Professional Documents
Culture Documents
By
A dissertation submitted to the Faculty of Engineering and the Built Environment in partial
fulfilment of the degree of
MASTER OF ENGINEERING
In
At the
UNIVERSITY OF JOHANNESBURG
SUPERVISOR:
Dr F.F. Pieterse
CO-SUPERVISOR:
MAY 2015
Declaration
I, Jules David de Ponte, hereby declare that this dissertation is wholly my own work and has
not been submitted anywhere else for academic credit by myself or another person. I
understand what plagiarism implies and declare that this dissertation is my own ideas, words,
phrases, arguments, graphics, figures, results and organisation except where reference is
explicitly made to another’s work.
I understand further that any unethical academic behaviour, which includes plagiarism, is
seen in a very serious light by the University of Johannesburg and is punishable by
disciplinary action.
i
“If we knew what we were doing, it wouldn’t be called research, would it?”
– Albert Einstein
ii
Abstract
The requirements which a wind tunnel balance must meet are becoming increasingly
stringent. The wind tunnel testing community is calling for balances which; offer higher
resolution, are stiffer, are immune to electro-magnetic interference (EMI) and provide
compensation for thermal effects. It is proposed that in order to meet these requirements,
balance design philosophy needs to be further expanded to include different manufacturing
methods, materials and sensor technology. This study investigates the design and
development of a six component wind tunnel platform balance incorporating Optical fibre
Bragg grating (OFBG) sensors.
Under an applied load, conventional balances measure strain at the surface of a material, by
means of a foil strain gauge. For this reason, sections of material in the balance are purposely
made thin, in order for the strain in that section to be sufficiently high to offer adequate
resolution. This may compromise the stiffness of the balance.
A platform balance is designed which incorporates OFBG sensors using the two-groove
method of strain measurement. The optical fibres are spanned between two probes. One probe
protrudes from the metric end of the balance, the other protrudes from the fixed end. Under
an applied load, the gap between the two protrusions will change, which will induce a strain
in the fibre Bragg grating spanning it. This strain in the fibre Bragg grating will cause the
Bragg wavelength to shift proportionally to the magnitude of the strain in the fibre.
Therefore, the balance is designed around the idea of measuring displacements within the
structure of the balance. This displacement is comparatively larger than the deformation at
the surface of a material. Therefore, strain induced in the fibre spanning the gap would be
larger than the strain at the surface of the material. The balance to which these fibres are
bonded can be made stiffer, while still offering a comparable relative resolution.
The two-groove method uses two fibres to measure one load component. Each fibre is part of
a pair for strain measurement. Each fibre Bragg grating has a different reference Bragg
wavelength. These are spanned across two separate gaps. The balance has been designed such
that, under an applied load, one fibre of a strain measurement pair experiences a tensile strain,
and the other experiences a compressive strain. The final output is determined by calculating
the change in difference between the two fibres’ respective Bragg wavelengths. This method
compensates for unwanted force and thermal interactions.
iii
The balance design was analysed using the finite element method (FEM). In the FEM study,
the design shows a high degree of sensitivity and stiffness. The FEM also indicated that the
design would have small interactions for all components. The design was simple and
affordable to manufacture. The decision to manufacture the balance was made in light of the
FEM analysis.
The balance was manufactured and tested. In the load tests, the balance was loaded to a
quarter of the design loads to avoid damaging it or the fibres. The output of the balance was
linear for each component individually. The repeatability of the balance was within 0.868%
for all of the components (based on a 2𝜎 standard deviation from the mean output), under the
application of the load set to which the balance was subjected. This value of repeatability is
larger than is required by the wind tunnel testing community. The sources of the uncertainty
in the measurements are attributed to misalignments in the bonding of the fibres, loading
errors and the phenomenon of fibre creep. When a load is applied or removed, the output of
the fibre tends to drift with a time dependent logarithmic trend, which adds to the uncertainty
of the output. The interactions of the physical balance were larger than predicted by the FEM.
This is due to the challenges of aligning and bonding the fibres with the required degree of
accuracy. A large interaction was due primarily to a manufacturing error on one of the
platform’s supporting members.
The balance was designed to prove the concept that a wind tunnel balance can be
manufactured using optical fibre sensors to measure the loads. It is concluded that the balance
is cost effective to manufacture and the use of optical fibre sensors offer the benefit of EMI
immunity. The balance is not yet ready to be used in practice; its repeatability is not within
the range sought by the wind tunnel testing community. The fibre creep characteristic must
be addressed and eliminated in future designs. Nevertheless, this balance does serve as a
useful research platform to continue the work in the field of optical fibres and their potential
use in wind tunnel balances.
iv
Acknowledgements
I would like to extend my gratitude to the following people, without whom this project would
not be possible.
v
Table of Contents
Declaration................................................................................................................................. i
Acknowledgements .................................................................................................................. v
vi
2.5 Summary ................................................................................................................................................ 44
5.2 Locating the Bragg grating and bonding the fibre .................................................................................. 90
vii
6.2 Signal noise band ................................................................................................................................. 105
Appendix E - Bonding of the Fibres, and Acrylate Coated Fibres .................................. E-1
E.1 Acrylate fibres ......................................................................................................................................... E-1
E.3 Observations made during the bonding of the fibres ............................................................................... E-5
ix
List of Figures
xiii
Figure B-11 – Yawing moment load in FEM model ........................................................... B-30
Figure B-12 – Isometric view of the mesh........................................................................... B-32
Figure B-13 – Normal force von Mises Stress..................................................................... B-33
Figure B-14 – Normal force deformation ............................................................................ B-34
Figure B-15 – Normal force displacement, front view ........................................................ B-35
Figure B-16 – Axial force von Mises Stress ........................................................................ B-35
Figure B-17 – Axial force displacement .............................................................................. B-36
Figure B-18 – Axial force displacement, front view ........................................................... B-36
Figure B-19 – Side force von Mises Stress .......................................................................... B-37
Figure B-20 – Side force displacement ................................................................................ B-38
Figure B-21 – Side force displacement, left view................................................................ B-38
Figure B-22 – Pitching moment von Mises Stress............................................................... B-39
Figure B-23 – Pitching moment displacement..................................................................... B-40
Figure B-24 – Pitching moment displacement, front view .................................................. B-40
Figure B-25 – Rolling moment von Mises Stress ................................................................ B-41
Figure B-26 – Rolling moment displacement ...................................................................... B-41
Figure B-27 – Rolling moment displacement, right view.................................................... B-42
Figure B-28 – Yawing moment von Mises Stress ............................................................... B-42
Figure B-29 – Yawing moment displacement ..................................................................... B-43
Figure B-30 – Yawing moment displacement, front view ................................................... B-43
Figure B-31 – Fixture load for all of the load studies .......................................................... B-49
Figure B-32 – Fixture condition for the pitching moment load study ................................. B-50
Figure B-33 – Fixture condition for the normal force load study ........................................ B-50
Figure B-34 - Side force load............................................................................................... B-51
Figure B-35 – Pitching moment load ................................................................................... B-52
Figure B-36 – Normal force load ......................................................................................... B-53
Figure B-37 – Isometric view of the calibration body mesh ............................................... B-54
Figure B-38 – Side force von Mises stress .......................................................................... B-55
Figure B-39 – Side force deformation ................................................................................. B-55
Figure B-40 – Pitching moment von Mises stress ............................................................... B-56
Figure B-41 – Pitching moment deformation ...................................................................... B-56
Figure B-42 - Normal force von Mises stress ...................................................................... B-57
Figure B-43 - Normal force deformation ............................................................................. B-57
Figure E 1 – Acrylate fibre bonded to 17-4PH using X-60, after being prepared by
cyanoacrylate super glue. ....................................................................................................... E-3
xiv
Figure E 2 – Output of the fibre with the constant pre-tension load applied ......................... E-4
Figure F 1 – The proposed sensor pair to measure the normal force ..................................... F-1
Figure F 2 - Normal force output was different for different loads ....................................... F-2
Figure F 3 - Balance with the retest of the normal force, loaded to 400kg ........................... F-3
Figure F 4 - Normal force retest ............................................................................................ F-4
xv
List of Tables
Table 2-1 – Resolution and accuracy comparison between sensors [18] ................................ 36
Table 2-2 – Signal performance of the sensors ........................................................................ 38
Table 3-1 – Design loads ......................................................................................................... 46
Table 3-2 – Dimension requirements ....................................................................................... 47
Table 3-3 – Concept 6 OFBG sensor wavelength shift [nm]................................................... 51
Table 3-4 – Concept 6 interactions [% of FS] ......................................................................... 52
Table 3-5 – Concept 6 maximum stress per load ..................................................................... 52
Table 3-6 – Concept 7 OFBG sensor wavelength shift [nm]................................................... 62
Table 3-7 – Concept 7 interactions [% of FS] ......................................................................... 62
Table 3-8 – Concept 7 maximum stress per load ..................................................................... 63
Table 3-9 – Comparison of sensor sensitivities between Concept 6 and Concept 7 ............... 63
Table 3-10 – Selection criteria weighting ................................................................................ 66
Table 3-11 – Concept selection – Manufacturability (4.5) ...................................................... 67
Table 3-12 – Concept selection – Low interactions (3) ........................................................... 67
Table 3-13 – Concept selection – Sensitivity and resolution (3) ............................................. 67
Table 3-14 – Concept selection – Strength (2.5) ..................................................................... 67
Table 3-15 – Concept selection – Stiffness (2) ........................................................................ 68
Table 3-16 – AISI 4340 material composition [34] ................................................................. 71
Table 3-17 – Mechanical properties of AISI 4340 [34] ........................................................... 71
Table 3-18 – 17-4 PH material composition [35] .................................................................... 72
Table 3-19 – Mechanical properties of 17-4 PH condition H900 [35] .................................... 72
Table 3-20 – Relative resolution .............................................................................................. 77
Table 5-1 – Equations inputted into the DAQ to give the desired output for each load
component ................................................................................................................................ 95
Table 6-1 – Load point repeatability of normal force sensor, n = 2, t = 12.71 ...................... 111
Table 6-2 – Hysteresis for medium normal force loads, averaged over two load/unload cycles
................................................................................................................................................ 112
Table 6-3 – Load Point Repeatability of axial force sensor, n = 4, t = 3.18 [38] .................. 114
Table 6-4 – Hysteresis, averaged over four load/unload cycles ............................................ 115
Table 6-5 – Load Point Repeatability of side force sensor, n=3, t=4.3 ................................. 117
Table 6-6 – Hysteresis, averaged over three load/unload cycles ........................................... 118
Table 6-7 – Load point repeatability of pitching moment sensor, n = 4, t = 3.18 [38].......... 121
xvi
Table 6-8 – Hysteresis for the medium pitching moment loads, averaged over four
load/unload cycles .................................................................................................................. 122
Table 6-9 – Load point repeatability of rolling moment sensor, n = 4, t = 3.18 [38] ............ 125
Table 6-10 – Hysteresis for the medium rolling moment loads, averaged over four
load/unload cycles .................................................................................................................. 126
Table 6-11 – Load point repeatability of yawing moment sensor, n = 4, t = 3.18 [38] ......... 129
Table 6-12 – Hysteresis for the medium yawing moment loads, averaged over four
load/unload cycles .................................................................................................................. 130
Table 6-13 – Interaction data for the balance [% FS] ............................................................ 131
Table 6-14 – Platform pitch (or roll) under the application of an axial (or side) force ......... 133
Table 6-15 – Repeatability ..................................................................................................... 133
Table 6-16 – Summary of results ........................................................................................... 142
xvii
Table B-8 – Displacement of gaps, due to moments [mm] ................................................. B-45
Table B-9 – Individual fibre strain, due to forces [µε] ........................................................ B-45
Table B-10 – Individual fibre strain, due to moments [µε] ................................................. B-45
Table B-11 – Resultant fibre strain [µε] .............................................................................. B-46
Table B-12 – Bragg wavelength shift [nm] ......................................................................... B-46
Table B-13 – Interactions [% of FS] .................................................................................... B-46
Table B-14 – Relative resolution ......................................................................................... B-47
Table B-15 – Material properties used in the calibration body FEM study......................... B-48
Table B-16 – Calibration body mesh information ............................................................... B-53
Table C-1 – Normal force, medium loads, raw data .............................................................. C-1
Table C-2 - Axial force, medium loads, raw data .................................................................. C-2
Table C-3 – Side force, medium loads, raw data ................................................................... C-4
Table C-4 – Pitching moment, medium loads, raw data ........................................................ C-5
Table C-5 – Rolling moment, medium loads, raw data ......................................................... C-7
Table C-6 – Yawing moment, medium loads, raw data ........................................................ C-9
xviii
List of Nomenclature
xix
List of Abbreviations
Abbreviation Description
ABS Acrylonitrile Butadiene Styrene
AIAA American Institute of Aeronautics and Astronautics
BCE Back Calculated Error
BMC Balance Moment Centre
CAD Computer Aided Design
CF Correction Factor
CFD Computational Fluid Dynamics
CSIR Council for Scientific and Industrial Research
DAQ Data Acquisition
EFPI Extrinsic Fabry-Perot Interferometry
EMI Electro Magnetic Interference
FEM Finite Element Method
FS Full Scale
OFB Optical Fibre Bracket
OFBG Optical Fibre Bragg Grating
UJ University of Johannesburg
xx
Chapter 1 - Introduction and Problem
Statement
1.1 Introduction
The Council for Scientific and Industrial Research (CSIR) has formed a collaborative
research effort with the University of Johannesburg (UJ) to design and develop wind tunnel
balances. In the field of aerodynamics, wind tunnels are still used to test the aerodynamics of
a body, and to verify the results obtained from computational fluid dynamics (CFD).
In the field of wind tunnel testing, aerodynamic loads will be imparted onto a model due to
the flow of fluid around it. The goal of wind tunnel testing is to measure what loads are being
imparted onto the model, and their magnitude. This is done through the use of a wind tunnel
balance.
At present, foil strain gauges are the strain measuring sensors being used in wind tunnel
balances. There are several problems with using these sensors, which will be discussed in
depth in later sections. To the end of meeting the stringent requirements for new wind tunnel
balances, the University of Johannesburg, and Dr Pieterse [1], have developed a new type of
balance incorporating OFBG sensors. There are several advantages to using these sensors.
They measure strain using light and not electric current. They are therefore immune to
electro-magnetic interference. Their use (using the method described by Pieterse [1]) allows
for balances to be made stiffer, while not compromising on the sensitivity of the balance. A
careful design of balance will also allow it to be made much simpler and less expensive to
manufacture.
“Design a six component wind tunnel platform balance which incorporates optical fibre
Bragg grating strain sensors. The balance should be tested to evaluate its performance to
determine if further research in this area is warranted.”
OFBG sensors are being investigated in wind tunnel balances in order to meet the needs set
out by the wind tunnel testing community in 1996 [2], and 2006 [3]. These needs include:
2
- Stiffer and stronger balances with high natural frequencies.
- Thermal stability.
- Temperature compensation.
- Temperature insensitive balances.
- Non strain gauge balances. That is, the use of different types of strain sensors.
- Balances which are simpler in design, which will reduce manufacturing lead time and
costs.
- Minimisation of electro-magnetic interference.
The use of fibre optic strain sensors may serve these needs, as shown by Pieterse in [1].
Optical fibres are not susceptible to electro-magnetic interference, and the balance can be
designed to be stiffer than those incorporating foil strain gauges.
This chapter will discuss some aspects of internal and external balances. It will also provide
outline to the research methodology, along with an outline of the layout of this thesis.
Internal Sting
Balance External Balance
(a) (b)
Figure 1-2 – Internal sting balance [4] (a), external 2 component balance [5] (b)
3
1.3.1 Internal balance
A sting balance (Figure 1-3) is an example of an internal balance. This balance, with all of the
load sensors, is inserted into the model. These balances are based on bending beam principles.
The loads are sensed using strain gauges arranged in a Wheatstone bridge configuration. The
balance is designed to reduce the interactions between load components. Sting balances are
generally manufactured from a single piece of material. This serves to reduce hysteresis.
Platform balances can either be made from a single piece of material, or it can be constructed
using multiple pieces, and can measure loads in six components. A single piece platform
balance is shown in Figure 1-4. Flexures are employed to decouple the load components, such
that each load transducer measures only the load in its respective direction.
4
(a)
(b)
5
(a)
(b)
6
1.4 Research methodology
There are several distinct sections of research for this project. The project focusses on the
design aspect of a new balance in some detail. As such, a substantial portion of this report
will detail the design process of the balance. Appendix A will cover the design evolution, and
Appendix B will look at the FEM analysis of the chosen concept. The rest of the document is
concerned with the setting up of the load test, and analysis and discussion of the results. This
outline should provide a good idea of the scope of the project.
1. Literature review: This section will discuss the pertinent points concerning balance
design, calibration and the background to the OFBG sensors and their use on wind
tunnel balances. The use of OFBG sensors in balances is of particular interest, as they
have been used in various ways, with varying degrees of success.
2. Detailing the full design brief for the required platform balance: The specifics of
the new platform balance will be fully defined. Included in these specifications will be
the physical dimensions of the balance full scale loads.
3. Design and development of a six component balance which incorporates OFBG
sensors: A six component platform balance will then be designed. The thought
process behind the design will be detailed. The evolution of the design concepts will
be analysed and discussed. The appropriate material will be selected, and FEM
simulations will be conducted.
4. Manufacturing of the Balance: Once the design has been approved, the balance will
be manufactured and the sensors will be bonded to it.
5. Design and development a calibration body for the platform balance: A load
applicator will have to be designed which can apply known loads to the balance. This
calibration body would have to be able to subject the balance to individual, as well as
combined loads.
6. Load testing of the balance with the OFBG sensors: Testing the balance will be an
exercise in applying known loads to the balance, and measuring the output response.
This will be done for each of the six components. Thus, the input/output relationship
will be known for each component of the balance. The exercise of performing a full
calibration of the prototype will be deferred to future work. The intention is to
measure the input/output relationship for the balance, and identify any problems.
These problems can then be addressed and rectified in future work, and then a full
calibration can be conducted.
7
7. Writing the dissertation: In this phase, the research will be documented. All of the
literature which was reviewed in preparation for this project will be included. The
results obtain during the calibration and testing phases will be included and discussed
in depth. In the end, conclusions will be drawn, and possibilities for future work will
be discussed.
8
Chapter 2 - Literature Review
2.1 Introduction
The literature review will cover topics, the knowledge of which is essential in order to design,
manufacture and calibrate a wind tunnel platform balance. The review will be broken down
into different areas, such as the concepts of balance design, strain measuring techniques and
previous work in the field of optical fibre strain sensors in wind tunnel balances. Where
applicable, at the end of each section, a paragraph will be presented to discuss the relevance
of the just surveyed literature to the present study.
The first step in designing a balance is to determine the coordinate system and nomenclature
which will be used. The AIAA recommends the following system [11].
The coordinate system shown in Figure 2-1 represents the coordinates for the balance itself.
Due to the common practice of assigning forces in North American wind tunnel tests, the
positive normal and axial forces are in opposite directions to the X and Z axes, respectively.
The positive axial force is downstream of the flow, and the positive normal force is upwards.
It is recommended that the BMC be located at the direct centre of the balance, for Force and
Moment types. In Direct-Read type balances, it is recommended that the BMC be placed at
the point where the pitching and yawing moments are resolved [11]. Nevertheless, the BMC
may be located at any point on the balance.
It can be seen from Figure 2-1 that there are in essence, two coordinate systems, one for the
balance, and one for the forces. The balance coordinate system presented in Figure 2-1 is
intended for sting balances. The balance being designed in this study will use the same force
coordinate system. That is, the positive axial force will be located downstream of the flow,
the positive normal force will be upwards, and the positive side force will be towards the
starboard side of the model.
10
The Moment balance measures two pitching moments (fore and aft of the BMC, in the X-Z
plane). It resolves these two pitching moments into the net pitching moment and the normal
force. It measures two yawing moments (fore and aft of the BMC, in the X-Y plane), and
resolves the two yawing moments into the net yawing moments, and the side force. It also
measures the axial force and the rolling moment [11].
2.2.2 Calibration
2.2.2.1 Interactions
The accepted source regarding calibration of wind tunnel balances is the AIAA [11]. As such,
this reference will be used almost exclusively in this sub-section. The main goal in multiple
component load measurement is to reduce, or if possible, eliminate, interactions. If a load is
applied in a particular direction, then the output values of all of the transducers should remain
unchanged, except the transducer intended to measure that load. In order for this to be
possible, there should be at least one load sensor per component. For an external balance,
which has plenty of room outside the wind tunnel, this is more easily achieved than in an
internal balance, where space is limited [6]. External balances have room to incorporate
flexures, which allow for load components to be decoupled. For internal balances, this
decoupling can only be fulfilled in part due to the limited amount of space. Thus, it is
impossible to completely remove all of the interactions in internal balances [6].
The manner in which the strain gauges are wired may also serve to limit the influence of
interactions. The balance could be wired such that interactions on the fore bridge will be
opposite in sign to the interactions on the aft bridge. If the gauges are wired in parallel, then
the interaction effects will tend cancel each other out [11].
Interactions form part of the systematic errors of the measurement. Systematic errors are
repeatable, and as such, they can be accounted for in the calibration matrix. When calibrating
the balance, it has to be subjected to a number of known loads, and load combinations, and
the output of each of these inputs are recorded. The point of doing this is to obtain the
calibration matrix. This matrix, once fully defined, will relate an input to an output by the
following equation [6, 11]:
11
{𝑭} = [𝑪]{𝑺} (𝟐. 𝟏)
Where:
If all of the interactions of the balance were completely eliminated, then the calibration matrix
would be a diagonal matrix, whose diagonal entries would relate a single input component to
a single output component. However, since interactions do occur, the interactions will be
represented as entries on either side of the diagonal. If nonlinearities are taken into
consideration, then the matrix size increases considerably. If third order nonlinearities are
taken into account, for a six component balance, then the matrix size would be 6×33, or 198
elements [6].
𝑛 𝑛 𝑛
Where:
The calibration load datum is an important concept. It can be defined as the sensor outputs
when the load condition of the balance is zero. Any sensor outputs which result from an input
load will only be valid with respect to the calibration load datum. Therefore, if the balance is
used in a different environment, the same calibration load datum should be used if the results
are to be meaningful. The choice of this datum is arbitrary; the salient point is that the same
12
datum be used in every environment. The AIAA recommends however, that a zero load
condition be used for this datum [11].
It is not of necessity that all of the terms in the mathematical model be accounted for. In
certain tests, the balance may not experience a significant load in a particular direction. These
terms can be eliminated from the mathematical model, thus significantly reducing the effort
in calibration. The AIAA suggests that the calibration mathematical model be chosen based
on the dataset. Should certain terms in the data be undefined, they should be excluded from
the rest of the calibration process [11].
How well the calibration matrix relates inputs to outputs depends on the experimental design
of calibration load process [12]; the order in, and rate at which loads are applied and the
magnitude of the loading have an influence. The AIAA has set out some guidelines to which
the wind tunnel engineer should adhere when designing the experimental loading. The
method in which a balance is calibrated, and the method by which the calibration matrix is
obtained are areas of on-going research, which is beyond the scope of this study [11, 12]:
- The allowable geometry of the balance must be established, along with the expected
loads.
- Obtain the optimum location of the load sensing elements within the allowable
geometry.
- Account for interactions and the effects of ambient conditions by means of flexures
and other decoupling elements.
- Calibration of the balance by subjecting it to loads similar to those which are expected
during wind tunnel testing.
Multi-piece balances have their components welded together. This presents a disadvantage,
because when a balance must be welded together, there is a greater likelihood of the
hysteresis increasing [1, 6]. The disadvantage of mono-piece balances is that they are more
difficult, and thus, more expensive to manufacture. In addition, decoupling of the load
components becomes more difficult when designing a mono-piece balance. Multi-piece
13
balances can have their components manufactured separately, and this reduces costs [6]. In
order to optimise the balance while it is still in the design phase, finite element analysis is
used. These simulations are used to determine if the balance can withstand the loads, where
interactions exist and if there is a need for additional decoupling elements. This section will
focus on these aspects and will demonstrate these concepts using FEM simulations of a
replica platform balance. The balance used in the FEM simulations is a replication of the
balance shown in Figure 1-4 (b). Since the exact dimensions of this balance are proprietary
information of the balance manufacturer, the dimensions and loads have been guessed. The
salient point is to take note of the operation of the balance in terms of load sensing elements,
interactions and component decoupling.
Figure 2-2 shows the CAD model of a mono-piece balance. The FEM simulations in this
section are not complete; they are provided merely for explanation purposes only. Figure 2-3
shows the front view of the balance. This figure shows the flexures and load measuring
elements.
14
Flexures Load Sensing Elements
Figure 2-4 shows the strain contour plot of the balance under a normal load; only the front of
the balance (one set of flexures and loads sensing elements) is shown. The balance’s fixture is
located at the bottom. The load sensing elements in the centre are experiencing significant
strain, while the ones on either side of it are not.
Flexures
Strained Load Measuring Elements
15
Tensioned Load Measuring Element Compressed Load Measuring Element
(a)
(b)
Flexures
(c)
Figure 2-5 – Balance under an axial force [8]
16
Figure 2-5 (a) and (b) show the contour plots for the balance under an axial load. In Figure
2-5 (b), the load sensing elements on the left and right hand side are clearly in tension and
compression, respectively. The flexures are decoupling the load, such that the load sensing
elements in the centre experience negligible strain. Figure 2-5 (c) shows the load sensing
elements on the side of the balance. The flexures take up the deformation, thus, very little
strain is detected by its load sensing elements. The load sensing elements of the balance are
the same on each side, thus, the balance would react in the same manner under a side load.
Figure 2-4 and Figure 2-5 serve to demonstrate how the loads in a balance are decoupled. By
locating load sensing elements and flexures correctly, the balance can almost eliminate
interactions. The mechanism by which balances work in general, is by measuring the
deformation of the material at its surface (strain). For this reason, some sections of the
balance have to be made relatively thin (the load sensing elements in Figure 2-4 and Figure
2-5, for example) in order for the strain to be high enough for it to be adequately resolved by
a strain gauge.
