Optimizing the design of brake disc for all-terrain vehicle using the Taguchi and

multi-criteria decision-making method AHP-TOPSIS

Arrown Dalsaniya1, Dhananj Modi1, Kishan Fuse1 Maulik Kanakhara2

2
1
Department of Industrial Engineering, Department of Mechanical Engineering
Pandit Deendayal Petroleum University, Pandit Deendayal Petroleum University,
Gandhinagar, India Gandhinagar, India
Email:[email protected], Email: [email protected]
[email protected],
[email protected]

Abstract- The need for transportation vehicle has become A. Designing of Brake Disc
dominant in the recent few years. The braking system in the
vehicle is concerned with the safety of the people travelling in The modelling of brake disc is done in Siemens-NX
it. Thus, to make a robust design of the brake disc serves as software keeping various parameters in mind. The initial
prime importance in the vehicle. Braking system consists of modelling was done keeping the connected parts i.e. the
various components like brake disc, clutch sleeve, brake wheel hub and tire, which limited the internal and external
callipers etc out of which the study is focused on the design radii. Then, the internal trusses were developed to reduce the
and selection of brake disc. There are numerous input weight. The thickness of the disc was kept as per the
parameters like material, loading condition, design models of distance between the 2 pads of the brake calliper. The
brake disc are studied each at three levels. The experimental
parameters varied for the 3 designs (As shown in Fig. 1) are
design was made by using the Taguchi’s L9 orthogonal array
to find the best optimum level of the process parameters so the contact area of the brake pads and disc, no. of points
that they are least affected by noise factors for obtaining a connecting wheel hub and brake disc and internal trusses.
robust design of parameters. Further, the analysis of variance The less the contact area, less is the weight and more cooling
was carried out to check the significance of the model. AHP [4]. While, for a larger contact area the force acting on the
and TOPSIS method are used to find the best combination of brake disc is increased which in turn would generate more
input parameters. torque and effectiveness [5]. The trusses developed were
based on the number of points of contact between wheel hub
Index terms- Braking system, Taguchi orthogonal array, and brake disc while the number contact points selected as
AHP, TOPSIS, ANOVA
per the hub design from the all-terrain vehicle. Also, the
I. INTRODUCTION pedal ratio, master cylinder and brake calliper were as per
the requirement of ATV calculation made for the braking
The evolution and need of automobile are facing a system.
drastic change recently. To match the change in growth of
evolution, the braking system must be robust as it is the B. Structural Simulation of Brake Disc
integral part of safety for passengers in an automobile. The The structural simulation carried out in Siemens- NX
efficiency of the braking system depends on the several and the solver used was NASTRAN. The face of brake disc
components like brake pedal, master cylinder, brake was divided as per the brake pads of the calliper used in
calliper, brake disc, brake pads etc. with each of them Suzuki GS-150R. The material of the disc is selected
depending on numerous factors. In the present scenario considering correct range of chemical composition,
various research works are being conducted on
hardness, mechanical properties especially tensile strength,
optimization of these factors. C.L.L [1] used Taguchi
approach coupled with grey relational analysis for multiple and other properties necessary for the intended use. Thus, 3
performance characteristics akin tool life, cutting force, different materials were selected for this purpose i.e. SS
and surface roughness during turning operations. AISI 304 annealed, AISI 1005 and AISI 4340. The mesh
Seyedmohammadi et. al. [2] developed a model using size was kept constant for all the cases i.e. 4mm. The loading
analytic hierarchy process and GIS techniques to estimate conditions were varied based on the range of pedal forces
the suitability of land in agriculture. Ananthakumar et. al. applied by the driver as mentioned in the ARAI rules i.e.
[3] employed the Technique for order of preference by 300-600 N.
similarity to ideal solution (TOPSIS) to calculate the
optimal cutting conditions in plasma arc cutting using
Monel 400 TM alloy. In the present study, braking system
for an ATV is considered and the customized design of
brake disc is optimized using multi criteria decision making
methods namely Analytical hierarchical process (AHP) for
weight calculation and TOPSIS for ranking of alternatives.
II. EXPERIMENTAL FRAMEWORK Figure 1. Three different designs of the disk
 The selected values of pedal force are 300, 450 & 600N. nine experiments with specific levels and combination of
This gave us 3 values of the force applied by the brake pads parameters as shown in the Table 1. In the L9 orthogonal
on the brake disc. The output parameters selected were the array it is assumed that there is no interaction between any
machining time, stress developed (Von-Mises Stress) and of the two factors. Furthermore, the sequence of the
weight of the disc. Von mises stress is considered as the experiments was randomized in order to prevent any kind of
criteria as the materials are ductile and von mises stress subjective or personal bias.
theory yields the best result for ductile materials [6]. The
stress values is given by the software while the weight of the Table 1. Taguchi L9 OA
disc is calculated such that the volume of the disc as per Material Loading
Alternatives Design of disc
software and density as per ASTM standard. The machining (AISI) condition (N)
process to be performed for the manufacturing of the brake 1 1 304 300
2 1 1005 450
disc was selected as the laser cut machining. The SIL-F- 3 1 4340 600
PRO series laser cut machine having class 4 laser protection 4 2 304 450
covers, laser powers up to 10Kwatt, acceleration 2.5G, and 5 2 1005 600
cutting speed up to 40 meters/min was considered while 6 2 4340 300
conducting the experiments. 7 3 304 600
8 3 1005 300
III. METHODOLOGY 9 3 4340 450

