Al-Khwarizmi 

Engineering   
Journal 

Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54  (2014) 

 

 

Optimization of Wear Parameters in AISI 4340 Steel 
 

Abbas Khammas Hussein 
Department of Materials Engineering/ University of Technology  

Emial: abbas2000x@yahoo.com 

 
(Received 26 May 2014; accepted 22 September  2014) 

 
 

 

Abstract 

  
This study investigated the optimization of wear behavior of AISI 4340 steel based on the Taguchi method under 

various testing conditions. In this paper, a neural network and the Taguchi design method have been implemented for 

minimizing the wear rate in 4340 steel. A back-propagation neural network (BPNN) was developed to predict the wear 

rate. In the development of a predictive model, wear parameters like sliding speed, applying load and sliding distance 

were considered as the input model variables of the AISI 4340 steel. An analysis of variance (ANOVA) was used to 

determine the significant parameter affecting the wear rate. Finally, the Taguchi approach was applied to determine the 

optimum levels of wear parameters. The results show that using the optimal parameter setting (load3, sliding speed1, 

and sliding distance2) a lower wear rate is achieved. The error between the predicted and experimental values is only 

3.19%, so good agreement between the actual and predicted results is observed.  

 

Keywords: AISI 4340, wear, pin on disc, Taguchi method, neural network. 

 

1. Introduction 
 
AISI 4340 steel have been used for a great 

number of applications, like aircraft landing gear, 

power transmission gear, and in the machine 

constructions industry with hardness between 28 

and 38 Rockwell C (HRC). This kind of low 

alloyed steel is very cheap, compared with the 

expensive high alloyed tools steel, and it has an 

appropriated hardness combined with a very high 

toughness and tensile strength [1,2]. Moreover, 

the cutting tools used for hard turning are 

relatively costly as compared to grinding and 

hence there is a need to investigate the effect of 

wear parameters on tool life. It has also been 

reported that the properties and the composition of 

tool materials are critical to the behavior of 

machining forces which in turn affect the tool life. 

Therefore, it is necessary to study the influence of 

process parameters on tool wear in hard turning 

process. Several investigations have been carried 

out to study the tool wear and improve tool life as 

well as productivity. Luo et al studied wear 

behavior in turning of AISI 4340 hardened alloy 

steels using cubic boron nitride (CBN) and 

ceramic tools. Oliviera et al investigated the hard 

turning of AISI 4340 steel in continuous and 

interrupted cuts with whisker reinforced cutting 

tools [3, 4, 5, and 6].  

In the present work, two of the techniques, 

namely, the neural network (NN) with a back-

propagation network (BPN) and the Taguchi 

design method have been employed. MATLAB 

(2013) and MINITAB16 programs were used for 

the NN modelling and the Taguchi optimization 

technique, respectively. Sliding speed, applying 

load, and sliding distance were selected as the 

input factors, whereas wear rate was selected as 

the response. A BPN model was developed for the 

prediction of the wear rate. An analysis-of-

variance (ANOVA) table was used to determine 

the significant wear parameters affecting the wear 

rate. An approach to determine the optimum wear 

parameter setting was proposed on the basis of the 

Taguchi design method. 

 

 

 

 

 

mailto:abbas2000x@yahoo.com


 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
46 

 

2. Experimental Set-up and Test 
Procedure 

 
In the present investigation, a pin on disc wear 

tests were performed on AISI 4340 steel which 

has been selected as work piece material. Table 1 

shows the spectrochemical analysis (Measured 

Results) of this steel. All experiments were 

performed by using a pin on disc wear apparatus 

shown in Figure 1.  

 
 Table 1, 

 Chemical composition (wt%) of AISI 4340. 

Si Mo Ni Cr Cu P S Mn Al C 

0.2541 0.2892 1.6991 0.9112 0.0811 0.0221 0.0132 0.5913 0.0167 0.3534 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

Fig.1. Pin on disc wear apparatus. 

 
This machine facilitates study of wear 

characteristics in sliding contact under desired 

conditions. Sliding occurs between the stationary 

pin on a rotating disc. Normal load, sliding 

distance and sliding speed can be varied to suit the 

test conditions. The pin specimen was tested in 

pin on disc apparatus. To perform the test 

specimen was clamped in jaw. Pin weight losses 

were measured using an electronic balance having 

an accuracy of ±0.001 mg.  Weight losses were 

converted to volume losses by dividing the weight 

to the density of steel (7.85 g/cm
3
). Dry sliding 

wear tests were carried out using pin-on disk type 

wear tester at different parameters, where three 

levels of three parameters were selected as shown 

in Table 2. The tests were conducted at a constant 

time of 30 min. 