The present study aims to build upon the so-called ‘two-groove method’, as developed by
Pieterse [1] (see Section 2.4.7). This method does not measure the deformation of material at
its surface; it measures the deformation of a groove, spanned by an OFBG sensor. The
relative deformation of the groove is comparatively much greater than the deformation at the
surface of the material. Thus, much higher apparent strains are produced, which can be
detected more easily [1]. Chapter 3 of this report will demonstrate the use of the two-groove
method in the design of a platform balance.
17
{𝐹} = [𝑘]{𝑥} (2.3)
Where:
The three points which will be discussed in this section are those of the loading and boundary
conditions, the mesh sensitivity analysis, and interpolation functions.
Loading and boundary conditions are crucial aspects to consider in order to ensure an
accurate solution. In terms of loading the object, care should be taken to not subject the load
to a single node. If this is done, the solution will indicate that that point will experience a very
large (infinite) stress; a force placed on an infinitesimally small area. For the platform balance
analysed in this study, no rotation or translation is permissible at the fixed end. Therefore, a
fixed condition will be used. As will be elaborated upon in Section 3.7, although a multi-
piece balance will be developed further, the FEM analyses will simplify the balance as being
a single piece.
The balance will be subjected to six load components; three forces, and three moments, and
different combinations thereof. The loadings in the FEM study will be applied in such a
manner so as to replicate reality as closely as possible.
Mesh sensitivity is an important aspect in FEM analyses. It dictates how accurate the solution
can potentially be, as well as the amount of time which is taken to complete the simulation.
Initially, the designer should run the simulation, using a fairly coarse mesh, to gain an
approximate idea of where the highest stresses are. The designer can then reduce the mesh
size sufficiently in those areas to increase the accuracy of the results. There is generally a
certain threshold where further mesh refinement does not affect the end result to a significant
degree. Therefore, the cost of additional computing time is not matched by the reward of a
more accurate solution. Thus, it is advisable to not refine the mesh past this threshold [13].
Various FEM packages provide an adaptive mesh refinement feature, which can significantly
reduce the amount of time required to run a mesh sensitivity study.
Interpolation functions are a means by which stress distributions can be approximated. The
FEM will calculate the stress state of a component at particular discrete points. Thus, the
designer will have an idea of the stress conditions at those points, but will not necessarily
18
have information about the stress distribution between those points. An interpolation function
is an attempt to describe the stress distribution between sets of discrete points [13]. A p-type
adaptive refinement technique will increase the order of the interpolation function, in so
doing, making the stress distribution appear more continuous.
Delving into more detail of how FEM works at a fundamental level is beyond the scope of
this review. The interested reader can consult a variety of sources, [13, 14] for a more in-
depth explanation.
The gauge factor is an empirically defined quantity which relates the change in resistance to
the experienced strain [15]. It is defined by the following equation [16]:
∆𝑹
= 𝑮𝑭𝜺 (𝟐. 𝟒)
𝑹
The gauge factor is a quantity which is defined by the manufacturer of the gauge, but its value
is usually approximately 2 [15, 16]. The gauge factor may become smaller, however, when
the size of the strain gauge decreases. This is because the change in resistance in a smaller
strain gauge is lower than it is in a larger gauge [16].
Figure 2-6 shows a conventional foil strain gauge. The gauge is manufactured in a similar
way as printed circuit boards. The loops in the conductor increase the effective length of the
gauge. This makes the change in resistance easier to detect. The gauge length determines the
resolution of the gauge. Since the strain is averaged out over the gauge length, the smaller the
gauge length, the higher the resolution will be [15].
19
Figure 2-6 – Strain gauge [15]
Strain gauges are often wired in a Wheatstone bridge configuration. Figure 2-7 shows a
Wheatstone bridge with a derivation which indicates the voltage, G, which is measured across
the bridge [17]. It was mentioned earlier that a strain gauge is a resistor. The sensitivity of a
strain gauge will typically be around 10-6Ω/(kNm2) [15]. As such, a highly sensitive device
should be used to measure this change. A Wheatstone bridge accomplishes this goal. Four
gauges can be wired in a Wheatstone bridge circuit, with only one of them measuring strain.
The four gauges will have the same initial resistance, therefore the bridge will be balanced,
and the voltmeter reading G, should be zero. Once the strain gauge experience strain (in this
case,𝑅2 ), its resistance will change, and it will cause the circuit to become unbalanced. Thus
the voltage G will change according to the equation [15]:
(𝑹𝟐 + 𝜹𝑹)𝑹𝟐 − 𝑹𝟏 𝑹𝟒
𝑮 + 𝜹𝑮 = 𝑽 (𝟐. 𝟓)
(𝑹𝟏 + 𝜹𝑹 + 𝑹𝟑 )(𝑹𝟐 +𝑹𝟒 )
But the initial resistances are equal, thus Equation 2.5 reduces to [15]:
𝜹𝑹
𝜹𝑮 𝜹𝑹/𝑹
= ≈ 𝑹 (𝟐. 𝟔)
𝑽 𝟒 + 𝟐(𝜹𝑹/𝑹) 𝟒
𝛿𝑅
For the condition where ≪ 1 , the relation in Equation 2.4 can be substituted into
𝑅
𝜹𝑮 𝑮𝑭𝜺
= (𝟐. 𝟕)
𝑽 𝟒 + 𝟐𝑮𝑭𝜺
20
Equation 2.7 provides a practical means to measure strain for a quarter bridge strain gauge
circuit. The advantage of a Wheatstone bridge is that it can be balanced by the use of a
variable shunt resistor. Thus, once the gauges have been applied to the test specimen, the
bridge can be balanced by varying the values of the shunt resistor for the no-load condition.
Using this bridge configuration, the measured output can be amplified or unwanted ambient
effects can be accounted for [15].
Advantages:
- The output is continuous. These gauges therefore have an infinite resolution, and a
relative resolution which is governed by the sophistication of the data acquisition
system.
- They are the commonplace strain measuring sensor; the technology is well
understood.
- They can be made very small.
- They are cost effective
Disadvantages:
21
- They are affected by side force loads as a function of Poisson’s ratio of the material to
which the gauge is bonded
- Depending on their configuration, they can have complex wiring, which makes
troubleshooting more difficult.
Input light
Reflected light
Fuse
The Bragg grating is written onto a Germanium doped fibre core by means of a UV light,
through a diffraction grating mask (Refer to Section 2.4.3.2). This causes a periodic change in
22
the refractive index of the core. Light is transmitted through the fibre core. When the light
reaches the Bragg grating, a certain wavelength of light is reflected (Bragg wavelength), and
the rest of the light is transmitted. Figure 2-9 illustrates this concept. When the fibre is
deformed due to strain, the Bragg grating deforms (is stretched or contracted). This alters the
wavelength of light that is reflected. An interrogator will sense the change in the wavelength
of the reflected light. This change is proportional to the strain experienced by the sensor.
The use of OFBG sensors advantages over other types of fibre optic strain measurement
devices are [20]:
23
- The grating can be written directly onto the fibre optic core, making it suitable for
applications where the small diameter of the fibre is required.
- The manufacturing technique lends itself to mass production of the fibres, making
them commercially competitive with electrical sensors.
- The fibre optic strand can be multiplexed, allowing for multiple Bragg gratings to be
written onto one fibre.
The advantages mentioned above could mean that the shortcomings of convention strain
gauges could be mitigated through the use of the OFBG sensor. Four instances of the use of
optical fibre strain sensors used in wind tunnel balances are:
- In 2000, Edwards [18] used EFPI sensors as strain gauges, bonded to the material.
Figure 2-10 shows the experimental setup used by Edwards. Refer to Section 2.4.4.
Figure 2-10 – Comparison of the EFPI sensor and a foil strain gage [18]
- In 2007, Vasudevan et al. [21] used EFPI sensors as strain gauges, bonded to the
material. Figure 2-11 shows the balance used by Vasudevan et al. in their experiment.
Refer to Section 2.4.5.
24
Figure 2-11 - A six component sting balance with both EFPI and strain gauge sensors
[21]
- In 2010, Pieterse [1] used OFBG sensors spanned over a groove. The groove was used
as a mechanical strain amplifier. The experimental setup to prove the concept is
shown in Figure 2-12. The two groove method was used as a means of temperature
compensation. Refer to Section 2.4.7.
Mechanical leverage
Throat
30X30mm
Optical fibre
25
- In 2014, Burger [10] retro-fitted OFBG sensors to a balance designed to use foil strain
gauges. This was done by designing a bracket which could be bolted to the balance,
onto which the fibres would be bonded.
If the manufacturer were to change the distance between the Bragg grating planes, different
wavelengths of light could be reflected at different positions along the fibre. Thus, OFBG
sensors lend themselves to multiplexing [16], and up to 100 Bragg gratings could be read
from a spectrum analyser [1].
Strain measurement using the OFBG sensor is simple in concept. The fibre is bonded to the
surface being examined. The fibre will experience the same strain as the body once loaded
(assuming zero slippage in the bond between the fibre and the surface). This will stretch or
contract the fibre, thus changing the inter-planar spacing in the Bragg grating. This will cause
a change in the wavelength of light which is reflected. A spectrum analyser will measure this
difference and relate it to strain by the following equation [20]:
𝒏𝟐
𝝆𝒂 = [𝝆 − 𝝂(𝝆𝟏𝟏 − 𝝆𝟏𝟐 )] (𝟐. 𝟏𝟎)
𝟐 𝟏𝟐
Where 𝜌11 and 𝜌12 are the components of the fibre optic’s strain tensor, and 𝜈 is Poisson’s
ratio [20].
Temperature measurement can also be achieved using OFBG sensors. Temperature changes
cause thermal strains in any physical body. In terms of the OFBG sensor, thermal strains will
effectively change the reflected wavelength for reasons just mentioned. The shift in the Bragg
wavelength is given by [20]:
27
due to strain and temperature will be indistinguishable. The total shift in wavelength is
illustrated in Figure 2-14 [1].
28
effects this method has on the outputs of the OFBG sensors in terms of the reflected
wavelengths of light.
Figure 2-16 – Temperature and side force compensation – ‘two groove method’ [1]
29
Strain is measured by noting the change in displacement of the two reflected wavelengths,
relative to each other. Notice in Figure 2-16, the initial zero load condition is illustrated by
the solid black lines. When a load is applied to the system, the Bragg wavelength in the top
fibre will increase (move toward the right), and the bottom groove’s Bragg wavelength will
decrease (move toward the left). This load condition is illustrated by the broken blue lines in
the figure. Thus it can be said that the difference in wavelengths of the two Bragg gratings
changed by a net amount of ∆𝜆𝐿 . If the temperature of the system is increased, both fibres will
experience a positive thermal strain. Thus both Bragg wavelengths will increase (move
toward the right), by the same amount (indicated by the red lines in Figure 2-16). However,
the difference between their two respective Bragg wavelengths remains unchanged, at ∆𝜆𝐿 .
Therefore, this difference, ∆𝜆𝐿 , can be related to the net strain experience by the system due
to the applied load only, free from interactions. Thus, in order to make a strain measurement,
a pair of readings must be taken. One sensor of the pair must experience a compressive strain,
and the other sensor must experience a tensile strain. If both sensors experience the same
tensile (or compressive) strain, ∆𝜆𝐿 will remain unchanged, and no net strain will be
measured. Figure 2-17 shows a CAD rendering of a balance with the top grooves, which was
developed by Pieterse [1].
Metric end
Top Groove
Non metric
end
Figure 2-17 – CAD rendering of the OFBG four component balance [1]
Other temperature compensation techniques have been proposed. One method is to use a
reference OFBG sensor. This reference OFBG sensor is free from strain and is placed in the
same thermal environment as the strain sensor. The wavelength shift experienced by this fibre
is subtracted from the wavelength shift measured by the strain sensor [20].
30
The so-called dual-wavelength superimposed FBG method [20] has two Bragg gratings of
different Bragg wavelengths written at the same location. Under strain and temperature, these
two Bragg wavelengths will change independently of each other, based on the values of the
sensitivities of the fibre Bragg grating due to strain (𝐾𝜀 ) or temperature(𝐾𝑇 ). The Bragg
wavelength shift can due to both temperature and strain is given as [1, 20]:
∆𝜆𝐵 = 𝐾𝜀 Δ𝜀 + 𝐾𝑇 Δ 𝑇 (2.12)
Equation 2.12 shows the Bragg wavelength shift for one Bragg grating. Thus, for two Bragg
gratings superimposed at the same position, a system of simultaneous equations can be set up
and solved for the strain and temperature [20]:
[𝚫𝝀
𝚫𝝀
]
𝑩𝟏
=[
𝑩𝟐
𝑲
𝑲
𝜺𝟏
𝜺𝟐
𝑲𝑻𝟏 𝚫𝜺
]
𝑲𝑻𝟐 𝚫𝑻
[ ] (2.13)
Three temperature compensation techniques have been discussed; each has its own merits.
With the two groove method it is simple to obtain the strain value. It also allows for the
compensation of side force interactions, which is advantageous. It does however require a
very specific balance design, and using this technique may make designing a platform
balance challenging. The reference fibre technique it is simple to obtain the strain value, but
is does require the routing of an additional fibre, and requires an extra communications port
on the interrogator. The dual-wavelength superposition method requires the routing of only
one fibre, but there may by additional costs involved in obtaining a fibre written in this
manner. In addition, side force interactions cannot be compensated for using this technique.
31
the Bragg grating. This can be coupled to precisely moving the mirrors to yield yet more
flexibility to the range of Bragg gratings which can be manufactured [20].
32
Figure 2-19 – Phase mask method to manufacture OFBG’s [23]
2.4.3.3 Resolution
Resolution is the smallest increment which a measuring instrument can record [15]. In the
present case, resolution would refer to the smallest value of strain which can be measured by
either a strain gauge or by an OFBG sensor. In order for OFBG sensors to be competitive
with convention foil gauges, they should provide a higher relative resolution. Strain gauges
provide a continuous output; they have an infinite resolution. Thus, their relative resolution is
governed by the data acquisition (DAQ) system used, and is generally less than 1µ𝜀 [1]. Due
to the wind tunnel testing environment, however, the effects of EMI due to testing equipment
and in lead wires, and temperature effects, the relative resolution is decreased to ±1µ𝜀 [24].
Relative resolution refers to the smallest resolvable part which can be measured, divided by
the full scale load. The smallest resolvable part is defined as the two sigma standard deviation
of the signal noise. The signal noise is obtained by continuously logging the output of the
balance under a zero load condition. A good DAQ system can measure output signals of 1µV
[24]. Axial forces, generally being the smallest load component, could induce a strain which
would produce such a small output signal. The largest loads could produce a strain which
produces an output signal of approximately 10mV [1]. Balance stiffness requirements usually
cap the output strain signal to closer to 5mV. The relative resolution of the system just
described may be calculated as follows [1]:
1 × 10−6 𝑉
× 100 = 0.02%
5 × 10−3 𝑉
33
This is the relative resolution of full scale. If the full scale load for that particular direction is
1 000N, then the smallest load which will be detectable by the equipment is 0.02% of 1 000N,
or 0.2N. In order for OFBG sensors to remain competitive with foil gauges, they must offer a
relative resolution which is comparable or higher than 0.02%. Two DAQ systems were
available for this dissertation. The Micron Optics si425 interrogator can detect a wavelength
change of 2pm (1.66µ𝜀 of fibre strain) [25], while the National Instruments NI PXIe-4844
can measure up to 1pm [26]. Under a full scale load, the fibres are expected to be strained to
approximately 6 000 µ𝜀 each. Refer to Section 3.7 for fuller details of how the strain readings
are calculated.
1.66µ𝜀
× 100 = 0.028%
6 000µ𝜀
Thus:
0.028% − 0.02%
× 100 = 28.57%
0.028%
The relative resolution of the OFBG sensors is 30% higher than the conventional strain
gauges. This is another point in favour of fibre Bragg grating sensors.
In the work of Pieterse [1], the relative resolution was found to be 0.0104%. The reason for
this is due to the fact that the balance in that study was designed to strain the fibres to
16 000 µ𝜀 . The fibres were pre-strained to 20 000 µ𝜀 , and could experience a strain of
±16 000µ𝜀. Thus, a 4 000µ𝜀 buffer was available on the lower end, meant to prevent the fibre
from becoming slack, and a 24 000µ𝜀 buffer on the upper end, to prevent the fibre from
breaking [1]. In order to have the fibre work in its linear output region, it was decided to
lower the strain which would be experienced by the fibre. Thus, the maximum operating
strain which the fibre will experience in this study will be 12 000µ𝜀 + 14 000µ𝜀 = 26 000µ𝜀;
10 000µ𝜀 lower than the maximum operating strain of the fibres used in [1]. In a further
study, Pieterse and Bidgood [27] pre-strained the fibres to 1% (10 000µ𝜀), with a maximum
34
operating band of 2% (20 000µ𝜀). This can be put down to the fibre’s long term operating
range is at 1% strain [28].
The uncertainty and hysteresis effects of the OFBG sensors used in wind tunnel balances are
an area of on-going research. The possible sources of hysteresis are:
35
- The interaction between the cladding and the adhesive.
- The hysteresis inherent in the balance material.
The EFPI sensors were much small than the foil gauges. This proved to be an advantage in
applications where space is a concern. Using EFPI sensors, Edwards suggests that smaller
sting balances could be designed. This would make testing models in supersonic wind tunnels
– where the cross-sectional area of the tunnel is small – easier.
Edwards noted that some difficulties and practical limitations of using optical fibre sensors.
They include [18]:
- Difficulty in applying the sensor to the balance. This is in part due to the bending
radius limitation of the fibre optic strands.
- Difficulty in routing the fibre optic leads out of the wind tunnel.
- The testing conditions within the supersonic wind tunnel were too harsh to measure
precise and accurate results. Both the foil sensors and the EFPI sensors measured this
unsteadiness.
36
- The calibration procedure was inadequate. It is postulated that the sensors may have
been damaged during this procedure which may have caused them to stop working
properly.
The future work recommended by Edwards is to design and build a balance which
incorporates the EFPI sensors exclusively. This would allow the designer to make allowances
for issues such as routing of the fibre optics. The balance could be made smaller, lighter and
be less expensive to produce [18].
The lessons learnt by Edwards will be taken into account in this study. The issues of routing
of the fibres, which is a major concern, will be borne in mind. The platform balance which
will be designed will not be used in a supersonic wind tunnel, which eliminates another
difficulty. While instructive, the results obtained in Edwards’ study, particularly those shown
in Table 2-1 are read with some scepticism.
With this EFPI sensor, Vasudevan et al. were able to accurately measure very small skin
friction loads on the order of 20 to 30 milligrams on a floating flat surface at speeds of 20m/s
[21]. This demonstrates, once again, the efficacy of using fibre optic sensors in wind tunnel
balances. It was noted that the bonding layer thickness for the EFPI sensor is thicker than that
of the strain gauges, and this needs to be accounted for. This was evident in some errors
which arose in the EFPI readings, as compared to those of the strain gauge [21].
The study found that the number 1 and 2 EFPI sensors’ results differed from each other by
10%. It was postulated that the gauge factor of the sensor was faulty [29]. The noise in the
output signals was measured, and Table 2-2 was complied.
The standard deviation is defined as the bounds which contain 68.2% of the acquired data
[29]. Thus, it can be seen that the electrical gauge has a much lower noise level than either of
the optical fibre sensors. The precision of the optical fibre sensors could be increased if by
using more advanced interferometer techniques. Temperature effects were examined, and it
was found that all of the sensors operate well at ambient temperatures [29].
- All of the sensors used are capable of recording strains at ambient temperature.
- The electrical strain gauge is the best choice, as it has the lowest standard deviation in
its results.
- The fibre Bragg gating sensors provide more reliable results than the Fabry-Perot
sensors.
The EWA study mentioned that some major problems were encountered with regards to the
EFPI DAQ system. This issue was not encountered in either study performed by Edwards
[18], or by Vasudevan et al. [21]. Additionally, Edwards stated that the EFPI sensors used in
his test provided better accuracy and resolution than the foil gauges [18]. The reasons for this
38
are unknown. Nevertheless, the findings of the EWA study will be taken into account in this
investigation, so as to minimise the potential for the problems which they encountered.
There are two novel aspects to this research. The first is using a beam with lateral grooves as
mechanical strain amplifiers, and the second is using two grooves (top and bottom) for
temperature and side force interaction compensation (see Section 2.4.3.1). In order to prove
the concept, an aluminium cantilever beam was modelled using FEM. Figure 2-21 shows the
FEM model of the cantilever with the upper and lower grooves, which is being subjected to a
bending stress.
Figure 2-21 – Mechanical strain amplification using the two groove method [1]
The black circle in Figure 2-21 shows the bottom of the groove. This is where the strain
gauges were placed during the lab tests of this concept. As it can be seen from this image, the
39
base of the groove has not displaced much; it therefore has not experienced much strain,
around 270µ𝜀 [1]. The fibre optic strain sensor traverses the top of the groove. The top of the
groove has clearly displaced a comparatively large amount, as highlighted by the red circle in
Figure 2-21. The dimension of the groove in this image is 10×10mm. Once loaded, the
groove width changes by 0.009324mm, which corresponds to a strain measured by the OFBG
of 932.4 µ𝜀, which is a 344.70% increase in strain (compared to the strain in the base of the
groove) [1]. Different groove shapes were simulated next; Figure 2-22 shows the different
groove shapes which were tested. It was found that the strain groove with the rounded ends
effectively increased the measured strain (in the OFBG) by 837.52% [1] over the strain
measured in the base of the groove. Mechanical strain amplification increases the full-scale
strain, thus the resolution of the measuring sensor is effectively increased. Using the concept
of mechanical strain amplification, a low-cost interrogator could be used as the DAQ system,
and a sufficiently high resolution can still be achieved. This offers a cost saving alternative to
using foil strain gauges which will need a more expensive DAQ system to measure strain at
the same resolution as the OFBG sensor [1]. This study will take mechanical strain
amplification into account when the platform balance is designed. The advantage of this
solution is that a balance with high stiffness can be designed, while still having a high degree
of accuracy.
Upper Grooves
Lower Grooves
41
2.4.8 University of Johannesburg, strain gauge side wall balance with retro-fitted
OFBG sensors
The University of Johannesburg completed a study in 2014 in which OFBG sensors were
retro-fitted to a sidewall balance [10]. This was a comparative study, to evaluate the
performance of OFBG sensors again the performance of strain gauges. In order to use OFBG
sensors, brackets had to be designed and manufactured which could accommodate the fibres.
These brackets were designed using the same concept as described by Pieterse [10]; that is,
have the fibre span a gap, whose length will change under the application of a force.
Figure 2-24 – Optical fibre bracket used to retro-fit fibres to the side wall balance [10]
In Figure 2-24, the optical fibre bracket (OFB) can be seen. Beneath this bracket is the shear
plate, onto which the strain gauges are bonded. The balance is designed such that, under the
application of a load, these shear plates will be subjected to high strains. The OFB was
designed such that any displacements will be amplified; this mechanical amplification
induces an increase strain in the fibre, thereby increasing its resolution.
42
Figure 2-25 – Deformation of a OFB due to a side force load [10]
It was found that the fibres were outperformed by the strain gauges in this side wall balance.
The back calculated error (BCE) was a metric employed by Burger to evaluate the
performance of the two sensor types. The back calculate error is the difference between the
predicted balance response, calculated using the calibration curve, to the actual balance
response for a given load [10]. The BCE’s for the strain gauges were 0.12% – 0.14% and for
the optical fibre sensors, the BCE was 0.46%.
Another performance metric was the relative resolution. The strain gauges used by Burger in
the comparison were Tungsten-Platinum gauges, with a gauge factor of 4.7 [10], which is
2.35 greater than the gauge factor of the conventional constantan gauges which used on the
balance originally. The relative resolution of the Tungsten-Platinum gauges used was
0.016%, while the optical fibres had a relative resolution of 0.039%. The relative resolution
of the constantan gauges was 0.09%. Thus, the optical fibres performed better than the
constantan gauges, by 2.3 times. They performed 2.4 times worse than the Tungsten-Platinum
gauges.
Finally, repeatability of the sensors was considered. The balance was loaded with a rolling
moment of 264.4 Nm, and unloaded to zero 11 times. The output response was taken for each
load step, and it was found that the scatter in the strain gauges’ output had a standard
deviation of 0.030 – 0.043. The scatter in the output of the response of the optical fibres had a
standard deviation of 0.095.
43
In all three performance metrics, it was seen that the fibres were out performed by the
Tungsten- Platinum strain gauges. It should also be noted that one of the fibres had slipped its
bond during the tests, and this could not be replace due to time and financial constraints. It
was also noted that the fibre experienced a characteristic of stress relaxation, or creep. That is,
once the fibre is strained (due to a load being applied to it), the output of the fibre creeps over
time. The stress in the fibre becomes less, with what appears to be a logarithmic relationship
with respect to time. The University is conducting a study at present to understand the cause
of this creep, and possible solutions to it.
The final point to note is that the test was unfairly weighted against the fibres. The fibres
were retro-fitted to a balance which was designed to use strain gauges. Thus, a bracket had to
be designed and fitted to the balance. This increases the uncertainties, over and above the
uncertainties inherent in the fibres themselves.
2.5 Summary
The literature review presented a brief overview of balance design in general, with more
focus being given to platform type balances. The fundamentals of balance calibration were
reviewed, along with finite element analysis. Attention was given to strain measurement
techniques in balance design, with particular focus given to the state-of-the-art in using
optical fibre sensors.
The wind tunnel testing community has called for balance designs which employ optical
fibres specifically. The potential advantage they bring in terms of EMI immunity are
attractive. However, it was found that most current methods of employing optical fibre
sensors in balances generally have inferior performance to strain gauges, which are the more
technically established sensor technology in this field.
Using OFBG sensors with the two groove method offers a potential solution; the sensor
performance may be comparable, or better than foil strain gauges when using this method of
strain measurement. It was also seen that, even while using the two groove method, fibres
retro-fitted onto an existing balance will offer inferior performance to the strain gauges. As
such, in order to develop the technology further, a balance should be designed to employ
optical fibre sensors a priori.
44
Chapter 3 - Design of a Six Component Wind
Tunnel Platform Balance
3.1 Introduction
The balance design process was an iterative one. From the outset, there were some design
goals which were set. The design of each concept aimed to meet the specifications discussed
in the following section. A major design driver was the minimisation of interactions. The idea
with the first five concepts that were designed was to decouple interactions through the use of
the mechanics of cantilevers. These designs are detailed in Appendix A. In these concepts, a
cantilever was mounted vertically, with a platform placed on top. The cantilever had grooves
near the base, and near the free end. Optical fibres were to span these grooves, in a manner
similar to that discussed in Section 2.4.7.
