Master's Degree Thesis

ISRN: BTH-AMT-EX--2017/D13--SE

Development of Tool in MATLAB

for the Durability Prediction of
Radiators

Akhil Mora
Raghavendra Machipeddi

Department of Mechanical Engineering

Blekinge Institute of Technology
Karlskrona, Sweden
2017

Supervisors: Tobias Karlsson, TitanX Engine Cooling AB, Mjällby

Ansel Berghuvud, BTH
 Development of Tool in
MATLAB for the Durability
Prediction of Radiators
Akhil Mora
Raghavendra Machipeddi

Department of Mechanical Engineering

Blekinge Institute of Technology
Karlskrona, Sweden
2017

Thesis submitted for completion of Master of Science in Mechanical

Engineering with emphasis on Structural Mechanics at the department of
Mechanical Engineering, Blekinge Institute of Technology, Karlskrona,
Sweden.
Abstract:
Durability is the most important factor in the design of heat-exchangers
to meet the specifications of the customers. To predict the durability,
endurance tests are carried out. In this thesis, one of the endurance tests,
thermal cycling is performed for three different internal coolant flows
until failure which is known as Wöhler test. A tool is developed in
MATLAB that could import data from all kinds of endurance tests and
visualizes time histories of every channel of the test (test parameters such
as temperature, pressure, flow). An algorithm is developed for Level
Crossing counting method which works based on the Markov cycle
counting method. This produces Level crossing curves (LCC) for all the
channels of the test which says how well a test has been performed and
the total number of cycles of the test. It was observed that the LCCs
obtained from the tool gives accurate results when compared with those
obtained from LMS software whose approach of producing LCCs is
Rainflow cycle counting method. Strain measurements are performed for
the same flowrates as that of the Wöhler tests. The results from strain
measurements and Wöhler tests are used in the determination of
Basquin’s coefficient of the Wöhler curve. It was observed that the
optimal value of Basquin’s coefficient is 3.4.
Keywords:
Basquin’s coeffcient, Cycle counting, Level Crossing, Thermal Cycling,
Wöhler curve.

2
Acknowledgements

This work was carried out at the Department of Mechanical Engineering,

Blekinge Institute of Technology (BTH), Karlskrona, Sweden and TitanX
Engine Cooling AB, Mjällby, Sweden from January 2017 to October 2017
under the supervision of Dr. Ansel Berghuvud, and Tobias Karlsson.

First and foremost, we would like to thank our parents for their endless love
and support over the years.
We wish to express our sincere gratitude to our main promoter Anders
Brorsson, RD&E Director OU ESA, our industrial supervisor Tobias
Karlsson and Magnus Nilsson, Test Center Manager, for their guidance and
professional engagement throughout the work.
We want to express our deep gratitude to our academic supervisor Dr. Ansel
Berghuvud for his guidance, valuable suggestions and support.
We would also like to express our gratitude to our friend, Rohan Appilla
Chakravarthula, Developer, Ericsson, who helped us in difficult situations in
the coding part of MATLAB.

Karlskrona, October 2017

Akhil Mora
Raghavendra Machipeddi

3
Contents
List of figures 6
List of Tables 8
1 Notations 9
2 Introduction 11
2.1 Background 11
2.2 Problem description 12
2.3 Aim and objectives 12
2.4 Research Questions 13
2.5 Delimitations 14
2.6 Disposition 14
3 Related work 15
3.1 Literature study 15
3.1.1 Strain gauges 15
3.1.2 Cycle counting methods 21
4 Studied system 24
4.1 Material data and Material Model 25
5 Overview of the Method 26
6 Experimentation process 27
6.1 Strain Measurements 27
6.2 Wöhler Testing 32
6.3 Results from experimentation 33
6.3.1 Strain measurements 33
6.3.2 Wöhler testing 35
7 Development of tool 41
7.1 Data importing 41
7.2 Level Crossing Cycle Counting 44
7.2.1 Implementation of Level Crossing Cycle Counting 44
7.2.2 Real data from test bench 48
7.2.3 Level Crossing counting for Real data 52
7.3 Graphical user interface (GUI) 54
8 Investigation of Basquin’s Coefficient 55
9 Results 61
10 Summary and Conclusions 64

4
11 Future works 65
12 References 66
13 Appendix 67
A. Numbering of tubes on radiator 67
B. Level crossing curves from Wöhler Testing 67

5
List of figures
Figure 2.1. A typical radiator. .................................................................... 11
Figure 3.1. A foil strain gauge. ................................................................... 16
Figure 3.2. Gauge nomenclature [2]. ......................................................... 16
Figure 3.3. Wheatstone bridge. ................................................................... 17
Figure 3.4. Wheatstone bridge with Shunt resistor. ................................... 18
Figure 3.5. Two-wire quarter-bridge circuit. ............................................. 19
Figure 3.6. Three-wire quarter-bridge circuit. ........................................... 19
Figure 3.7. Half-bridge circuit. .................................................................. 20
Figure 3.8. Full-bridge circuit. ................................................................... 20
Figure 4.1. Studied Radiator....................................................................... 24
Figure 4.2. Thermal cycling Test Bench. .................................................... 25
Figure 5.1. Method overview. ..................................................................... 26
Figure 6.1. Kyowa Strain gauge. ................................................................ 28
Figure 6.2. Strain gauges installed on Radiator. ........................................ 29
Figure 6.3. Strain gauges with the solder terminal and bridge circuit....... 30
Figure 6.4. Bridge with Shunt resistance. ................................................... 31
Figure 6.5. Two radiators mounted in the test bench. ................................ 32
Figure 6.6. Radiator with cut breakaways. ................................................. 33
Figure 6.7. Strain Range. ............................................................................ 35
Figure 6.8. Failure positions of object 1..................................................... 36
Figure 6.9. Failure position of object 2. ..................................................... 36
Figure 6.10. Failure position of object 4. ................................................... 37
Figure 6.11. Failure position of object 5. ................................................... 38
Figure 6.12. Failure position of object 6. ................................................... 38
Figure 6.13. Failure position compared with thermographic image of object
5. .................................................................................................................. 39
Figure 6.14. Failure position compared with thermographic image of object
6. .................................................................................................................. 40
Figure 7.1. Synthetic signal 1. .................................................................... 45
Figure 7.2. Level crossing curve for synthetic signal 1 from MATLAB. .... 46
Figure 7.3. Level crossing curve for synthetic signal 1 from LMS. ............ 46
Figure 7.4. Synthetic signal 2. .................................................................... 47
Figure 7.5. Level crossing curve for synthetic signal 2 from MATLAB. .... 47
Figure 7.6. Level crossing curve of synthetic signal 2 from LMS. ............. 48

6
Figure 7.7. Time history of temperature (1 day). ....................................... 49
Figure 7.8. Time history of pressure (1 day). ............................................. 49
Figure 7.9. Time history of flow (1 day). .................................................... 50
Figure 7.10. Time history of temperature (5 cycles). ................................. 50
Figure 7.11. Time history of pressure (5 cycles). ....................................... 51
Figure 7.12. Time history of flow (5 cycles). .............................................. 51
Figure 7.13. Level crossing curve of Temperature history......................... 52
Figure 7.14. Level crossing curve of Pressure history. .............................. 53
Figure 7.15. Level crossing curve of Flow history. .................................... 53
Figure 7.16. GUI of tool developed. ........................................................... 54
Figure 8.1. Wöhler curve (log-log scale). .................................................. 58
Figure 8.2. Wöhler curve (semi-log scale). ................................................ 59
Figure 9.1. Temperature history of one day data. ...................................... 62
Figure 9.2. Level crossing spectrum of Temperature data for one day...... 62
Figure 13.1. Engine side (Tubes 1-88). ...................................................... 67
Figure 13.2. LCC of Inlet and Outlet Temperature (Object 1). .................. 68
Figure 13.3. LCC of Pressure (Object 1). .................................................. 68
Figure 13.4. LCC of Flow (Object 1). ........................................................ 68
Figure 13.5. LCC of Inlet and Outlet Temperature (Object 2). .................. 69
Figure 13.6. LCC of Pressure (Object 2). .................................................. 69
Figure 13.7. LCC of Flow (Object 2). ........................................................ 69
Figure 13.8. LCC of Inlet and Outlet Temperature (Object 3)................... 70
Figure 13.9. LCC of Pressure (Object 3). .................................................. 70
Figure 13.10. LCC of Flow (Object 3). ...................................................... 70
Figure 13.11. LCC of Inlet and Outlet Temperature (Object 4)................. 71
Figure 13.12. LCC of Pressure (Object 4). ................................................ 71
Figure 13.13. LCC of Flow (Object 4). ...................................................... 71
Figure 13.14. LCC of Inlet and Outlet Temperature (Object 5)................. 72
Figure 13.15. LCC of Pressure (Object 5). ................................................ 72
Figure 13.16. LCC of Flow (Object 5). ...................................................... 72
Figure 13.17. LCC of Inlet and Outlet Temperature (Object 6)................. 73
Figure 13.18. LCC of Pressure (Object 6). ................................................ 73
Figure 13.19. LCC of Flow (Object 6). ...................................................... 73