 
Table 2, 

Parameters with level values.  

Parameters Level 1 Level2 Level3 

Load (N) 20 25 30 

Sliding speed (r.p.m) 250 350 400 

Sliding distance (cm) 6 8 10 

 

 

In the present investigation an L9 orthogonal 

array was chosen as shown in Table 3. The 

experiment consists of 9 tests (each raw in the L9 
orthogonal array) and the columns were assigned 



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
47 

 

with parameters). The orthogonal array table in 

the Taguchi design method was applied to BPN as 

testing data. BPN was developed to predict the 

wear rate. The optimum wear-parameter 

combination was obtained by using an analysis of 

the signal-to-noise (S/N) ratio.  

 
Table 3, 

L9 orthogonal array and a neural-network training set.  

 

 

The signal-to-noise (S/N) ratio is a measure of 

the magnitude of the data set relative to the 

standard deviation. If the S/N is large, the 

magnitude of the signal is large relative to the 

noise , as measured with the standard deviation 

[7,8] . There are several S/N ratios available 

depending on the types of characteristics. The 

nominal ratio is the best, higher is better and 

lower is better [8] . We would select the S/N if the 

system is optimized when the response is as small 

as possible. The S/N ratio for the LB (lower is 

better) characteristic is calculated by using the 

following equations [9]: 

 

𝐿𝑗 =
1

𝑛
 𝑦𝑖

2                                                        …  1 

𝑛

𝑘=1

 

𝜂𝑗 = −10𝑙𝑜𝑔𝐿𝑗                                                    … (2) 

 
Where 𝑦𝑖   is the response value, 𝐿𝑗  is the loss 

function and 𝜂𝑗  is the S/N ratio.   

 

 

3. Statistical Analysis 
 

The results of the experiments were evaluated 

by the analysis of variance (ANOVA) . The main 

objective of the analysis was to determine the 

influence of every parameter on the variance of 

results, regarding the total variance of all the 

parameters. This is defined by the sum of squares. 

The calculation of ANOVA was made on the 

basis of the recommendations in [8]: 

 

𝑆𝑆 =   𝑦𝑖 − 𝑦  
2

𝑁

𝑖=1

=  𝑦𝑖
2 − 𝐶𝐹              … (3)   

𝑁

𝑖=1

 

𝐶𝐹 =
𝑇2

𝑁
                                                              … (4) 

 
To calculate the effect of an individual parameter 

on the variance a more useful equation is used: 

 

𝑆𝑆𝐴 =
𝐴1

2

𝑁𝐴1
+

𝐴2
2

𝑁𝐴2
+ ⋯ +

𝐴𝑛
2

𝑁𝐴𝑛
−

𝑇2

𝑁
              … (5) 

 

The following equations are also needed for the 

calculations: 

𝑀𝑆𝐴 =
𝑆𝑆𝐴
𝑓𝐴  

                                                          … (6) 

𝐹𝐴 =
𝑀𝑆

𝐸
                                                                … (7) 

𝑃𝐴 =
𝑆𝑆𝐴
𝑆𝑆

× 100                                                  … (8) 
 

Where: 

SS: sum of squares. 

𝑦𝑖 : Value of each results. 
CF: correction factor. 

T: the sum of all results. 

N: the total number of results. 

𝐴1 : The sum of results (𝑦𝑖 ) where parameter  

(𝐴1 ) is present . 
𝑁𝐴1 : number of experiments where parameter 

(𝐴1 ) is present . 

MSA: mean square where parameter (𝐴1 ) is 
present. 

𝑓𝐴: degree of freedom of factor A . 
𝐹𝐴: F ratio. 
𝑃𝐴: Percentage of contribution for factor A . 



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
48 

 

The degree of freedom are an important part of 

the statistical analysis because they provide us 

with additional information about the process. The 

degree of freedom for the Taguchi array is defined 

as follows [9]: 

DOF parameter = number of factor levels-1  

DOF experiment= number of results-1 

DOF error = number of all DOFs-number of 

DOFs of all parameters. 