Under the application of an axial force, according to the orientation indicated in Figure 3-1,
the bottom grooves were expected to deform more than the top grooves. Under the
application of a pitching moment, the top and bottom grooves should deform by equal
amounts. It was thought that using the mechanics of cantilevers in this way could decouple
the loads, and minimise interactions. While these early concepts were able to decouple some
load interactions, certain interactions were still on the order of 100% or more. This was
deemed unacceptable. The main contributor to these interactions was the fact that the
platform tends to pitch (or roll) under the application of an axial (or side) force.
AF
(a)
(a) (b)
(b)
1. The loads for which the balance has to be designed are presented Table 3-1:
46
2. The balance should conform to the following outer dimensions, as shown in Table
3-2:
47
Balance
Moment
Centre
Concept 6 is a platform which is supported by sets of ‘cradles’. These cradles are flexible
enough to provide an allowable deformation for all of the load components, while keeping the
platform horizontal during axial and side force loads. That is, minimizing sagging of the
platform when an axial or side force load is applied. In this concept, each corner has two
48
cradles to govern the movement of the platform. Figure 3-4 to Figure 3-7 show this concept
from various views.
Metric End
Sensor Pillar
Cradles
Non Metric End
49
OFBG
Sensor
Locations
OFBG Sensor
Locations
50
Figure 3-8 shows how the platform of Concept 6 moves under a side force. The maximum
displacement of the platform is 0.0474mm under a side force load of 2 500N. This indicates
that the balance is sufficiently stiff. The image also shows that the platform move horizontally
under a side force; no rolling of the platform is present. This design thus eliminates the
problems of interactions which had plagued the previous design concepts.
SF
Table 3-4 shows the interactions in terms of percentages for each of the sensors. All of the
interactions – apart from two – are below 0.1%. The largest interactions which exist are those
of the axial force sensor, when a pitching moment is applied and of the side force sensor
51
when a rolling moment is applied. These interactions are 5.374% and 4.038% respectively.
These interaction data bode well for the design; these are the lowest interactions of any
concept considered thus far.
Table 3-5 shows the maximum stress experienced by the balance when an individual load is
applied. All of these values are well below the 250MPa threshold which was established in
the product design specification. The only drawback of this design is that it may prove to be
somewhat complex to manufacture.
Advantages:
Disadvantages:
52
The interactions of this design are small; however, there are two interactions which are
relatively large. The axial force sensor is affected by a pitching moment load, and the side
force sensor is affected by a rolling moment load. Some discussion as to where these
interactions come from is warranted.
The interaction present in the axial force sensor due to the application of a pitching moment
load is due to the location of the axial force sensor. The axial force sensor is bonded to the
sensor pillar and a probe protruding from the platform. The sensor pillar remains stationary
when a load is applied, and the probe moves in unison with the platform. This setup can be
seen in Figure 3-9.
Sensor Pillar
53
The selected area in Figure 3-9 is magnified in Figure 3-10. This figure shows the
deformation which is present when a pitching moment is applied; this deformed result is
superimposed onto an image of an undeformed balance (areas shown in red). Notice the
movement of the probe relative to the sensor pillar. This is the movement which causes the
interaction of the axial force sensor when a pitching moment is applied. The same situation is
true for the side force sensor when a rolling moment is applied.
Probe
Sensor pillar
54
3.3.2 Concept 7 (Single cradle design)
The final design which was investigated was a derivative of the previous design, Concept 6.
The present design operates in exactly the same manner, but instead of having two cradles per
corner, allowing for motion in the x- and y-directions, this design has a single cradle design,
which allows for motion in both directions. Figure 3-11 shows the single cradle design. The
purpose of a single cradle design was to make the balance simpler to manufacture. The sensor
locations for this concept are exactly the same as those for Concept 6. This section will
present the full explanation of the operation of this balance, along with FEM data which were
captured for this design. The advantages and disadvantages of this design will be presented. A
comparison given in the following sub-section will demonstrate why this concept was chosen
for further development.
55
tension, one in compression) will cause a relative displacement between the two reflected
wavelengths for each OFBG sensor (this is the Δ𝜆𝐿 mentioned in Section 2.4.3.1).
Compression Tension
Normal
force
The normal force sensors are experiencing compression (left sensor) and tension (right sensor)
The pitching moment sensor pair is encircled in Figure 3-14. Under a pitching moment acting
in the direction indicated by the arrow, the sensor on the left will experience a compressive
56
strain, and the sensor on the right will experience a tensile strain. This yields the measurement
pair which is required to determine the strain. Figure 3-15 shows the deformed contour plot
for the balance under a pitching moment. This image shows how interactions are
compensated for. The pitching moment sensors act in tension and compression, yielding a
strain result, while the normal force sensors both act in tension. So the pitching moment
sensor pair will experience a Bragg wavelength shift of Δ𝜆𝐿 . The Δ𝜆𝐿 for the normal force
sensor pair will be zero, since both fibres experience an equal tensile strain.
Pitching moment
Compression Tension
Compression Tension
Normal force
sensors both
act in tension
Compression Tension
Compression
Axial
force
Tension
Compression
Tension
Side
Force
Tension Compression
Yawing
moment
Compression Tension
61
Table 3-6 – Concept 7 OFBG sensor wavelength shift [nm]
Sensor
nf af sf pm rm ym
Load
NF 13.413 0.009 0.006 0.001 0.001 0.003
AF 0.000 11.725 0.001 0.009 0.002 0.000
SF 0.002 0.003 14.021 0.005 0.009 0.003
PM 0.000 1.271 0.029 9.712 0.002 0.040
RM 0.000 0.015 0.427 0.001 7.617 0.001
YM 0.002 0.001 0.002 0.002 0.002 15.179
Table 3-6 shows the resultant wavelength shift for each of the sensors under a full scale load
for each component. Comparing the wavelength shift for the single cradle design, to the two
cradle design, it can be seen that the sensitivities of five of the six sensors has increased. The
sensitivity of the rolling moment sensor has decreased. The change in the sensitivities of the
sensors between these two designs will be quantified shortly. This means that the balance is
less stiff with a single cradle, than it was with two cradles per corner. However, the
deformations under full scale loads are still acceptably small.
The interactions in terms of percentages are shown in Table 3-7. Overall, the interactions
have remained largely unchanged. However, the axial force sensor now picks up a 10.840%
interaction when a pitching moment is applied.
62
Table 3-8 – Concept 7 maximum stress per load
The maximum stresses for this concept are all below the 250MPa threshold which was
applied in the Product Design Specification. The stresses experienced by the balance under
individual loads all fall within the expected stresses for a block type balance made from 17-4
PH or maraging steel [30, 31]. With the maximum combination load applied, the balance
experiences a stress which still nets a safety factor of 3.
The comparison of sensitivities of the sensors between the two designs is presented in Table
3-9. The rightmost column shows the percentage change (increase or decrease) in terms of the
resultant wavelength shift of Concept 7 (Single Cradle Design), with respect to Concept 6
(Two Cradle Design). A positive value indicates that the sensitivity has increased, and a
negative value shows that the sensitivity has decreased. Table 3-9 shows that the single cradle
design has increased sensitivities for five components. This is advantageous; the sensor for
the rolling moment has decreased in sensitivity by 37.0%, nevertheless, it is still sensitive
enough to provide sufficient resolution to the measurement.
Advantages:
- The single cradle design makes the balance simpler, and therefore less expensive to
manufacture than Concept 6.
63
- The single cradle design increases the sensitivity of five of the six load sensors, when
compared to Concept 6.
- Sufficient resolution is expected in the sensors for each component.
- The maximum stresses are well below that of the 250MPa threshold established in the
PDS.
- The factor of safety is maintained.
Disadvantages:
With this concept, a single cradle is required to provide the desired motion of the platform
(horizontal motion in the x- and y-directions) under an axial or a side force. Thus, one cradle
has to support the forces which were supported by two cradles in the previous concept. The
cradle design which was selected was a compromise between stiffness, desired motion of the
platform and interactions. The selected design provides the planar motion of the platform
when subjected to an axial or side force, but it also results in the increased interactions, as
discussed previously. If the cradle had been designed to be stiffer, the platform would no
longer move horizontally (without pitching or rolling) under an axial or side force.
64
in the pairwise selection process to determine which criteria are more important, and which
are less so. A description of the pairwise comparison process follows:
A certain number of criteria are selected; these must be pertinent to the goal of the project.
These criteria can be: manufacturability, accuracy and stiffness. Essentially, any performance
characteristic of the product can be chosen. Once a list of criteria has been determined, the
importance, or weight of each criterion can be evaluated. A 𝑛 × 𝑛 matrix is set up with each
criterion plotted along the topmost row and leftmost column (assuming n criteria). Each
criterion is compared against every other criterion. When two criteria are compared against
each other, the criterion with the higher priority is assigned a value of 1, and the lower
priority criterion is assigned a value of 0. If the priority of the two cannot be distinguished,
they are each awarded a value of 0.5 each. This way, the most important criteria are
determined, and they are assigned the most weight.
After the criteria have been selected, the alternative design options can be compared. An
m× 𝑚 matrix is set up, for each criterion (Assuming m alternative designs). The design
concepts are plotted along the topmost row and leftmost column. The designs are compared
against one another – in a similar fashion as mentioned above – in terms of a particular
criterion. This is done for each criterion.
The alternative design which has obtained the highest score for a particular criterion is
multiplied with the weight of the criterion. Thus, one design solution may have fared the best
in one particular comparison, but that criterion may not have the highest priority. This way,
the alternatives are compared on a level playing field, and the design solution that fared the
best, in the most important comparisons will end up being the design that is chosen.
- Low interactions: Any good balance design aims to eliminate, or reduce the
interactions as much as possible; this can be very challenging. Interactions, being
systematic errors can be calibrated out; nevertheless, reducing interactions is very
important.
- Manufacturability: The complexity of the manufacturing required is directly
proportional to the cost. Thus, the balance should be as simple to manufacture as
possible. This is one of the main goals of using OFBG sensors in wind tunnel
balances.
65
- Stiffness: The ideal situation in a wind tunnel test is for the model to remain perfectly
stationary while under load. This is impossible to achieve; some movement of the
model is required in order for the balance to take measurements. Thus, in order to
approximate the ideal, the movement of the model should be as small as possible;
therefore, the balance itself should be as stiff as possible. The use of a different type
of sensors in wind tunnel balances permits a variation in the possible designs; this
could potentially result in stiffer balances with the same measurement resolution as
their predecessors.
- Sensitivity and resolution: OFBG sensors could potentially offer comparable or
higher resolution than conventional strain gauges using less expensive data acquisition
systems [1]. In order to reap the benefits of using the OFBG sensors, the balance
should have the highest resolution possible.
- Strength: Strength is a vitally important aspect. Failure of the balance during a wind
tunnel test is not an option. Failure could result in millions of Rands in damage to the
tunnel as well as weeks of down time. Therefore, a safety factor has to be adhered to.
A safety factor of 3 is used in this design process.
Once the selection criteria have been established, they have to each be assigned a weighting.
This weighting process is presented in Table 3-10.
Table 3-10 shows that the criterion “Manufacturability” was deemed to be the most
important. This is because one of the main reasons of using OFBG sensors was to develop a
balance which is simpler to manufacture than conventional strain gauge balances. There
would be little point in trying to use new types of sensor technology, if one cannot reap
benefits such ease of manufacture. “Low Interactions” was ranked as the second most
important criterion. This balance must have comparable performance characteristics to
conventional platform balances, and thus should have low interactions. “Sensitivity” was tied
for second. In order to compete with conventional balances, this design must be sufficiently
66
sensitive. The higher the sensitivity, the better the resolution of the measurement will be.
“Strength” and “Stiffness” were ranked next. The balance should be stiff, to prevent the
model from grounding during a test, and the balance should be strong enough to not fail. Both
concepts were designed to meet the safety factor of 3, to there will be little to choose between
the designs in terms of strength.
67
Table 3-15 – Concept selection – Stiffness (2)
The concepts’ score for each comparison is now multiplied with the weight for that criterion.
The scores will then be added up, and the concept with the highest score will be developed
further. In the case of a tie, the concept with the highest score for the highest priority
criterion, manufacturability, will be chosen.
Concept 6 and Concept 7 have both scored 7.5 points in the pairwise comparison. Therefore,
the chosen concept will be Concept 7 (single cradle design), based upon its relative ease of
manufacture. This selection process was instructive, as it neatly quantifies the strengths and
weaknesses of each design. It also shows the compromises one must make during a design
process. What one gains in terms of manufacturability, one loses in terms of interactions, for
example. Therefore, as this project is the first prototype of an OFBG platform balance, it
makes sense to select a concept which is simpler to manufacture, while still offering
acceptably low interactions and high sensitivity and resolution.
68
Figure 3-24 – Single piece balance
Although this can be manufactured, there are yet other issues which require the design be
refined. The balance shown in Figure 3-24 would require a billet of material with the
dimensions of at least 218×218×150mm. A block of 17-4PH (refer to Section 3.6) of this size
would be very expensive. Most of the material from this block would have to be cut away and
discarded/recycled. This is a very wasteful proposition. The manufacturing costs would also
be very high. A skilled machinist would have to see to it that no mistakes are made. If a
manufacturing error were made to one of the crucial components, such as the cradles, then the
entire balance would have to be discarded. This would waste a lot of material, time and
money. With these considerations in mind, the following modifications were made to the
design.
The balance would now become a multi-piece balance; the cradles would each be
manufactured separately, as will the platform, and the base. An illustration can be seen in
Figure 3-25. The cradles would be press fitted into the platform and the base. Once the pieces
have been fitted together, they will be welded, to reduce hysteresis. These modifications serve
to reduce manufacturing costs.
69
Metric end
Cradles
Non metric
end
It was decided that the ideal material for the balance was one with a martensitic
microstructure. Good corrosion resistance was also required, as the balance cannot be painted.
Thus, a material containing chromium, nickel, copper or a combination thereof would be
required. A high degree of hardenability, or the ability to form martensitic microstructures
within the material is desirable, thus manganese should be an alloying agent. Finally, as
silicon improves the toughness of a material, this should be included as one of the alloying
agents [32, 33].
Two materials have been considered; AISI 4340 (EN24) and 17-4 PH. The properties of each
will be presented.
70
Table 3-16 – AISI 4340 material composition [34]
Element wt%
Carbon 0.37 - 0.43
Silicon 0.15 - 0.30
Manganese 0.60 - 0.80
Nickel 1.65 - 2.0
Chromium 0.70 - 0.90
Molybdenum 0.20 - 0.30
Sulphur <= 0.040
Phosphorus <= 0.035
Iron Balance
The mechanical properties of AISI 4340 meet the criteria set out in the Product Design
Specification. The material composition of 17-4 PH is shown in Table 3-18.
71
Table 3-18 – 17-4 PH material composition [35]
Element wt%
Carbon 0.04
Manganese 0.40
Phosphorous 0.02
Sulfur 0.005
Silicon 0.05
Chromium 15.5
Nickel 4.5
Columbium + Tantalum 0.3
Copper 3.5
Iron Balance
The mechanical properties of this material are shown in Table 3-19. These are the mechanical
properties which were used in the FEM analyses.
The mechanical properties of 17-4 PH are desirable. This material has a yield strength of
1 240 MPa. This material meets the requirements set out in the Product Design Specification.
It has a slightly higher yield strength than AISI 4340. In addition, the University of
Johannesburg has previously developed a balance using 17-4 PH. Therefore, it shall be used
as the material to construct the platform balance.
72
3.7 Finite element method analysis
In order to proceed with the development of the balance, it had to first be simulated in FEM.
This section will outline the calibration process, and the finite element analysis report will be
presented in Appendix B, along with the complete FEM calibration data.
A cursory review of the FEM setup will be discussed first. As was mentioned in Section 2.3,
a mesh sensitivity analysis must be performed. The general idea is to run the analysis using a
fairly coarse mesh, to observe where the highest stresses are. Thereafter, the mesh can be
refined in those areas. The mesh is refined to a point at which the solution does not vary by
more than a certain threshold between FEM trials. This can often be a time consuming
exercise. Therefore, to save time, various FEM packages offers the designer an adaptive mesh
refinement feature.
The user is required to generate a global mesh. The study will run, and the areas of maximum
stress will be located. The software will then refine the mesh in those areas automatically;
areas which experience very little stress comparatively will have their mesh coarsened; saving
on simulation time. The other type of mesh refinement does not vary the density of the mesh.
Rather, it changes the order of the interpolation polynomial for the solution. The FEM
process generates discrete solutions for discrete elements within the domain. In the case of
stress, for example, the interpolation polynomial will generate an approximate stress
distribution between two elements [13]. In terms of the mesh sensitivity analysis, the standard
practice was followed. The mesh was initially set to be coarse. This was refined iteratively,
until the resultant displacement (which is the key factor in this design) did not vary by more
than 1% between FEM analyses. The mesh that was used for the simulations had an element
size of 3mm, with 99.7% of all the elements having an aspect ratio of less than 3.
The model was tested by noting the nodal displacement between two points of the balance
under applied loads. An example of this is shown in Figure 3-26.
73
Figure 3-26 – Probe result
This view shows the deformed result of the balance under a full scale normal force. A Probe
Plot Tool available in the software package will output the displacement between two
selected points on the model. In this case, one of the pitching moment sensors of the pair was
determined.
The sensor locations for Concept 6 and Concept 7 are the same, and can be seen in Figure
3-12 to Figure 3-23 on pages 56 to 61. These sensor locations were used in the FEM
simulations. The unloaded nodal displacement between the two points in Figure 3-26 is 8mm.
Under a full scale normal force, this distance increased from 8mm to 8.043541mm, which is a
net deformation of 0.043541mm. If a fibre optic strand were spanned across this gap, it would
experience a strain of:
∆𝑙 0.043541
𝜀= = = 5 442𝜇𝜀
𝑙 8
The Bragg wavelength shift for a fibre experiencing this strain would be:
1.2 × 10−12 𝑚
∆𝜆𝐿 = 𝐶𝐹 × 𝜀 = ( ) (5 442 × 10−6 𝜀) = 6.5304𝑛𝑚
10−6 𝜀
CF is the correction factor of the fibre, which is approximately 1.2pm/𝜇𝜀. That is, for every
one micro-strain experienced by the fibre, the Bragg wavelength shift will be 1.2pm.
74
In order to compensate for interactions, a pair of readings must be taken. The method by
which the FEM data was obtained is best explained using an example and an image (refer to
Figure 3-27). If a full scale normal force was applied to the balance, one fibre optic sensor of
the pair would experience a tensile strain and the other would experience a compressive
strain. The platform in Figure 3-27 is displaced downwards by approximately 0.0435mm. In
terms of this FEM model, one side would experience a positive deformation – its length
would increase (right hand side of the image), and the other side would experience a negative
deformation – its length would decrease (left hand side of the image). The values of the
respective deformations were divided by the unloaded reference dimension (8mm) to obtain
the respective strains in each of the two fibres. The two strain values were then subtracted
from each other. This value is the resultant strain. The order of magnitude being used for the
strain is 10-6. OFBG sensors relate the change in Bragg wavelength to applied strain by the
following relation: The Bragg wavelength will shift 1.2pm, for 1µ𝜀 [20]. This will give the
total Bragg wavelength shift in nanometres (or ∆𝜆𝐿 , as discussed in Section 2.4.3.1), for the
sensor pair, under the applied load.
Deformation of Deformation of
- 0.043540mm 0.043562mm
∆𝑙 −0.043540
𝜀= = = −5 442𝜇𝜀
𝑙 8
75
∆𝑙 0.043562
𝜀= = = 5 445𝜇𝜀
𝑙 8
1.2 × 10−12 𝑚
∆𝜆𝐿 = 𝐶𝐹 × 𝜀 = ( ) (10 887 × 10−6 𝜀) = 13.06𝑛𝑚
10−6 𝜀
This is the Bragg wavelength shift for the normal force sensor, when a normal force is applied
in the z-direction (downwards). The sign of the Bragg wavelength shift indicates the
directionality of the force. With the normal force still applied, the outputs for all of the other
sensors were determined, using exactly the same steps described above. The other load
sensors’ outputs were then divided by the normal force sensor’s output, and the result was
multiplied by 100. This gave the output response of all of the other component’s sensors as a
percentage of the normal force sensor, when a normal force is applied; i.e. the interactions.
This method was repeated for each load component. Using this data, the component
interactions of full scale loads could be determined. This method was repeated using
intermediary loads, in order to test the response of the balance under different load
conditions. Finally, the balance was tested using all of the possible two load combinations,
and its responses were recorded. The response the balance would have to a load in a particular
direction would be the same if the direction of the load were reversed (albeit, the sign would
be different).
The relative resolution of the balance is an important aspect; one of the main reasons for this
design is to determine if OFBG sensors can be a viable alternative to strain gauge sensors in
wind tunnel balances. As such, a table was compiled, showing the relative resolution for each
load component, as a percentage of full scale (See Table 3-20). The smallest strain value
which can be measured by the interrogator is approximately 2µ𝜀. The results shown in Table
3-20 represent the values which are expected, based upon the results obtain from the FEM
analyses. It can be seen that all but one of the sensors offer a resolution which is higher than
0.02% (The relative resolution commonly available from strain gauge balances) [1]. These
results further show the merit of using OFBG sensors in wind tunnel balances.
76
Table 3-20 – Relative resolution
Relative
Induced fibre Smallest
FS Load resolution
strain resolvable load
[% of FS]
Pitching
400Nm 8 093 µ𝜀 0.08Nm 0.020%
moment
Rolling
260Nm 6 347 µ𝜀 0.06Nm 0.026%
moment
Yawing
225Nm 12 649 µ𝜀 0.03Nm 0.013%
moment
There are some shortcomings of using FEM to analyse a model. The material properties are
approximated to be homogeneous and isotropic, which is not necessarily the case in reality.
The FEM model also assumes no manufacturing defects are present. Finally, the FEM model
does not account for mechanical hysteresis which may be present due to the material
properties, or due to the geometry of the balance. Nevertheless, the FEM data which has been
collected for the balance shows that the interactions are acceptably low. The balance also
behaves in a predictable manner under combination loads. It is also easy to manufacture,
compared to the alternatives. This one piece balance design is meant to prove the concept that
the measurement of the displacements within the structure of the balance can be used to
measure the applied forces and moments. As such the FEM data is promising enough to
warrant the manufacture of this design.
3.8 Summary
This chapter described the evolution of Design Concepts 6 and 7. The earlier concepts aimed
to use the mechanics of cantilevers to try decouple load interactions. It was found, however,
that the platform tends to pitch (or roll) under the application of an axial (or side) force.
Therefore, to eliminate this, Concepts 6 and 7 employed cradles to govern the motion of the
77
platform in such a way that the amount of pitch (or roll) under an axial (or side) force is
minimised. The cradles are also designed to offer sufficient stiffness in all directions.
It was determined that the Concept 7 should be the design to be prototyped, as it is the most
practical to manufacture. It also offers higher sensitivity than Concept 6. The manufacturing
considerations were investigated, and it was determined to manufacture the balance using
multiple pieces of material, and then weld them together and heat treat the final assembly.
Finally, a brief outline of the finite element analysis was given. A full report of the FEM for
the balance is given in Appendix B.
78
Chapter 4 - Manufacturing
4.1 Introduction
There were several components which had to be manufactured to complete the project. The
balance itself was 3D printed in order to refine the design. Once the design was finalised, the
stainless steel model was manufactured. A new calibration body was designed and
manufactured in order to apply the loads to the balance, and a mounting bracket was made
such that the balance could interface correctly with the calibration rig at the CSIR.
79
4.3 Stainless steel prototype
After noting and correcting a few flaws in the design by studying the printed prototype, the
steel balance was manufactured using 17-4PH steel. This can be seen in Figure 4-2. The
intricate cuts, such as the cradles and sensor pillars were wire cut. The base and platform were
milled. There was a manufacturing defect during the wiring cutting of the cradles. The slit,
shown in Figure 4-3, was cut 0.1mm too wide. This was a systemic error, and all four cradles
have the same defect.
80
Figure 4-3 - This slit was cut 0.1mm too wide
The balance cradles were MIG welded to the platform and base. The components were
manufactured such that there was a 2mm fillet which could be filled by the weld. It was first
considered to use electron-beam welding or laser welding, but this would have added to the
complexity and cost of the design. The decision was therefore taken to design the balance to
be MIG welded. Welding causes the material to deform, due to the heat, and the part will be
left with a heat affected zone, in which the material properties differ from the original
condition [35]. To minimise the effect of thermal distortion during welding, the balance was
welded in sections. One short section would be welded, and left too cool. Another section was
welded and left to cool, and so on. The balance was welded before being taken for heat
treatment.
The steel was heat treated to Condition H900 after welding. The material properties are
altered as it is heat treated. The material was supplied in Condition A, which has a yield
strength of 760MPa [37]. Heat treating to Condition H900 increases the yield strength of
1 240MPa [37]. Table 3-19 shows the mechanical properties of 17-4PH in Condition H900.
81
The heat treatment changed the physical appearance of the steel from a silver colour with a
bright sheen to a dull brown. This is to be expected with this heat treatment. Welding also left
some beads which had to be machined flat.
Heat treatment relieves some of the stresses caused due to welding, but the process also
causes thermal expansion. The part must therefore be cooled uniformly to ensure that the
amount of deformation caused by the thermal expansion and contraction is minimal. It was
recommended by the manufacturer that the part be manufactured to the rough dimensions,
heat treated, and then go back for finishing work. Thus, the part will have the fine tolerances
called for in the design, after the part has been for heat treatment. This was not done in this
case, as time and financial constraints were present.
(a)
82
(b)
Figure 4-4 – The balance after heat treatment (a) before having the weld beads
machined flat, and (b) after.
83
Figure 4-5 – Sting extension
Figure 4-6 shows the rear isometric view of the mounting bracket, with threaded holes, to
mount the bracket to the sting extension, and counter bore holes, to bolt the balance to the
mounting bracket using hexagon socket screws. The available material was 6082-O
aluminium. In order to interface correctly with the sting extension, the piece of material
needed for the bracket had to be quite large. The thickness of the material is 50mm, with an
outer diameter of 240mm. Twelve threaded holes were drilled on a pitch circle diameter of
210mm, to interface with the sting extension, and 9 counter bore holes were drilled to
interface with the balance, as per drawing 001-002 in Appendix G.
