Available online at www.sciencedirect.

com

ScienceDirect
Materials Today: Proceedings 5 (2018) 23945–23954 www.materialstoday.com/proceedings

IConAMMA_2017

Analysis of Radiator Mounting Bracket Design and Derivation of

Transfer Function to Predict Modal Frequencies Based on
Parametric Study
Pooja Morea*, B Santhosha, Amol Jainb
a
Department of Mechanical Engineering, Amrita School of Engineering,Coimbatore,Amrita Vishwa Vidyapeetham,Amrita University, India
b
Power Systems Business Unit, Cummins India Office Campus, Survey No.21, Balewadi, Pune-411045, India.

Abstract

Radiator and engine, the major components of the generator are mounted on the base frame with the help of mounting brackets.
These brackets are subjected to vibration during the working of the generator. Noise related problems that occur in mechanical
structures are mostly due to vibrations. Excessive vibration results in structural damage. The structure itself has certain internal
properties and it is important to understand its characteristics. This paper focuses on the derivation of transfer function of radiator
brackets to express their dynamic vibration characteristics. The approach to derive the same comprises of finite element
modelling of the radiator bracket, model validation, parameter identification based on sensitivity study and design of experiments
to derive the transfer function. In this paper, transfer function for the first two bending frequencies of the bracket is developed
with parameters that would significantly affect the fundamental frequency as inputs. The first two bending frequencies obtained
from analytical method are compared with that from the finite element models and found to have good agreement. The transfer
function developed will avoid expensive testing and time consuming simulations. It will also form a base to develop a more
robust design.

© 2018 Elsevier Ltd. All rights reserved.

Selection and/or Peer-review under responsibility of International Conference on Advances in Materials and Manufacturing Applications
[IConAMMA 2017].

Keywords:Modal analysis; Parameter identification; Transfer function; Design of experiments.

* Corresponding author. Tel.: +0-940-436-1279

E-mail address: [email protected]

2214-7853© 2018 Elsevier Ltd. All rights reserved.

Selection and/or Peer-review under responsibility of International Conference on Advances in Materials and Manufacturing Applications
[IConAMMA 2017].
23946 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954

Nomenclature

AVMs Anti-vibration mounts Me Engine mass

B Number of bolts Mr Radiator mass
DOE Design of experiment T Overall thickness of the bracket
F1 First bending frequency of the bracket TFF1 Transfer function for first bending frequency F1
F2 Second bending frequency of the bracket TFF2 Transfer function for second bending frequency F2
FE Finite element
Lb Bracket length
Lt Length of tie rod
MAC Modal assurance criterion

1. Introduction

A generator is a device that converts mechanical energy obtained from an external source to electrical energy.
The main components of a generator are the engine, alternator, control system and the cooling system of which the
radiator is a part. The radiator and the engine are mounted on a base frame or skid with the help of mounting
brackets. The mounting bracket is generally designed in such a way that it withstands the weight of the radiator and
the engine. However, it is subjected to vibration during the working of the generator. Various rotating and
reciprocating parts of the engine serve as major sources of vibration in a generator. Structural failure of radiator
mounting bracket has been continuously a concern when the vibration and stresses are severe and excessive.
Resonance is often the cause of, or at least a contributing factor to many of the vibration and noise related problems
that occur in structures and operating machinery [1]. Thus, to have a better understanding of any structural vibration
problem, the resonant or the natural frequency of a structure need to be identified and quantified. The natural
frequency of the radiator bracket should not coincide with the engine critical frequency for safe operation.
Critical Engine frequency = (Engine speed in RPM/60) × (Engine order)
Engine order is the number of disturbances generated in one revolution of crankshaft. Engine firing or
combustion produces major vibrations which are relative to the number of cylinders in the engine. The engine order
caused due to engine firing is generally one half the numbers of cylinders [2]. Table.1 shows the critical engine
frequency for selected engine speed and number of cylinders.