 

 

4. Experimental Results and Data Analysis 
 
A neural network based on back propagation is 

a multilayered architecture made up of one or 

more hidden layers placed between the input and 

output layers [10] .  The components of the input 

pattern consisted of the control variables of 

applied load, sliding speed and sliding distance 

whereas the output pattern component represented 

the wear rate. Table 3 shows a Taguchi L9 

orthogonal-array plan of the experiment and a 

training set for the neural-network application. 

The orthogonal-array table used in the Taguchi 

design method was applied to BPN as testing data 

.The network structure was selected to be the 

3:4:1 type .The used BNP model is shown in 

Figure 2.  

 

 

 
 

Fig. 2. 3:4:1 (3 inputs, 1 hidden layer with 4 neurons and 1 output) type of the BPN algorithm used for modeling. 
 

 

The testing validity of the regression analysis 

and the neural network results was achieved by 

using the input parameters according to the design 

matrix given in Table 4 The performance of each 

BPN was calculated with the absolute error (%) of 

the tested subset using MATLAB ToolBox of 

neural network. The average absolute error (%) 

was calculated as 1.02 %. The wear rate can be 

predicted in a quick and accurate manner, the 

BPN results show that the predicted values were 

very close to the experimental values. The value 

of the multiple coefficients R
2
 is 0.9441, which 

means that the explanatory variables explain 

94.41% of the variability in the response variable. 

The predicted values of the wear rate were 

compared with experimental values as shown in 

Figure 3.  

 

 
Table 4, 

Test set for the validity of the constructed neural network. 

 

Sliding Distance (cm)Sliding Speed (r.p.m)Load (N)Exp. No.

PredictedExperimental

0.00410.004310250201

0.00390.00416250252

0.00440.004610400303

0.00430.00428350254

0.00410.00416400305

wear rate ( mm
3
/mm)



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
49 

 

 
 

Fig.3. Comparison of experimental and predicted values (BPN Model). 

 

 

The effect and optimization of wear settings 

for the minimum wear rate was investigated 

experimentally. The optimum wear-parameter 

combination was obtained by analyzing the S/N 

ratio . ANOVA was used to consider the effects of 

the input factors on the response and was 

performed on experimental data. The confidence 

level was selected as 0.95%. The results of 

ANOVA for the wear rate are shown in Table 5.  

The results of contribution percentage (%) are 

shown Table 6. 

 

 
Table 5, 

Results of ANOVA for the wear rate.  

 
 
Table 6, 

Results of contribution percentage (%) of wear parameters. 

 
 

 

It was observed that the applying load and 

sliding speed have a great influence on the 

obtained wear rate after analyzing Table 4 and 5. 

The sliding distance do not effect significantly the 

obtained wear rate . The plot of the main-factor 

effects is shown in Figure 4. The S/N graph for 

the wear rate is shown in Figure 5. It is evident 

that the applying load (66.67%) and sliding speed 

(30.56%) have the most significant effect on the 

wear rate. Sliding distance (2.78%) has little 

effect on the wear rate.   

Percentage contribution of each variable on wear rate

Variables Percentage Contribution(%)

Load 66.66666667

Speed 30.55555556

Distance 2.777777778

Total 100



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
50 

 

 
 

Fig.4. Main effect plot for wear rate vs. load, sliding speed and sliding distance. 

 

 

 
 

Fig.5. Main effect plot for SN ratios vs. load, sliding speed and sliding distance. 

                     
 

Optimum factor levels and S/N ratios obtained 

at the end of the Taguchi design technique are 

summarized in Table 7.  Based on the S/N ratio 

plot in Figure 5, the optimum wear parameters for 

the AISI 4340 Steel are applying load at level 3, 

sliding speed at level1 , and sliding distance at 

level 2  . The applying load and sliding speed are 

two parameters affecting the wear rate compared 

to sliding distance. 

 
Table 7, 

Optimum factor levels and their S/N ratios. 

 



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
51 

 

According to Figure 6 (a), higher applying 

load at higher  sliding speed   gives a greater  

wear rate compared to the combination effect of 

applying load and sliding distance as shown in  

Figure 6 (b) and the combination effect of sliding 

speed   and sliding distance as shown in  Figure 6 

(c) .  Table 8 shows a multiple linear regression 

equation developed for wear rate. This developed 

model gives the relationship between 

independent/ predictor variable and a response 

variable. Figure 7 shows comparison of predictive 

performance when using the developed model 

with experimental data. 
It can be see there is a better model fit. As 

these values are closely resembling the actual data 

with minimum error (R
2
=0.9049), design of 

experiments by Taguchi method was successful 

for calculating wear rate from the regression 

equation.  