7
List of Tables
Table 4.1. Material properties of Aluminum............................................... 25
Table 6.1. Gauge dimensions. ..................................................................... 28
Table 6.2. Position of strain gauges. .......................................................... 29
Table 6.3. Channel set-up analysis. ............................................................ 31
Table 6.4. Maximum strain range. .............................................................. 34
Table 6.5. Cycles to failure of each object. ................................................. 39
Table 7.1. Sample data from day 1. ............................................................ 42
Table 7.2. Sample data from day 2. ............................................................ 42
Table 7.3. Concatenated data of the whole test. ......................................... 43
Table 8.1. Wöhler results. ........................................................................... 55
Table 8.2. Stress and strain values for each flow. ...................................... 56

8
1 Notations

E Young’s Modulus
N Number of cycles
ܴ଴ Unstrained Gauge Resistance
ܴଶ Resistance in Wheatstone bridge
ܴଷ Resistance in Wheatstone bridge
ܴସ Resistance in Wheatstone bridge
ܴ௅ Lead wire resistance
ܴௌ Shunt resistance
ܴ௫ Unknown resistance in Wheatstone bridge
οܴ Change in resistance in the gauge
S Stress (Basquin’s relation)
ߙ Constant based on material (Basquin’s relation)
ߚ Basquin’s coefficient
ߝ Strain
ߪ Stress

9
Abbreviations
GF Gauge Factor (Strain gauge)
GUI Graphical User Interface
LCC Level crossing curve
TP Turning points

10
2 Introduction
TitanX Engine Cooling AB in Mjällby is the leading global company that
develops and manufactures heat-exchangers. Before introducing a new
product into the market, different kinds of tests are carried out such as
corrosion, vibration, pressure pulsation and thermal cycling. This chapter
introduces the description of heat-exchangers, problem description, aim and
scope of the thesis work.

2.1 Background
Engine cooling is an important feature that helps in increasing the efficiency
of an automotive engine. During the working of an automobile engine, part
of the heat generated by the engine is converted into power and some of it is
released through the exhaust system. The remaining waste heat is rejected
from the engine into the atmosphere by the cooling system.
In automobile engines, radiators are used as heat exchangers for cooling the
engine. A typical radiator can be seen in Figure 2.1. A cooling system
consists of series of channels in the engine block that allows the coolant
liquid to flow through them, a radiator, a pump and a thermostat.

Figure 2.1. A typical radiator.

A cooling system works by sending a liquid coolant through passages in the
engine block and heads. During this phase, heat from the engine is absorbed
by the coolant. This hot fluid goes to the radiator through a rubber hose. As

11
the coolant flows through the thin tubes present in the radiator, it is cooled
by the air stream. Once the fluid is cooled, the same cycle is repeated. The
pump does the job of keeping the fluid moving throughout the system.
At TitanX, Radiators are manufactured according to the customer’s
specifications and are tested to see if the desired specifications have been
achieved or not. The Quality assurance of a product that is manufactured is
given the utmost priority in any industry. It’s a way of preventing unwanted
mistakes in manufacturing process and avoiding any problems when
delivering solutions to the customers.

2.2 Problem description

The analysis of the data such as temperatures, pressures and flows, obtained
from the validation tests is quite cumbersome task. A data file is produced in
the logging system at the end of every day of the test. A Wöhler test generally
takes one to two months depending upon the flow of the coolant. There is no
coordinated approach in TitanX to analyze the data of the whole test. It is
most important to observe the quality of the test and to check if the test has
been performed at the targeted level.
It is also important to accurately predict the durability of the radiators which
depends on the Basquin’s coefficient. There is no validation for the
Basquin’s coefficient considered in TitanX which is equal to 5.

2.3 Aim and objectives

The main aim of the thesis work is as follows:
x Develop a tool in MATLAB that could import measured data from
tests in a structured way, visualize time histories of the test
parameters such as temperature, pressure and flow and perform cycle
counting method to produce Level Crossing curves (LCC) for all
parameters.
x Investigation of Basquin’s coefficient from strain measurements and
Wöhler tests and determination of an optimal value.

12
The objectives of the work are:
x Literature study regarding the strain gauges.
x Installation of strain gauges at critical positions that are found from
the previous tests and perform strain measurements for different
flows.
x Wöhler testing: Thermal cycling for three different flows.
x Data importing, development of an algorithm for Level Crossing
cycle counting and development of tool (GUI) in MATLAB.
x Analysis of data from Wöhler tests using the tool developed in
MATLAB to see if the test has been performed at a targeted level.
x Determination of an optimal value of Basquin’s coefficient from
strain measurements and Wöhler test.
x Construction of S-N curve.

Scope of the work

x The tool developed can be extended to the other validation tests that
are performed like that of TitanX and can visualize LCCs to check
the quality of the test.
x The optimal value of Basquin’s coefficient determined helps in the
accurate prediction of the damage of the radiators.

2.4 Research Questions

Can a tool be developed in MATLAB that could import measured data from
the tests, perform Level Crossing cycle counting and produce LCCs to see if
the test has been performed at a targeted level?
Is the Basquin’s coefficient considered now, an optimal value?

13
2.5 Delimitations
The Wöhler testing has been performed only on pure radiators, standing
freely in the test bench to minimize the effect from surroundings. Hence, the
Basquin’s coefficient obtained is valid for pure radiators but not for those
mounted in a whole module.

2.6 Disposition
This thesis report starts with the introduction of radiators, problem
description, related work, literature survey and the studied system. It then
moves on to the overview of the method, the experimentation process, the
development of the tool and the investigation of Basquin’s coefficient. The
next chapters contain the results, discussions and conclusions that are drawn
from the work. The report ends with the future works and the references.

14
3 Related work
3.1 Literature study
The literature on cycle counting methods, strain gauges and their
functionality has been carried out.

3.1.1 Strain gauges

A strain gauge is a device which is used for the measurements of surface

strains. It can also be defined as an electric sensor which measures the strain.
Carbon gauges are the first type of strain gauges that were used. They were
used by Charles Kearns who measured the surface strains to improve the
blade design of propeller blades. Carbon composition resistors were used in
these strain gauges. Resistance stability with time and temperature is low in
these type of strain gauges. These gauges were restricted to measurement of
dynamic strains because of the lack the resistance stability [1].
Bonded wire strain gauges are extensively used after the carbon gauges as
they produced smaller resistance changes than the carbon strain gauges. In
this strain gauges a small diameter wire made of any electrical resistant
materials is attached to a structure to find the surface strains. Because of the
small resistance changes, these strain gauges have an advantage over carbon
gauges when measuring the static and dynamic strains.
Based on the same operating principle as the bonded wire strain gauges, foil
strain gauges were introduced. Foil strain gauges are the most extensively
used strain gauges in the present market. One of the main reasons of its
extensive usage is because of its ability to be used at very high temperatures.
Foil gauges are produced by etching or cutting the desired gage pattern into
a thin sheet of metal foil of the appropriate alloy [1]. A typical foil strain
gauge is shown in Figure 3.1. The gauge nomenclature of a foil strain gauge
is shown in Figure 3.2.

15
 Figure 3.1. A foil strain gauge.

Figure 3.2. Gauge nomenclature [2].

The foil strain gauges have more advantages compared to the bonded wire
strain gauges. The difference in them is because:
x The exposed surface of the foil conductor is greater for the same
cross-sectional area resulting in a complete strain transmission to the
grid.
x Because of a better thermal path from the foil conductor to the
substrate, they can operate at higher power levels.
x Manufacturing process of the foil strain gauges is carried out by a
photoetching process where as the wire gauges are manufactured by
hand.
Strain gauge bonding
Adhesives are used to attach the strain gauge to the structure that is analysed.
Adhesives must be compatible with the gauge, the specimen material and the
requirements of the test. The adhesive used must be chosen carefully since
the operating characteristics are not the same in different situations. Due to
this several adhesive systems are developed by the manufacturers. There are
several types of adhesives such as epoxy adhesives, cyanoacrylate adhesives,
thermoplastic adhesives, thermosetting adhesives, phenolic adhesives,
polyimide adhesives etc.