A. Taguchi Orthogonal Array C. Analytic Hierarchy Process

Genichi Taguchi – a Japanese quality engineer, widely The Analytic hierarchy process (AHP) was developed
recognized as a quality engineer constructed a special set by Saaty [10] is a analysis technique for solving complex,
of general guidelines for the screening of process unstructured and multi-criteria decision problems. The main
parameters so that high quality can be achieved without objective of the process is to break down the complex
increasing the total numbers of experiments and cost decision problems into the hierarchical structure.
expenditure of the experiment. This technique is used to Considering the typical hierarchy, the overall goal is
determine the effect of different process parameters such as situated at the top of the structure, criterions at the levels and
temperature, pressure, and time by comparing the mean and sub-levels of the hierarchical structure, and the decision
variance of the process performance [7]. The ideal process variables at the lowest level of the structure. This
parameters obtained from the Taguchi technique are not methodology is successful to solve the decision-making
biased towards the variation of various noise factors and problems where the decision alternatives can be correlated
environmental conditions. Since most of the industrial with one another by establishing mathematical functions
experiments comprise of vital number of elements, full [11]. With the help of pair wise comparison matrix, AHP
factorial design results in the large number of experiments consolidates all the experts assessments, into the final
[8]. In order to decrease the number of experiments to decision, without altering their utility functions [12]. The
practical number, only a small fraction of all the outcomes linguistic terms of fundamental scale of 1-9 is utilized to
is selected with minimum loss of information. Here, compute the intensity of preference between the two
Taguchi orthogonal array technique is employed because elements. The basic steps involved in AHP are [13].
unlike the factorial design, where all possible combinations
are taken into consideration for performing the test, Step 1: The pair wise comparison matrix is formulated
Taguchi reduces the number of trials to a minimum number according to the criteria.
keeping intact the minimum possible loss of information by Step 2: The pairwise comparison matrix is then converted
testing pairs of combinations and determines the best level to the normalized matrix by dividing each elements of the
factor combination [7]. This method employs special column by its column summation.
design of arrays called orthogonal arrays. These orthogonal Step 3: The criteria weights are then obtained by the
arrays arrange the parameters in a way such that the matrix summation of each elements in each row.
obtained is a balanced matrix. This results in saving of both Step 4: The weighted sum value is obtained by the
time and resources. summation of each row elements which are the product of
the criteria weights (obtained from step 3) and the original
B. Selection of an Orthogonal Array values of each cells (values of cells before normalization of
In order to define an orthogonal array, one must identify the matrix).
various experimental conditions viz. the number of factors Step 5: The eigen value vector is obtained by the ratio of
to be studied, level of each factor, special difficulties that the weighted sum value to the criteria weights of each
might be encountered while performing the experiments [9]. criteria. Then the principle eigen value vector (λmax) is
In the present study, the three process parameters considered obtained by the average of the sum of the eigen value
are type of material, loading condition and brake disc obtained for every criterion.
design. Each of these parameters was selected at three Step 6: Calculate the consistency index by using [1]
different levels each, as mentioned in Table 1. Three
parameters, at three levels each, would give twenty - seven 𝐶. 𝐼 =
(&'() *+)
1
experiments (33) in full factorial design. However, using (+*-)
Taguchi’s L9 orthogonal array, we only need to perform
Step 7: The consistency ratio is obtained by the ratio of the n 2
consistency index to the random index number for the same
size matrix. The random index number is chosen from the
D = -
j å (vi =1
ij
-
- v ) , j = 1,..., J
i
8
standard value table based on the number of responses
Step 5: Determine the relative closeness to the ideal
considered in study. With the help [2] consistency ratio is
solution. It can be calculated by using [9]
obtained.
0.1
𝐶. 𝑅 = 2.1 2 +
D -j
C = j
9
The pair wise comparison matrix is only acceptable if the D +j + D -j
value of consistency ratio is less than 0.1. Where j = 1,...,J
D. TOPSIS Step 6: Rank the preference order from the highest value
The TOPSIS method was revised by Chen and Hwang to the lowest.
(1992) which was initially developed by Hwang and Yoon IV. RESULTS AND DISCUSSIONS
(1981) [14]. TOPSIS is a productive method to solve
problems existing multi criteria decision-making with finite The result of the responses obtained using orthogonal
alternatives [15]. The basic objective of TOPSIS method is array designed by Taguchi are normalised using [3] and are
to determine the optimum alternative, by ranking the shown in Table 2. Analysis of variance (ANOVA) is a
distance of each alternative. This distance is calculated from method of comparing the means by which the significant
the positive ideal solution and negative ideal solution. The terms of the model can be determined. Thus, the validation
highest value obtained for the alternative indicated that for checking the model adequacy can be done using
either the alternative is nearest to the theoretical positive ANOVA. The stepwise ANOVA was carried out for each
ideal solution or the farthest from the positive ideal solution. output response to check the significance of each term in
Similarly, it also accounts the negative ideal solution in the the model. The volume of the designed brake disc was
same way for ranking the alternatives [16]. In the process of obtained which was then multiplied with the density of the
TOPSIS, the performance ratings and the weights of the material used. The weights of the design is measured in
criteria are assigned crisp values obtained from the field grams. As mentioned above, the machining time obtained
experts. The TOPSIS procedure consist of the following by laser cut machining process is measured in seconds.
steps [17]. With help of the Nastran solver of NX software the values
Step 1: Convert decision matrix into normalized decision of stress are obtained in MPa.
matrix. The normalized value of rij is calculated as
Table 2. Normalised decision matrix
fij
rij = 3 Alternative Weight Machining time Stress
J