 

 
           

Fig. 6. 3D surface plot for wear rate vs. load, sliding speed and sliding distance. 

 

 
Table 8, 

Regression equation of wear rate. 

 

 

 

 

 

 

 

 

 

 

 

 

(a) (b) (c)



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
52 

 

 
 

Fig.7. Comparison of experimental and predicted values (Regression Model). 

 

 

5. Confirmation Tests 
 

A confirmation experiment is the final step in 

first step iteration of designing an experiment 

process i.e., is the final step in Taguchi method 

and it is used to verify the optimal combination of 

the factor settings [9].  The purpose of the 

confirmation experiment is to validate the 

conclusions drawn during the analysis phase. The 

confirmation experiment is performed by 

conducting a test with a specific combinations of 

the factors and levels previously evaluated [9] . In 

this study, after determining the optimum 

conditions and predicting the response  (wear rate) 

under these conditions , a new experiment was 

designed and conducted with the optimum levels 

of the wear rate parameters . The final step is to 

predict and verify the improvement of the 

performance characteristic. The predicted value of 

wear rate (𝜂 ) at the optimum levels is calculated 
by using the relation given as [8]: 

𝜂 = 𝜂𝑚 +   𝜂 𝑖 − 𝜂𝑚  

𝑛

𝑖=1

                                    … (9) 

Where 𝜂𝑚  is the total mean of response, 𝜂 𝑖  is the 
mean of the response at the optimum level, and n 

is the number of the main wear parameters that 

that significantly affect the performance. The 

result of the experimental confirmation using the 

optimum wear rate parameters is shown in Table 

9. It can be seen that the results are consistent, i.e. 

a good agreement between the predicted and 

actual wear rate being observed. 

 
Table 9, 

Predicted values and confirmation test results for wear rate. 

  
 

 

6. Conclusions 
 

This study presents a prediction, optimization 

and modelling of the wear behavior of the AISI 

4340 steel based on the Taguchi-based neural 

network with the back-propagation algorithm 

method . The following conclusions can be drawn 

from this study: 

1. The main wear parameters that affect the wear 
rate of the AISI 4340 steel were determined as 

applying load (66.67%) and sliding speed 

(30.56%) among three controllable factors 

influencing the wear rate using ANOVA.  

                     Taguchi optimal parameters settings 

Level Prediction Experiment Error(%)

Load (N) 3

Sliding Speed (r.p.m) 1 0.003789 0.00391 3.193454737

Sliding distance (cm) 2



 Abbas Khammas Hussein                Al-Khwarizmi Engineering Journal, Vol. 10, No. 4, P.P. 45- 54 (2014) 

 

 
53 

 

2.  A neural network based on the back-
propagation network (BPN) algorithm was 

constructed for predicting the wear rate. The 

predicted values were found to be very close to 

the experimental values.  

3. The optimum parameter combination for the 
minimum wear rate was obtained by using the 

Taguchi design method with analysis of the 

S/N ratio. 

4. The confirmation test supports the finding that 
the wear rate is greatly decreased by using the 

optimum design parameters. From 

confirmation test, the error (%) associated with 

wear rate is 3.193454737% resulting in the 

conclusions that the design of experiments by 

the Taguchi method was successful for 

calculating wear rate from the regression 

equation. 

5. The obtained results indicate that the BPN 
model agreed with the Taguchi analysis.  

 

 

7. References 
 

[1] Gilberto Bearano, Julio Caicedo, Juan Munzo 
Saldana, " Mechanical and Tribological 

Properties Enhancement of Heat Treated 

AISI4340 Steel by using a TiN/TiAlN", 

Rev.Fac.Ing.Univ.Antioquia No.44.pp.36-

42.Jun.2008. 

[2] Jitendra . M.Varma, Chirag. P. Patel, " 
Parametric Optimization of Hard Turning 

AISI 4340 Steel by Solid Lubricant with 

Coated Carbide Insert", Vol3, Issue3, 

pp.1011-1015, May-Jun2013. 

[3] R.Suresh, S.Basavarajappa, V.N.Gaitonde, 
G.L.Samuel, "Machinability investigations 

on Hardened AISI 4340 Steel Using Coated 

Carbide Insert", int.journal of Refractory 

Metals and Hard Materials 33(2012) pp.75-

86. 

