial interests or personal relationships that could have appeared 2014, IIT Guwahati, Assam, India.
to influence the work reported in this paper.

You might also like