åf
j =1
2
ij
1
2
0.381
0.371
0.295
0.324
0.172
0.232
3 0.374 0.365 0.331
Where j=1,...,J and i=1,...,n 4 0.337 0.321 0.267
Compute the weighted normalized decision matrix. The 5 0.328 0.346 0.567
weighted normalized value is calculated as: 6 0.330 0.382 0.290
7 0.292 0.297 0.507
vij = wij rij 4 8 0.284 0.330 0.165
where j = 1,...,J ; i = 1,...,n 9 0.286 0.330 0.214
Step 2: Determine the positive ideal solution and negative
The Table 3 shows the ANOVA result conducted for
ideal solution.
the weight response. The major contribution factor that
A+ = {v1+ ,..., v +n }={( max vij | i Î I ¢ ), ( min vij |i Î affects the output response is the design of the brake disc.
j j
The experiments were carried out using the 95%
I ¢¢ )} 5 confidence interval which indicates that there is only 5%
A- = {v1- ,..., v -n }={( min vij | i Î I ¢ ), ( max vij |i Î chances that the Model F value could occur due to noise.
j j
The model summary value of analysis of variance for
I ¢¢ )} 6 weight for R2 adjusted and R2 predicted are 99.98% and
Where I’ is associated to the benefit criteria and I” is 99.94% respectively. The R2 predicted value suggests that
associated with the cost criteria. there is only 0.06% chances that the noise could affect the
Step 3: Compute the separation measures by using n- response data. The graph obtained from the ANOVA is
dimension Euclidean distance. The distance between each shown in Fig. 2. In the same way the ANOVA shown in
alternative and positive ideal solution is calculated using Table 4. is conducted for the machining time. The values
[7] for adjusted R2 and predicted R2 are 81.90% and 54.19%
n 2 respectively. The adjusted R2 value suggests that there is an
D =+
j å (v
i =1
ij - v ) , j = 1,..., J
+
i
7 only 19.64% chance that the noise could affect the response
data. The graph for machining time is as shown in Figure
Step 4: The distance between each and every alternative 3. Similarly, the analysis of variance was conducted for the
and negative ideal solution is calculated using [8] stress output. The Table 5 shows the ANOVA result condu-
 Table 3. Stepwise analysis of variance for weight
Sum of
Source DOF Contribution Adj SS Adj MS F-Value P-Value
squares
Design 2 5054.29 99.01% 5054.29 2527.15 17038.85 0.00
Material 2 50.12 0.98% 50.12 25.06 168.97 0.00
Error 4 0.59 0.01% 0.59 0.15
Total 8 5105.01 100.00%