[4] Arun.S, T.Sivakumar, Viswanathan P. and 
R.Subramanian , "Study of Hardness and 

Wear Properties of Boronized AISI 4340 

Steel", Vol.3,Issue4,Jul-Aug 2013, pp.1927-

1929 . 

[5] Luo SY, Liao YS, Tsai YY, " Wear 
Characteristic in Turning High Hardness 

Alloy Steel by Ceramic and CBN Tools", 

Jou.Mater Process Technol , 1995, pp.1009-

1019. 

[6] Oliverira AJ, Diniz AE, Ursolino Dj. " Hard 
Turning in Continuous and Interrupted Cut 

with PCBN and Whisker Reinforced Cutting 

Tools",  Jou.Mater Process Technol , 2009, 

209:5262-70 . 

[7] Allen, Theodore T., "Introduction to 
Engineering Statistics and Six Sigma: 

statistical quality control and design of 

experiments and systems", Springer-Verlag 

London Limited 2006. 

[8] Peter Goos and Bradley Jones, " Optimal 
Design of Experiments: a case study 

approach " , Wiley & Sons, Ltd 2011 . 

[9] Douglas C.Montgomery , "Design and 
Analysis of Experiments", John Wily &Sonc, 

Inc., 2009 . 

[10] Daniel Graupe , " Priciples of Artificial 
Neural Networks ",(2nd Edition), World 

Scientific Publishing Co. Pte. Ltd.2007 . 

 

 

 

 

 

 

 

 

 

 

 

 

  



 (2014)45- 54  ، صفحة4، العذد10دجلة الخوارزمي الهنذسية المجلعباس خماس حسين                                                          م 

 

 
54 

 

 

 

 AISI4340تحذيذ العوامل المثلى للبلى في الفوالر نوع 
 

 عباس خماس حسين
 اٌحىٕىٌىخٍةاٌداِعة / لسُ هٕذسة اٌّىاد

abbas2000x@yahoo.com : االٌىحشؤً اٌبشٌذ  

 
 

 

 الخالصة

 
و . طشٌمة جاوىخً جحث ظشوف اإلخحباس اٌّخحٍفة  عٍى وفك  AISI 4340-جحضّٓ هزٖ اٌذساسة جحذٌذ اٌعىاًِ اٌّثٍى ٌسٍىن اٌبٍى فً اٌفىالر ٔىع

و جُ إعحّاد إسٍىب اإلٔحشاس اٌعىسً ٌٍشبىة .  AISI 4340فً هزا اٌبحث اٌشبىة اٌعصبىٍٔة و طشٌمة جاوىخً ٌخفض ِعذي اٌبٍى فً اٌفىالر  تأُسحخذَ

اٌحًّ اٌّسٍظ و  و أِا ِحغٍشات اإلدخاي أي ِعاِالت اٌبٍى اٌحً جُ إسحخذاِها ٌٍحصىي عٍى ّٔىرج اٌحٕبأ فهً سشعة اإلٔزالق. اٌعصبىٍٔة ٌٍحٕبأ بّعذي اٌبٍى 

 .و أخٍشاً، جُ جطبٍك طشٌمة جاوىخً ٌححذٌذ اٌّسحىٌات اٌّثٍى ٌعىاًِ اٌبٍى . وّا أسحخذَ جحًٍٍ اٌحبآٌ ٌححذٌذ جأثٍش اٌعىاًِ عٍى ِعذي اٌبٍى . ِسافة اإلٔزالق 

وّا أْ لٍّة اٌخطأ ِابٍٓ اٌمٍُ . ي اٌى خفض ِعذي اٌبٍى جؤد(  2، و ِسافة اإلٔزالق  1سشعة اإلٔزالق  و 3اٌحًّ )و أظهشت إٌحائح ، أْ اٌعىاًِ اٌّثٍى 

 .أي هٕاٌٍه جىافك خٍذ ِابٍٓ إٌحائح اٌحً جُ اٌحٕبأ بها و إٌحائح اٌحدشٌبٍة  %3.19اٌحً جُ اٌحٕبأ بها و اٌمٍُ اٌحدشٌبٍة هً فمظ 

       
 

 

 

mailto:abbas2000x@yahoo.comالبريد
mailto:abbas2000x@yahoo.comالبريد