16
Gauge factor
The gauge factor of a strain gauge is defined as the ratio of the resistance
change in the gage to the unstrained gauge resistance and the engineering
strain. The gauge factor is obtained from the manufacturer.
οோൗ
ோೀ
‫ ܨܩ‬ൌ  ఌ
 (3.1)
Where, οܴ is the change in resistance in the gauge.
ܴை Is the unstrained gauge resistance and ߝ is the strain.
Measurement of resistance
A strain gauge is a resistor. Its resistance changes as the strain changes. As
the gauge wire is of very small diameter, the change in resistance is
effectively small. So, it is difficult to measure these small changes. And, the
device that is being used to measure the strain gauge’s resistance will have
the resistance of its own. To overcome this problem a Wheatstone bridge is
used. It is an electrical circuit which is used to measure an unknown
resistance. It works on balancing the legs of a bridge circuit. The unknown
resistance is placed in one leg of the circuit. A typical Wheatstone bridge is
shown in Figure 3.3.

Figure 3.3. Wheatstone bridge.

In Figure 3.3, the unknown resistance Rx can be found out if resistances of
other resistors, supply voltage and the output voltage are known. If the output
voltage is zero volts, the value of resistance Rx can be found out using the
following relation:

17
 ோೣ ோయ ோయ
ൌ െ൐ ܴ௫ ൌ  ‫ܴ כ‬ଶ (3.2)
ோమ ோర ோర

If all the values of resistors are equal, then the circuit is known as a perfectly
balanced bridge. However, during the working operation as the resistance of
the strain gauge changes, the output voltage 2 will not be zero. Hence, to
measure the resistance, the output voltage and the strain value must be
known. To calculate this output voltage, a voltmeter is used.
For calibration, a shunt resistor may be added to the bridge. A Wheatstone
bridge with a shunt resistor can be seen in Figure 3.4. Rs is the resistance of
the Shunt resistor.

Figure 3.4. Wheatstone bridge with Shunt resistor.

To relate the measured voltage to the resistance change, the voltage between
the two measuring points is calculated with the shunt resistor across Rx first
and is also calculated without the shunt resistor.
After calculating the resistance change, it is related to strain using the gauge
factor.
Strain gauge configurations
The most common strain gauge configurations are the Quarter bridge circuit,
the half bridge circuit and the full bridge circuit.
In a quarter-bridge circuit, one of the resistor is replaced by a strain gauge.
The ratio of the two arms of the circuit (on the other side) are set to equal.
So, the bride is balanced when there is no usage of strain gauge. When the
strain gauge is either compressed or under tension, the bridge is unbalanced
because of the change in the resistance. This results a change in the output
voltage leading to the measurement of strain. There are two configurations
in quarter-bridge circuit, the two-wire circuit and three wire circuit.

18
The typical two wires and three wire quarter bridge circuits can be seen in
Figure 3.5 & Figure 3.6.

Figure 3.5. Two-wire quarter-bridge circuit.

The wire resistance plays an important role in the operation of the circuit.
The lead wires that connect the strain gauge to the circuit also have a
resistance (RL). If this resistance is negligible then the bridge remains
balanced. But the lead wires have some internal resistance. So, the gauge arm
resistance becomes the sum of the gauge resistance and twice the lead wire
resistance. The main problems of this type of circuit are that it influences the
sensitivity, it causes an initial imbalance. Another major problem arises when
the temperature of the lead wires changes during the strain measurement
process. This may lead to a change in the resistance of the lead wires. To
overcome these problems, a third wire is added to the circuit. It is shown in
Figure 3.6. The two-wire quarter bridge circuit is mainly used for dynamic
strain measurements.

Figure 3.6. Three-wire quarter-bridge circuit.

In the three-wire quarter bridge, the first lead wire is in series with the gauge.
The second wire is in series with the resistor (R4- dummy resistor). The third
wire is a voltage sensing wire. The advantages of this configuration is that it

19
offers greater measurement sensitivity, the effect of change in temperature in
lead wires is resolved and it offers intrinsic bridge balance.
Half-bridge circuit consists of two strain gauges in place of two resistors. In
this type, both the strain gauges can be placed on the test piece. These strain
gauges are placed perpendicular to each other which results in more
responsive measurement. Because of this arrangement, the effects of change
in temperature are nullified. This is the main advantage of this type of
configuration. A typical half-bridge circuit can be seen in Figure 3.7.

Figure 3.7. Half-bridge circuit.

Full-bridge circuit consists of four strain gauges. All the arms of the circuit
are replaced by strain gauges. This type of configuration offers the highest
degree of accuracy. Another main advantage of this configuration is that the
output voltage is proportional to the applied force. A typical full-bridge
circuit can be seen in Figure 3.8.

Figure 3.8. Full-bridge circuit.

The main disadvantage of the half-bridge and full-bridge circuits is that it is
difficult to install complementary pairs of strain gauges to the test piece. This
leads to frequent usage of quarter-bridge circuit in strain measurement
systems.

20
3.1.2 Cycle counting methods

Cycle counting methods are applied on load signals. Load signals in this
context refers to any physical quantity that is changing over time. Some
typical loads may be temperatures, flows, pressure etc., most load signals can
be described as a function varying over time. The measurement process is in
such a way that the samples are recorded at a specified sampling rate. This
set of data may be referred as a time history or load history. Cycle counting
methods reduce the data from the load histories in different ways which have
their own advantages [3].
The cycle counting methods are divided into two categories:
x Amplitude-based methods
In amplitude-based methods, the turning points of the signal are taken
into consideration [3]. They do not depend on the rate of the load signal.
They are the fundamental methods for the durability prediction.
x Frequency-based methods
In frequency-based methods, the rate of the load signal is taken into
consideration. They are mostly used for system load analysis.
However, each method has its own specialty and can reduce the load signal
in a different way in different scenarios.
Amplitude-based methods: As mentioned, the amplitude based methods only
focus on the local maxima and minima of the signal, which are the turning
points of the signal. The counting methods are again classified into cycle
counting methods and level crossing methods. The cycle counting methods
counts the number of cycles in the load signal using an algorithm.
Cycle counting
Cycle counting methods count the number of cycles in the load signal. There
are different methods for cycle counting, each having its own algorithm.
There are different methods of cycle counting such as range-pair cycle
counting, rainflow cycle counting, markov counting etc.
Before starting the cycle counting procedure, pre-processing needs to be
performed on the signal [3].

21
 i) Pre-processing:
In this step, unnecessary information is filtered from the load signal. This is
done in the following way.
a) Determination of turning points
In this step, the turning points of the signal are found out. This step reduces
the data points in the signal to a very large extent. The intermediate points in
the signal are neglected. This step does not affect the durability estimation as
the intermediate points do not affect the durability in a large extent.
b) Discretization
In this step, the load range in the signal is divided into a number of levels
(bins) and each load value is identified with the level it belongs to. The
discretized points may be less than the turning points of the signal as two
consecutive turning points are considered in the same bin. Normally, the
signal is discretized with a bin size of 128.

ii) Cycle counting

A cycle is counted if the load value of the signal goes from a maximum value
to a minimum value and again a maximum value (or vice versa). If the signal
goes from maximum to minimum (or vice versa) it is known as a half cycle.
Two half cycles constitute to one full cycle of the signal.
a) Markov cycle counting [3]
The Markov cycle counting method uses a simple algorithm for counting the
cycles in the signal. In this method, the local minima and local maxima are
paired into a set. The half cycles are defined as the local minima and local
maxima are paired.
For a discretized load signal, the markov cycle counting results are stored in
a matrix. In the matrix, the number of transitions from bin i to bin j are stored.
The whole number of transitions in the markov matrix are also stored in the
matrix.

22
 b) Level crossing counting [4]
The level crossing counting procedure counts the number of times the load
crosses a level. In this method, the up-crossings and the down-crossings are
counted each time the load crosses a load level. The level crossing count can
be obtained from Markov matrix. From the Markov matrix, looking at the
load levels, the up-crossing and down-crossings are obtained. The obtained
results are plotted in a graph to obtain the level crossing curve. This plot
gives an overview of the distribution of the signal across the range of the
signal.