cted for the stress response. The model summary value adjusted value suggests that there are only 19.64%
of analysis of variance for stress for R2 adjusted and R2 chances that the noise could affect the response data. The
predicted are 80.66% and 51.06% respectively. The R2 graph of ANOVA for stress is shown in Figure 4.
Table 4. Stepwise analysis of variance for machining time
Sum of
Source DOF Contribution Adj SS Adj MS F-Value P-Value
squares
Design 2 133.78 22.56% 133.78 66.89 4.99 0.082
Material 2 405.55 68.39% 405.55 202.77 15.12 0.014
Error 4 53.66 9.05% 53.66 13.41
Total 8 592.98 100.00%

Table 5. Stepwise analysis of variance for stress

Sum of
Source DOF Contribution Adj SS Adj MS F-Value P-Value
Squares
Loading
1 274887 62.02% 25164 25164 2.35 0.200
Conditions
Design 2 69812 15.75% 69812 34906 3.26 0.145
Loading
Conditions*
1 55673 12.56% 55673 55673 5.20 0.085
Loading
Conditions
Error 4 42852 9.67% 42852 10713
Total 8 443225 100.00%

Figure 2. ANOVA for weights. Figure 3. ANOVA for machining time

After checking the adequacy of the model, the AHP and shown in Table 8. The output normalised data from Table
TOPSIS was used for picking of the best combination of 3 is then converted to the weighted normalised decision
the input parameters on the brake disc. The pairwise matrix using [4] as shown in Table 10. The positive ideal
comparison matrix was formulated by asking the experts in and negative ideal solution was calculated using the [5]
the field and the linguistic variables were converted to the and [6] respectively. Considering all the responses as cost
crisp values using the Table 6. The criteria weights are criteria, the positive ideal values obtained are 0.021,0.084.
obtained by normalizing the column followed by the 0.106 for weight, machining time, stress respectively.
summation of each cells along the row. The pairwise Similarly, the negative ideal solution values obtained are
comparison matrix is shown in the Table 7. Further, the 0.028, 0.108, 0.362 for weight, machining time, stress
principle eigen value is calculated and is shown in Table 9. respectively. The Euclidean separation measure for the
The consistency ratio is calculated using the [2] and is positive ideal and negative ideal solution can be obtained
 coefficient index can be calculated using the [9]. The
values are shown in Table 10. Arranging the values in the
descending order and the first obtained highest value is the
first rank optimized output for the combination of input
parameters. The Taguchi’s S/N ratio was calculated by
using the values of closeness coefficient index and is
shown in Table 10. The graph is shown in Fig. 5.

V. CONCLUSION
The present study emphasised more upon the details
understanding the dependency of the selected parameters
like weights, machining time on the design of the brake
disc. The effective results as the combination of the input
Figure 4. ANOVA for stress parameters are obtained as follows:
• Taguchi orthogonal array technique was found
Table 6. Scale of relative importance
effective in conducting the experiments with a
Crisp values Linguistic variable minimal loss of resources and information
9 Perfect • Analysis of Variance confirmed that the model is
8 Absolute significant and from the result it can be inferred
that material is one of the major affecting factor in
7 Very good
the response recorded.
6 Fairly good
• The analytical hierarchy process resulted into
5 Good deciding the weights of the output response and
4 Preferable thus resulting into importance of each response.
3 Not bad • TOPSIS method ranked the alternatives obtained
from the response recorded while conducting the
2 Weak advantage
experiment which is optimized combination of the
1 Equal input parameter resulting into desired output.
The proposed method is found to be significantly
effective in the selection of the brake disc and thus it can
Table 7. Pairwise comparison matrix
be applied to the various industries by giving the fit of
Weight Machining time Stress curve and effective solution for best alternative from the
Weight 1 1/5 1/7 available options.
Machining time 5 1 1/3 Table 10. Closeness co-efficient index
Stress 7 3 1 Di+ Di- CCi S/N ratio
Alternative
1 0.008 0.255 0.968 -0.2780
Table 8. Consistency Ratio
2 0.044 0.216 0.830 -1.6236
Weighted Criteria
λmax CR
sum value weight 3 0.109 0.152 0.582 -4.7039
Weight 0.22 0.07
4 0.066 0.194 0.746 -2.5472
Machining
0.87 0.28 3.07 0.06
time 5 0.259 0.011 0.041 -27.7944
Stress 2.01 0.64
6 0.084 0.178 0.679 -3.3666
Table 9. Weighted normalized decision matrix 7 0.220 0.046 0.173 -15.2583