84
Threaded holes
(Sting extension
mounting holes)
The design of the calibration body was done such that its centre of mass would be located on
the centre of the face of the platform of the balance. This is the BMC. The centre of mass of
the calibration body can be seen in Figure 4-8. Thus, all the forces and moments which act
upon the calibration body should act through this point.
Holes were machined into the ends of the arms of the calibration body, as seen in Figure 4-9
(a). These holes are the positions through which loads were applied. Precision made inserts
were fitted in these holes, as seen in Figure 4-9 (b). These inserts have dimples which will
give a precise point onto which the load hanger can hang.
86
(a)
(b)
87
Figure 4-10 –Calibration assembly
Sting Extension
Bull’s-Eye
Bubble Mounting Bracket
Level
Balance
Calibration Body
88
The balance was manufactured from multiple pieces of material, which were then welded
together, and heat treated. The sensor pillars were bolted to the balance, but not welded. As
further work is done using this balance, new sensor pillars can be used on the design, thus
increasing its flexibility as a research tool. The welding and heat treating were done in such a
way as to minimise the unwanted effects of thermal distortion.
The calibration body was designed and manufactured such that its centre of gravity would
coincide with the balance moment centre. Therefore, all forces acting through the calibration
body would act through the balance moment centre. The general shape was laser cut, and the
rest of the work was done according to the technical drawing with a 5-axis milling machine.
In order for the balance to interface with the sting extension, a mounting bracket was
designed and manufactured. The assembly is shown in Figure 4-11.
89
Chapter 5 - Experimental Setup
5.1 Introduction
This section will focus on how the load tests were set up. This will include a description of
how the fibres were bonded to the balance, the manner in which the loads were applied and
the computer software used to log the data.
A full calibration of the balance is not in this scope of work. Therefore, the load test will
purely serve to demonstrate whether or not the balance conforms to what was predicted in the
FEM study. Therefore, the plan was to load the balance in one component to some percentage
of the full scale design load for that component, unload it again, and measure the results. This
was done for each of the six components. The balance was loaded to about a quarter of the
full scale design loads to avoid damaging it, or the fibres. This was decided as this is a new
design. The balance was loaded and unloaded in each component between three and five
times to determine repeatability, and also to determine and quantify the presence of hysteresis
as a percentage of full scale.
90
This carrier moves the hot wire
along the length of the fibre.
When the output shows an
increase in the Bragg
Wavelength, the Bragg grating
has been found, and its location is
marked.
Figure 5-1 – Hot wire moving over the region of the Bragg grating to determine its exact
location
The fibres were bonded to the balance while it was bolted to the calibration rig. This made the
task of bonding fibres much easier, as the rig could be pitched or rolled so as to allow for easy
access to the points to which the fibres were to be bonded. The adhesive used was X-60, an
epoxy. The Ormocer coated fibres are the same type as used by Burger [10]. The fibre was
bonded to a probe on the sensor pillar, and the X-60 left to set for 15 to 20 minutes. The fibre
was then clamped in a clamp, to pre-tension the fibre such that the Bragg wavelength shift is
approximately 10nm. The fibre is connected to the interrogator during the bonding process to
monitor the amount of pre-tension being applied, and also to identify any problems which
may arise during the process. Refer to Appendix E for an explanation of some of the
problems that arose.
The clamp that was used to add the pre-tension is shown in Figure 5-2. A cable is passed
through the eye of the clamp and then through a pulley. A load pan is hung from the cable,
and applies a load in the direction indicated by the arrow in the figure. This allows for a
constant load to be applied to the fibre during pre-tensioning; even if the fibre slips in the
clamp, the load on the fibre will remain unchanged. While the fibre is pre-tensioned, it is
bonded to the probe on the balance. This method of tensioning also produces a sources of
error, as the load pan was not perfectly still while the pre-tension was being applied. This
may have affected the quality of the bond, and this process should be refined for future work.
91
2. The fibre is then placed into
this clamp. A load is applied, 1. The fibre is bonded here first
which then tensions the fibre. and the adhesive is allowed to
set.
The positions and names of the fibres on the balance are shown in Figure 5-4 to Figure 5-6.
93
Figure 5-5 – Side view, showing two sensor locations
The outputs for each load component are equal to the difference in the responses of two
individual fibres. What is physically being measured is the change in difference in the
wavelengths between two fibres. The equations used to give the desired outputs were:
Table 5-1 – Equations inputted into the DAQ to give the desired output for each load
component
The equation to measure the Normal Force component was later changed to:
The reason for this is explained in Appendix F. In essence, the equation for the Normal Force
shown in Table 5-1 is not robust enough to given an accurate response under the application
of a normal force load. Another point to note is that the location of the fibre used to measure
the rolling and pitching moment is different in reality, than it was in the FEM model. The
reason simply is: it is easier to bond the fibres in the locations shown in Figure 5-4 and Figure
5-5.
95
Figure 5-7 (a) – Calibration body orientation
96
Figure 5-7 (b) – Calibration body load configurations
Figure 5-7 shows the orientation of the calibration body, and indicates the directions of the
forces and moments applied to it. If it is stated that the balance was rolled to, say, 90 degrees,
then the arm labelled 90° in Figure 5-7 (a) would be at the top. Figure 5-7 (b) shows the
positive orientations of the forces and moments which are applied to the balance through the
calibration body. Moments were added to the balance by applying masses on only one arm.
Thus, applying moments included a combined force in that direction, which must be taken
into account. The lever arm, from the centre of the calibration body, to any of the points on
which the loads are applied, is 250mm.
97
5.4 Bull’s-eye bubble level
An assembly called the bull’s eye bubble level was attached to the calibration body, shown in
Figure 5-8. The calibration rig was moved such that the calibration body and platform were
vertical. After the application of a load, the bubble was observed to see if it still centred. This
provided a visual means of ascertaining whether the cradles are effectively cancelling out any
pitching of the platform which may occur under the application of an axial force. The balance
can be rotated on the calibration rig, such that side forces can be applied. The calibration rig
has very precise encoders and stepper motors which can measure the angle of rotation and
rotate the rig to within 0.001°. The bull’s-eye bubble is bonded to a movable collar, which
can rotate around a shaft which is centrally located on the calibration body. Once the
calibration rig has rotated through 90°, the collar is rotated such that the bubble is level with
respect to the ground. The rig itself can then pitch the balance with the calibration assembly
to ensure the bubble is level, with respect to the ground, and thus, the load applicator and
platform, are vertical, with respect to the ground.
Bull’s-eye bubble
level
Movable collar
Shaft
98
force, and a load into/out of the page will be a normal force. On the calibration rig, a pure
axial/side force could be applied to the balance. This is seen in Figure 5-10.
Figure 5-11 shows the application of a yawing moment. Note that the application of a yawing
moment includes a downward combined load as well. Therefore, the yawing moment data
includes a response from the axial (or side) force sensors (depending on the orientation of the
balance). Therefore, to obtain a result for the yawing moment exclusively, the response of the
axial (or side) force sensors must be removed.
99
Figure 5-10 – Application of an axial/side force
100
(a) (b)
Figure 5-12 – (a) Hangers applying a normal force, (b) belt over the pulley.
Note in Figure 5-12 (b) that the belts from both hangers, are loaded with one load pan, so as
to minimise the possibility of applying a pitching/rolling moment combined load when
applying a normal force. A 10kg mass was applied to take up the slack in the belt wrapped
around the pulley.
A pitching or rolling moment is applied when the rod/hanger of only one side has a load
applied. With the application of these loads, there is a normal force combined load present,
and should be accounted for when reviewing the data. Note also that the normal force (and
pitching and rolling moment) components all have the combined load of the mass of the
calibration body, rods and hangers acting downwards. These must be taken into account when
reviewing the interaction data.
The calibration room is air conditioned, and has a positive pressure, so as to reduce the
amount of dust present. The air vents in the room create air currents which can interfere with
the fibres’ responses, as the air moves over them. Therefore, the balance was wrapped with
paper, so as to block out the air moving over the fibres. The paper was positioned such that it
would not touch the balance itself, or touch any of the Bragg gratings.
101
Figure 5-13 – Paper wrapped around the sting extension, to protect the fibres from air
currents
102
1.
2.
3.
The output file generated by the DAQ shows the outputs of each fibre Bragg grating sensor,
as well as the outputs for the user generated equations. The program also allowed for the
continuous logging of samples. These continuous logs were taken over long periods of time,
such as overnight, to determine if there is any drift in the outputs. A repeatability test was also
conducted, to see if the output response changes over time, when a load is applied or
removed. The data presenting in the following section is data captured as single points.
5.7 Summary
This Chapter describes how the Bragg gratings were located and how the fibres were bonded.
There were several problems which arose during the bond process, which are detailed in
103
Appendix E. The notation of the fibres was given, along with the equations used by the DAQ
software to output the desired results. The normal force equation had to be changed after the
data was captured and analysed. The reasons for this are detailed in Appendix F. The type of
force which is applied to the balance depends on its orientation in the calibration rig. This
orientation convention was presented in Figure 5-7. Figure 5-10 to Figure 5-12 shows the
balance being subjected to various loads. The number and type of loads to which the balance
was subjected will be covered in Chapter 6. Finally, a brief description of the software was
given.
104
Chapter 6 - Experimental Results and
Discussion
6.1 Introduction
The aim of this section of the report is to evaluate the performance of the balance design,
based on various metrics which are: linearity, repeatability, hysteresis and interactions. There
will undoubtedly be areas where the balance can be improved, and the results shown in this
section will highlight where those areas are. The goal was to subject the balance to pure
loads1, or as close to pure loads as could be achieved, and measure the sensor outputs. Once
the sensor readings are tabulated, the performance of the balance can be ascertained, with
respect to the abovementioned metrics. This is an important step; it serves to confirm that the
cradles perform as intended. It also serves as the first steps towards a full calibration of the
balance. The full results data for the tests can be found in Appendix C.
The signal noise was found by logging data overnight for each component. For each
component, a noise data sample for a randomly selected one hour duration was taken, and the
standard deviation for this sample set was found. Eleven such random sample sets were
selected, and the standard deviations were calculated. This was done for each component; as
it is assumed that each load set would not take more than one hour. Random samples of more
than one hour in duration will start displaying other effects, such as any temperature
fluctuations in the room.
As an example, Figure 6-1 shows the overnight signal noise for the normal force sensor over
the course of one hour; 3600 discrete data points exist in this sample set. The average mean
output (µ) for the data in Figure 6-1 was found to be -0.021nm, and its standard deviation (𝜎)
1
A pure load refers to a load which is applied exclusively in one direction or orientation.
105
was calculated to be 0.001nm. The noise band for this figure is defined as the two sigma
standard deviation of the noise signal output, so 2 𝜎 = 0.002nm.
-0.016
0 500 1000 1500 2000 2500 3000 3500 4000
-0.017
-0.018
2 𝜎 noise band Mean output Interrogator resolution
Wavelength [nm]
-0.019
-0.02
-0.021
-0.022
-0.023
-0.024
Time [s]
𝑛
1
𝑠=√ ∑(𝑋𝑖 − 𝑋̅)2 (6.1)
𝑛−1
𝑖=1
Where:
- s is the sample standard deviation
- n is the sample size
- 𝑋𝑖 is the ith data point
- 𝑋̅ is the sample mean
106
The population mean of the output response for a load input is defined as µ. It is desired to
find a margin of error, E, such that the probability that µ will lie within the interval of
[𝑋̅ – E, 𝑋̅ + E] is 95%, or
The margin of error for E can be found multiplying the sample standard deviation by the
critical t-value, corresponding to the confidence level of 0.95, and the number of degrees-of-
freedom (n, number of samples in the set) [38]. This is then divided by the square root of the
number of samples in the set. Therefore, the margin of error is given by [38]:
𝑡×𝑠
𝐸= (6.3)
√𝑛
In the tables to follow, the mean outputs will be sample means. The calculations of the
repeatability and back calculated errors will be done using the sample mean. The population
mean for each load set will fall within a given interval, with a 95% confidence. The value of t
is found using a t-distribution table.
The repeatability for each load point (for each component) is defined as the two sigma sample
standard deviation of the output for that load point. The sample standard deviation is divided
by the full scale load response, to give the repeatability as a percentage of full scale load
response.
The back calculated error is a measure of how well the linear regression curve fits the data.
An example of a back calculated error is shown in Figure 6-2. An applied load [for example,
3N] will yield a certain output response [of 0.035nm]. A linear regression curve through the
data would predict that for that output [of 0.035nm], a slightly different load [of 3.33N] had
been applied. The difference between the actual applied load, and the load predicated by the
linear regression curve is the back calculated error.
107
0.045 Sample mean output of
0.035nm for a 3N load.
0.04
0.035
Wavelength Shift [nm]
0.03
Predicted load for a
y = 0.0105x
0.025 0.035nm output
based on the
0.02 equation of the line
0.015 of best fit.
0.01
Back calculated
0.005 error.
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Load [N]
For each component that was tested, both a low and a medium load set were applied. The low
load set was intended to observe the balance response without risk of damaging the balance or
the fibres. The medium load sets that were applied (up to about 25% of the full scale design
load) were intended to observe the balance response with more representative loads.
The low normal force set was applied to the balance while it was positioned in all
orientations, 0°, 90°, 180° and 270° (according to the convention shown in Figure 5-7 on page
97). The loads were applied using calibrated masses. The gravitational acceleration in the
CSIR calibration room is 9.7860992m/s2. The balance was tested first with low loads, from 0-
98N, increasing in 19.5N increments. The balance was subjected to a loading-up and loading-
down process. This was done twice for the balance in the 0° and 180° position, and once each
108
for the balance in the 90° and 270° positions. Once each load was applied, the load pan was
steadied, and the datum point was captured.
In the second (or medium) load set, the balance loads were increased to 1 566N, in 196N
increments, to ascertain if the output response is consistent for both low loads and higher
loads. The medium loads were only up to 15% of the full scale design load. The balance
would have been loaded with more masses, but there simply was not enough space on the
load pan. The normal force was retested later; it was placed on a flat and level surface, and
masses were carefully placed on it. In this case, the balance was loaded to 3 914N. This is
covered in Appendix F. The balance was loaded using medium loads while it was positioned
in the 0° orientation.
Table 6-1 shows the data for the normal force output response. The normal force data was
found by using equation 5.7.
This output is a sum of the averages of the pitching moment and rolling moment sensors’
responses.
NF
3.5
AF y = 0.00197x
SF R² = 0.9989
3 PM
RM
2.5 YM
Linear (NF)
Wavelength Shift [nm]
2 Linear (AF)
Linear (SF)
1.5 Linear (PM)
Linear (RM)
1 Linear (YM)
0.5
0
0 200 400 600 800 1000 1200 1400 1600 1800
-0.5
-1
Load [N]
109
Figure 6-3 shows there are interactions present, especially with the rolling and yawing
moment sensing elements. The rolling moment interaction can be explained by the fact that
the balance was loaded in the 0° orientation. If an imbalance were present in loading the
normal force, one arm of the calibration body may have been loaded more than the other one;
in which case, a rolling moment load would also have been applied. The large yawing
moment can be explained by a fibre misalignment. The errors present in the measurement
(including fibre misalignment) are discussed further in Section 6.12. Figure 6-20 and Figure
6-21 on page 140 shows the fibres for the yawing moment sensor.
Table 6-1 below shows the normal force output response for both low and medium loads. For
each load that was applied, the data point was captured. The average of these data points was
taken, and the sample standard deviation was found for each load point. As the standard
deviation has the same units as the data itself, the repeatability at each point can be found, as
a percentage of the full scale load to which it was subjected for that test. The average
repeatability for the normal force component was found to be 0.691%. For Table 6-1, and for
the tables to follow, the repeatability will be calculated by using the output response for the
largest positive load that was applied. This will be referred to as the full scale load response.
The back calculated errors for each load point were averaged. The average BCE was found to
be 0.505% of the full scale load. The BCE was found using the largest positive load. This is
the full scale applied load. Table 6-1 also shows the 95% confidence interval for the
population mean output response for each load point. This confidence interval is found using
Equations 5.8 to 5.10. The value for n (number of samples in the set) is 2, the value for t is
found to be 12.71, from the t-distribution table in [38]. The full set of data is available in
Appendix C.
110
Table 6-1 – Load point repeatability of normal force sensor, n = 2, t = 12.71
Mean 95%
confidence interval
Sample Repeatability
Lower Upper BCE % of
Load mean 2𝝈 std. dev. % of max
bound bound max load
[N] output [nm] output
[nm] [nm] (1 565.776N)
[nm] (3.029nm)
0 0.001 0.002 -0.002 0.004 0.066 0.032
19.572 0.038 0.006 0.029 0.047 0.198 0.018
39.144 0.075 0.008 0.063 0.087 0.264 0.069
Low loads
58.717 0.112 0.012 0.094 0.130 0.396 0.119
78.289 0.149 0.014 0.128 0.170 0.462 0.170
97.861 0.185 0.016 0.161 0.209 0.528 0.252
111
Table 6-2 – Hysteresis for medium normal force loads, averaged over two load/unload
cycles
Load [N] 0 195.722 391.444 587.166 782.888 978.61 1174.332 1370.054 1565.776
Load up [nm] 0 0.393 0.781 1.165 1.546 1.922 2.291 2.661
Load 3.03
Unload[nm] 0.004 0.483 0.809 1.203 1.583 1.964 2.324 2.685
set 1
Delta [nm] -0.004 -0.091 -0.028 -0.039 -0.037 -0.042 -0.033 -0.023 -
Load up [nm] 0.008 0.397 0.786 1.174 1.558 1.934 2.305 2.669
Load 3.028
Unload [nm] 0.005 0.408 0.803 1.199 1.583 1.963 2.329 2.686
set 2
Delta [nm] 0.004 -0.011 -0.018 -0.025 -0.025 -0.029 -0.025 -0.017 -
Average delta
0 -0.051 -0.023 -0.032 -0.031 -0.036 -0.029 -0.02 -
[nm]
Hysteresis
-0.008 -1.675 -0.751 -1.04 -1.015 -1.172 -0.949 -0.66 0
[%FS]
The hysteresis data are presented in Table 6-2. The hysteresis characteristics are not going to
be discussed in any further detail as it is beyond the scope of this work.
The relative resolution of the balance for the normal force load is calculated by taking the
smallest resolvable part, which is the two sigma standard deviation of the signal noise,
divided by the full scale signal of the applied load (𝑂𝑓𝑠 ,in this case 3.029nm). The highest
load to which the balance was subjected was 160kg, which is 16% of the full scale design
load. The one sigma standard deviation of the signal noise (𝑂𝑛 ) for the normal force sensor
was found to be 1pm. Therefore, the two sigma standard deviation is 2pm. Thus:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.066%
𝑂𝑓𝑠 (3.029 × 10−9 )
As the full scale load was 16% of the full scale design load, the relative resolution for a
normal force of 11 500N (1 175kg) can be estimated. Using the equation of the line of best fit
through the data, the output response is extrapolated to be
The relative resolution for this full scale output response would be:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.009%
𝑂𝑓𝑠 22.655 × 10−9
Therefore, if the balance were loaded to its full scale design load, the relative resolution is
estimated to be 0.009% of full scale.
112
6.5 Axial force
Figure 5-10 on page 100 shows the application of an axial force (this looks the same as the
application of a side force; the orientation of the balance would be different). Figure 5-7
shows the orientation of the calibration body when an axial force is applied. As with the
normal force, the axial force component was loaded using low, and medium loads. A negative
axial force is applied in the 0° position and a positive axial force is applied when the balance
is in the 180° position. The tares in this load test were the calibration body and the load pans.
In the low load test, the axial force component was loaded up to 49N, in 9.8N increments. For
the medium load tests, the balance was loaded in the positive direction up to 196N, in 39N
increments. In the negative direction, the balance was loaded to 235N, in 39N increments. For
each of the load tests, for both low and medium loads, the balance was loaded up and down
twice. For the positive load set, an inclinometer was used to measure the angle of the
platform; to determine if it pitches under the application of an axial force. This is explained
further in Section 6.10.2.
The axial force component was measured using fibres C2R2 and C2R4. The data for the axial
force component are shown graphically in Figure 6-4 and presented in Table 6-3.
4
y = 0.01694x
R² = 0.9999 k
3
2
Wavelength Shift [nm]
0
-300 -250 -200 -150 -100 -50 0 50 100 150 200 250
-1 NF
AF
-2 SF
PM
RM
-3 YM
Linear (NF)
-4 Linear (AF)
Linear (SF)
Linear (PM)
-5
Linear (RM)
Load[N] Linear (YM)
113
Table 6-3 – Load Point Repeatability of axial force sensor, n = 4, t = 3.18 [38]
Mean 95%
confidence interval
Sample Repeatability
2𝝈 std. Lower Upper
Load mean % of max BCE % of max
dev. bound bound
[N] output output load (195.722N)
[nm] [nm] [nm]
[nm] (3.305nm)
-48.930 -0.844 0 -0.844 -0.844 0 0.456
-39.144 -0.675 0.002 -0.677 -0.673 0.061 0.359
-29.358 -0.506 0.004 -0.509 -0.503 0.121 0.262
-19.572 -0.339 0.004 -0.342 -0.336 0.121 0.225
-9.786 -0.171 0.006 -0.176 -0.166 0.182 0.158
Low loads 0 -0.010 0.015 -0.022 0.002 0.454 0.302
9.786 0.146 0.022 0.129 0.163 0.666 0.596
19.572 0.313 0.020 0.297 0.329 0.605 0.559
29.358 0.479 0.018 0.465 0.493 0.545 0.553
39.144 0.643 0.016 0.630 0.656 0.484 0.606
48.930 0.813 0.012 0.803 0.823 0.363 0.479
114
Table 6-4 – Hysteresis, averaged over four load/unload cycles
Load [N] 0 4 8 12 16 20
Load up [nm] 0 0.665 1.327 2.024 2.652
3.311
Load set 1 Unload [nm] 0.030 0.639 1.303 1.969 2.639
Delta [nm] -0.030 0.026 0.024 0.055 0.013 0
Load up [nm] 0.030 0.640 1.306 1.972 2.636
3.299
Unload [nm] 0.0410 0.625 1.291 1.959 2.628
Load set 2 Delta [nm] -0.011 0.015 0.015 0.013 0.008 0
Average Delta -0.021 0.021 0.020 0.034 0.011 0
% of FS -0.525 0.525 0.500 0.85 0.275 0
Table 6-3 shows the data for the low and medium loads. The repeatability was averaged to be
0.574% and the average magnitude of the back calculated error was found to be 0.426%. The
graphs shown in Figure 6-4 indicate that the balance indeed acts linearly in the axial force
direction. The R2 value for the axial force medium load case is 0.9999.
The interactions are much smaller for the axial force test than they were for the normal force.
This is due to the fibre misalignment not being so critical in this test, and the fact that a pure
axial force load could be applied. There is a pitching moment interaction present. This is due
to the platform pitching slightly as an axial force is applied. This is a pure interaction,
characteristic of the balance.
The hysteresis data for the medium loads are shown in Table 6-4. The balance was loaded up
and down four times to obtain this data. The difference between the output responses were
determined for each load point during a load/unload cycle. These differences were averaged.
This average was divided by the average of the full scale load response in order to get the
percentage hysteresis (of full scale) for each load. The hysteresis data is in line with the
expected value of 0.5% of full scale
115
The one sigma signal noise response for the axial force sensor was found to be 1pm. Thus, the
two sigma standard deviation is 2pm. The full scale output for the axial force is 4.006nm :
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.050%
𝑂𝑓𝑠 (4.006 × 10−9 )
If the balance were subjected to a full scale axial force load, the output response is estimated
using the equation of the line found using linear regression. The full scale load would be
900N. The full scale output response is estimated to be
The relative resolution for the axial force for the full scale design load is estimated to be:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.013%
𝑂𝑓𝑠 15.246 × 10−9
8
y = 0.00906x
6 R² = 0.9999
4
Wavelength Shift [nm]
0
-1000 -800 -600 -400 -200 0 200 400 600 800 1000
-2 NF
AF
-4 SF
PM
RM
-6 YM
Linear (NF)
-8 Linear (SF)
Linear (SF)
Load [N] Linear (PM)
Table 6-5 – Load Point Repeatability of side force sensor, n=3, t=4.3
117
Table 6-6 – Hysteresis, averaged over three load/unload cycles
Load [N] 0 195.722 391.444 587.166 782.888
Load up [nm] 0 1.792 3.584 5.348 DNE
Load set 1 Unload [nm] 0.037 1.761 3.552 5.333
Delta [nm] -0.037 0.031 0.032 0.015 -
Load up [nm] 0.037 1.747 3.556 5.332
7.116
Load set 2 Unload [nm] 0.036 1.746 3.546 5.334
Delta [nm] 0.001 0.001 0.01 -0.002 0
Load up [nm] 0.036 1.781 3.537 5.32
7.060
Load set 3 Unload [nm] 0.012 1.765 3.540 5.294
Delta [nm] 0.024 0.016 -0.003 0.026 0
The average repeatability for the balance under the application of a side force load is 0.181%
of full scale. The average of the magnitudes of the back calculated errors is 0.153%.
The hysteresis data are presented in Table 6-6. Three load/unload cycles were conducted; two
for the negative direction, one for the positive direction. The hysteresis data is in line with
what was expected for OFBGs based on [1].
The relative resolution is found for the side force in the same ways as it was for the axial
force. The signal noise response for the side force sensor was found to be 1pm. Thus, the two
sigma standard deviation is 2pm.
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.028%
𝑂𝑓𝑠 7.116 × 10−9
The anticipated full scale load response for 2500N is 22.650nm. The anticipated relative
resolution is:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.009%
𝑂𝑓𝑠 22.650 × 10−9
118
6.7 Pitching moment
The setup for the pitching moment test is similar to that of the normal force. The difference is
that the weight was applied to one rod/hunger pulling only one arm of the calibration body.
Refer to Figure 5-7 (a) of page 97 for the orientation in which the loads are applied. As a
result, the pitching moment has a combined normal force associated with it. This combined
normal force is positive, for both positive and negative pitching moments, and is compensated
for when the interactions are calculated.