23
4 Studied system
The function of a radiator is to cool the engine by sending the coolant through
passages in the engine block. The coolant absorbs heat from the engine and
is sent into the radiators that flows into the tubes where the heat is absorbed
thereby cooling the coolant. The fins in between the tubes adds more
efficiency to the heat absorption. The radiator studied is shown in Figure 4.1,
which consists of 88 tubes, 2 rows thick (named as engine side and grill side).
The tubes are counted from left to right when observed from the engine side
and the numbering can be seen in Appendix A.

Figure 4.1. Studied Radiator.

A thermal cycling test bench is used for the Wöhler testing. In the test bench,
the hot cycle temperature and the cold cycle temperature, the pressure and
the flow rate are defined. The test bench can handle temperatures up to 105°C
(±5°C) in the hot cycle and -10°C (±5°C) in the cold cycle. It can handle a
maximum pressure of 150kPa (±15kPa) and a maximum flow of 200 l/min.
The test bench simulates cold start conditions. The coolant consists of 50%
glycol mixture. Thermocouples are used to measure the temperatures at inlet
and outlet. The pressure is measured at inlet and flow is measured at outlet
using a pressure gauge and flow gauge respectively. A logging system
records the data from the test at a specified sampling rate. The thermal
cycling test bench used is shown in Figure 4.2.

24
 Figure 4.2. Thermal cycling Test Bench.

4.1 Material data and Material Model

The inlet and outlet tanks of the radiator are made from a mixture of plastic.
The tubes and the fins of the radiator are made of Aluminum. The properties
of Aluminum are shown in Table 4.1.

Table 4.1. Material properties of Aluminum.

Property Value in SI units
Density 2700 Kg/m3
Modulus of Elasticity 69 GPa

25
5 Overview of the Method

Figure 5.1. Method overview.

26
6 Experimentation process
6.1 Strain Measurements
The experimentation process starts with measurement of strain on the
radiator for different flows. For this purpose, strain gauges are equipped on
the radiator. The selection of strain gauge is the first important step in this
process.
The requirements that the selected strain gauges should meet are:
x The size of the strain gauge should be small.
x The backing material must be the same as that of the material of the
test piece. This is to facilitate the same rate of thermal expansion as
that of the test piece.
x Uniaxial strain gauge, as the strains are measured in a single direction
Based on the above required properties, Kyowa strain gauges are used for
the measurement of strains on the radiator. The model number of the strain
gauges used is KFG-1N120-C1-23N15C2. The first half part of the model
number gives us description about the pattern, the gauge resistance and the
backing material whereas the second part of the model number describes the
type of lead wire used and its length.

KFG- 1N120 -C1-23N15C2

Uniaxial Base material

Gauge Resistance

Lead wire cable

From the above serial number of the strain gauge, it can be deduced that a
uniaxial strain gauge with a gauge resistance of 120Ω is used. The base

27
material of the strain gauge is aluminum and the type of lead wire is
Polyester-coated 2-wire copper cable of a length 15cm.
The gauge dimensions are shown in Table 6.1.

Table 6.1. Gauge dimensions.

Grid(mm) Base(mm)
Length Width Length Width
1 0.65 4.2 1.4

The strain gauge has a gauge factor (GF) of 2.1 and a gauge resistance (R0)
of 120Ω. The lead wires of length 15cm are pre-attached to the gauge. A
picture of the stain gauge used is shown in the Figure 6.1.

Figure 6.1. Kyowa Strain gauge.

The standardized distance for the placing the strain gauges is 7mm from the
header-tube joint to the center of the gauge. The strain gauges are placed on
the flat side of the tubes and on the tube noose. This is because the flat side
has the highest strain levels. The positioning of strain gauges is based on the
previous failure positions of the radiator. Looking at the data from the
previous tests, the radiator always fails at the outer most tubes. Taking this
as a reference, strain gauges are placed on the outermost tubes of the radiator.
The positions at which strain gauges are installed are shown in Table 6.2.

28
 Table 6.2. Position of strain gauges.
Position of strain gauges
Engine Inlet Outlet Outlet Outlet Outlet Outlet Outlet Outlet
side
88 1 (left 2 (left 87 88 88 40 40
(Tube)
(left and and (left (left) (right- (right) (tube
and right) right) and 3,7,11 noose)
right) right) mm)
Front Outlet
side
40(Tube noose)
(Tube)

The pictures of strain gauges installed on the radiator are shown in Figure
6.2.

Figure 6.2. Strain gauges installed on Radiator.

The adhesive used to attach the strain gauges to the test piece is HBM-X280.
It is a two component epoxy resin adhesive. The set-up is now placed in an
oven at 95°C for a duration of 1 hour. For accurate measurement of strains
there must be no gap between the material and the strain gauge. To achieve
this, for the strain gauges on the tube noose, silicon rubber is placed on the
strain gauges and heavy weights are placed on the rubber, while for the strain
gauges on flat side of the tubes, a special clamping device is used. This
process removes any gap between the material and the strain gauges. For the
final step in the installation of strain gauges, protective coating (HBM-
SG250) is applied on strain gauges.

29
A quarter bridge 3 wire circuit is used for the measurement of the change in
resistance. This type of circuit is used because it offers high measurement
sensitivity. The strain gauges mounted on the radiator along with the solder
terminal and the bridge circuit can be seen in Figure 6.3. The strain gauges
are then connected to the data acquisition unit, LMS Scadas recorder. It can
directly be connected to a personal computer for calibration and analysis
purposes.

Figure 6.3. Strain gauges with the solder terminal and bridge circuit.
The next step is to mount the radiator in the test bench. Air is removed from
the radiator before starting the test. This is done by filling the radiator with
the coolant while having a small pipe in the top of the radiator to remove air.
When there are no bubbles observed, it is concluded that there is no air in the
radiator. The resistance of the strain gauge after mounting and at the
connector are also measured. The Scadas recorder is connected to a computer
which has LMS Test Lab v14.8. The first step is to perform channel set up in
the software. In this step, the details of the strain gauges used, information
about the position of strain gauges are specified. An excitation voltage of 2V
is used for the bridge. A bandwidth of 32Hz is used and a sampling frequency
of 64Hz is used. The details of channel set-up analysis can be seen in Table
6.3.

30
 Table 6.3. Channel set-up analysis.

The second step is to perform calibration in the software. This is the most
important step and it is performed before starting the measurements. In the
calibration step, the offset for pressure and flow are calculated. In the offset
calculation, the mA (4mA to 20mA) signal of the test bench is converted into
voltage signal (2V to 10V), which is used by LMS Test Lab.
The third step is to perform bridge nulling and shunt calibration. Bridge
nulling sets all the values of the bridge to zero. After this, shunt calibration
is performed to establish a relation between the millivolt signal and strain.
The shunt resistor has a resistance of 80kΩ. A strain gauge with a shunt
resistance is shown in Figure 6.4. In Figure 6.4, Rs is the shunt resistance
and Rx is the gauge resistance.

Figure 6.4. Bridge with Shunt resistance.

The relation between the strain and the shunt resistance is as follows:
ோ
ߝ ൌ ீி‫כ‬ሺோభ ‫כ‬ோ ሻ ‫଺Ͳͳ כ‬ (6.1)
భ ೞ

where ܴଵ is the gauge resistance and ܴ௦ is the shunt resistance.

The strain measurements are performed on the radiator for the following flow
rates: 50 l/min, 70 l/min, 85 l/min, 100 l/min and 120 l/min. The breakaways

31
are cut in the beginning to facilitate uniform distribution of strain. As the test
is performed in a thermal cycling test bench, the temperature of the coolant
varies from hot cycle to cold cycle. The flow rate and the pressure are set to
the specified value. But it is observed that the flow rate changes to some
extent form the prescribed value during the hot and cold cycles of the test.
This is because of different density of the coolant at different temperatures.
The pressure of the coolant also varies because of this reason.
The test specifications are as follows:
x Inlet temperature during hot cycle: 95°C (±3°C)
x Inlet temperature during cold cycle: -10°C (±3°C)
x Coolant flow: 50 l/min, 70 l/min, 85 l/min, 100 l/min and 120 l/min.
x Pressure (relative): 1 bar (±0.2bar)
For each flow rate, the strain measurements are made for up to 10 cycles.
The strain measurements are saved in an asci format. The results obtained
from the strain measurements are discussed in the Chapter 6.3.1.

6.2 Wöhler Testing

For Wöhler testing, three tests are performed with different internal coolant
flows until failure where two radiators are used for each test. The test
specifications are same as that of strain measurements and the flow rates are
50 l/min, 70 l/min and 85 l/min. The set-up of Wöhler testing can be observed
in Figure 6.5.