Table 1. Critical engine frequency for selected engine speed and number of cylinders
RPM 4 Cylinder ( 2nd order ) 6 cylinder ( 3rd order ) 8 cylinder ( 4th order )
750 25 Hz 37.5 Hz 50 Hz
1500 50 Hz 75 Hz 100 Hz
3000 100 Hz 150 Hz 200 Hz

Experimental modal analysis (EMA) is widely used to identify a modal model through measured dynamic
excitations and vibration responses [3]. However, this method heavily depends on experiments and testing, which
require high costs and long preparations for ground tests. Simulation based analysis tool can also be used to predict
the behavior of the structure and understand its dynamic behavior well in advance before the experimental test.
However, the development of simulation model for the entire structure under consideration and long computation
time restricts its use during the preliminary design to explore design alternatives at the system level [4].
In this paper, a different approach to identify the dynamic characteristic of the radiator brackets by developing a
transfer function is presented. The paper focuses on developing the transfer function for the first two bending
frequencies of the bracket with the help of design of experiment (DOE) which is a systematic method to determine
the relationship between factors affecting a process and its output. In this case, the factors affecting the bending
frequencies significantly are identified by carrying out sensitivity study of all the parameters of the bracket. The
transfer function developed is a simple mathematical equation relating the bending frequencies and the parameters
of the bracket, which is easy to solve and would thus avoid long simulation hours as well as expensive tests.
 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954 23947

2. Methodology

In this section, the methodology used for the derivation of transfer function is explained in different subsections.

2.1 Finite element modelling of the structure and validation

Radiator mounting bracket is subjected to vibration during the working of generator. It is therefore necessary to
understand the dynamic characteristic of the structure. Modal analysis is the study of the dynamic properties of
structures under vibration excitation [5]. Modes are inherent properties of a structure, and are determined by the
material properties, and boundary conditions of the structure. If either the material properties or the boundary
conditions of a structure change, its modes will change. Each mode is defined by a natural frequency, modal
damping, and a mode shape known as modal parameters [2]. Modal analysis using ANSYS software was carried out
to have a clear understanding of the mode shape and modal frequency of the bracket. Fig.1 shows the FE model of
the bracket along with engine and the alternator modelled as point masses at their respective centres of gravity and
connected to each other with the help of beams. Fig.2 (a) shows the FE model of the radiator which is assembled
with the FE model of the bracket to carry out the modal analysis. The FE model of the assembly is shown in Fig.2
(b). Boundary conditions in terms of bolts, rigid connections and anti-vibration mounts (AVMs) are also created
within the assembly.

Fig.1 Finite element model of the bracket along with engine and alternator point mass

Fig.2. (a) Finite element model of radiator; (b) Assembly of finite element model of radiator and the bracket
The frequencies obtained from analysis of the full assembly were validated with experimental results and are
tabulated below.
Table 2. Comparison of simulation and experimental results
Modal frequency (Hz)
Modes
Simulation Experimental
1. 6.03 5.79
AVM modes 2. 8.40 7.91
3. 9.53 8.66
Support bracket mode 21.09 22.59
Radiator global mode and localized support bracket mode 30.23 32.78
23948 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954

The modal analysis carried out for full assembly show various mode shapes. However, as the mode shapes and
modal frequencies of the radiator bracket are of major concern, reduced equivalent finite element models for the full
assembly were created. In this, model 1 considered only the radiator point mass at its centre of gravity location
whereas model 2 considered the radiator point mass at its centre of gravity location along with the radiator frame
structure which accounted for the stiffness of the radiator and are shown in the Figs.3.(a) - (b).

Fig.3. (a) Model 1; (b) Model 2

On carrying out the modal analysis for both the models it was found that modal frequencies of model 2 were in
close approximation with that of full assembly model which are shown in Table.3.

Table 3. Comparison of frequencies between full assembly model and model 2

Frequency Full model Model 2 Frequency error
1st bending frequency 21.0926 Hz 21.3157 Hz 1.057 %
2nd bending frequency 30.237 Hz 27.0023 Hz -10.7 %

For further acceptance, modal assurance criterion (MAC) was carried out between the full assembly model and
model 2 to determine the similarity between the mode shapes. The MAC values of 0.961 and 0.953 are obtained for
the mode shapes of first and second bending frequencies. Thus, model 2 is further used to obtain the frequency
response data required to derive the transfer function using DOE instead of the full assembly, which will reduce the
computational time.