The pitching moment tares include the hanger, rod, belt which wraps around the pulley, load
pan and a 6kg mass to take up the slack in the belt. This was taken as the zero point. Loads
were applied to the balance which induced a moment of 12Nm for the low load test,
increasing in 2.5Nm increments, for the low load tests, and up to 49Nm for the medium load
test, increasing in 4.9Nm increments.
For the low load test, the pitching moment response was tested by applying both a positive
and negative pitching moment for the balance orientated in positions 90° and 270°. For the
medium load tests, the balance was tested by applying both positive and negative moments
for the balance orientated at 90° only. The results for the tests are presented in Table 6-7 to
Table 6-8.
2.5
y = 0.03971x
R² = 0.9999
2
1.5
1
Wavelength Shift
0.5
0
-60 -40 -20 0 20 40 60
-0.5
NF
AF
-1 SF
PM
-1.5 RM
YM
-2 Linear (AF)
Linear (SF)
Linear (PM)
-2.5
Linear (RM)
Load [Nm] Linear (YM)
119
2.5
y = 0.0397x
R² = 9999
2
1.5
1
Wavelength Shift
0.5
0
-60 -40 -20 0 20 40 60
-0.5 AF
SF
PM
-1 RM
YM
-1.5 NF Comp
Linear (AF)
Linear (SF)
-2
Linear (PM)
Linear (RM)
-2.5 Linear (YM)
Load [Nm] Linear (NF Comp)
Figure 6-7 – Pitching moment, having compensated for the combined normal force load
120
Table 6-7 – Load point repeatability of pitching moment sensor, n = 4, t = 3.18 [38]
121
Table 6-8 – Hysteresis for the medium pitching moment loads, averaged over four
load/unload cycles
Load [Nm] 0 4.893 9.786 14.679 19.572 24.465 58.717 68.503 78.289 88.075 97.861
Load up
0 0.197 0.395 0.591 0.784 0.983 1.178 1.375 1.570 1.767
[nm] 1.959
Load set 1 Unload [nm] 0.005 0.205 0.406 0.603 0.800 0.995 1.191 1.384 1.577 1.770
Delta [nm] -0.005 -0.008 -0.011 -0.012 -0.016 -0.012 -0.013 -0.009 -0.007 -0.003 0
Load up
0.005 0.204 0.397 0.596 0.792 0.988 1.181 1.371 1.569 1.764
[nm] 1.956
Load set 2 Unload [nm] 0.002 0.202 0.402 0.598 0.794 0.991 1.186 1.379 1.573 1.766
Delta [nm] 0.003 0.002 -0.005 -0.002 -0.002 -0.003 -0.005 -0.008 -0.004 -0.002 0
Average
-0.001 -0.003 -0.008 -0.007 -0.009 -0.007 -0.009 -0.008 -0.006 -0.003 0
delta
% of FS -0.057 -0.170 -0.454 -0.397 -0.511 -0.397 -0.511 -0.454 -0.340 -0.170 0
Load [Nm] 0 -4.893 -9.786 -14.679 -19.572 -24.465 -58.717 -68.503 -78.289 -88.075 -97.861
Load up
0.002 -0.197 0.387 0.581 0.772 0.964 1.156 1.348 1.541 1.732
[nm] 1.924
Load set 3 Unload [nm] 0.006 -0.187 0.379 0.571 0.765 0.957 1.149 1.342 1.534 1.728
Delta [nm] -0.004 0.010 0.008 0.010 0.007 0.007 0.007 0.006 0.007 0.004 0
Load up
0.006 0.186 0.377 0.571 0.765 0.957 1.150 1.343 1.535 1.726
[nm] 1.917
Load set 4 Unload [nm] 0.001 0.186 0.379 0.570 0.763 0.955 1.148 1.340 1.531 1.725
Delta [nm] 0.005 0 -0.002 0.001 0.002 0.002 0.002 0.003 0.004 0.001 0
Average
-0.001 0.005 -0.003 -0.005 -0.005 -0.005 -0.004 -0.004 -0.005 -0.003 0
delta
% of FS -0.057 0.284 -0.170 -0.284 -0.284 -0.284 -0.227 -0.227 -0.284 -0.17 0
Figure 6-6 shows the pitching moment with the clear combined normal force load. The
normal force load was compensated out and this result is shown in Figure 6-7. The equation
for the normal force response was estimated in Figure 6-3; the slope of the least squares line
through the data was found to be -0.0019x. To obtain the graph in Figure 6-7, this slope was
multiplied by the load (-0.0019 × 9.786N, for example), and this was subtracted from the
normal force response for that load. This was done for each load.
Table 6-7 shows the results of the pitching moment tests. The behaviour is linear, with the R 2
value of 0.9999. The repeatability of the balance was averaged to be 0.716% of full scale, and
the back calculated error had an average of 0.326%.
The hysteresis for the pitching moment is calculated in Table 6-8. The hysteresis is in the
region of about 0.5% of full scale; this is in line with expectations from [1]. There seems to
be an interaction with respect to the axial force. This was predicted by the FEM model with a
value of 13.05%, while the actual interaction is 16.79% of full scale.
122
The one sigma standard deviation for the signal noise for the pitching moment is 1pm.
Therefore, the two sigma standard deviation is 2pm. The relative resolution for the pitching
moment is
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.104%
𝑂𝑓𝑠 1.921 × 10−9
The anticipated full scale load response (for a pitching moment of 400Nm) is 7.942nm. The
anticipated relative resolution for this output response is:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.025%
𝑂𝑓𝑠 7.942 × 10−9
2
y = 0.04630x
R² = 0.9998
1.5
1
Wavelength Shift [nm]
0.5
0
-40 -30 -20 -10 0 10 20 30 40
-0.5 NF
AF
SF
-1 PM
RM
-1.5 YM
Linear (AF)
Linear (SF)
-2 Linear (PM)
Load [Nm] Linear (RM)
Linear (YM)
1
Wavelength Shift [nm]
0.5
0
-40 -30 -20 -10 0 10 20 30 40
-0.5
NF Comp
AF
SF
-1 PM
RM
YM
-1.5 Linear (NF Comp)
Linear (AF)
Linear (PM)
-2 Linear (RM)
Linear (RM)
Load [Nm}] Linear (YM)
Figure 6-9 – Rolling moment, having compensated for the combined normal force load
124
Table 6-9 – Load point repeatability of rolling moment sensor, n = 4, t = 3.18 [38]
Mean 95%
confidence interval
Sample Repeatability
Lower Upper BCE % of
mean 2𝝈 std. % of max
Load [Nm] bound bound max load
output dev. [nm] output
[nm] [nm] (34.251Nm)
[nm] (1.569nm)
-12.233 -0.554 0.020 -0.570 -0.538 1.275 0.774
-9.786 -0.440 0.006 -0.445 -0.435 0.382 0.820
-7.340 -0.330 0.008 -0.336 -0.324 0.510 0.616
-4.893 -0.220 0.008 -0.226 -0.214 0.510 0.410
-2.447 -0.096 0.072 -0.153 -0.039 4.589 1.089
Low
0 0.001 0.004 -0.002 0.004 0.255 0.063
loads
2.447 0.121 0.014 0.110 0.132 0.892 0.487
4.893 0.233 0.006 0.228 0.238 0.382 0.410
7.340 0.348 0.006 0.343 0.353 0.382 0.519
9.786 0.462 0.008 0.456 0.468 0.510 0.568
12.233 0.573 0.004 0.570 0.576 0.255 0.425
125
Table 6-10 – Hysteresis for the medium rolling moment loads, averaged over four
load/unload cycles
As with the pitching moment, there is a combined normal force, with the application of a
rolling moment. This was compensated for using the same method as that used for the
pitching moment. There is an interaction, with respect to the side force sensor, which was
predicted by the FEM. The graph for the normal force load, with the normal force combine
load taken out is shown in Figure 6-9. The output of the rolling moment sensor is linear, with
an R2 value of 0.9998.
The repeatability of the rolling moment sensors is averaged to be 0.868% of full scale. The
back calculated error is averaged to be 1.011%, which is much higher than the BCE’s
obtained for the components covered in the previous sections. The hysteresis data is shown in
Table 6-10.
The one sigma standard deviation for the noise signal is 1pm. Thus the two sigma standard
deviation is 2pm. Therefore, the relative resolution for the rolling moment is:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.130%
𝑂𝑓𝑠 1.542 × 10−9
126
The estimated full scale load response (for a rolling moment of 250Nm) is: 5.789nm. The
estimated relative resolution for the application of a full scale load is:
𝑂𝑛 2 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.035%
𝑂𝑓𝑠 5.789 × 10−9
The medium load tests were conducted for the balance in position 0°. For the medium load
test, the balance was loaded to 49Nm, increasing in 4.9Nm increments. The balance was
loaded with both a positive and a negative yawing moment. As the medium load test was
done for the balance in position 0°, Figure 6-10 shows only a large axial force combined load.
This was compensated for using the same method as described above. The slope of the line of
best fit for the axial force data was found, and is shown in Figure 6-4. This slope is multiplied
with the applied combined load, in Newton’s, to obtain the output wavelength shift, in
nanometres. This value was then subtracted from the axial force response obtained from the
application of a yawing moment. Thus, the axial force load was compensated out.
127
NF
5
AF y = 0.08553x
SF R² = 0.9998
PM 4
RM
YM 3
Linear (NF)
Linear (SF) 2
Wavelength Shift [nm]
Linear (PM)
Linear (RM) 1
Linear (YM)
0
-60 -40 -20 0 20 40 60
-1
-2
-3
-4
-5
Load [Nm}
5
y = 0.08553x
R² = 0.9998
4
1
Wavelength Shift [nm]
0
-60 -40 -20 0 20 40 60
-1 NF
SF
PM
-2
RM
YM
-3 AF Comp
Linear (NF)
-4 Linear (SF)
Linear (PM)
-5 Linear (RM)
Linear (YM)
Load [Nm] Linear (AF Comp)
Figure 6-11 - Yawing moment, having compensated for the combined axial force load
128
Table 6-11 – Load point repeatability of yawing moment sensor, n = 4, t = 3.18 [38]
Mean 95%
confidence interval
Sample Repeatability BCE % of
Lower Upper
mean 2𝝈 std. % of max max load
Load [Nm] bound bound
output dev. [nm] output (34.251Nm
[nm] [nm]
[nm] (4.208nm) )
-12.233 -1.039 0.032 -1.064 -1.014 0.760 0.380
-9.786 -0.831 0.030 -0.855 -0.807 0.713 0.356
-7.340 -0.624 0.022 -0.641 -0.607 0.523 0.261
-4.893 -0.417 0.014 -0.428 -0.406 0.333 0.166
-2.447 -0.208 0.010 -0.216 -0.200 0.238 0.119
Low loads 0 -0.007 0.064 -0.058 0.044 1.521 0.760
2.447 0.204 0.006 0.199 0.209 0.143 0.071
4.893 0.412 0.004 0.409 0.415 0.095 0.048
7.340 0.619 0.004 0.616 0.622 0.095 0.048
9.786 0.828 0.006 0.823 0.833 0.143 0.071
12.233 1.039 0.038 1.009 1.069 0.903 0.452
129
Table 6-12 – Hysteresis for the medium yawing moment loads, averaged over four
load/unload cycles
Load [Nm] 0 4.893 9.786 14.679 19.572 24.465 58.717 68.503 78.289 88.075 97.861
Load up
0 0.424 0.847 1.271 1.693 2.111 2.534 2.953 3.373 3.795
Load [nm] 4.211
set 1 Unload [nm] 0.016 0.406 0.831 1.251 1.679 2.101 2.525 2.949 3.371 3.792
Delta [nm] -0.016 0.018 0.016 0.02 0.014 0.01 0.009 0.004 0.002 0.003 0
Load up
0.016 0.408 0.832 1.257 1.68 2.102 2.525 2.948 3.368 3.785
Load [nm] 4.205
set 2 Unload [nm] 0.022 0.401 0.824 1.249 1.674 2.097 2.518 2.933 3.363 3.785
Delta [nm] -0.006 0.007 0.008 0.008 0.006 0.005 0.007 0.015 0.005 0 0
Average
-0.003 0.004 0.004 0.004 0.003 0.003 0.004 0.008 0.003 0 0
delta
% of FS -0.072 0.096 0.096 0.096 0.072 0.072 0.096 0.191 0.072 -0.072 0
Load [Nm] 0 -4.893 -9.786 -14.67 -19.57 -24.46 -58.71 -68.50 -78.28 -88.07 -97.86
Load up
0.022 0.417 0.833 1.246 1.664 2.079 2.494 2.911 3.329 3.741
Load [nm] 4.167
set 3 Unload [nm] 0.004 0.675 0.836 1.252 1.667 2.084 2.5 2.915 3.332 3.748
Delta [nm] 0.018 -0.258 -0.003 -0.006 -0.003 -0.005 -0.006 -0.004 -0.003 -0.007 0
Load up
0.004 0.423 0.839 1.255 1.673 2.086 2.502 2.918 3.329 3.745
Load [nm] 4.156
set 4 Unload [nm] 0.001 0.411 0.826 1.24 1.66 2.075 2.49 2.905 3.32 3.739
Delta [nm] 0.003 0.012 0.013 0.015 0.013 0.011 0.012 0.013 0.009 0.006 0
Average
0.002 0.006 0.007 0.008 0.007 0.006 0.006 0.007 0.005 0.003 0
delta
% of FS 0.143 0.167 0.191 0.167 0.143 0.143 0.167 0.119 0.072 0.143 0
The highlighted value in the Table 6-12 is out of line with the rest of the data for a yawing
moment of magnitude 4.9Nm. The magnitude of the value output for the yawing moment
sensor for this yawing moment is about 0.4nm. This value of 0.675nm can be attributed to a
loading error.
The repeatability for the yawing moment component is averaged to be 0.251% of the full
scale load response. The back calculated error’s magnitudes for the linear regression is
averaged to be 0.292% of the full scale load. The hysteresis for the yawing moment sensor is
in the region of 0.2%, which is better than the expected value.
The noise signal response one sigma standard deviation is 2pm. The two sigma standard
deviation is 4pm. The yawing moment relative resolution is:
𝑂𝑛 4 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.096%
𝑂𝑓𝑠 4.162 × 10−9
The estimated full scale load response for the application of 225Nm is: 9.622nm.Therefore,
the anticipated full scale relative resolution is:
130
𝑂𝑛 4 × 10−12
𝑅𝑅𝑒𝑠 = × 100 = × 100 = 0.042%
𝑂𝑓𝑠 9.622 × 10−9
6.10 Interactions
With the normal force load, the interaction of the rolling moment sensor is due to loading
error; the one arm of the calibration body was carrying more load than the other. This loading
error can be eliminated by placing the balance on a flat and level surface and stacking the
weights on top of the balance, as was done in Appendix E. The yawing moment interaction is
due to a fibre misalignment.
When an axial force is applied, there is a pitching moment interaction of 16.7%. This is a
pure interaction which is a characteristic of the balance. The cradle design aimed to minimise
this interaction as much as possible. The same is true for the rolling moment interaction under
the application of a side force, although, this interaction is more significant, at 75.9%. This
interaction is much larger than was expected based on the FEM results. It is possible that the
manufacturing defect mentioned in Sections 4.3 and 6.12.1 could have been the main
contributor to this large interaction. This manufacturing defect was on the cradle that governs
the motion of the platform under the application of a side force. The yawing moment
interaction, under the application of a side force is due to a fibre misalignment.
131
Under the application of a pitching moment, a large normal force interaction is sensed. This is
due to a loading error, as the pitching moment is applied with a combined normal force load.
The large axial force was predicted by FEM, and is explained in Section 3.4.2.
Figure 6-12 – Test to determine platform pitch (or roll), under the application of an
axial (or side) force.
132
Table 6-14 – Platform pitch (or roll) under the application of an axial (or side) force
6.11 Repeatability
Repeatability is the measure of the ability of an instrument to produce the same output for the
same, but independently applied input [15]. In order to ascertain the repeatability of the
balance, an axial force load of 195.7N was applied and removed ten times, and the value
output was recorded each time. The result of this can be seen in Table 6-15. This output
response is for the axial force sensor. The output response was averaged, and the standard
deviation was found. The weight was put on for the eleventh time, and left for roughly five
minutes. The output was recorded continuously. This data produced a time-response plot as
seen in Figure 6-13. The weight was then removed and the output was recorded again. This
produced a time-response plot which can be seen in Figure 6-14.
Load Point Out response for 0N [nm] Out response for 195.7N [nm]
1 -4.019 -7.618
2 -3.996 -7.622
3 -3.997 -7.625
4 -3.990 -7.616
5 -3.994 -7.622
6 -3.994 -7.618
7 -3.989 -7.626
8 -3.988 -7.616
9 -3.98 -7.623
10 -3.982 -7.627
11 - -7.620
Mean response [nm] -3.993 -7.621
Standard deviation [nm] 0.006 0.004
Repeatability % max output 0.150 0.052
133
Table 6-15 shows the axial force sensor output with the application of a 195.7N load in the
axial force direction. The data indicates that balance output is repeatable to within 0.15%. The
first data point for zero load was taken to be -4.019nm. This is an outlier of the data. If this
point were excluded from the data set (the reason for this will be explained in the following
paragraphs), then the data would have a mean of -3.990nm and a standard deviation of
0.074% of full scale. This is slightly better than the repeatability found by Burger [10]. An
explanation for the possible cause of the initial zero load output being so far removed from
the rest of the data will be explained.
The data shown in Figure 6-13 and Figure 6-14 show the same problem noticed by Burger
[10]. This data indicates that once the mass is applied, and left applied, the reading tends to
drift, over a course of about five minutes. This drift exhibits a time dependent logarithmic
decay. This drift is also noticeable when looking at the continuous plot of the axial force
load/unload cycle. For each load that is applied, there is a distinct step in the graph of Figure
6-15. On closer inspection of one of these steps, the drift is observed. This same output was
noticed by Burger as well [10]. It is for this reason that acrylate coated fibres were to be used
for this balance. Appendix E explains the reasons why Ormocer fibres were used instead.
The fact that the initial zero load output was an outlier from the rest of the data suggests that
the problem of fibre creep may be the cause. During the test, the load was applied, removed
and applied again without giving the fibre a chance to relax into the position it was in before
the test began. Looking at Figure 6-14 serves to strengthen this hypothesis; as the load is
removed, and left for a few minutes, the sensor output tends to creep back to the value it was
at before the test began.
This fibre creep characteristic is unusual in the following sense. When a load is applied, the
fibre is strained to some value. When the load is left on the balance for some time, the strain
in the fibre relaxes. When the load is removed, the fibre strain decreases to some value. When
left unload for some time, the strain in the fibre tends to increase with a time dependent
logarithmic trend. This is an interesting phenomenon which should be investigated further.
134
-7.5
0 50 100 150 200 250 300 350 400
-7.52
y = 0.0164ln(x) - 7.656
-7.54 R² = 0.9496
Wavelength Shift [nm]
-7.56
-7.58
-7.6
-7.62
-7.64
Figure 6-13 – Axial Force sensor response, with a 20kg applied load over 300 seconds
-3.93
0 50 100 150 200 250 300 350 400
-3.94
Wavelength Shift [nm]
-3.95
Axial force sensor output
Log. (Axial force sensor output)
-3.96
-3.97
y = -0.009ln(x) - 3.9333
R² = 0.9073
-3.98
-3.99
Time [s]
Figure 6-14 – Axial Force sensor response, with the 20kg load removed, over 300
seconds
135
0.5
0
0 100 200 300 400 500 600 700 800
-0.5
-1
Wavelength Shift [nm]
-1.5
-2
-2.5
-3
-3.5
-4
-4.5
Time [s]
-3.75
0 10 20 30 40 50 60 70 80
-3.8
Wavelength Shift [nm]
-3.85
-3.9
-3.95
-4
-4.05
Time [s]
136
6.12 Error analysis
There are several problems which led to errors in the results, and they will be explained in
this section. For the most part, these errors can be linked to precision in manufacturing,
alignment of fibres, loading errors and fibre creep.
6.12.1 Manufacturing
It was mentioned in Chapter 4 that there was a machining error with one of the slits in the
cradles. Figure 6-17 show a close up view of one of the cradles. The encircled slit was cut
0.1mm too wide. It was decided to accept the part in this condition for the reason of time and
financial constraints.
A FEM study was done to ascertain what influence this machining defect would have on the
performance of the part. The main influence, it was thought, would be the interaction of a
rolling moment, under the application of a side force. The FEM was conducted using a full
scale side force for the balance with the machining defect, and for the one without. It was
found that without the machining defect, the rolling moment sensor picked up an interaction
of 0.003% of full scale. With the machining defect, the rolling moment sensor picked up an
interaction of 0.098%. This manufacturing defect may well have contributed the increased
rolling moment interaction experienced by the rolling moment sensor under the application of
a side force.
137
6.12.2 Fibre bonding errors
The FEM model has protrusions on the balance and sensor pillar which closely resemble the
probes onto which the fibres were bonding on the physical balance. With the FEM model,
however, these protrusions were taken as datum points, and they line-up with each other
perfectly; any movement between these datum points would then be due to the forces acting
on the balance. Refer to Figure 6-18.
On the physical balance, however, provision must be made to accommodate the fibre. The
fibre is 0.4mm in diameter. Once the fibre is bonded, the thickness of the adhesive will also
have an influence on the eventual position of the fibre. Applying the adhesive with a
consistent thickness is very difficult. Even more challenging is bonding the fibre once the pre-
tension has been applied. The adhesive must completely encase the fibre; there must be glue
between the fibre and the balance. This is crucial for a good quality bond.
Figure 6-18 – These datum points are perfectly aligned in the FEM model
A further difficulty is the fact that two sensor locations had fibres bonded directly next to one
another, as seen in Figure 6-19 and Figure 6-20. This meant that one fibre was bonded and
pre-tensioned. The fibre was then looped around, and the second Bragg grating was bonded in
the gap directly adjacent to the first one. This creates two problems. The fibre is now being
bonded to adhesive, and not steel. This adds another variable, the interaction of the X-60
bonding to already cured adhesive; it is unwanted, generally, to add more variables to the
138
measurement than necessary. The second problem with bonding fibres between adjacent gaps
is that it causes a large misalignment. In Figure 6-19, the second fibre to be bonded was 3mm
away from the first fibre. This would contribute to the interaction picked up by the normal
force sensor, under a pitching and rolling moment. In Figure 6-20 and Figure 6-21, there is
another example of fibre misalignment which would lead to the yawing moment interaction,
under the application of a normal force.
Figure 3-9 and Figure 3-10 further serve to illustrate the point that misalignments will cause
large errors. Therefore, the misalignments shown in Figure 6-19 Figure 6-20 are significant.
Figure 6-19 – Fibres bonded next to one another. The Bragg grating is marked in red.
139
1.
2.
140
Figure 6-21 – Close-up view of the fibres bonded on top of one another
Figure 6-21 shows yet another source of uncertainty. There is a slight amount of X-60 bonded
on the fibre itself (encircled in yellow). This may have an influence on how the fibre reacts
under the application of a load. This may serve to decrease the gap width of the Bragg
grating, such that the strain experienced by the Bragg grating is increased. The large
interaction experience by the rolling moment sensor pair under a normal force load indicates
that the rolling moment fibres are not being strained by the same amount when a normal force
is applied.
Misalignment of the bonded fibres aside, a large source of uncertainty is the fact that the
interaction between the fibre and the coating, and the coating and the adhesive is unknown. It
was shown that there is a significant amount of fibre creep in Section 6.11. Notes were taken
during the process of bonding fibres to the balance. Refer to Appendix E. These notes may
serve to explain from where some of the errors may have originated.
141
type of error would be to perform multiple loading tests. Also, it would be beneficial to load
the balance in a random sequence. This would eliminate drift in the measurement [15].
The CSIR calibration rig was designed primarily for sting and side wall balance calibration,
not for a platform balance. It would be best to calibrate the balance in the orientation that
most closely resembles how it would be used in practice. This balance was calibrated with the
platform vertical. This increased the challenge in terms of obtaining clean data. However, the
CSIR calibration room was the best available location; it has the tools and equipment
necessary to perform the tests. For future work, it is recommended that a calibration rig be
designed to accommodate a platform type balance.
6.13 Summary
This chapter started off by defining the signal noise output, and the smallest resolvable part;
that is essentially determining the noise band for each sensor. It then covered the results for
the balance for each component. The balance loaded a small number of times for each
component. Therefore, the mean output for each load point for each load set was found within
a 95% confidence interval. This was done using the t-distribution table.
If was found that the interactions for each component were larger than those predicted by the
FEM. This is hypothesised due to the contributions of fibre misalignment, loading errors and
pure interactions in the balance itself. The repeatability of each component was found to be
within 0.868% for all components; higher than the sought-after value of 0.1%. The hysteresis
data was presented, but it is beyond the scope of this study to investigate the sources and
remedies for the hysteresis present.
NF AF SF PM RM YM
Repeatability [%FS] 0.691 0.574 0.181 0.716 0.868 0.501
Average BCE [%FS] 0.505 0.426 0.153 0.326 1.011 0.292
R2 Value 0.9989 0.9999 0.9999 0.9999 0.9998 0.9998
142
Chapter 7 - Conclusions and Future Work
Burger [10] concluded that fibres retro-fitted onto a strain gauge balance could not out-
perform strain gauges; thus, for OFBG sensor to be considered as a viable alternative to strain
gauges, a balance should be designed to use these sensors a priori. Literature was surveyed to
gain an understanding of balance design and the sensor technology. A large part of the survey
included the current research in terms of using optical fibres on wind tunnel balances.
The design phase started; it was hypothesised that the use of the two groove method of strain
measurement, described by Pieterse [1] would be used as the main principle of the design. It
was also thought that the design could, in essence, have a sting orientated vertically, with a
platform on top. The fibres could then be spanned over grooves in the sting, and thus measure
forces.
The problem, it was found, was that the platform tends to sag/be displaced vertically (pitch or
roll) under the application of a horizontal axial/side force load. This resulted in predicted
interactions on the order of 100% or more. It is poor practice to design a balance with such
large interactions. A new design was tried, whereby the platform is supported by so-called
cradles. These cradles control the motion of the platform, such that, under the application of
an axial/side force, the balance would not pitch/roll. Also, these cradles would provide
enough stiffness for the intended purpose for force and moment measurement in a wind
tunnel setting. A FEM analysis was conducted, after which, it was decided to manufacture a
prototype.