Figure 6.5. Two radiators mounted in the test bench.

32
Objects 1&2 are tested until failure for a flow of 85 l/min, 3&4 for a flow of
50 l/min, 5&6 for a flow of 70 l/min. The breakaways are cut before starting
the test which can be observed in Figure 6.6.

Figure 6.6. Radiator with cut breakaways.

There is no automatic failure detection for small leakage. It is determined
manually by observing the radiator. In case of heavy leakage, a test bench
stopping system is available which activates when the coolant falls in the
plate below the radiator. The test data is acquired using a National instrument
logging system. The logging system stores a text file for each day of the test.
The number of cycles can be seen in the test bench control panel. The results
from the Wöhler testing will be discussed in Chapter 6.3.2.

6.3 Results from experimentation

6.3.1 Strain measurements

The results from strain measurements are obtained as data files which contain
strain values (με) at every position on the radiator wherever the strain gauges
are installed. LMS Tecware is used for analyzing the data. For each flow, the
data is considered for 10 cycles such that the flow rate is constant for all the
cycles. Maximum strain values are considered for each flow to compare
strains between different positions for a given flow rate which can be
observed in Table 6.4.

33
 Table 6.4. Maximum strain range.
Flow (l/min)

F50 F70 F85 F100 F105 F120 F50_ F85_new

Position (με) (με) (με) (με) (με) (με) new (με) (με)
EngIn_88_left 422.5 375.9 379.9 420.2 390.3 382.4 498.3 468.5
EngIn_88_right 344.8 388.2 425.7 452.4 462.2 489.0 333.6 443.9
EngOut_1_left 672.4 795.8 757.9 762.8 741.8 599.5 622.0 740.1
EngOut_1_right 1329 1496 1549 1523 1535 1389 1319 1502
EngOut_2_left 539.2 616.8 664.3 714.7 709.4 745.6 541.6 660.5
EngOut_2_right 639.1 833.7 815.6 762.9 800.2 562.6 588.8 770.8
EngOut_87_left 362.0 391.2 384.4 380.8 353.2 351.6 371.0 372.4
EngOut_87_rig
ht 543.3 599.5 637.6 684.8 695.1 747.0 519.9 621.0
EngOut_88_left 908.9 935.0 909.8 906.8 882.6 874.5 915.0 892.5
EngOut_88_rig
ht_7mm 298.9 253.1 215.5 220.4 195.5 203.3 301.5 216.4
EngOut_88_rig
ht_3mm 314.1 298.8 311.5 304.8 323.8 323.8 341.2 335.5
EngOut_88_rig
ht_11mm 302.8 273.8 250.9 259.2 227.7 238.2 309.2 247.7
EngOut_40_rig
ht 234.5 264.5
EngOut_40 171.7 274.8
FrOut_40 341.9 335.9
Pressure_inlet 50.8 34.9 36.3 54.1 42.3 57.2 48.8 44.1
Flow_outlet 32.7 22.0 16.3 20.5 23.7 30.6 29.5 26.2
Temp_inlet 102.8 104.2 106.6 106.6 107.4 106.7 103.5 106.7
Temp_outlet 94.7 99.0 102.4 102.9 104.0 103.8 96.2 101.6

The maximum strain values for all the positions are plotted as bar graphs as
presented in Figure 6.7.

34
 Strain range
1800.0
1600.0
1400.0
1200.0
1000.0
800.0
600.0
400.0
200.0
0.0

F50_10c F70_10c F85_10c F100_10c

F105_8c F120_10c F50_pos1_10c F85_pos1_10c

Figure 6.7. Strain Range.

It can be observed that tube 1 on Engine outlet side to the right of the tube
has the highest strain of 1329με. Measurements at tube 40 on the Engine side
and the grill side are performed only for two flow rates, 50 l/min and 85
l/min. From the strain data, the expected failure position from Wöhler test
would be the first tube on Engine outlet side.

6.3.2 Wöhler testing

As mentioned earlier, Wöhler test is carried out until failure. The test is
stopped when any leakage is detected.
a) Flow rate of 85 l/min:
For object 1, leakage is detected after 6675 cycles. The type of failure is
Tube-header leakage. The leakage positions are as follows:
x Engine side outlet tube 39,40,43,44 and Front side outlet tube 39, 42.
The failure positions can be observed in Figure 6.8.

35
 Figure 6.8. Failure positions of object 1.
For object 2, leakage is detected after 7380 cycles. The type of failure is
Tube-header leakage. The leakage position is as follows:
x Engine side outlet tube 40. The failure position can be observed in
Figure 6.9.

Figure 6.9. Failure position of object 2.

36
 b) Flow rate of 50 l/min:
For object 3, leakage is detected after 12240 cycles. The type of failure is
Tube-header leakage. The leakage position is as follows:
x Grill side outlet tube 1 (right hand flat side of tube).
For object 4, leakage is detected after 12054 cycles. The type of failure is
Tube-header leakage. The leakage position is as follows:
x Grill side outlet tube 1 (right hand flat side of tube). The failure
position can be observed in Figure 6.10.

Figure 6.10. Failure position of object 4.

c) Flow rate of 70 l/min:
For object 5, leakage is detected after 6794 cycles. The type of failure is
Tube-header leakage. The leakage position is as follows:
x Grill side outlet tube 39. The failure position can be observed in
Figure 6.11.

37
 Figure 6.11. Failure position of object 5.
For object 6, leakage is detected after 6794 cycles. The type of failure is
Tube-header leakage. The leakage position is as follows:
x Engine side outlet tube 49. The failure position can be observed in
Figure 6.12.

Figure 6.12. Failure position of object 6.

38
The number of cycles to failure for each object can be seen in Table 6.5.

Table 6.5. Cycles to failure of each object.

Object Flow rate Number of Failure Type
number (l/min) cycles

1 85 6675 Tube-header leakage

2 85 7380 Tube-header leakage
3 50 12240 Tube-header leakage
4 50 12054 Tube-header leakage
5 70 6794 Tube-header leakage
6 70 6794 Tube-header leakage
From the strain measurement results, the expected failure position is assumed
to be at the tube with the highest strain i.e., Engine side outlet tube 1, but
looking at the results from Wöhler test, the failure positions did not correlate
with the expected position. Only objects 3&4 fail at the first tube on the grill
side outlet. The failure positions of the other objects are almost at the center
tubes of the radiator.
To investigate the results, thermographic images of objects 5&6 are
produced using an infrared camera, shown in Figure 6.13 & Figure 6.14.
From the images, it can be observed that there is a temperature lag in the
failure position. This is because of clogging in the tubes with the
accumulation of dirt.

Figure 6.13. Failure position compared with thermographic image of

object 5.

39
 Figure 6.14. Failure position compared with thermographic image of
object 6.

The positions of failure of the two objects are shown in a red circle. Because
of the clogging, the flow rate is reduced at the failure tube which leads to
large temperature gradient. Due to this, shear stresses are produced which is
the reason failure is observed on the center tubes of the radiator instead of
the outermost tubes. The thermo-graphic images for objects 1&2 are not
observed, but looking at the failure positions, it is concluded that this is also
due to the clogging in the tubes of failure.

40
7 Development of tool
The tool is developed in MATLAB’s Graphical User Interface design
environment. It allows to create new GUI’s or edit the existing GUI’s
interactively from fig files. In this case, a new GUI is created that can import
data from all kinds of endurance tests, visualize time histories and produce
Level crossing curves for all the parameters in the data.

7.1 Data importing

The first step in the development of tool is handling and importing the data
from different test benches. The data pattern of each test bench is different
because of the variation in number of channels from one test bench to
another. Hence, different import functions are created, one for each test
bench using the default import tool in MATLAB. This allows to import
multiple files from the same test bench.
During the test, the logging system imports data and creates a text file at the
end of every day. If a test is carried out for 15 days, 15 data files are produced,
one for each day. The data consists of text at the beginning of the file,
followed by date, time and the parameters recorded at a sampling frequency.
This part of the developed tool handles multiple files from a single test bench,
imports data from all the files, ignoring the text at the beginning and produces
a concatenated file for the whole test which contains header of the test bench.
The header contains the name of the channels, its units, maximum and
minimum values and the sampling rate at which the data is recorded.
The examples of two logfiles from a test containing 5 samples each can be
seen in Table 7.1 and Table 7.2.