2.2 Parameter identification based on sensitivity study

System properties such as mass, damping and stiffness along with boundary conditions affect the mode shape and
frequency of any structure. Thus, all the possible parameters that would affect the radiator bracket frequency were
noted and presented in the form of a tree diagram as shown in Fig.4.The tree diagram shows the possible parameters
such as radiator and engine mass, bracket parameters, AVM between bracket and radiator, general AVM and
junction stiffness in terms of bolts in a broader manner along with the breakdown of bracket parameters, AVMs on
bracket and general AVMs to a lower level.

Fig.4. Tree diagram showing all the parameters affecting the bracket frequency
 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954 23949

Sensitivity study for all these parameters was carried out using model 2 to understand the effect of each
parameter on the bracket frequency. Table 4 provide results for parameters that have significant effect on the bracket
frequency.

Table 4. Critical parameters affecting the frequency of the bracket

Parameter values Frequency (Hz)
Parameters Frequency shift (Hz)
Initial value Changed value Initial After change
21.3157 23.7793 2.4582
Radiator mass (kg) 101.66 72.099
27.0023 29.4940 2.4917
21.3157 22.0390 0.7233
Engine mass (kg) 417 280
27.0023 29.4823 2.48
21.3157 14.4230 -6.8927
Number of bolts 3 1
27.0023 26.5977 -0.4046
21.3157 22.3510 1.0353
Overall thickness of bracket (mm) 8 10
27.0023 28.5090 1.5067
21.3157 11.1495 -10.1662
AVMs between radiator and bracket Absent Present
27.0023 22.6716 -4.3307
21.3157 23.4305 2.1148
Bracket length (mm) 605 505
27.0023 30.6608 3.6585
21.3157 18.0046 -3.3111
Tie rod length (mm) 340 312
27.0023 30.0983 3.0960

The sensitivity study narrowed down the parameters to only a few significant parameters of all the possible
parameters. These significant parameters are highlighted in the tree diagram shown in Fig.5, which are engine mass,
radiator mass, length of tie rod, overall thickness of bracket and bracket length as bracket parameters, presence or
absence of AVMs and junction stiffness in terms of bolts. From Fig.4 it can be seen that the stiffness value of
neither the AVMs between bracket and radiator nor general AVMs affect the bracket frequency. Also the length and
the thickness of the lateral and vertical stiffener and the thickness of tie rod do not have major effect on bracket
frequencies. Fig.5 provides a clear illustration of the design parameters to be considered in further analysis.

Fig.5. Design parameters for the bracket

2.3 Derivation of transfer function using Design of Experiment (DOE)

In this paper, a transfer function for first and second bending frequencies of the bracket need to be derived with
the identified seven critical parameters as inputs. Design of experiment is one of the method that helps to investigate
the effects of input variables on an output response at the same time. These experiments consist of a series of runs,
or tests, in which purposeful changes are made to the input variables [6]. Data collected for each run is then used to
evaluate the effect of different factors on the response variable. Design of experiments was carried out using the
MINITAB software. Here, factorial design is used to study the effect of each factor as well as their interaction on a
23950 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954

response. The factorial design is further divided into full factorial and fractional factorial design. A full factorial
analysis considers all the possible level combinations for a given number of factors increasing the number of
experiments exponentially with the number of factors in a design, thereby dramatically increasing the experimental
cost [7]. Thus, a fractional factorial is used wherein only a fraction of experimental conditions are considered based
on sparsity of effect principle. Table.5 shows a factorial design with the seven identified critical parameters as
factors having two levels.

Table 5. Factorial design with high and low values in un-coded units
Factors High Low
Radiator mass (kg) 101.66 72.099
Engine mass (kg) 417 280
Overall thickness (mm) 10 6
Number of bolts 3 1
AVMs With (1) Without (0)
Bracket length (mm) 605 505
Tie rod length (mm) 340 312

A half fractional factorial design for seven parameters with two levels provides 64 experimental runs. The first
two bending frequencies for these 64 runs are obtained by carrying out analysis in ANSYS using the finite element
model 2 which was previously created. Table.6 shows experimental runs with F1 and F2 as the first and second
bending frequencies of the bracket for a few combinations of input parameters for reference.