The prototype was manufactured, along with a calibration body and mounting plate, such that
it could be tested at the CSIR calibration room. Bonding of the fibres proved to be a difficult
task, which may have contributed to in errors in the measurement. Time constraints limited
143
the amount of testing which could be conducted. The data was captured and analysed. The
report was written, which included the results, and the conclusions drawn from them.
This balance is the first of its kind; no other platform balance could be found which operates
using a similar principle, that of having the platform supported by cradles or pillars, and
measuring strain in the manner this balance does, using OFBG sensors. As with any new
technology, there will be teething troubles, and this was the case with this project as well.
This balance is not fit for test usage in a wind tunnel as yet. However, further research in this
field will develop the technology, and solutions to the problems will be found. This balance
serves to prove the concept that a platform balance can be manufactured using OFBG sensors.
7.2 Conclusions
The problem statement for this dissertation was formulated in Section 1.1. The aim was to
design a wind tunnel platform balance which uses OFBG sensors, and to test and evaluate the
performance of the prototype. Chapter 3 details the design process, Chapter 4 describes the
manufacturing and Chapter 5 explains the test set-up and the results are presented in Chapter
6. With respect to the results, the following conclusions can be put forward:
1. All of the components individually showed a linear relationship between the input
load and the output response of the Bragg grating.
2. The hysteresis for all components was in the region of the expected range of 0.3% of
full scale, as found by Pieterse [1].
3. Repeatability of the balance was found to be within 0.868% of the full scale output
response for any component. This is based on a 2𝜎 standard deviation of the mean
outputs. The back calculated errors for the linear regressions through the scatter plots
of the data for each component are in the region of 0.5%. The side force linear
regression BCE was 1.011% of full scale. The uncertainty of the system should
generally be no larger than 0.1% of full scale [1, 10, 24]. The back calculated errors
mentioned in [1, 10, 24] are referring to the back calculated error for the calibration
matrix of the balance itself. The repeatability information indicated in the results
indicate that the balance does not yet provide accurate enough results to be used in a
wind tunnel. Part of the uncertainty is attributed to the fibre creep discussed in Section
6.11.
4. The interactions were higher than the results predicted by FEM. This is to be
expected, as FEM works in a perfect environment; devoid of error. The largest
144
interaction – that of the rolling moment sensor – when a side force is applied, is due to
the manufacturing error on the cradle.
5. There were errors associated with respect to the bonding of the fibres. Misalignment
of the fibres add to the uncertainty, and increase their susceptibility to interactions.
The physical interaction between the bonding agent and the fibre coating, and the
interaction between the coating and the glass fibre itself are unknown. This is the
subject of an on-going investigation.
6. The fibre creep detected by Burger [10] was present in this test as well. This creates a
problem in terms of increased uncertainty and hysteresis, and decreased repeatability.
1. Investigate the use of different fibre coatings and adhesives for use on wind tunnel
balances. This was done in this project (see Appendix E), but not in a rigorous enough
setting to draw final conclusions. The aim of this research would be to identify the
source, and possible remedy for the fibre creep reported in Section 6.11.
145
2. Investigate the thermal characteristics of the balance. In particular, to see if the two-
groove method adequately compensates thermal effects, of if another compensation
technique is required.
3. Investigate methods by which the fibres can be pre-tensioned accurately. Clamping
the fibre in a clamp, and hanging a load pan from the clamp does not provide the
accuracy required for a precision measuring instrument of this kind.
4. Investigate methods by which the fibre can be positioned accurately. Fibre
misalignment results in errors in the output and this was perhaps the major contributor
for not meeting the original design goals.
5. Test and/or calibrate the balance using a rig that was designed specifically for
platform balances. This may serve to reduce loading error seen in this test.
In conclusion, the project set out to determine if a wind tunnel balance can be manufactured
to make use of OFBG sensors. There were problems that arose, and that need to be addressed
before optical fibre sensors can become competitive against strain gauges. The balance
designed does offer distinct advantages, which makes the prospect of further research
attractive. As a proof of concept, this project was successful, and with further funding and
research, this concept of balance design may become a viable alternative to strain gauge
platform balances.
146
Bibliography
[1] F. Pieterse, "The application of optical fibre Bragg grating sensors to an internal wind
tunnel balance", Johannesburg: University of Johannesburg, 2010.
[2] “Summary report of the First International Symposium on Strain Gauge Balances and
Workshop on AoA/Model Deformation Measurement Techniques,” International
Symposium for Strain Gauge Balances, Hampton, Virginia, 1996.
[3] U. Jensen and J. Quest, “Balance related Activities within EWA,” International
Symposium of Strain Gauge Balances, Mondane, France, 2006.
[5] M. Morris and S. Post, "AC 2010-393: Force balance design for educational wind
tunnels", Peoria: Bradley University, 2010.
[7] K. Hufnagel, “Mono Piece Box Balance Compensation And Calibration Results,”
Internation Symposium for Strain Gauge Balances, Zwolle, 2008.
[8] D. Booth and D. King, “Developmentof the Platform Type Balances,” Internation
Symposium for Strain Gauge Balances, Modane, France, 2006.
[10] A.Burger, "Evaluation of optical fibre Bragg grating sensors on a sidewall wind
tunnel balance", Johannesburg: University of Johannesburg, 2014.
[12] P. Bidgood, "Internal balance calibration and uncertainty estimation using Monte
Carlo simulation", Johannesburg: University of Johannesburg, 2013.
[13] M. Fagan, "Finite Element Analysis", 1st ed., Harlow: Pearson Prentic Hall, 1992.
[14] M. Onishi, “Basic Concept and a simple example of FEM,” University of Arizona,
Tucson, 2007.
[15] R. Figiola and D. Beasley, "Theory and Design for Mechanical Measurements",
Fourth ed., Clemson University: John Wiley & Sons, Inc., 2006.
[16] T. Kleckers and B. Gunther, “Optical versus electrical strain gages: A comparison,”
HBM Measurements, Bavaria, 2007.
[17] National Aeronautics and Space Administration, October 2009. [Online]. Available:
http://www.grc.nasa.gov/WWW/k-12/airplane/tunwheat.html. [Accessed 23 May
2015].
[18] A. Edwards, “Comparison of Strain Gage and Fiber Optic Sensors On A Sting
Balance In A Supersonic Wind Tunnel,” VirginiaTech, Blacksburg, 2000.
[19] S. Jiang, B. Zeng, Y. Liang and B. Li, “Optical fiber sensor for tensile and
compressive strain measurements by white-light Fabry–Perot interferometry,” Opt.
Eng., vol. 46, no. 3, March 2007.
[20] Y. Rao, “In-fibre Bragg grating sensors,” Measurement Science and Technology, vol.
8, no. 4, pp. 355-375, November 1997.
[21] B. Vasudevan, S. Padbidri and M. C. Kishore, “Flow diagnostics using fibre optics,”
Sadhana, Indian Academy of Sciences, vol. 32, no. 1-2, pp. 7-18, February-April
2007.
[22] K. P. Chuang, L. G. Sheu, M. C. Tsai and Y. Lai, “Fabrication of true apodized fiber
Bragg grating using a new two-beam interferometer with polarization control”.
United States of America Patent 6,963,432, 8 November 2005.
148
[23] S. Mihailov, D. Grobnic, C. Smelser, P. Lu, R. Walker and H. Ding, “Bragg grating
inscription in various optical fibers with femtosecond infrared lasers and a phase
mask,” Optical Materials Express, vol. 1, no. 4, pp. 754-765, August 2011.
[25] Micron Optics, “Optical Sensing Interrogator,” Micron Optics, Century Place.
[28] FBGS Technologies, “Strain Gauge Chain SGC-01 Data Sheet,” 2012.
[29] B. Kaul, “EWA Fiber Optical Balance,” European Wind Tunnel Association,
Cologne, 2012.
[30] C. Zimmermann, et al. and , “Mono-bloc balance for high loads,” 5th International
Symposium on Strain-Gauge Balances, 9-12 May 2006.
[31] K. Hufnagel, “Mono Piece Box Balance,” 5th International Symposium on Strain-
Gauge Balances, 9-12 May 2006.
[32] K. Budinski and M. Budinski, "Engineering Materials - Properties and Selection", 9th
ed., Pearson Prentice Hall, 2010.
[33] W. Callister, "Materials Science and Engineering - An Introduction", 7th ed., John
Wiley & Sons, Inc., 2007.
[35] ATI Allegheny Ludlum, Stainless STeel AL 17-4 Precipitation Hardening Alloy
149
Technical Data Blue Sheet, 2006.
[38] R. Budynas and J. Nisbett, Shigley's Mechanical Engineering Design, 8th ed., New
York: McGraw Hill, 2008.
[39] FBGS, “DTG coating Ormocer-T for temperature Sensing Applications,” Fiber Bragg
Grating Sensors, Bell Telephonelaan 2H, B-2440 Geel, 2015.
150
Appendix A - Design Concepts
A.1 Introduction
This Appendix will detail the design evolution of the platform balance. The idea was to use
the mechanics of cantilevers to decouple loads from one another. These designs were
developed, and refined as much as possible. It was found, however, that the design
philosophy used in Concepts 6 and 7, as discussed in Chapter 3 produced much better results.
A-1
A.3 Bending due to a moment
Bending due to a moment is fundamentally different than bending due to a force. The same
cantilever beam as the preceding section is considered here, except that in this section, the
force has been replaced by a moment. From Figure A-2 it can be seen that the stress
experienced by the beam under the application of a moment load is consistent throughout its
length. Therefore, the deformation along the length will be constant as well.
Moment
A.4 Concept 1
Concept 1 was the first prototype which was designed using SolidWorks. This balance was in
essence, a five component balance. It could measure all of the loads except for the normal
force. The groove design in use in Concept 1 is a strain groove, which was discussed in
Section 2.4.7. In terms of practicality, this design has many flaws. The groove width which is
A-2
indicated in Figure A-4 is 3mm. This presents two problems: Firstly, the smallest Bragg
grating which can be written onto a fibre is 3mm in length, and the resolution of the Bragg
grating decreases when its length is that short. For this reason, it is recommended that a 5mm
Bragg grating be used instead. The second problem is that the room for error in terms of
manufacturing is unreasonably small; it would be very difficult to bond the fibre to the
balance with the entire Bragg grating being directly in the groove, if the groove and Bragg
grating length are the same. It is therefore recommended that the groove width be at least
2mm wider than the length of the Bragg grating.
A-3
3mm
Table A-2 shows the interactions which exist in terms of percentages. The diagonal entries’
values are all 100%, as that sensor is designated to measure that particular load. Interactions
which are higher than 40% have been highlighted. There are four interactions higher than
40%; two of which are close to 100%. Interactions are systematic errors, and can be
compensated for in the calibration matrix. However, it would be poor practice to design a
balance with such high interactions, and thus, this design is unacceptable.
A-4
Table A-2 – Concept 1 interactions [% of FS]
Sensor
nf af sf pm rm ym
Load
NF - 0 0 0 0 0
AF - 100 0.064 38.179 0 0.033
SF - 0.058 100 0.03 94.645 0.663
PM - 127.264 0.107 100 0.087 0.033
RM - 0.077 48.361 0.121 100 1.142
YM - 0.058 0.171 0.121 0.435 100
Table A-3 shows the maximum predicted stress which would occur under each of the full
scale loads. The maximum stress experienced under any of the loads is 76.354MPa, which is
relatively small; well below the 250MPa limit which was imposed in the Product Design
Specification.
Advantages
- As the first concept, it provides a starting point for the design process.
- The stresses present in this concept are all reasonably low.
Disadvantages
A-5
A.5 Concept 2
The issues of practicality present Concept 1 were addressed to some degree in Concept 2.
Here the groove width has been increased to 7mm, as seen in Figure A-5. This allows for a
Bragg grating of 5mm, with a 2mm room for error, in terms of bonding the fibre to the
balance. The disadvantage of increasing the groove width is that the strain experienced by the
fibre is decreased. According to the equation to measure strain [31]:
∆𝑙
𝜀= (31)
𝑙
When l, the original length, is increased, the induced strain is decreased. This was an
underlying problem for the subsequent designs. The axial force (the smallest of the six loads),
is required to produce a strain in its fibre sensor of approximately 3 500µε, in order to make a
measurement with sufficient resolution. Later concepts tried to solve this problem through the
use of mechanical strain amplifiers. This concept has a set of grooves cut through its base.
This was meant to provide a means for measuring the normal force.
7mm
A-6
the stress up to a point where the safety factor 3, which is a design requirement, would be
violated.
NF
AF
A-7
Table A-4 – Concept 2 OFBG sensor wavelength shift [nm]
Sensor
nf af sf pm rm ym
Load
NF 18.207 0.003 0.454 0.002 0.173 0.004
AF 2.487 5.438 0.005 2.860 0.001 0.003
SF 0.000 0.009 3.725 0.000 1.764 0.425
PM 2.530 7.413 0.008 7.286 0.002 0.003
RM 0.000 0.004 1.834 0.000 1.806 0.210
YM 0.003 0.001 0.012 0.001 0.001 12.088
Advantages
Disadvantages
- The yawing moment sensor is strained to well beyond its breaking point when a full
scale normal force is applied.
- Very large interactions exist.
A-8
- The normal force sensor is too sensitive, and the rolling moment, side force and axial
force sensors are not sufficiently sensitive.
- The stress present when a normal force is experienced (315MPa) is too large; the
safety factor is not adhered to.
A.6 Concept 3
This concept overcame the problem of measuring the normal force. The base design allowed
for motion in the vertical direction, but it restricted motion in the two horizontal directions.
The portion encircled Figure A-9 is a mechanical strain amplifier, meant to amplify the strain
produced by a normal force. As this was also a preliminary concept, no provision was made
to resolve the yawing moment just yet; this concept was merely a trial to determine if the
base design could resolve a normal force; which it can. This base design was centred on the
concept of combined loadings [30].
A-9
Figure A-9 – Concept 3, front view
There is a fundamental flaw with this design which can be seen by comparing the deformed
result shown Figure A-10 to the one shown in Figure A-11. It is almost impossible to
decouple the pitching moment from the axial force (note how similar the deformed results
appear to be). The same is true for the side force and rolling moment. The mechanics of
cantilevers predicts that the grooves should deform by different amounts when a pitching
moment is experienced, as opposed to an axial force. This is indeed that case; however,
interactions of the order of 100% still exist. Since one of the main aims of a balance design is
to minimise interactions, such large interactions are unacceptable. Table A-8 shows the
interactions in terms of percentages for Concept 3.
A-10
AF
PM
A-11
Table A-7 – Concept 3 OFBG sensor wavelength shift [nm]
Sensor
nf af sf pm rm ym
Load
NF 9.946 0.005 0.010 0.002 0.006 0.000
AF 0.000 3.752 0.011 1.995 0.002 0.000
SF 0.051 0.097 6.287 0.001 3.738 0.000
PM 0.221 28.391 0.037 28.362 0.003 0.000
RM 0.056 0.093 11.107 0.016 12.505 0.000
YM 0.000 0.000 0.000 0.000 0.000 0.000
A-12
Table A-9 – Concept 3 maximum stress per load
Advantages
- The concept offered the most feasible solution so far, in terms of measuring the
normal force.
Disadvantages
A-13
A.7 Concept 4
This concept explored the idea of using the mechanics of cantilevers a bit further. In this
design, the thickness of the cantilever was changed along its length, as seen in Figure A-12.
The reason for this will become apparent shortly.
Therefore, the bottom grooves deform more significantly under an axial force (and thus are
the axial force sensors), and the top grooves deform more under a pitching moment (and thus
are the pitching moment sensors). This design was meant to reduce the axial/pitching and
side/rolling load interactions present in the previous designs. Therefore, as a trial run, this
concept could only measure the axial force and pitching moment. The side force and rolling
moment can be measured in the same way as the axial force and pitching moment. If this
A-14
design was successful, it would be easy to use the base design of Concept 3 to measure the
normal force. The yawing moment load could also be measured quite simply, if the design
showed any promise.
AF
A-15
PM
A-16
Table A-12 – Concept 4 maximum stress per load
Advantages
- None
Disadvantages
A.8 Concept 5
This concept was a radical change from the previous four discussed, and tried using the
mechanics of cantilevers in a more explicit fashion.
Axial Force
Sensor
Location
AF
PM
A-19
Advantages
- The axial force fibre was strained to a sufficiently high strain, netting a reasonable
resolution in the measurement.
- Some of the interactions were minimised.
Disadvantages
- The use of the strain amplifiers makes the implementation of a yawing moment sensor
difficult.
- A full scale pitching, rolling and yawing moment imparted a stress which was higher
than that allowed in the Product Design Specification.
- The normal force could not be resolved using this design.
A.9 Summary
The first five concepts which were discussed in this Appendix intended to minimise
interactions using the mechanics of cantilevers. It was found, however, that large interactions
still do exist with all of these designs. Moreover, most of these designs would experience
stresses which were too large, or were insensitive when certain loads were applied. Therefore,
this line of investigation was abandoned in favour of the design philosophy discussed in
Chapter 3.
A-20
Appendix B - Finite Element Analysis
B.1.1 Introduction
The design process of the balance was an iterative one, as discussed in Chapter 3 and
Appendix A. Every design was tested in a FEM environment to determine its viability of the
design as a working prototype. Therefore, all of the seven concepts had their own FEM
analysis. However, only the last design, Concept 7, showed real promise as a working
prototype. As such, a full study of this concept was conducted, and is presented here.
There are a few things to bear in mind when reading this Appendix. This study is meant to
determine whether or not the design would be appropriate to manufacture; would the actual
prototype balance meet the requirements set out? In Chapter 4, the manufacturing of the
balance was discussed. It was stated that in order to save on costs, and to aid in
manufacturability, that the balance would be manufactured using several pieces. These pieces
would be press-fitted and welded together. The FEM model however, is one where the
balance is a single unit. This simplification was made for the following reasons:
- This is not meant as an exhaustive FEM study of the balance. An exhaustive study
would include simulating the press-fits of the multiple pieces, and the welds. This
would increase the computing cost, simulation time and complexity. A FEM study of
this nature is not in the scope of this work.
- This FEM analysis serves to demonstrate that the design is feasible, and that is would
be worth manufacturing a prototype for further testing.
B-21
Figure B-1 – FEM model of Concept 7
Figure B-2 shows the locations of the normal force (orange dots) and pitching moment (blue
dots) datum points used in this study. One can see the correlation between these datum points
and the location of the force and moment sensor pairs shown in Figure 3-12 and Figure 3-14.
Take note that there is a reference datum shared by the normal force and pitching moment, on
the right. These dots are coloured, half orange and half blue. The reason behind having a fibre
serving as one half of the sensor pair for two different loads reduces the amount of fibre
Bragg gratings required in the balance. This also reduces routing complexity. Take note that
the coordinate triads used in the FEM models do not necessarily represent the coordinate
system of the balance as used in the wind tunnel. The coordinate system as used in the wind
tunnel is shown in Figure 3-3.
B-22
Figure B-2 – Front view showing the normal force and pitching moment datum points
Figure B-3 shows the left view of the FEM model, this time, the rolling moment datum points
which were used are shown.
Figure B-3 – Left view showing the rolling moment datum points
The top view of the FEM model is shown in Figure B-4. This figure shows the axial force
(blue dots), side force (orange dots) and yawing moment (red dots) reference datum points.
There is a fibre (reference datum in the case of the FEM model) which is used by both the
side force sensor, and yawing moment sensor.
B-23
Figure B-4 – Top view showing the axial and side force, and yawing moment datum
points
As this is a six component balance, six different FEM studies were conducted. Each study
yielded results detailing the stress experienced by the component, as well as the nodal
displacements. Once the idea of the cradles was settled upon, the task of fine tuning the
dimensions of the cradle began. The cradles have to be strong enough to offer a safety factor
of at least 3, while being flexible enough to strain the fibres sufficiently (to offer decent
resolution); while being stiff enough so as to not over strain the fibres. Further, as discussed
in Chapter 3, the cradles’ dimensions have to govern the motion of the platform in such a way
as to eliminate vertical deflections of the platform under the application of a horizontal force.
The task of designing the cradles was thus an iterative one.
In Chapters 2 and 3, mesh refinement techniques were discussed. FEM software often offers
p-type and h-type mesh refinement. While conducting these FEM studies, both mesh
refinement techniques were tried in order to ascertain whether they offer any real benefit.
From these early tests, the solutions found when using these refinement techniques did not
vary significantly when compared against a study that did not use mesh refinement.
Therefore, in the interest of time saving, these mesh refinements were not employed. The
B-24
FEM studies presented no real challenges in terms of being solved. The computer used to
solve the studies had the following specifications:
B-25
B.1.4 Units
The SI unit system was used in all six studies. This is shown in Table B-2.
B-26
Figure B-5 – Fixed boundary condition on the underside of the base
The loading conditions for each of the six tests will be dealt with separately. The normal
force load was applied to the top face, the orientation was normal, the direction was the
negative Y direction, and the magnitude was 11 500N.
B-27
The axial force load was applied to the top face, the ‘Selected direction’ option was chosen,
and the option ‘Along Plane Dir 2’ was selected to coincide with the negative Z direction.
The magnitude of the load was set to 900N.
B-29
The face about which the rolling moment acts is shown in Figure B-10. The torque magnitude
was set to 260Nm and is applied to the top face of the platform.
B-31
Figure B-12 – Isometric view of the mesh
B.1.8 Study results
The material used in this study has a yield strength of 1 240 MPa. Therefore, in order to
achieve a safety factor of at least 3, no individual stress should exceed 414 MPa. In addition,
the displacement of the points mentioned in Appendix B.1.2 should not be such that it would
strain a fibre spanning the gap to beyond 6 000 µ𝜀, for a single load. Therefore, once the
strength requirement for the balance was met, the design was tuned such that the
displacements would yield the desired outputs. In the figures to follow, the deformation scale
is stated on the figure. All of the concepts in these figures are called ‘Concept PM 4’, but for
the purpose of this dissertation, is concept is officially called ‘Concept 7’.
B-32
Figure B-13 – Normal force von Mises Stress
The probe tool was used to measure the displacement of the points discussed in Section 3.7.
A figure depicting the use of this tool can be seen in Figure 3-26. Screen shots of this process
will be omitted here, as it does not present any useful detail to the reader. Figures showing
the displacements of the model from useful or interesting angles will be presented.
B-33
Figure B-14 – Normal force deformation
B-34
Figure B-15 – Normal force displacement, front view
B-35
Figure B-17 – Axial force displacement
B-36
Figure B-19 – Side force von Mises Stress
B-37
Figure B-20 – Side force displacement
B-39
Figure B-23 – Pitching moment displacement
B-40
Figure B-25 – Rolling moment von Mises Stress
B-42
Figure B-29 – Yawing moment displacement
The displacements of each of the gaps discussed in Appendix B.2 are presented in Table B-7
and Table B-8. Note that there are two gaps per sensor.
B-44
Table B-8 – Displacement of gaps, due to moments [mm]
Sensor
pm rm ym
Load
NF 0.045 0.045 0.047 0.047 0 0
AF 0 0 0 0 0 0
SF 0 0 0 0 0.053 0.053
PM -0.032 0.032 0 0 0 0
RM 0 0 -0.025 0.025 -0.002 -0.002
YM 0 0 0 0 0.057 -0.057
The strain experienced by each individual fibre due to an applied load is presented in
Table B-9 and Table B-10.
As was stated in Chapter 3.8, the resultant fibre strain for each load sensor will be found by
taking the difference between the two individual fibre strains for per load sensor pair.
Therefore, if both fibres are strained in tension (or compression), then the resultant fibre
strains will cancel each other out, and the resultant fibre strain will effectively be zero. If one
fibre of a pair is strained in tension, and the other in compression, then the resultant fibre
B-45
strains is equal to the sum of the magnitudes of the individual fibres. The sign of the resultant
strain will indicate the direction of the load. This information can be seen in Table B-11.
B-46
Table B-14 – Relative resolution
Relative
Induced fibre Smallest
FS Load resolution
strain resolvable load
[% of FS]
Pitching
400Nm 8 093 µ𝜀 0.08Nm 0.020%
Moment
Rolling
260Nm 6 347 µ𝜀 0.06Nm 0.026%
Moment
Yawing
225Nm 12 649 µ𝜀 0.03Nm 0.013%
Moment
B.1.9 Conclusion
This Appendix Eealt with the finite element analysis which was performed on this design in
detail. Some simplifying assumptions were made, concerning the geometry of the design. It
was determined that, in spite of these assumptions, the results which were obtained are
representative of what could be expected in reality. The nodal points (across which fibres
would be spanned in the working prototype) were identified. It was said that the displacement
of these points would be taken into consideration when determining the magnitude of strain a
fibre would experience in the working prototype.
After tabulating the results, it was found that the design did not experience excessively high
stresses, and thus, the safety factor of three was maintained. The displacements of the
platform in each direction were sufficient (to strain a fibre to produce a decent resolution),
but were not excessive, so as to break the fibre. The relative resolution of the balance is
comparable to that of contemporary strain gauge balances.
In light of the FEM results, is can be said that the resultant balance would be comparable to
modern strain gauge balances in terms of performance, while offering the advantages which
OFBG sensors provide. The impetus to manufacture this design was taken in light of this
FEM analysis.
B-47
B.2 Calibration body finite element analysis
B.2.1 Introduction
The calibration body was not designed to accommodate the full normal force load of 11 500N
for two reasons. It was decided that if a full scale normal force load were to be applied to the
balance, then it would be done when the balance is sitting on a flat and level surface, with the
weights being placed on top of it, as seen in Appendix F. Secondly, there is physically not
enough space for the calibration rig to accommodate a load that large. Therefore, the
calibration body was designed to accommodate a normal force load of 2 500N, the full scale
side force load of 2 500N, and the pitching moment of 400Nm.
The analysis of the calibration body was performed primarily to ascertain the amount of
deformation the arms would undergo when subjected to a load. A large deflection in the arms
under the application of a load would cause the load to not be aligned, and this would
contribute to the loading error. The part being analysed is the same as the part shown in
drawing 001-00 in Appendix G.
Table B-15 – Material properties used in the calibration body FEM study
B-48
B.2.4 Loads and boundary conditions
The boundary condition used in all three studies was a fixed condition of the bolt holes of the
calibration body, as seen in Figure B-31. For the pitching moment load study, the fixed
condition included half of the back face, as seen in Figure B-32. For the normal force load
test, the entire back face was used as a fixture, as shown in Figure B-33. The reasons for
these boundary conditions are due to the manner in which the loads are applied. The fixture
of the back face of the calibration body represents the platform of the balance, which the
fixture of the bolt holes represents the bolts.