41
 Table 7.1. Sample data from day 1.
Thermal cycling test

Pressure in 1

Pressure in 2
Temperature

Temperature

Temperature

Temperature

Flow out 1

Flow out 2
outlet 1

outlet 2
inlet 1

inlet 2
Time
Date

20-3- 00:00 86.14 91.63 97.3 96.26 48.28 112.4 1.54 96.14
2017 :0.0 3
20-3- 00:00 86.04 91.46 97.4 96.23 48.32 112.3 1.34 95.79
2017 :2.0 7
20-3- 00:00 85.9 91.44 97.45 96.33 48.79 111.4 1.29 96.38
2017 :4.0 3
20-3- 00:00 85.59 91.38 97.5 96.41 49.12 111.8 1.45 96.53
2017 :6.0 7
20-3- 00:00 85.53 90.83 97.6 96.42 48.92 112.2 1.56 96.32
2017 :8.0 7

Table 7.2. Sample data from day 2.

Thermal cycling test
Pressure in 1

Pressure in 2
Temperature

Temperature

Temperature

Temperature

Flow out 1

Flow out 2
outlet 1

outlet 2
inlet 1

inlet 2
Time
Date

21-3- 00:00 83.73 89.84 -6.04 -5 53.64 128.8 1.65 84.01

2017 :0.0 1
21-3- 00:00 82.87 89.69 -6.15 -4.87 53.81 128.9 1.72 84.06
2017 :2.0
21-3- 00:00 82.76 89.58 -6.33 -4.92 54.3 128.8 1.56 84.31
2017 :4.0 3
21-3- 00:00 85.63 89.45 -6.39 -4.88 54.12 130.4 1.63 84.19
2017 :6.0 9
21-3- 00:00 85.39 89.37 -6.49 -4.77 53.99 129.9 1.58 84.69
2017 :8.0

42
It can be observed from the Table 7.1 and Table 7.2 that the sampling
frequency at which the data is recorded is 2 seconds.
The data importing part of the developed tool handles these files, imports
data and produces a concatenated file which can be seen in Table 7.3.

Table 7.3. Concatenated data of the whole test.

BEGIN
CHANNELNAME = ['Temp in LV1','Temp out LV1','Temp in LV2','Temp out
LV2','Presure LV1','Pressure LV2','Flow LV1','Flow LV2']
MINIMUM = [-20,-20,-20,-20,-20,-20,0,0]
MAXIMUM = [150,150,150,150,400,400,200,200]
UNIT=['°C','°C','°C','°C','kPa','kPa','l/min','l/min']
START = 0.0
RATE = 0.5
END
Pressure in 1

Pressure in 2
Temperature

Temperature

Temperature

Temperature

Flow out 1

Flow out 2
Outlet 1

Outlet 2
Inlet 1

Inlet 2
Time

0 86.14 91.63 97.3 96.26 48.28 112.43 1.54 96.14

2 86.04 91.46 97.4 96.23 48.32 112.37 1.34 95.79
4 85.9 91.44 97.45 96.33 48.79 111.43 1.29 96.38
6 85.59 91.38 97.5 96.41 49.12 111.87 1.45 96.53
8 85.53 90.83 97.6 96.42 48.92 112.27 1.56 96.32
10 83.73 89.84 -6.04 -5 53.64 128.81 1.65 84.01
12 82.87 89.69 -6.15 -4.87 53.81 128.9 1.72 84.06
14 82.76 89.58 -6.33 -4.92 54.3 128.83 1.56 84.31
16 85.63 89.45 -6.39 -4.88 54.12 130.49 1.63 84.19
18 85.39 89.37 -6.49 -4.77 53.99 129.9 1.58 84.69

43
In Table 7.3, the starting text is known as header, which varies from one test
bench to another. The original date and time formats are removed and a new
time vector is inserted with the same sampling frequency but is simplified in
such a way that it is supported by MATLAB.

7.2 Level Crossing Cycle Counting

7.2.1 Implementation of Level Crossing Cycle Counting

An algorithm is developed for the Level crossing cycle counting which

works based on the Markov cycle counting procedure. Before implementing
the method to the real data, synthetic signals are considered whose LCCs are
compared to those obtained from LMS Tecware software. The approach of
producing LCCs in LMS Tecware is Rainflow cycle counting method.
The following is the algorithm developed for producing LCC.
x Initially, the turning points (peaks and valleys) of the signal are
determined using extrema function [5].
x The next step is to define the Markov pairs. Starting with the first
turning point (TP), the first and second turning point makes one pair,
second and third makes another one and it is continued until the last
value.
x Once the Markov pairs are defined, the unique matrix of the turning
points is considered which is a singleton set of the turning points
matrix, arranged in the ascending order.
x The values of the Unique matrix are found out in the first column of
Markov matrix, arranged in the order of Unique matrix.
x The respective pairs corresponding to these values are determined in
the second column of Markov matrix, named as Unique pairs and
arranged in the same order.
x Starting with the first value of Unique matrix, the respective Unique
pairs are compared, if they are greater than or equal to the next value
of Unique matrix. If so, one cycle is counted for every such positive
case.
x Likewise, the procedure is carried out for all the values in the Unique
matrix. These gives ‘Upcrossings’.

44
 x The Unique matrix is considered in the reverse order.
x The values in this matrix are found out in the first column of Markov
matrix. The respective pairs corresponding to these values are
determined in the second column of Markov matrix and are
compared, if they are less than or equal to the next value of Unique
matrix in reverse order. If so, one cycle is counted for every such
positive case.
x Likewise, the procedure is carried out for all the values in the reverse
Unique matrix. These gives ‘Downcrossings’.
x The Upcrossings and Downcrossings together gives Level Crossing
curve of the signal.
This algorithm is applied for synthetic signals to produce LCC’s and
compare them to LCC’s obtained from LMS Tecware.
The following are the two synthetic signals considered.
Signal 1: A sine wave is considered with two different amplitudes, four
cycles each as shown in Figure 7.1.

Figure 7.1. Synthetic signal 1.

The algorithm developed is applied and the Level crossing curve of the signal
is produced as shown in Figure 7.2.

45
 Figure 7.2. Level crossing curve for synthetic signal 1 from MATLAB.
The LCC of the signal is compared to that obtained from LMS Tecware,
shown in Figure 7.3.

Figure 7.3. Level crossing curve for synthetic signal 1 from LMS.
It can be observed from Figure 7.2 & Figure 7.3 that the LCC obtained from
the developed algorithm is exactly equal to that obtained from LMS Tecware.
Hence, the algorithm is generalized and applied for another signal.

Signal 2: A sine wave is considered with two different amplitudes, ten cycles
each as shown in Figure 7.4.

46
 Figure 7.4. Synthetic signal 2.
The Level crossing curve of the signal obtained using the developed
algorithm is shown in Figure 7.5.

Figure 7.5. Level crossing curve for synthetic signal 2 from MATLAB.
The LCC of the signal is compared to that obtained from LMS Tecware,
shown in Figure 7.6.

47
 Figure 7.6. Level crossing curve of synthetic signal 2 from LMS.
It can be observed from Figure 7.5 & Figure 7.6 that the LCC of the signal
obtained from the generalized algorithm is exactly equal to that computed
from LMS Tecware.

7.2.2 Real data from test bench

With the algorithm working for synthetic signals, the next step is to apply it
to real data obtained from the test bench. The real data obtained from the
thermal cycling test bench are shown in Figure 7.7, Figure 7.8 & Figure 7.9.
From thermal cycling tests, the data of temperature, flow and pressure is
obtained.
The time history of temperature for one day of data is shown in Figure 7.7.

48
 Figure 7.7. Time history of temperature (1 day).
The time history of pressure for one day of data is shown in Figure 7.8.

Figure 7.8. Time history of pressure (1 day).

The time history of flow for one day of data is shown in Figure 7.9.

49
 Figure 7.9. Time history of flow (1 day).
From these figures, it can be observed that the data obtained from the tests
looks like a lump of mass. The same data for a limited number of cycles is
shown in Figure 7.10, Figure 7.11 & Figure 7.12.

Figure 7.10. Time history of temperature (5 cycles).

50
 Figure 7.11. Time history of pressure (5 cycles).

Figure 7.12. Time history of flow (5 cycles).

The interpretation of one day data by looking at the time histories is difficult.
By looking at a limited number of cycles, there is a better understanding of
the process undergoing in the test bench. By producing the level crossing
curves of the time histories, a much better understanding of the test data can
be obtained.