Table 6. Design of experiment with F1 and F2 as the output response

Run Radiator Engine Overall Number of AVMs Bracket length Tie rod F1 (Hz) F2 (Hz)
order mass (kg) mass (kg) thickness (mm) bolts (mm) length (mm)
1 72.099 280 10 3 0 505 340 27.7184 37.4086
2 72.099 417 10 1 1 605 340 11.3307 20.5829
3 72.099 280 6 1 0 505 340 18.4253 32.4904
4 72.099 417 6 1 1 505 340 9.9303 20.7618
5 72.099 280 10 3 1 605 340 13.8841 27.409
6 101.62 417 10 1 1 605 312 10.2433 20.0633
7 101.62 280 10 3 1 505 340 14.3775 27.1018
8 101.62 417 10 1 0 505 312 17.0982 36.3782

3. Results and discussions

The analysis based on the factorial design is carried out and the results are presented for a significance value α =
0.05 with null hypothesis as parameters do not affect the bending frequencies. On carrying out the analysis of
factorial design, P-value higher than the set significance value (α) of 0.05 was obtained. P-value is the probability of
rejecting the null hypothesis in favor of the alternative hypothesis. P-value less than the significant value α indicate
considerable effect of parameters on the bending frequencies and so the model was refit by removing the terms
having P-value higher than 0.05. Main and interaction effect plots for the terms included in the transfer function are
plotted to understand their significant effect on the bending frequencies. The Pareto chart plotted highlight
magnitude and importance of the terms included in the transfer function based on the reference line drawn on the
chart. All the terms beyond the reference line have potential effect on the output response.

2.4 Results for analysis of variance for first and second bending frequencies
Table.7 (a) and (b) highlights the P-values and statistical information like degree of freedom (DF), adjusted sum
of squares (Adj SS), adjusted mean squares (Adj MS) for linear terms and its 2-way and 3-way interactions, which
forms a basis for test of significance and transfer function for bending frequencies F1 and F2 respectively. 2-way and
 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954 23951

3-way interactions indicate that the effect of the independent linear terms included in the model will differ
depending on the level of the other linear term in combination. Thus, P-value less than the significance value
indicate that the interactions between two and three linear terms in the model will have significant effect on the
bending frequencies.

Table 7. (a) Analysis of variance for frequency F1; (b) Analysis of variance for frequency F2
(a) (b)
Adj Adj F- P- Adj Adj F- P-
Source DF Source DF
SS MS value Value SS MS Value Value
Model 33 1509.8 45.75 3224.6 0 Model 27 2198.5 81.43 947.37 0
Linear 7 1403.9 200.56 14135.4 0 Linear 7 2000.9 285.85 3325.7 0
Mr 1 33.39 33.39 2353.5 0 Mr 1 84.46 84.46 982.61 0
Me 1 6.93 6.93 488.11 0 Me 1 76.9 76.9 894.7 0
T 1 86.05 86.05 6064.9 0 T 1 102.51 102.51 1192.6 0
B 1 166.86 166.86 11760.3 0 B 1 125.58 125.58 1461.1 0
AVMs 1 1039.1 1039.1 73242.3 0 AVMs 1 1327.2 1327.2 15442.0 0
Lb 1 38.63 38.63 2722.3 0 Lb 1 159.56 159.56 1856.4 0
Lt 1 32.87 32.87 2316.3 0 Lt 1 124.69 124.69 1450.6 0
2-Way Interactions 18 92.74 5.15 363.13 0 2-Way Interactions 15 188.1 12.54 145.89 0
Mr×Me 1 0.15 0.15 10.57 0.003 Mr×T 1 0.94 0.94 10.89 0.002
Mr×T 1 0.11 0.11 7.74 0.009 Mr×B 1 0.83 0.83 9.65 0.004
Mr×B 1 0.21 0.21 14.57 0.001 Mr×AVMs 1 4.54 4.54 52.77 0
Mr×AVMs 1 8.41 8.41 592.77 0 Mr×Lb 1 1.34 1.34 15.6 0
Mr×Lb 1 0.17 0.17 12.07 0.002 Mr×Lt 1 0.48 0.48 5.56 0.024
Mr×Lt 1 0.04 0.04 2.56 0.12 T×B 1 1.87 1.87 21.8 0
Me×T 1 0.06 0.06 4.03 0.054 T×AVMs 1 20.35 20.35 236.77 0
Me×AVMs 1 0.31 0.31 21.72 0 T×Lb 1 0.92 0.92 10.68 0.002
Me×Lt 1 0.07 0.07 5.05 0.032 T×Lt 1 0.86 0.86 9.96 0.003
T×B 1 0.06 0.06 3.89 0.058 B×AVMs 1 70.27 70.27 817.55 0
T×AVMs 1 0.07 0.07 5.11 0.031 B×Lb 1 2.75 2.75 31.94 0
T×Lt 1 1.93 1.93 135.86 0 B×Lt 1 0.65 0.65 7.61 0.009
B×AVMs 1 50.75 50.75 3577.0 0 AVMs×Lb 1 40.47 40.47 470.79 0
B×Lb 1 0.04 0.04 2.59 0.118 AVMs×Lt 1 37.58 37.58 437.28 0
B×Lt 1 9.74 9.74 686.15 0 Lb×Lt 1 4.26 4.26 49.57 0
AVMs×Lb 1 7.17 7.17 505.22 0 3-Way Interactions 5 9.49 1.9 22.08 0
AVMs×Lt 1 11.94 11.94 841.72 0 Mr×B×AVMs 1 0.59 0.59 6.86 0.013
Lb×Lt 1 1.53 1.53 107.63 0 T×B×AVMs 1 1.34 1.34 15.56 0
3-Way Interactions 8 13.16 1.64 115.92 0 B×AVMs×Lb 1 1.32 1.32 15.42 0
Mr×Me×T 1 0.11 0.11 7.49 0.01 B×AVMs×Lt 1 1.44 1.44 16.74 0
Mr×T×AVMs 1 0.18 0.18 12.83 0.001 AVMs×Lb×Lt 1 4.8 4.8 55.81 0
Mr×AVMs×Lt 1 0.1 0.1 7.31 0.011 Error 36 3.09 0.09
T×B×AVMs 1 1.62 1.62 114.48 0 Total 63 2201.62
T×AVMs×Lt 1 0.67 0.67 47.38 0
B×AVMs×Lb 1 2.21 2.21 155.93 0
B×AVMs×Lt 1 7.79 7.79 548.82 0
AVMs×Lb×Lt 1 0.47 0.47 33.15 0
Error 30 0.43 0.01
Total 63 1510.2
23952 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954