B-49
Figure B-32 – Fixture condition for the pitching moment load study
Figure B-33 – Fixture condition for the normal force load study
B-50
The loading conditions for each of the three load tests will be shown in order: the side force
test, the pitching moment test and the normal force test.
NF
AF
SF
B-51
(a)
AF
NF
SF
(b)
B-52
Figure B-36 – Normal force load
B.2.5 Mesh Information
Table B-16 – Calibration body mesh information
B-53
Figure B-37 – Isometric view of the calibration body mesh
B.2.6 Study results
The two metrics that are of interest are the whether the material can withstand the loads
applied to it, and what the deformations these loads cause are. The forces applied to the
calibration body which would induce a side force on the balance is in total 2 500N. This
causes a von Mises stress of 82.128MPa, which is well below the 215MPa yield limit for the
material. This is seen in Figure B-38. The deformation due to this load is also acceptably
small, 0.06mm.
B-54
Figure B-38 – Side force von Mises stress
B-55
The stress caused by the pitching moment also falls below the 215MPa limit for the material.
The application of forces which would induce a pitching moment on the balance cause a
deformation of 0.7mm, which is acceptably low.
B-57
B.2.7 Conclusion
The design of the calibration body is acceptable. The material selected is desirable, as it has a
low density, yet having a high enough yield so as to comfortably withstand the loads placed
on it. The design which places its centre of gravity on the balance moment centre should aide
in the process of capturing clean data.
B-58
Appendix C - Raw Data
C-1
Table C-2 - Axial force, medium loads, raw data
C-2
0 3082.241 -3.962 0.863 -0.05 0.12 -1.445
180
degrees -4 3082.239 -4.633 0.841 -0.105 0.12 -1.438
-8 3082.238 -5.305 0.819 -0.16 0.122 -1.432
-12 3082.238 -5.974 0.798 -0.215 0.124 -1.424
-16 3082.238 -6.643 0.777 -0.27 0.124 -1.418
-20 3082.237 -7.306 0.754 -0.324 0.126 -1.411
-24 3082.237 -7.972 0.733 -0.379 0.125 -1.404
-20 3082.239 -7.284 0.756 -0.322 0.126 -1.412
-16 3082.24 -6.613 0.777 -0.268 0.126 -1.418
-12 3082.24 -5.939 0.801 -0.213 0.127 -1.428
-8 3082.243 -5.268 0.821 -0.159 0.125 -1.434
-4 3082.243 -4.596 0.843 -0.103 0.124 -1.445
-4 3082.245 -4.62 0.84 -0.106 0.119 -1.443
-8 3082.242 -5.28 0.82 -0.159 0.126 -1.437
-12 3082.241 -5.952 0.798 -0.213 0.126 -1.429
-16 3082.24 -6.624 0.779 -0.268 0.126 -1.423
-20 3082.241 -7.295 0.756 -0.323 0.128 -1.416
-24 3082.241 -7.961 0.735 -0.378 0.128 -1.409
-20 3082.242 -7.286 0.757 -0.324 0.129 -1.414
-16 3082.241 -6.611 0.779 -0.268 0.128 -1.422
-12 3082.241 -5.937 0.801 -0.213 0.126 -1.43
-8 3082.243 -5.265 0.822 -0.158 0.125 -1.437
-4 3082.245 -4.594 0.843 -0.104 0.124 -1.446
0 3082.246 -3.925 0.865 -0.049 0.123 -1.453
-20 3082.242 -7.292 0.757 -0.323 0.132 -1.415
0 3082.246 -3.933 0.865 -0.049 0.122 -1.454
C-3
Table C-3 – Side force, medium loads, raw data
C-4
Table C-4 – Pitching moment, medium loads, raw data
C-5
0 3082.141 -0.679 0.051 -0.446 0.253 -1.646
-4.893 3082.105 -0.724 0.047 -0.643 0.252 -1.645
-9.786 3082.072 -0.768 0.044 -0.833 0.252 -1.643
-14.679 3082.039 -0.812 0.04 -1.027 0.25 -1.64
-19.572 3082.006 -0.857 0.036 -1.218 0.247 -1.639
-24.465 3081.974 -0.901 0.032 -1.41 0.245 -1.635
-29.358 3081.943 -0.945 0.028 -1.602 0.243 -1.633
-34.251 3081.912 -0.989 0.024 -1.794 0.239 -1.63
-39.144 3081.879 -1.033 0.02 -1.987 0.237 -1.627
-44.037 3081.848 -1.077 0.016 -2.178 0.233 -1.623
-48.931 3081.818 -1.119 0.012 -2.37 0.229 -1.62
-44.037 3081.848 -1.069 0.014 -2.174 0.23 -1.622
-39.144 3081.881 -1.024 0.018 -1.98 0.232 -1.625
-34.251 3081.914 -0.978 0.021 -1.788 0.237 -1.629
-29.358 3081.944 -0.933 0.025 -1.595 0.237 -1.632
-24.465 3081.977 -0.888 0.028 -1.403 0.238 -1.634
0 3082.009 -0.843 0.032 -1.211 0.241 -1.637
-14.679 3082.045 -0.798 0.035 -1.017 0.244 -1.64
-9.786 3082.075 -0.754 0.039 -0.825 0.243 -1.642
-4.893 3082.108 -0.71 0.042 -0.633 0.245 -1.645
90
0 3082.142 -0.666 0.046 -0.44 0.246 -1.647
degrees
-4.893 3082.109 -0.71 0.043 -0.632 0.246 -1.645
-9.786 3082.075 -0.754 0.04 -0.823 0.245 -1.643
-14.679 3082.042 -0.798 0.037 -1.017 0.244 -1.641
-19.572 3082.008 -0.843 0.033 -1.211 0.242 -1.638
-24.465 3081.976 -0.887 0.029 -1.403 0.241 -1.636
-29.358 3081.943 -0.932 0.026 -1.596 0.237 -1.633
-34.251 3081.911 -0.977 0.022 -1.789 0.235 -1.63
-39.144 3081.88 -1.022 0.018 -1.981 0.232 -1.627
-44.037 3081.848 -1.066 0.014 -2.172 0.228 -1.624
-48.931 3081.817 -1.109 0.01 -2.363 0.224 -1.62
-44.037 3081.848 -1.064 0.014 -2.171 0.227 -1.623
-39.144 3081.881 -1.02 0.017 -1.977 0.231 -1.626
-34.251 3081.912 -0.974 0.02 -1.786 0.232 -1.628
-29.358 3081.943 -0.928 0.024 -1.594 0.234 -1.632
-24.465 3081.977 -0.883 0.027 -1.401 0.236 -1.635
-19.572 3082.009 -0.838 0.031 -1.209 0.237 -1.637
-14.679 3082.042 -0.793 0.035 -1.016 0.238 -1.64
-9.786 3082.075 -0.749 0.038 -0.825 0.24 -1.642
-4.893 3082.109 -0.706 0.042 -0.632 0.242 -1.645
0 3082.141 -0.663 0.046 -0.445 0.243 -1.647
C-6
Table C-5 – Rolling moment, medium loads, raw data
C-7
0 3082.085 -2.329 0.845 0.072 0.915 -1.293
4.893 3082.041 -2.325 0.821 0.072 1.149 -1.301
9.786 3081.999 -2.321 0.799 0.07 1.383 -1.309
14.679 3081.958 -2.317 0.776 0.07 1.613 -1.317
19.572 3081.914 -2.312 0.753 0.069 1.844 -1.326
24.465 3081.873 -2.308 0.729 0.067 2.076 -1.334
29.358 3081.835 -2.305 0.705 0.066 2.303 -1.341
34.251 3081.796 -2.301 0.681 0.064 2.537 -1.35
29.358 3081.832 -2.302 0.706 0.068 2.308 -1.343
24.465 3081.871 -2.305 0.729 0.07 2.081 -1.335
0 degrees
19.572 3081.913 -2.308 0.753 0.071 1.851 -1.327
14.679 3081.956 -2.313 0.776 0.071 1.619 -1.319
9.786 3081.998 -2.317 0.799 0.073 1.386 -1.311
4.893 3082.042 -2.322 0.822 0.074 1.153 -1.302
0 3082.084 -2.326 0.844 0.074 0.917 -1.293
4.893 3082.041 -2.325 0.822 0.073 1.148 -1.301
0 3081.997 -2.32 0.799 0.072 1.387 -1.31
14.679 3081.954 -2.316 0.775 0.071 1.62 -1.319
19.572 3081.912 -2.312 0.753 0.07 1.852 -1.327
24.465 3081.87 -2.308 0.728 0.069 2.083 -1.335
C-8
Table C-6 – Yawing moment, medium loads, raw data
RM NF AF SF PM RM YM
0 3082.246 -3.935 0.865 -0.049 0.124 -1.453
4.893 3082.244 -4.302 0.845 -0.08 0.124 -1.029
9.786 3082.242 -4.668 0.823 -0.11 0.122 -0.606
14.679 3082.239 -5.033 0.803 -0.139 0.12 -0.182
19.572 3082.236 -5.397 0.779 -0.169 0.119 0.24
24.465 3082.236 -5.758 0.757 -0.2 0.117 0.658
29.358 3082.234 -6.119 0.732 -0.229 0.114 1.081
34.251 3082.234 -6.478 0.708 -0.259 0.113 1.5
39.144 3082.234 -6.836 0.684 -0.29 0.109 1.92
44.037 3082.232 -7.196 0.658 -0.319 0.107 2.342
48.931 3082.233 -7.552 0.636 -0.349 0.1 2.758
44.037 3082.232 -7.19 0.656 -0.318 0.107 2.339
39.144 3082.234 -6.828 0.678 -0.288 0.109 1.918
34.251 3082.234 -6.465 0.701 -0.258 0.113 1.496
29.358 3082.235 -6.105 0.724 -0.227 0.115 1.072
24.465 3082.236 -5.741 0.748 -0.197 0.116 0.648
19.572 3082.238 -5.377 0.771 -0.167 0.119 0.226
14.679 3082.24 -5.009 0.793 -0.137 0.119 -0.202
9.786 3082.243 -4.647 0.816 -0.107 0.121 -0.622
4.893 3082.245 -4.281 0.837 -0.078 0.122 -1.047
0 degrees 0 3082.247 -3.919 0.858 -0.048 0.122 -1.469
4.893 3082.245 -4.287 0.84 -0.078 0.122 -1.045
9.786 3082.243 -4.654 0.82 -0.108 0.121 -0.621
14.679 3082.24 -5.021 0.798 -0.138 0.12 -0.196
19.572 3082.239 -5.385 0.776 -0.169 0.118 0.227
24.465 3082.234 -5.75 0.753 -0.196 0.115 0.649
29.358 3082.236 -6.112 0.728 -0.229 0.114 1.072
34.251 3082.235 -6.472 0.705 -0.258 0.11 1.495
39.144 3082.234 -6.833 0.682 -0.288 0.111 1.915
44.037 3082.234 -7.189 0.659 -0.318 0.106 2.332
48.931 3082.234 -7.546 0.632 -0.348 0.103 2.752
44.037 3082.233 -7.184 0.654 -0.318 0.108 2.332
39.144 3082.234 -6.823 0.678 -0.288 0.11 1.91
34.251 3082.235 -6.452 0.699 -0.256 0.113 1.48
29.358 3082.236 -6.099 0.723 -0.227 0.115 1.065
24.465 3082.238 -5.736 0.746 -0.196 0.116 0.644
19.572 3082.24 -5.373 0.77 -0.166 0.117 0.221
14.679 3082.241 -5.007 0.793 -0.137 0.118 -0.204
0 3082.243 -4.642 0.815 -0.107 0.121 -0.629
4.893 3082.246 -4.276 0.836 -0.078 0.122 -1.052
0 3082.248 -3.914 0.859 -0.048 0.123 -1.475
C-9
0 3082.248 -3.919 0.86 -0.048 0.123 -1.47
-4.893 3082.25 -4.228 0.861 -0.073 0.126 -1.887
-9.786 3082.25 -4.537 0.86 -0.098 0.128 -2.303
-14.679 3082.252 -4.844 0.865 -0.123 0.13 -2.716
-19.572 3082.255 -5.152 0.865 -0.149 0.134 -3.134
-24.465 3082.258 -5.461 0.868 -0.173 0.137 -3.549
-29.358 3082.261 -5.769 0.869 -0.199 0.143 -3.964
-34.251 3082.265 -6.074 0.872 -0.225 0.148 -4.381
-39.144 3082.269 -6.381 0.876 -0.25 0.153 -4.799
-44.037 3082.275 -6.682 0.873 -0.273 0.157 -5.211
-48.931 3082.28 -6.981 0.879 -0.299 0.164 -5.637
-44.037 3082.276 -6.669 0.88 -0.275 0.161 -5.218
-39.144 3082.273 -6.36 0.88 -0.25 0.155 -4.802
-34.251 3082.269 -6.053 0.879 -0.225 0.154 -4.385
-29.358 3082.268 -5.748 0.879 -0.2 0.151 -3.97
-24.465 3082.265 -5.444 0.879 -0.175 0.148 -3.554
-19.572 3082.263 -5.139 0.878 -0.149 0.143 -3.137
-14.679 3082.26 -4.834 0.88 -0.125 0.141 -2.722
-9.786 3082.259 -4.529 0.878 -0.1 0.138 -2.306
-4.893 3082.257 -4.409 0.882 -0.088 0.143 -2.145
0 degrees 0 3082.256 -3.92 0.874 -0.049 0.132 -1.474
-4.893 3082.257 -4.229 0.872 -0.075 0.135 -1.893
-9.786 3082.257 -4.535 0.871 -0.099 0.137 -2.309
-14.679 3082.259 -4.841 0.871 -0.124 0.139 -2.725
-19.572 3082.262 -5.146 0.873 -0.15 0.143 -3.143
-24.465 3082.263 -5.451 0.873 -0.174 0.144 -3.556
-29.358 3082.266 -5.756 0.874 -0.199 0.148 -3.972
-34.251 3082.269 -6.063 0.876 -0.223 0.15 -4.388
-39.144 3082.272 -6.365 0.877 -0.249 0.154 -4.799
-44.037 3082.276 -6.67 0.878 -0.274 0.159 -5.215
-48.931 3082.28 -6.973 0.881 -0.298 0.163 -5.626
-44.037 3082.278 -6.661 0.884 -0.275 0.161 -5.209
-39.144 3082.273 -6.352 0.88 -0.25 0.157 -4.79
-34.251 3082.27 -6.049 0.884 -0.223 0.153 -4.375
-29.358 3082.268 -5.743 0.883 -0.199 0.149 -3.96
-24.465 3082.265 -5.437 0.881 -0.174 0.147 -3.545
-19.572 3082.262 -5.133 0.882 -0.149 0.143 -3.13
-14.679 3082.261 -4.826 0.88 -0.124 0.14 -2.71
-9.786 3082.259 -4.524 0.877 -0.099 0.138 -2.296
-4.893 3082.256 -4.219 0.876 -0.074 0.134 -1.881
0 3082.255 -3.916 0.876 -0.05 0.132 -1.469
C-10
Appendix D - Data Sheets
D-1
Cert No: 14-1026
MATERIAL CERTIFICATE
SPECIAL ALLOYS & METALLURGICAL
SERVICES EN 10204 (2005), 3.1
P O Box 32245 Glenstantia 0010
(012) 305 3987 Fax (012) 305 3948
Material Specification: ASTM A564-72 TYPE 630 (17-4 PH Stainless steel) (22mm plate)
Heat No. : R384679 (160 x 165 x 22mm) (x1)
Heat Treatment : Condition A
Customer Order No : 13695 (GKS Engineering)
CHEMICAL COMPOSITION (Wt%)1)
Sample No. Description C Si Mn P S Cr Ni Mo Cu Al Nb V Ta -
R384679 ASTM A564-72 TYPE 630 0,039 0,580 0,55 0,025 0,001 15,70 4,08 0,16 3,24 - 0,24 - - -
MECHANICAL PROPERTIES 2)
Heat Yield Elon- Charpy “V” Notch
Sample UTS R/A Hardness Impact 20˚C Date Signature
Description Treatment 0,2% Offset gation
Id. (MPa) (%) (VPN) (JOULE)
Condition (MPa) (%)
R384679 ASTM A564-72 Type 630 Cond. A N/A N/A N/A N/A N/A N/A 2014/08/04
Note 1) Transferred from original test certificate 384679-07/05/14 2) Original material in Cond. A. Final mechanical properties can only be determined after precipitation hardening.
We hereby certify that the material described herein has been manufactured and tested in accordance with the specification concerned and were found to meet
the requirements.
Manufacturer : Special Alloys & Metallurgical Services 3rd Party Inspection : N/A
Q.A. Representative: FJ Koch Q.A. Representative: N/A
Signature : Signature : N/A
Date : 2014-05-07 Date : N/A
GKS Egn (Univ JHB) - MatCert - 17-4 PH - R384679 - Cond A - 14-1026 - Block.doc
Cert No: 12-1076
MATERIAL CERTIFICATE
SPECIAL ALLOYS & METALLURGICAL SERVICES EN 10204 (2005), 3.1
P O Box 32245 Glenstantia 0010
(012) 305 3987 Fax (012) 305 3948
MECHANICAL PROPERTIES 2)
Charpy “V”
Heat Yield Elon-
Sample UTS R/A Hardness Notch Impact
Description Treatment 0,2% Offset gation Date Signature
Id. (MPa) (%) (VPN) 20˚C
Condition (MPa) (%)
(JOULE)
R710898 ASTM A564-72 Type 630 Cond. A N/A N/A N/A N/A N/A N/A 2014/08/04
Note 1) Transferred from original test certificate 2/1/0026 – 710898 2) Original material in Cond. A. Final mechanical properties can only be determined after precipitation hardening.
We hereby certify that the material described herein has been manufactured and tested in accordance with the specification concerned and were found to meet
the requirements.
Manufacturer : Special Alloys & Metallurgical Services 3rd Party Inspection : N/A
Q.A. Representative: FJ Koch Q.A. Representative: N/A
Signature : Signature : N/A
Date : 2012-10-16 Date : N/A
GKS Eng (Univ JHB) - MatCert - 17-4 PH - R710898 - Cond A - 12-1076 - Blocks.doc
Cert No: 13-1025
MATERIAL CERTIFICATE
SPECIAL ALLOYS & METALLURGICAL SERVICES EN 10204 (2005), 3.1
P O Box 32245 Glenstantia 0010
(012) 305 3987 Fax (012) 305 3948
Material Specification: ASTM A564-72 Type 630 (17-4 PH Stainless steel) (Ø 160mm)
Heat No. : R711977 (175 x 175 x 40mm) (x1)
Heat Treatment : Condition A (45 x 40 x 110mm) (x4)
Customer Order No : 13695 (GKS Engineering)
CHEMICAL COMPOSITION (Wt%)1)
Sample No. Description C Si Mn S P Cr Ni Mo Cu Nb/C Nb - - -
R711977 ASTM A564-72 TYPE 630 0,037 0,42 0,66 0,001 0,018 15,72 4,43 0,22 3,22 7,297 0,27 - - -
MECHANICAL PROPERTIES 2)
Charpy “V”
Heat Yield Elon-
Sample UTS R/A Hardness Notch Impact
Description Treatment 0,2% Offset gation Date Signature
Id. (MPa) (%) (VPN) 20˚C
Condition (MPa) (%)
(JOULE)
R711977 ASTM A564-72 Type 630 Cond. A N/A N/A N/A N/A N/A N/A 2014/08/04
Note 1) Transferred from original test certificate 2/1/0097 – 711977 2) Original material in Cond. A. Final mechanical properties can only be determined after precipitation hardening.
We hereby certify that the material described herein has been manufactured and tested in accordance with the specification concerned and were found to meet
the requirements.
Manufacturer : Special Alloys & Metallurgical Services 3rd Party Inspection : N/A
Q.A. Representative: FJ Koch Q.A. Representative: N/A
Signature : Signature : N/A
Date : 2013-02-12 Date : N/A
GKS Eng (Univ JHB) - MatCert - 17-4 PH - R711977 - Cond A - 13-1025 - Blocks.doc
Technical Sales
(866) 531-6285
orders@ni.com
4 optical channels Compatibility with extrinsic Fabry-Perot, long period grating (LPG), and other optical
10 Hz full spectrum scan frequency sensors
1510 to 1590 nm wavelength range Rugged carrying case with shoulder strap
Automatic sensor detection and configuration utility for fiber Bragg gratings 1-year warranty
Overview
The NI PXIe-4844 optical sensor interrogator is a dual-slot 3U PXI Express data acquisition module for fi ber Bragg grating (FBG) optical sensors. It has four optical channels that
are simultaneously sampled at 10 Hz for measurement of FBG strain and temperature sensors. The NI PXIe-4844 features an optical core, which combines a high-power, low-
noise swept wavelength laser with fi ber Fabry-Perot tunable fi lter technology from Micron Optics. Each optical channel has an 80 nm wavelength range (1510 to 1590 nm), which
can typically scan up to 20 FBG sensors per channel (more than 80 FBG sensors per module, depending on sensor range and sensitivity). You can further extend the maximum
number of FBG sensors per module by connecting one or more optical channels to an external optical multiplexer or by adding more NI PXIe-4844 modules in the same PXI
chassis. The NI PXIe-4844 can also be integrated into the same chassis as other PXI modules for electrical data acquisition and control.
Back to Top
Back to Top
1/6 www.ni.com
You configure the NI PXIe-4844 and FBG sensors in the NI-OSI Explorer, a configuration manager that scans the optical wavelength range to identify all connected FBG sensors.
After you specify the range and scaling equations for each detected FBG, all scans automatically parse the channel data into individual sensor measurements and scale the data
into appropriate engineering units. You can export this scaling confi guration for use in the NI-OSI LabVIEW API.
The NI-OSI LabVIEW API has a similar look and feel to NI-DAQmx driver software. Data is returned as an array that you can directly wire to a LabVIEW chart to display scaled
data in a value-versus-time format.
Figure 1. Interrogation of an FBG with a broadband light source results in the reflection of a specific wavelength.
Theory of Operation
A fiber Bragg grating (FBG) is an in-fiber structure that reflects specific wavelengths of light while allowing all others to transmit (Figure 1). The reflected wavelength (Bragg
wavelength) is a function of both temperature and strain, enabling FBGs to be used as sensors. Each FBG has a unique Bragg wavelength set by the vendor during
manufacturing. You can daisy chain FBGs with different nominal Bragg wavelengths within a single optical fiber as long as each FBG occupies a unique wavelength range within
the 80 nm optical spectrum of the NI PXIe-4844.
Measure Anywhere
Ability to mix and match different measurement types on the same sensor array
Simplified cable management and lightweight fiber
Potential for faster response than thermocouples and longer fatigue life than foil strain gages
Potential for immunity to lightning, high voltages, and corrosion depending on sensor packaging
Ability to embed sensors in composite materials
PXI Express
Easy to Use
NI-OSI Explorer utility for automatic sensor detection and simplified sensor configuration for easy scaling to physical units
NI-OSI driver software with easy-to-use LabVIEW API including common optical measurement functions
Tight integration with LabVIEW for data analysis, viewing, and logging
Application Examples
Civionics – distributed measurements over long distances, including bridges, buildings, and other civil structures
Energy – monitoring wind turbine blades, pipelines, nuclear reactors, offshore platforms, power generators
Marine Vessel and Platforms – cargo containers, deck, mast, or any other corrosion-prone structure
Aerospace – airframe, composite structure, fuel tank, wind tunnels
2/6 www.ni.com
Research – big physics, material testing, vehicle prototyping
Back to Top
Ordering Information
For a complete list of accessories, visit the product page on ni.com.
NI PXIe-4844
Back to Top
Software Recommendations
NI LabVIEW Full Development Fully integrated graphical system design NI LabVIEW Real-Time Design deterministic real-time applications
System for Windows software Module with LabVIEW graphical programming
Support for a wide range of measurement Download to dedicated NI or third-party
hardware, I/O, and buses hardware for reliable execution and a wide
Custom, event-driven user interfaces for selection of I/O
measurement and control Take advantage of built-in PID control, signal
Extensive signal processing, analysis, and processing, and analysis functions
math functionality Automatically take advantage of multicore
Advanced compiler to ensure CPUs or set processor affinity manually
high-performance execution and code Includes real-time OS, development and
optimization debugging support, and board support
Includes SSP for professional technical Purchase individually or as part of a
support, online training, and software LabVIEW suite
upgrades
Back to Top
Calibration
NI measurement hardware is calibrated to ensure measurement accuracy and verify that the device meets its published specifications. To ensure the ongoing accuracy of your
measurement hardware, NI offers basic or detailed recalibration service that provides ongoing ISO 9001 audit compliance and confidence in your measurements. To learn more
about NI calibration services or to locate a qualified service center near you, contact your local sales office or visit ni.com/calibration.
Technical Support
Get answers to your technical questions using the following National Instruments resources.
Support - Visit ni.com/support to access the NI KnowledgeBase, example programs, and tutorials or to contact our applications engineers who are located in NI sales
offices around the world and speak the local language.
Discussion Forums - Visit forums.ni.com for a diverse set of discussion boards on topics you care about.
Online Community - Visit community.ni.com to find, contribute, or collaborate on customer-contributed technical content with users like you.
Repair
While you may never need your hardware repaired, NI understands that unexpected events may lead to necessary repairs. NI offers repair services performed by highly trained
technicians who quickly return your device with the guarantee that it will perform to factory specifications. For more information, visit ni.com/repair.
Classroom training in cities worldwide - the most comprehensive hands-on training taught by engineers.
3/6 www.ni.com
Classroom training in cities worldwide - the most comprehensive hands-on training taught by engineers.
On-site training at your facility - an excellent option to train multiple employees at the same time.
Online instructor-led training - lower-cost, remote training if classroom or on-site courses are not possible.
Course kits - lowest-cost, self-paced training that you can use as reference guides.
Training memberships and training credits - to buy now and schedule training later.
Visit ni.com/training for more information.
Extended Warranty
NI offers options for extending the standard product warranty to meet the life-cycle requirements of your project. In addition, because NI understands that your requirements may
change, the extended warranty is flexible in length and easily renewed. For more information, visit ni.com/warranty.