51
7.2.3 Level Crossing counting for Real data

The generalized algorithm is applied for the real data obtained from the test
bench for parameters such as temperature, pressure etc. The level crossing
results obtained were observed to be wrong. The algorithm is investigated
and observed that the determination of turning points (TPs) doesn’t work
correctly using extrema function in case of complex signals like Figure 7.10,
Figure 7.11 & Figure 7.12.
When these complex signals are compared with the simple synthetic signals,
it can be observed that there are a numerous local maxima and minima which
makes the extrema function difficult to determine the exact global maxima
and minima. To overcome this problem, the signal processing toolbox in
MATLAB is used to determine the TPs.
The signal processing toolbox in MATLAB works based on a function
named as ‘findpeaks.’ This is implemented in the algorithm leaving the initial
method of determination of peaks using extrema function and the level
crossing curves are produced for the real data.
The Level crossing curves obtained for the time histories Figure 7.10, Figure
7.11 & Figure 7.12 are shown in Figure 7.13Figure 7.14 & Figure 7.15.

Figure 7.13. Level crossing curve of Temperature history.

52
The level crossing spectrum obtained for pressure history is shown in Figure
7.14.

Figure 7.14. Level crossing curve of Pressure history.

The level crossing spectrum obtained for flow history is shown in Figure
7.15.

Figure 7.15. Level crossing curve of Flow history.

53
7.3 Graphical user interface (GUI)
A graphical user interface is created using MATLAB’s graphical user
interface design environment. As there are eight different test benches, eight
push buttons are created representing each test bench. Different pushbuttons
are added for each different test bench which show the time-histories of the
respective channel of the test bench. Push buttons are also generated for
plotting the level crossing spectrum of each channel of a test bench. The GUI
developed can be seen in the Figure 7.16.
Using the GUI developed, the level crossing curves are generated for all the
objects in Wöhler testing. The results can be seen in Appendix B.

Figure 7.16. GUI of tool developed.

54
8 Investigation of Basquin’s Coefficient
The Basquin’s coefficient is estimated from the results of strain
measurements together with those obtained from Wöhler testing. From the
Wöhler test, the number of cycles to failure obtained from each flow are
considered and from the strain measurements, the highest strain positions are
considered. Based on these results, Basquin’s coefficient is determined and
S-N curve is constructed.
From the results of Wöhler testing, the failure positions of both objects for
each flow are as follows: The failure location for a flow of 50 l/min is front
side, outlet tube 1, for a flow of 85 l/min the failure positions are on engine
side, outlet tubes 39,40,43,44 and front side, outlet tubes 39, 42, for a flow
of 70 l/min the failure positions are engine side, outlet tube 49 and grill side,
outlet tube 39. From the failure positions and the strain measurements it is
observed that strains at all the failure positions are not available. There is not
enough data to plot the S-N curve as the strains on all the failure positions
are not known. So, an evaluation from the available data is made in this study.
The highest stresses induced are calculated from the strain data. The average
number of cycles to failure for each flow rate can be seen in Table 8.1.

Table 8.1. Wöhler results.

S.no Flow Number of Average number
(l/min) cycles to of cycles
failure
1 85 6675 7028
2 85 7380
3 50 12240 12147
4 50 12054
5 70 6794 6794
6 70 6794

As the highest strains are used for the estimation of Basquin’s coefficient,
the strains recorded on Engine side (outlet) tube 1 right side are considered
for finding the stress values. The above tube is chosen because it has the

55
highest strain values when the maximum ranges of strain are observed (in
Figure 6.7).
The stress (ߪ) is found out using Hooke’s law. The elastic modulus, E of
Aluminum is 69 GPa.
ߪ ൌ ߝ‫ܧכ‬ (8.1)
where ߝ is strain and E is the modulus of elasticity
The values of flow, strain, stress and the number of cycles are shown in Table
8.2.

Table 8.2. Stress and strain values for each flow.

S.no Position Flow Strain Stress Average
(l/min) (micro (MPa) number
strain) of cycles
1 EngOut_1_right 50 1319.6 91.05 12147
2 EngOut_1_right 70 1496.3 103.245 6794
3 EngOut_1_right 85 1549.2 106.895 7028

The Basquin’s coefficient is found out from the Basquin’s relation which
provides an analytical expression for the S-N curve.
ܰ ൌ ߙ ‫ି ܵ כ‬ఉ (8.2)
where ߙ is a constant depending on material & ߚ is the Basquin’s coefficient
Using this analytical expression, the value of β is found out.
The estimation of Basquin’s coefficient is done by considering the flows 50
l/min & 85 l/min in one case, 50 l/min & 70 l/min in the other case.
For case (i)
The equation for Basquin’s coefficient is given by,
ܰ ൌ ߙ ‫ି ܵ כ‬ఉ
For the flow 50 l/min, the equation can be written as,
ͳʹͳͶ͹ ൌ ߙ ‫ כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ Æ (1)

56
For the flow 85 l/min, the equation can be written as,
͹Ͳʹ͹ ൌ ߙ ‫ כ‬ሺͳͲ͸Ǥͻ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺ͹Ͳʹ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͳͲ͸Ǥͻ ‫ ଺Ͳͳ כ‬ሻ Æ (2)

Solving equations (1) and (2),

Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ
Ž‘‰ሺ͹Ͳʹ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͳͲ͸Ǥͻ ‫ ଺Ͳͳ כ‬ሻ
ͳʹͳͶ͹ ͳͲ͸Ǥͻ ‫଺Ͳͳ כ‬
Ž‘‰ ൬ ൰ ൌ ߚ ‫ ‰‘Ž כ‬ቆ ቇ
͹Ͳʹ͹ ͻͳǤͲͷ ‫଺Ͳͳ כ‬
ͲǤʹ͵ͺ ൌ ߚ ‫Ͳ כ‬ǤͲ͸ͻ
ͲǤʹ͵ͺ
ߚൌ ؆ ͵ǤͶ
ͲǤͲͲ͸ͻ
ߚ ؆ ͵ǤͶ
For case (ii)
The equation for Basquin’s coefficient is given by,
ܰ ൌ ߙ ‫ି ܵ כ‬ఉ
For the flow 50 l/min, the equation can be written as,
ͳʹͳͶ͹ ൌ ߙ ‫ כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ Æ (1)
For the flow 70 l/min, the equation can be written as,
͸͹ͻͶ ൌ ߙ ‫ כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺ͸͹ͻͶሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻ Æ (2)

Solving equations (1) and (2),

Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ
Ž‘‰ሺ͸͹ͻͶሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻ
ͳʹͳͶ͹ ͳͲ͵Ǥʹͷ ‫଺Ͳͳ כ‬
Ž‘‰ ൬ ൰ ൌ ߚ ‫ ‰‘Ž כ‬ቆ ቇ
͸͹ͻͶ ͻͳǤͲͷ ‫଺Ͳͳ כ‬

57
 ͲǤʹͷʹ ൌ ߚ ‫Ͳ כ‬ǤͲͷͷ
ͲǤʹͷʹ
ߚൌ ؆ ͶǤͷͺ
ͲǤͲͷͷ
ߚ ؆ ͶǤ͸
The Wöhler curve is plotted for the two cases mentioned above. It consists
of the stress on the Y axis and the number of cycles to failure on X axis in
logarithmic scale. The obtained curves are shown in Figure 8.1.

Figure 8.1. Wöhler curve (log-log scale).

The Wöhler curve in semi-logarithmic scale is shown in the Figure 8.2.