Figs.6 (a) and (b) shows the main effect plot of all the terms included in the transfer function model. Out of all
the terms, the slope of line for AVMs is steeper indicating that the AVMs have higher effect on first as well as
second bending frequencies of the bracket.

Fig.6. (a) Main effect plot for F1; (b) Main effect plot for F2

Figs.7 (a) and (b) shows the interaction effect plot of the terms included in the transfer function model. Out of all
the interactions, few dominant interactions for bracket bending frequencies are number of bolts and AVMs, AVMs
and bracket length, AVMs and length of tie rod, number of bolts and length of tie rod based on the non-parallelism
between the lines and the difference in the slopes.

Fig.7. (a) Interaction effect plot for F1; (b) Interaction effect plot for F2

The Pareto chart of standardized effect of F1 and F2 are shown in Figs.8 (a) and (b). It provides information
regarding the extent of relevance of the terms and interactions on the frequencies. It is clearly evident that the extent
to which the AVMs affect the bending frequency is much higher than the engine mass. Taking into consideration the
interaction effect as well, it can be observed that AVMs have more effect on the frequencies than the interaction
between AVMs and number of bolts. This helps one to understand the relevance of the terms and its corresponding
effect and make necessary changes in the terms as per the requirements.
 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954 23953

Fig.8. (a) Pareto chart of standardized effect for F1; (b) Pareto chart of standardized effect for F2

After refitting the model such that only terms with P-value less than the significance value are taken into
consideration, the designs of experiment provides the transfer functions for the output response and are given below.