Alternative Weight Machining time Stress 8 0.010 0.259 0.964 -0.3202

1 0.028 0.084 0.110 9 0.033 0.228 0.874 -1.1685

2 0.027 0.092 0.149

3 0.028 0.103 0.213
4 0.025 0.091 0.172
5 0.024 0.098 0.365
6 0.024 0.108 0.187
7 0.022 0.084 0.326
8 0.021 0.093 0.106
9 0.021 0.093 0.137

using [7] and [8] respectively. Further, they can be used to

find the closeness coefficient index. The closeness Figure 5. Main effects plot for S/N ratio
 10. Saaty, T.L., A scaling method for priorities in
REFERENCES
hierarchical structures. Journal of Mathematical
1. Lin, C., Use of the Taguchi method and grey relational Psychology, 1977. 15(3): p. 234-281.
analysis to optimize turning operations with multiple 11. Tung, S.L. and S.L. Tang, A comparison of the Saaty's
performance characteristics. Materials and AHP and modified AHP for right and left eigenvector
manufacturing processes, 2004. 19(2): p. 209-220. inconsistency. European Journal of Operational
2. Seyedmohammadi, J., et al., Development of a model Research, 1998. 106(1): p. 123-128.
using matter element, AHP and GIS techniques to 12. Pohekar, S.D. and M. Ramachandran, Application of
assess the suitability of land for agriculture. Geoderma, multi-criteria decision making to sustainable energy
2019. 352: p. 80-95. planning—A review. Renewable and Sustainable
3. Ananthakumar, K., et al., Measurement and Energy Reviews, 2004. 8(4): p. 365-381.
optimization of multi-response characteristics in 13. Babu, K.A. and P. Venkataramaiah, Multi-response
plasma arc cutting of Monel 400™ using RSM and optimization in wire electrical discharge machining
TOPSIS. Measurement, 2019. 135: p. 725-737. (WEDM) of Al6061/SiCp composite using hybrid
4. Limpert, R., Brake design and safety. 2011: SAE. approach. Journal for Manufacturing Science and
5. Puhn, F., Brake handbook. 1985: HP Books. Production, 2015. 15(4): p. 327-338.
6. Bhandari, V., Design of machine elements. 2010: Tata 14. Chen, S.-J. and C.-L. Hwang, Fuzzy multiple attribute
McGraw-Hill Education. decision making methods, in Fuzzy multiple attribute
7. Shravani, D., P.K. Lakshmi, and J. Balasubramaniam, decision making. 1992, Springer. p. 289-486.
Preparation and optimization of various parameters of 15. Hwang, C.-L., Kwangsun Yoonet al. Multiple attribute
enteric coated pellets using the Taguchi L9 orthogonal decision making: methods and applications: a state-of-
array design and their characterization. Acta the-art survey. Springer-Verlag New York, 1981. 24: p.
Pharmaceutica Sinica B, 2011. 1(1): p. 56-63. 113.
8. Panneerselvam, R., Design and Analysis of Algorithms. 16. Triantaphyllou, E., Multi-criteria decision making
2016: PHI Learning Pvt. Ltd. methods, in Multi-criteria decision making methods: A
9. DONNELLY, P., DESIGN AND ANALYSIS OF comparative study. 2000, Springer. p. 5-21.
EXPERIMENTS, -MONTGOMERY, DC, 1984, AMER 17. Opricovic, S. and G.-H. Tzeng, Compromise solution
SOC QUALITY CONTROL-ASQC ASQC by MCDM methods: A comparative analysis of VIKOR
MEMBERSHIP MANAGER 611 E. WISCONSIN …. and TOPSIS. European journal of operational research,
2004. 156(2): p. 445-455.
18. https://www.araiindia.com/

Authors’ background (This form is only for submitted manuscript for review)
Your Name Title* Affiliation Research Field Personal website
Optimization
Pandit Deendayal
techniques, Quality
Arrown Dalsaniya Student Petroleum -
control, Statistical
University
analysis
Pandit Deendayal Entrepreneurship,
Dhananj Modi Student Petroleum Economics, -
University Operation research
Welding, Non-
Pandit Deendayal
Assistant conventional kishan148fuse.wixsite.com/mysite
Kishan Fuse Petroleum
professor machining,
University
Optimization
Pandit Deendayal Advanced
Maulik Kanakhara Student Petroleum manufacturing, -
University Automobile