58
 Figure 8.2. Wöhler curve (semi-log scale).

As there was clogging observed in the tubes of objects 5 &6 there was a
premature failure observed in both objects. The failure position of object 3&4
corresponds to failure at expected position (outer tubes). Also, for a flow of
85 l/min, there are no thermographic images. Assuming there is no clogging
in objects 1&2, the number of cycles to failure for the flow 70 l/min is
calculated from the Basquin’s coefficient obtained in case (i).
From case (i), the value of Basquin’s coefficient obtained is 3.4.
For the flow 50 l/min, the equation 8.2 can be written as,
ͳʹͳͶ͹ ൌ ߙ ‫ כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ͵ǤͶ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ
݈‫ ߙ݃݋‬ൌ Ž‘‰ሺͳʹͳͶ͹ሻ ൅ ͵ǤͶ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ
Ž‘‰ ߙ ൌ ͵ͳǤͳͶ
This ߙ value is used to find the number of cycles to failure for a flow of 70
l/min. As the values of stress, ߙ and ߚ are known, the actual number of cycles
to failure is estimated.
For the flow 70 l/min, the equation 8.2 can be written as,
ܰ ൌ ߙ ‫ כ‬ሺܵሻିఉ

59
 ܰ ൌ ሺͳͲሻଷଵǤଵସ ‫ כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻିଷǤସ
ܰ ൌ ͹ͺͳʹ
Using the number of cycles to failure from the above relation for a flow of
70 l/min, the Basquin’s coefficient value for case (ii) is estimated again.
For the flow 50 l/min, the equation can be written as,
ͳʹͳͶ͹ ൌ ߙ ‫ כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ Æ (1)
For the flow 70 l/min, the equation can be written as,
͹ͺͳʹ ൌ ߙ ‫ כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻିఉ
Ö Ž‘‰ሺ͹ͺͳʹሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻ Æ (2)

Solving equations (1) and (2),

Ž‘‰ሺͳʹͳͶ͹ሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͻͳǤͲͷ ‫ ଺Ͳͳ כ‬ሻ
Ž‘‰ሺ͹ͺͳʹሻ ൌ ݈‫ ߙ݃݋‬െ ߚ ‫‰‘Ž כ‬ሺͳͲ͵Ǥʹͷ ‫ ଺Ͳͳ כ‬ሻ
ͳʹͳͶ͹ ͳͲ͵Ǥʹͷ ‫଺Ͳͳ כ‬
Ž‘‰ ൬ ൰ ൌ ߚ ‫ ‰‘Ž כ‬ቆ ቇ
͹ͺͳʹ ͻͳǤͲͷ ‫଺Ͳͳ כ‬
ͲǤͳͻͳ ൌ ߚ ‫Ͳ כ‬ǤͲͷͷ
ͲǤͳͻͳ
ߚൌ ؆ ͵ǤͶ͹
ͲǤͲͷͷ
ߚ ؆ ͵Ǥͷ

The Basquin’s coefficient obtained from case (i) can be considered as

approximate or close to the optimal value whereas the result obtained from
case (ii) cannot be considered because there were impurities or dirt in the
tubes 39-42 for the test of flow 70 l/min. This leads to the clogging of the
flow in those tubes and results in premature failure of the specimen without
reaching the fatigue limit. When considering the clogging and estimating the
actual number of cycles to failure for a flow of 70 l/min, the value of
Basquin’s coefficient is found out to be 3.5.

60
9 Results
A tool has been developed in MATLAB to check if the test has been
performed according to the targeted level. The following are the functions of
the tool.
x Different functions are generated for importing the data for different
test benches.
x The tool will handle the data files using these functions, imports
multiple files and creates a concatenated file.
x The tool shows time histories of all the channels of the data such as
temperature, pressure, flow, etc. to see how the parameters are
varying with time.
x An algorithm is developed for Level crossing counting that works
based on Markov cycle counting method.
x The algorithm produces Level crossing spectra for all the channels of
the data which is used for the interpretation of the test.
x The Level crossing spectra also shows the number of cycles of the
test. For example, in the Figure 9.1, the time history of the
temperature is solid mass and is hard to interpret the behavior of the
temperature. Whereas, in the Figure 9.2, the Level crossing of the
time history of the temperature shows that there were 240 load cycles
of the temperature ranging from -10 ˚C to 100 ˚C.

61
 Figure 9.1. Temperature history of one day data.

Figure 9.2. Level crossing spectrum of Temperature data for one day.
x GUI is generated by integrating all the individual codes for importing
the data, algorithms for Markov cycle counting and Level crossing in
a structured manner. Different push buttons are implemented for
importing the data for different test benches. Every channel of a test
bench, two push buttons are implemented for producing time
histories and Level crossings.

62
x Strain measurements are performed for three different flows for 10
cycles. Flows: 50 l/min, 70 l/min and 85 l/min. It was observed that
the most fatigue tube is Engine side outlet tube 1 because of the
highest strains recorded at that location.
x From the Wöhler tests for the same flows as that of the strain
measurements, the failure locations did not coincide with the highest
strain position because of the clogging of dirt in the tubes at the
failure positions. This leads to the premature failure of the tubes.
x Basquin’s coefficient which is determined from the strain
measurements and Wöhler testing is found to be 3.4. By finding the
actual number of cycles to failure for the flow of 70 l/min, the optimal
value of Basquin’s coefficient is found to be 3.5

63
10 Summary and Conclusions
In TitanX Engine Cooling, various tests are performed when introducing a
new heat-exchanger into the market. The result of the test is a logfile
produced in the logging system that contains data of the channels such as
temperature, pressure, flow etc. The analysis of the test data is difficult and
its interpretation is very important.
The research in the thesis has accomplished this by developing a tool in
MATLAB for handling the test data. It performs cycle counting methods:
Markov cycle counting and Level crossing counting. Finally, Level crossing
spectra are produced for all the parameters in the test data that shows how
well a test has been carried out.
The optimal value of basquin’s coefficient is determined from the results of
strain measurements and Wöhler testing and is found to be 3.5 for a pure
radiator.

64
11 Future works
The estimation of Basquin’s coefficient is based on the highest strain
obtained in the tubes and not on the failure positions. As the strain gauges
cannot be placed on each tube of the radiator, a finite element model of the
radiator could help to achieve an accurate Basquin’s coefficient.
Wöhler testing on the whole radiator module can also be performed to know
how the Basquin’s coefficient varies when compared to a pure radiator.
To consider the effect of clogging on the overall failure position of the
radiator, strain gauges can be installed on the tubes of the radiator at which
failure occurs to record the strain range. A theory can be estimated that higher
strains will be recorded at the position near failure than the outermost tubes
of the radiator.

65
12 References
[1] R. L. Hannah, S. E. Reed, and Society for experimental mechanics,
Strain gage users’ handbook. Chapman & Hall, 1992.
[2] “Strain gauge coding system|Strain gauges|Products|Tokyo Sokki
Kenkyujo Co., Ltd.” [Online]. Available:
http://www.tml.jp/e/product/strain_gauge/gaugetype.html. [Accessed:
05-Oct-2017].
[3] P. Johannesson, M. Speckert, SP Bygg och Mekanik, SP – Sveriges
Tekniska Forskningsinstitut, and RISE, Guide to Load Analysis for
Durability in Vehicle Engineering, 1st ed. Somerset: John Wiley & Sons,
Incorporated, 2013.
[4] Y.-L. Lee, J. Pan, and R. Hathaway, Fatigue Testing and Analysis.
Butterworth-Heinemann, 2011.
[5] “extrema.m, extrema2.m - File Exchange - MATLAB Central.”
[Online]. Available:
http://se.mathworks.com/matlabcentral/fileexchange/12275-extrema-m-
-extrema2-m. [Accessed: 06-Dec-2017].

66
13 Appendix
A. Numbering of tubes on radiator

Engine side

Tubes (1-88)

Figure 13.1. Engine side (Tubes 1-88).

B. Level crossing curves from Wöhler Testing

The level crossing curves obtained from the test data in the thermal cycling
test bench for all the six objects.
For Object 1 (Flow-85 l/min):

67
Figure 13.2. LCC of Inlet and Outlet Temperature (Object 1).

Figure 13.3. LCC of Pressure (Object 1).

Figure 13.4. LCC of Flow (Object 1).

68
For Object 2 (Flow-85 l/min):

Figure 13.5. LCC of Inlet and Outlet Temperature (Object 2).

Figure 13.6. LCC of Pressure (Object 2).

Figure 13.7. LCC of Flow (Object 2).

69
For Object 3 (Flow-50 l/min):

Figure 13.8. LCC of Inlet and Outlet Temperature (Object 3).

Figure 13.9. LCC of Pressure (Object 3).

Figure 13.10. LCC of Flow (Object 3).

70
For Object 4 (Flow-50 l/min):

Figure 13.11. LCC of Inlet and Outlet Temperature (Object 4).

Figure 13.12. LCC of Pressure (Object 4).

Figure 13.13. LCC of Flow (Object 4).

71
For Object 5 (Flow-70 l/min):

Figure 13.14. LCC of Inlet and Outlet Temperature (Object 5).

Figure 13.15. LCC of Pressure (Object 5).

Figure 13.16. LCC of Flow (Object 5).

72
For Object 6 (Flow-70 l/min):

Figure 13.17. LCC of Inlet and Outlet Temperature (Object 6).

Figure 13.18. LCC of Pressure (Object 6).

Figure 13.19. LCC of Flow (Object 6).

73
Department of Mechanical Engineering Telephone: +46 455-38 50 00
Blekinge Institute of Technology E-mail: [email protected]
SE-371 79 Karlskrona, SWEDEN