Transfer function for first bending frequency F1 of the bracket is derived as,
TFF 2.44 0.0463. Mr 0.00051. Me 3.056. T 16.175. B 0.45. AVMs 0.0752. Lb 0.1272. Lt
0.000113. Mr . Me 0.00743. Mr . T 0.00385. Mr . B 0.0490. Mr . AVMs 0.000070. Mr . Lb
0.000310. Mr . Lt 0.001532. Me . T 0.002026. Me . AVMs 0.000035. Me . Lt 0.0650. T . B
2.348. T . AVMS 0.009858. T . Lt 17.317. B . AVMs 0.004198. B . Lb 0.05277. B . Lt
0.0516. AVMS . Lb 0.0733. AVMs . Lt 0.000343. Lb . Lt 0.000020. Mr . Me . T
0.00361. Mr . T . AVMs 0.000390. Mr . AVMs . Lt 0.1593. T . B . AVMs 0.00732. T . AVMs . Lt
0.007437. B . AVMs . Lb 0.04983. B . AVMs . Lt 0.000245. AVMs . Lb . Lt

Transfer function for second bending frequency F2 of the bracket is derived as,
TFF 266.8 0.3058. Mr 0.016002. Me 1.926. T 0.08. B 227.1. AVMs 0.3192. Lb 0.5863. Lt
0.00410. Mr . T 0.00121. Mr . B 0.0621. Mr . AVMs 0.000196. Mr . Lb 0.000418. Mr . Lt
0.0133. T . B 0.8530. T . AVMs 0.001198. T . Lb 0.00413. T . Lt 12.25. B . AVMs
0.00126. B . Lb 0.00349. B . Lt 0.2983. AVMs . Lb 0.5865. AVMs . Lt 0.000760. Lb . Lt
0.01300. Mr . B . AVMs 0.1446. T . B . AVMs 0.00576. B . AVMs . Lb 0.02142. B . AVMs . Lt
0.000782. AVMS. Lb . Lt

Table 8 provides results obtained from the FE model analysis and the transfer function of the first two bending
frequencies for parameter combination of run order 1 as given in Table.6.

Table 8 Comparison of frequencies obtained from FE model and transfer function

Frequency FE model TF model % error
F1 27.7184 Hz 27.4403 Hz -1.003
F2 37.4086 Hz 37.3084 Hz -0.277

From the values one can conclude that the frequencies obtained from the derived transfer function makes clear
agreement with the FE model and hence reduces the computational effort.

4. Conclusion

In this paper, the transfer functions to obtain the first two bending frequencies of radiator mounting bracket are
derived based on parametric study and design of experiments. The seven critical parameters affecting the dynamics
of the bracket are radiator mass, engine mass, number of bolts, overall thickness of bracket, bracket length, tie rod
23954 Pooja More/ Materials Today: Proceedings 5 (2018) 23945–23954

length and presence or absence of AVMs between radiator and bracket. The main and interaction effect plot indicate
that AVMs have a higher effect on the bending frequencies and the Pareto chart plotted provided information
regarding the extent of relevance of these main and interaction terms. The bending frequencies obtained from the
transfer function should not coincide with the critical engine frequency for safe operation. The transfer function will
thus help to make necessary changes in the parameters at the preliminary design level and also serve as a base for
much more robust bracket design. Also modal analysis carried out for reduced equivalent model added value to the
information that mass and stiffness affect the mode shape and frequency of the bracket and use of this model instead
of the full assembly would reduce finite element modelling and analysis time.

Acknowledgement

The first author would like to thank Cummins Inc, Pune for providing an opportunity to work on a project in the
engineering field.

References

[1] Patrick Guillaume, Modal analysis, Department of Mechanical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussel, Belgium.
[2] Fundamentals of noise vibration and harshness, Toyota technical training, section 1, pp-32.
[3] Wu X, Lei Y, Li D, Xie Y. Transfer function modeling of structural vibration of complex aerospace structures based on finite element
analysis. Journal of Mechanical Science and Technology. 2013;27(5):1245-1253.
[4] Nasser M, Jawad B. Developing a Transfer Function for Vehicle Ride Performance. SAE International Journal of Passenger Cars -
Mechanical Systems. 2008;1(1):1003-1008.
[5] Girish Dalvi, Pankaj Kumar, Tejas Vispute,et al., Modal Analysis of Beam Type Structures. International Journal of Engineering Research
and technology. 2015;V4(04).
[6] Minitab Inc. (2010). Designing and experiment, Getting Started with Minitab 17, pp.42-50. Available from: www.minitab.com.
[7] Xin Zhang, Mark E. Barkey, Yung-Li Lee, Charles J. Orsette ,et al., DOE Analysis of Factors Affecting Ultimate Strength of Multiple
Resistance Spot Welded Joints, SAE technical paper series,2007