OEM
NI offers design-in consulting and product integration assistance if you need NI products for OEM applications. For information about special pricing and services for OEM
customers, visit ni.com/oem.
Alliance
Our Professional Services Team is comprised of NI applications engineers, NI Consulting Services, and a worldwide National Instruments Alliance Partner program of more than
700 independent consultants and integrators. Services range from start-up assistance to turnkey system integration. Visit ni.com/alliance.
Back to Top
Detailed Specifications
The following specifications are typical for the NI PXIe-4844 operating at 25 °C unless otherwise noted.
Bus Interface
Lasers
Class 1
Min 0.06 mW
Max 0.25 mW
Number of lasers 4
Wavelength accuracy 1 pm
Wavelength repeatability 1 pm
Physical Characteristics
If you need to clean the NI PXIe-4844, use a soft, non-metallic brush. Make sure that the device is completely dry and free from
contaminants before returning it to the PXI Express chassis.
Note For two-dimensional drawings and three-dimensional drawings of the NI PXIe-4844 module and connectors, visit
ni.com/dimensions and search by module number.
Slot requirements Two side-by-side PXI Express slots, other than slot
one
Slot compatibility x4, x8, and x16 PXI Express or PXI Express hybrid
slots
4/6 www.ni.com
Safety Standards
This product is designed to meet the requirements of the following standards of safety for electrical equipment for measurement, control, and
laboratory use:
Note For UL and other safety certifications, refer to the product label or the Online Product Certification section.
Electromagnetic Compatibility
This product meets the requirements of the following EMC standards for electrical equipment for measurement, control, and laboratory use:
Note For EMC declarations and certifications, refer to the Online Product Certification section.
Laser Compliance
This product meets the requirements of the following laser compliance standards for electrical equipment for measurement, control, and laboratory
use:
Note For EMC declarations and certifications, refer to the Online Product Certification section.
CE Compliance
This product meets the essential requirements of applicable European Directives, as follows:
Refer to the product Declaration of Conformity (DoC) for additional regulatory compliance information. To obtain product certifications and the
DoC for this product, visit ni.com/certification, search by module number or product line, and click the appropriate link in the Certification column.
Mechanical shock
Operating (IEC 60068-2-7 Annex A, section A.4, Table A.1) 15 g peak, half-sine, 11 ms pulse Non-operating (IEC
60068-2-7)
Random vibration
Environmental
Caution Do not exceed the operating temperature, even when using the module in a chassis with a higher temperature range
Environmental Management
5/6 www.ni.com
National Instruments is committed to designing and manufacturing products in an environmentally responsible manner. NI recognizes that
eliminating certain hazardous substances from our products is beneficial to the environment and to NI customers.
For additional environmental information, refer to the NI and the Environment Web page at ni.com/environment. This page contains the
environmental regulations and directives with which NI complies, as
EU Customers At the end of the product life cycle, all products must be sent to a WEEE recycling center. For more information about
WEEE recycling centers, National Instruments WEEE initiatives, and compliance with WEEE Directive 2002/96/EC on Waste Electrical
and Electronic Equipment, visit ni.com/environment/weee.htm.
Back to Top
©2010 National Instruments. All rights reserved. CVI, DIAdem, LabVIEW, Measurement Studio, National Instruments, National Instruments Alliance Partner, NI, ni.com, NI-DAQ, and SignalExpress are trademarks
of National Instruments. The mark LabWindows is used under a license from Microsoft Corporation. Windows is a registered trademark of Microsoft Corporation in the United States and other countries. Other
product and company names listed are trademarks or trade names of their respective companies. A National Instruments Alliance Partner is a business entity independent from National Instruments and has no
agency, partnership, or joint-venture relationship with National Instruments.
My Profile | RSS | Privacy | Legal | Contact NI © 2014 National Instruments Corporation. All rights reserved.
6/6 www.ni.com
Quick Notes Page 2
Quick Notes Page 3
Quick Notes Page 4
Quick Notes Page 5
Bonding of the Fibres, and
Acrylate Coated Fibres
It was shown in Burger [10], that the coating of the fibre has an influence on the value
outputted. The fibre used by Burger [10] was a commercially available fibre with a coating
called Ormocer. Ormocer is a type of ceramic polymer [39]. The polymer coating on the fibre
has a lower elasticity than glass. One of the findings in Burger [10] showed that there was a
tendency for the fibres’ output to drift over time; similar to the result shown in Figure 6-13.
This adds to the uncertainty, and potentially also to the hysteresis. The trouble with this sort
of result is: what is the actual response to the load? Is the true response more similar to the
response outputted by the fibre when the load is initially applied, or is the true response more
similar to the output response after the fibre is left with the load for some time? The fact that
fibre’s output response drifts over time is a problem, and must be resolved before fibre Bragg
grating sensors can become a viable alternative to strain gauges for use in balances.
Two possible sources for this drift were identified. The drift could be attributed to the
interaction between the glass fibre, and its coating, or to the interaction between the coating
and the adhesive. Quantifying and characterising the hysteresis in the fibre, and the means by
which it is bonded is the subject of its own dissertation, currently being investigated at the
Department.
Two fibre coatings are under investigation; Ormocer, used by Burger [10], and acrylate
coated fibres. The idea was to use the acrylate fibres for this balance. The different coat may
have provided a different output response; whether the acrylate coating was better or worse
than the Ormocer was unimportant at the outset. The idea was to try a new fibre coating, and
document the results.
E-1
The first challenge was finding a suitable bonding method by which the acrylate fibre could
be bonded to the 17-4PH stainless steel. The first step was to try X-60, the same epoxy used
by Pieterse [1] and Burger [10] to bond the fibres in that test. The tests proved to be
unsatisfactory, as the fibre would slip through the bond when the equivalent of 2% strain was
applied to the fibre. For the present application, the bond would need to hold to at least 3%
induced fibre strain.
Cyanoacrylate super glue was tried next. The fibre was glued to a piece of 17-4PH (an off-cut
from the manufacturing process), and weights were hung from the fibre. The fibre once again
slipped the bond short of the 3% induced fibre strain goal. On close inspection of the fibre, it
was noted that there was some cyanoacrylate residue left on the fibre, and this had left a
rough texture on the fibre. It was then thought that if the fibre was prepared by applying some
cyanoacrylate to the surface of the fibre, and wiping it off (leaving behind the residue), that
the roughened its surface, once bonded using the X-60, would be able to carry a greater load.
A fibre was then prepared in this manner; applying a drop of cyanoacrylate to the fibre,
leaving it to dry for a few seconds, and wiping it off. This indeed left a roughened surface,
which was then bonded to a piece of 17-4PH steel using the X-60. The result of this test,
shown in Figure E-1 was positive. This fibre supported up to 3kg, which is equivalent to 3%
fibre strain. It was thus decided that the fibre to be used was acrylate coated, having been
prepared using cyanoacrylate super glue.
E-2
Figure E-1 – Acrylate fibre bonded to 17-4PH using X-60, after being prepared by
cyanoacrylate super glue.
The fibre tested in Figure E-1 was loaded to failure. Upon closer inspection, it was found that
the fibre had been stripped clean of its coating. This was encouraging, as it indicated that the
bond was strong. Thus, acrylate fibres were selected for the balance.
The fibres were bonded to the balance following the same procedure as stated in Section 5.2.
The fibres were prepared using the cyanoacrylate as discussed. The fibre was plugged into
the interrogator during bonding. After the first bond had set, the pre-tension was applied. It
was noted that, while the pre-tension was applied, that reading being outputted for that
particular fibre was declining; despite that fact that the load was constant. Figure E-2 shows
this trend. The sharp drop, shown in Figure E-2, at around 1 000 on the Time axis is the point
at which the pre-tension load was released. The downward trend continued after the load was
released.
E-3
1541.1
1541.08
Wavelength Shift [nm] 1541.06
1541.04
1541.02
1541
1540.98 C1R3
1540.96
1540.94
1540.92
1540.9
0 200 400 600 800 1000 1200 1400
Time [s]
Figure E-2 – Output of the fibre with the constant pre-tension load applied
It was thought that the bond was of a poor quality, and that the fibre was slowly slipping in
the bond. Another acrylate fibre was bonded, this time paying very close attention to the
quality of the bond. The same pre-tensioning method was followed, with the same result: The
fibre seemed to be slipping. As the quality of the bond was thought to be good for this second
bond, the attention turned to the adhesive itself. A third fibre was bonded using Q-Bond, a
commercially available superglue. Q-Bond is also a cyanoacrylate type superglue. The pre-
tension was applied, and the same results were obtained.
It was then postulated that the cyanoacrylate glue that was being used may in some way have
influenced the integrity of the acrylate coating. Therefore, a clean fibre was bonded to the
balance using only X-60. Once again, the same result was obtained; the output of the fibre
declined slowly and steadily. The thinking is that the adhesive has in fact formed a good bond
with the fibres’ coating, but the glass fibre is slipping within the coating itself. This
hypothesis has not been confirmed, and may become the subject of future research. It was
therefore decided to use the Ormocer coated fibres, as they had been used twice previously,
in [1] and [10].
E-4
conclusion was drawn by Pieterse [1]. The procedure by which the fibres were bonded is
discussed in Section 5.2.
The first issue to address is the fact that the Bragg grating location must be found, and
marked accurately. The Bragg gratings used in this test were 5mm long; the gaps which they
spanned were 8mm wide. Therefore, there is about 1.5mm tolerance on either side of the
Bragg grating in order to place the fibre. This first fibre which was placed was bonded such
that part of the Bragg grating was bonded. This renders the fibre unusable; as any tension is
applied, part of the Bragg grating expands, changing its wavelength, and part of the Bragg
grating does not move as much. The interrogator picks this up as two separate Bragg gratings
with two distinct wavelengths.
Another problem which must be addressed in future work is how to align the fibres
accurately. A fibre placed with a slight misalignment will produce an output which was not
predicted by the design. A final, crucial aspect to be further investigated is the means of
applying an accurate pre-tension to the fibres. If masses are hung from a clamp, it should be
ensured that the masses do not swing. This can cause the fibre to shift slightly while the
adhesive is setting, and thus result in a poor quality bond. In addition, a means must be
developed which will allow for a precision pre-tension; if the user wishes to add a 10.2nm
pre-tension, for example, then this should be possible.
1. The balance deforms under the application of the pre-tension. A fibre was bonded to
one of the probes on the balance, and the adhesive let to cure. Once cured, the fibre
was pulled so as to apply a pre-tension. When the fibre is pre-tensioned, it was noted
that some of the other sensors’ outputs changed, to register an increase in load. When
the second bond was applied and had cured, the pre-tensioning load was removed,
leaving the tensioned fibre in position. Once the pre-tension load was removed, that
sensor’s output registered a sharp decrease in load, as did the other sensors. Although
the load applied to pre-tension the fibres was 1.2kg, this caused the platform to shift
slightly. Therefore, the act of bonding the fibres causes misalignments and shift in the
platform. It is recommended that Fabry-Perot sensors are tried for future tests; these
E-5
sensors will not be bonded with a pre-tension, and thus the problem just discussed
will be eliminated.
2. It was noted that with each fibre that was bonded, the Bragg grating reading registered
an increase in strain, despite having a constant load applied to it. This could have one
of two causes. The first is that the fibre could be experiencing an increase in
temperature. When X-60 cures, it tends to warm up slightly; the curing is an
exothermic reaction. What is unusual is that the Bragg grating experiences an increase
in strain, which then levels off, and does not return to its previous value. This casts
doubt over the speculation that the Bragg grating was warming up; as it cooled, the
Bragg wavelength shift should return to its previous value. The second theory is that,
when the X-60 cures, it contacts marginally. This contraction could be pulling the
fibre slightly, and this could be registered as an increase in strain. If this is the case,
then attention must turn to the quality of the bond. Does this slight contraction effect
the bond quality in a significant way? Would this lead to an increase in hysteresis? It
is clear that a lot of work must go into the research of the bonds before these
questions can be adequately answered.
E-6
Problems Measuring Normal
Force
It was explained in Chapter 3 how the forces were to be measured; which sensor pair to use to
measure each component. The sensor pair that was proposed to be used for the normal force
is shown in Figure F-1. This sensor pair was bonded to the balance, and it seemed to work
when applying the loads; this is based on the output graphs on the screen. There was an
interesting problem which was noted on reviewing the data.
Compression Tension
Normal
Force
Figure F-1 – The proposed sensor pair to measure the normal force
In Chapter 5.6.1, the method of the load test for the normal force was described. The balance
was loaded with low loads, loaded up to 10kg, and it was loaded in positions 0°, 90°, 180°
and 270°. What was noted was that, when the balance was loaded in positions 0° and 180°,
the slope of the line of best fit through the data was 0.01. When the balance was loaded in
positions 90° and 270°, the slope of the line of best fit through the data was 0.0221. The
balance was responding differently to the load, depending on how the load was applied.
F-1
0.25
NF Loaded 0 and 180 degrees
y = 0.0221x - 0.0007
NF Loaded 90 and 270 degrees
0.2 R² = 0.9996
0.15
Wavelength Shift
y = 0.0099x - 0.0007
R² = 0.9953
0.1
0.05
0
0 2 4 6 8 10 12
-0.05
Load [N]
Figure F-2 - Normal force output was different for different loads
The first thought was that the load path through the calibration body had an influence on the
output; this would be a serious problem practically. The balance cannot have two different
output responses which where depended on how the load is applied. A FEM study was
conducted to determine from where this discrepancy was originating. The same result was
present in the FEM. Therefore, it could not be any of the errors mentioned in Section 6.12
that was causing this response. It was then proposed to retry the normal force test. This test
would be conducted by placing the balance on a flat and level surface, and stacking weights
on top of it. The balance was placed on a pedestal base which was levelled using an
inclinometer.
F-2
Figure F-3 - Balance with the retest of the normal force, loaded to 400kg
In this test, the balance was loaded up and down twice. The results were linear, as was the
case with the other results presented in this paper. The unsatisfying result was that the output
was different yet again. The slope of the line of best fit through the retest data was 0.0163, as
seen in Figure F-2. The other point to note is that this normal force was carefully placed on
the balance, such that there would be the minimal pitching/rolling moment combine load
during the test. This was done by lining up the centre of the weight with the centre of the
calibration body. The calibration body had a reference hole drilled through its centre of just
such alignments. Although it is impossible to align the weight perfectly, it was aligned very
carefully; which adds another problem. The balance would not be loaded in this manner
during a wind tunnel test. The load would be applied in a chaotic manner. What would one do
with the result of a wind tunnel test, when three different tests netted three different output
responses? Focus was then shifted to the sensors themselves, to see if a repeatable result
could be obtained for the normal force, with the data that was acquired.
F-3
7
y = 0.0017x + 0.0766
R² = 0.999
6
5
Wavelength Shift [nm]
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500
-1
Load [N]
The sensor pair to measure the normal force could no longer be the one proposed in Figure
F-1 Figure F-2. The best was to measure the normal force, is was concluded, was to
average the response of the pitching moment sensor, and add it to the averaged response of
the rolling moment sensor. As the only instance in which all four of these fibres would be
loaded in tension or compression, at the same time, is when a normal force is applied.
In terms of interactions, the normal force sensor would not pick up an interaction with respect
to a pitching/rolling moment. When there is a pitching (or rolling) moment applied, one of
the pitching (or rolling) moment sensors is strained in tension, and the other is strained in
compression. This strain, tensile and compressive, should be equal in magnitude, and
therefore, the averaged response should remain unchanged. This is indeed the case in reality,
at least with the pitching moment sensor. The rolling moment sensor interaction was large
(14.5%), when a normal force is applied. This rolling moment sensor was dealt with in
F-4
Section 6.12. Interestingly, the yawing moment interaction was large (14.5%), but the axial
force and side force interactions were lower (5.2% and 2.9%, respectively). This was, in a
large part, put down to loading errors and fibre misalignment errors.
F-5
Technical Drawings
G-1
DRAWING NAME DRAWING NUMBER MATERIAL QUANTITY
BALANCE ASSEM 000-000 1
MULTI BALANCE ASSEM 000-000 1 1
BALANCE BASE 000-001 17-4PH 1
SENSOR PILLAR 000-003 17-4PH 2
SENSOR PILLAR SF 000-004 17-4PH 2
RIGHT CRADLE 000-007 17-4PH 2
LEFT CRADLE 000-008 17-4PH 2
BALANCE PLATFORM 000-010 17-4PH 1
LEVER ARM 001-000 6063-T6 1
LOAD INSERT 001-001 AISI304 12
MOUNTING PLATE 001-002 6082-O 1
1
3
2
30
30
45
1
8X M8X1.25 THROUGH
30 30 45
106
208
NAME SIGNATURE DATE DWG NO.
DE PONTE, JD 2015/03/01
000-000
DRAWN
CHK'D
APPV'D
BALANCE ASSEM
PROJECTION Q.A TITLE:
A3
For Academic Use Only.
WEIGHT: SCALE:1:2 SHEET 1 OF 1
150
30 30 45 8X M8 X 1.25, THROUGH
30
150
30
45
2
2 5
3 2
2 2
106
9
59.500 59.500
1 2X M8 X 1.25 HOLE 20mm DEEP
30 30 45 8X M8 X 1.25, THROUGH
THREAD 18mm DEEP, SEE NOTE 2
45
2
2
5 NOTES:
30
1. ITEMS 2 AND 3 HAVE A TRANSITION FIT WITH ITEMS 1 AND 4.
2. HOLES ARE COMMON ON ALL FOUR SIDES OF THE BALANCE.
3. THE TAPERED CUTOUTS ON ITEM 1 HAVE COMMON DIMENSIONS ON ALL
FOUR SIDES.
15
10
30
ITEM NO. DESCRIPTION PART NUMBER QTY.
1 PLATFORM BALANCE BASE 000-001 1
67.500
DE PONTE, JD 2013/08/05
000-000 1
DRAWN
2 2 APPV'D
MFG
26.800 - 0.013 4
+0.020
45
30
150
30
+0.1
30.8 0
+0.1
38.3 0 67.500
5
+0.250
10 0
+0.250
15 0 SEE NOTE 2
NOTES:
1. THESE HOLES HAVE COMMON DIMENSIONS ON ALL FOUR SIDES OF
59.500 59.500 THE COMPONENT.
9
U.O.S.
TOLERANCES:
LINEAR: 0.5mm U.O.S.
ANGULAR: 0 30'00" U.O.S.
DE PONTE, JD 2013/08/03
000-001
DRAWN
CHK'D
APPV'D
BALANCE BASE
PROJECTION Q.A MATERIAL: TITLE:
A3
17-4 PH
For Academic Use Only.
WEIGHT: 2745 g SCALE:1:2 SHEET 1 OF 1
4
+0.023
15 0
B
5 25°
10
8
15 5
13
5
A
18
25°
13.5
21
5
5
DETAIL B
SCALE 2 : 1
105
2X R5 5.5
2X 8.10 THROUGH R11 2X R10
32
CBORE 15 X 8.5 DEEP
2.25 ±0.05
1
18
9
6
10 10 29
51
34
1
4
DETAIL A
SCALE 2 : 1 UNLESS OTHERWISE SPECIFIED: FINISH: DEBUR AND
DO NOT SCALE DRAWING REVISION
6.4
DIMENSIONS ARE IN MILLIMETERS BREAK SHARP
SURFACE FINISH: 6.4 U.O.S. EDGES
TOLERANCES:
LINEAR: 0.5mm U.O.S.
0 30'00" U.O.S.
U.O.S.
ANGULAR:
APPV'D
000-003
SolidWorks Student Edition.
MFG
SENSOR PILLAR
PROJECTION Q.A MATERIAL: TITLE:
A3
For Academic Use Only. 17-4 PH
WEIGHT: 231 g SCALE:1:1 SHEET 1 OF 1
+0.023
2.25 ±0.05
15 0
10
1
6
B
DETAIL A 5
SCALE 2 : 1 1
4
15
3.5
36° A
18
13.5
5
21
11.5
5
105
2X R5 5.5
2X R10
32
1
11.5
R11
5
18
36°
DETAIL B
9
SCALE 2 : 1
10 10
2X 8.100 THROUGH 51 29
CBORE 15 X8.5DEEP
34
UNLESS OTHERWISE SPECIFIED: FINISH: DEBUR AND
DO NOT SCALE DRAWING REVISION
6.4
DIMENSIONS ARE IN MILLIMETERS BREAK SHARP
SURFACE FINISH: 6.4 U.O.S. EDGES
U.O.S.
TOLERANCES:
LINEAR: 0.5mm U.O.S.
ANGULAR: 0 30'00" U.O.S.
APPV'D
000-004
SolidWorks Student Edition.
MFG
SENSOR PILLAR SF
PROJECTION Q.A MATERIAL: TITLE:
A3
For Academic Use Only. 17-4 PH
WEIGHT: 240 g SCALE:1:1 SHEET 1 OF 1
38.3 -0.1
0
0
30.8 -0.1
1 1 1 1
0.8
15
0.8
0.8
0.8
8 ±0.10
3
8 ±0.1
13 ±0.1
8.30 ±0.05
15 ±0.1
8 ±0.10
5.80 ±0.05
106
70
61
8 ±0.1
34.50 ±0.05
15 ±0.1 R4(TYP)
3
13 ±0.05
0.8
0.8
0.8
0.8
15
0 0
2 X 45° (TYP) 34.300 -0.100 26.800 -0.084
APPV'D
000-007
SolidWorks Student Edition.
MFG
RIGHT CRADLE
PROJECTION Q.A MATERIAL: TITLE:
A3
17- 4PH
For Academic Use Only.
WEIGHT: 496 g SCALE:1:1 SHEET 1 OF 1
26.800 -0.084
30.8 -0.1
0
0
0
34.300 -0.100
0
38.3 -0.1
2 X 45° (TYP) 1
1
1 1
0.8
15
3
0.8
0.8
0.8
3
13 ±0.1
8 ±0.1
8 ±0.1
8.30 ±0.05
5.80 ±0.05 15 ±0.1
8 ±0.1
15 ±0.1
61
70
34.50 ±0.05
R4(TYP)
15 3
13 ±0.05
0.8
0.8
0.8
0.8
0
34.300 -0.100
0
26.800 -0.084 UNLESS OTHERWISE SPECIFIED: FINISH: DEBUR AND
DO NOT SCALE DRAWING REVISION
6.4
DIMENSIONS ARE IN MILLIMETERS BREAK SHARP
SURFACE FINISH: 6.4 U.O.S. EDGES
U.O.S.
TOLERANCES:
LINEAR: 0.5mm U.O.S.
ANGULAR: 0 30'00" U.O.S.
DE PONTE, JD 2013/08/03
000-008
DRAWN
CHK'D
APPV'D
LEFT CRADLE
PROJECTION Q.A MATERIAL: TITLE:
A3
17- 4PH
For Academic Use Only.
WEIGHT: 496 g SCALE:1:1 SHEET 1 OF 1
164
150
2 X 45 CHAMFER (TYP) +0.024
4 34.300 - 0.015 75 8X M8 X 1.25 THROUGH
SEE NOTE 1
4
SEE NOTE 2
+0.020
26.800 - 0.013 4.5
5
3.25 ±0.05
164
150
1
1
7
SEE NOTE 2
75
27
14
5
SEE NOTE 1 5 72.500 5 1.5
8
A 7
3
18
DETAIL A
SCALE 1 : 1
SEE NOTE 4
10
32 86 SEE NOTE 3
32
7.500 ±0.05
67.5 ±0.1
NOTES:
1. THESE PROBES ARE DIRECTLY IN LINE WITH EACH OTHER.
2. THESE PROBES ARE OFFSET FROM EACH OTHER AROUND THE CENTRE LINE.
3. TAKE CAREFUL NOTE OF THE ORIENTATION OF THESE PROBES.
4. THE PROBES INDICATED IN DETAIL A HAVE COMMON DIMENSIONS.
86
15
U.O.S.
TOLERANCES:
DE PONTE, JD 2013/08/03
000-010
DRAWN
CHK'D
APPV'D
BALANCE PLATFORM
Q.A
A3
For Academic Use Only. SEE NOTE 3 17-4 PH
WEIGHT: 3110 g SCALE:1:2 SHEET 1 OF 1
5 CBORE 20 X 7 DEEP
8.1 CBORE 15 X 9 DEEP
30
30
17
530
A A
1
30
DETAIL B
125 SCALE 1 : 1
15
30
15 4X 12 THORUGH
30 DETAIL A
60 125 SCALE 1 : 1
530 SEE NOTE 1
17
40
SECTION A-A
250 B DETAIL C
A
C SCALE 1 : 2
SEE NOTE 2.
40
232 232
25
500
135
NOTES:
1. ALL HOLES HAVE COMMON DIMENSIONS AS INDICATED IN DETAIL A.
2. THE LOAD INSERT (DRAWING 001-001) WILL BE PRESS FITTED INTO EACH OF
135
APPV'D
001-000
SolidWorks Student Edition.
MFG
LEVER ARM
PROJECTION Q.A MATERIAL: TITLE:
A3
150
For Academic Use Only. 270
6063-T6
WEIGHT: 5 000g SCALE:1:7 SHEET 1 OF 1
16
A A
130°
12 3.500
6
DE PONTE, JD 2015/03/01
001-001
DRAWN
CHK'D
APPV'D
LOAD INSERT
PROJECTION Q.A MATERIAL: TITLE:
A3
AISI304
For Academic Use Only.
WEIGHT: 8g SCALE:5:1 SHEET 1 OF 1
240
8X 8.4 CBORE 15 X 9 DEEP
-0.100
140 -0.150
12XM8 ON PCD 210, THROUGH 50
A
15°
60
30
+0.100
A 30 0
SECTION A-A
30
60
UNLESS OTHERWISE SPECIFIED: FINISH: DEBUR AND
DO NOT SCALE DRAWING REVISION 3 - 2015/01/14
12
DIMENSIONS ARE IN MILLIMETERS BREAK SHARP
SURFACE FINISH: EDGES
U.O.S.
TOLERANCES:
LINEAR: 0.5MM U.O.S.
ANGULAR: 1 U.O.S.
APPV'D
001-002
SolidWorks Student Edition.
MFG
MOUNTING PLATE
PROJECTION Q.A MATERIAL: TITLE:
A3
For Academic Use Only. ALUMINIUM 6082 T6