University of Huddersfield Repository

University of Huddersfield Repository
University of Huddersfield Repository
Clough, David A
Non-Contact Measurement and Analysis of Machine Tool Spindles
Original Citation
Clough, David A (2012) Non-Contact Measurement and Analysis of Machine Tool Spindles.
Masters thesis, University of Huddersfield.
This version is available at http://eprints.hud.ac.uk/18101/
The University Repository is a digital collection of the research output of the
University, available on Open Access. Copyright and Moral Rights for the items
on this site are retained by the individual author and/or other copyright owners.
Users may access full items free of charge; copies of full text items generally
can be reproduced, displayed or performed and given to third parties in any
format or medium for personal research or study, educational or not-for-profit
purposes without prior permission or charge, provided:
•
•
•
The authors, title and full bibliographic details is credited in any copy;
A hyperlink and/or URL is included for the original metadata page; and
The content is not changed in any way.
For more information, including our policy and submission procedure, please
contact the Repository Team at: [email protected]
http://eprints.hud.ac.uk/
Non-Contact Measurement and Analysis of Machine Tool Spindles
Submitted to
School of Computing and Engineering
of
University of Huddersfield
in partial fulfilment of the requirements for
the degree of MSc by Research
David Clough
October 2012
Abstract
An increasing demand on the manufacturing industry to produce tighter tolerance parts at a
consistent rate means it is necessary to gain a greater understanding of machine tool capabilities,
error sources and factors affecting asset availability. The machine tool spindle can be a significant
contributor to both machine tool errors and failures, resulting in a requirement for spindle error
measurement.
This work is an investigation into the most significant static and dynamic errors associated with a
variety of common machine tool spindles and the issues around spindle metrology in hostile
manufacturing environments.
Various non-contact measurement sensors are investigated and assessed for their suitability for use
in a spindle analysis system. Based on the sensor selection, new measurement hardware and
analysis methods were designed and developed for affordable, robust and efficient characterisation
of machine tool spindle errors.
Details of practical spindle analysis are presented and reviewed, discussing potential issues and
methods of minimisation of measurement error.
2
Acknowledgements
I would like to acknowledge the assistance of the following people whose help made this research
possible.
A special thank you goes to my supervisor, Simon Fletcher for his encouragement, support and
direction.
Thanks to Andrew Longstaff, john Richards and Peter Willoughby for their continued support
throughout.
I would also like to thank all other member of technical and academic staff at the University of
Huddersfield who aided me during the course of this research.
3
Table of Contents
Abstract ................................................................................................................................................ 2
Acknowledgements .............................................................................................................................. 3
Table of Contents .................................................................................................................................. 4
Table of Figures ..................................................................................................................................... 7
Table of Tables .................................................................................................................................... 10
1.0
Introduction ............................................................................................................................ 11
1.1.
Spindle Error Sources .......................................................................................................... 13
1.1.1.
Geometric Spindle Errors ................................................................................................ 13
1.1.2.
Non-Rigid Spindle Errors ................................................................................................. 14
1.1.3.
Spindle Thermal Errors .................................................................................................... 14
1.2.
2.0
Chapter Conclusions............................................................................................................ 15
Literature Review .................................................................................................................... 16
2.1.
ISO Standards ...................................................................................................................... 16
2.1.1.
Determination of Thermal Affects................................................................................... 16
2.1.2.
Geometric Accuracy of Axes of Rotation ......................................................................... 17
2.1.3.
Determination of Vibration Levels .................................................................................. 19
2.1.4.
ISO Standards Conclusions .............................................................................................. 19
2.2.
Non-Contact Sensing Technology........................................................................................ 20
2.2.1.
Capacitance Sensors ........................................................................................................ 20
2.2.1.
Laser Triangulation Sensors ............................................................................................ 21
2.2.2.
Eddy Current Sensors ...................................................................................................... 22
2.2.3.
Eddy Current Sensor Calibration ..................................................................................... 23
2.2.4.
Electrical Run-out ............................................................................................................ 24
2.3.
Related Work ...................................................................................................................... 24
2.4.
Conclusions ......................................................................................................................... 27
2.4.1.
3.0
Aims and Objectives ........................................................................................................ 27
Sensor Testing & Selection ...................................................................................................... 29
3.1.
Sensor Specification Requirements ..................................................................................... 29
3.2.
Laser Triangulation Testing ................................................................................................. 31
3.2.1.
Warm Up Period Test ...................................................................................................... 32
4
3.2.2.
Thermal Stability Test...................................................................................................... 34
3.2.3.
Static Resolution Test ...................................................................................................... 36
3.2.4.
Dynamic Resolution Test ................................................................................................. 37
3.2.5.
Conclusions of Laser Testing ........................................................................................... 39
3.3.
Eddy Current Testing ........................................................................................................... 40
3.3.1.
Thermal Stability ............................................................................................................. 40
3.3.2.
Linearity and Calibration ................................................................................................. 45
3.3.3.
Sensor Sensitivity Test ..................................................................................................... 46
3.3.4.
Static Resolution Test ...................................................................................................... 47
3.3.5.
Dynamic Resolution Test ................................................................................................. 48
3.3.6.
Conclusions of Eddy Current Sensor Testing ................................................................... 50
3.4.
Capacitance Sensor Testing ................................................................................................. 51
3.5.
Non-Contact Sensor Solution .............................................................................................. 52
4.0
System Design ......................................................................................................................... 57
4.1.
Equipment Design ............................................................................................................... 57
4.1.1.
Test-bar Design ............................................................................................................... 57
4.1.2.
Fixture Design ................................................................................................................. 60
4.1.3.
Data Acquisition .............................................................................................................. 61
4.2.
Data Processing ................................................................................................................... 63
4.2.1.
Thermal Data Processing ................................................................................................. 63
4.2.2.
Geometric Data Processing ............................................................................................. 65
4.2.2.1.
4.2.3.
4.3.
5.0
Donaldson reversal method ........................................................................................ 66
Vibration Data Processing ............................................................................................... 68
Chapter Conclusions............................................................................................................ 70
Practical Spindle Analysis ........................................................................................................ 71
5.1.
In situ Calibration ................................................................................................................ 71
5.2.
Thermal Analysis ................................................................................................................. 72
5.2.1.
Thermal Imaging ............................................................................................................. 74
5.3.
Geometric Analysis ............................................................................................................. 75
5.4.
Vibration Analysis................................................................................................................ 78
6.0
Conclusions ............................................................................................................................. 79
7.0
Future Work ............................................................................................................................ 80
7.1.
Test-bar Design and Holding ............................................................................................... 80
7.2.
Fixtures ............................................................................................................................... 81
5
7.3.
Data Processing Software ................................................................................................... 82
7.4.
Test Methodology ............................................................................................................... 82
8.0
References .............................................................................................................................. 83
9.0
Appendix ................................................................................................................................. 86
9.1.
Micro Epsilon Eddy Current Testing .................................................................................... 87
9.2.
Further Examples of Practical Spindle Analysis ................................................................... 89
6
Table of Figures
Figure 1.1 – Sectional View of a motorised spindle [1] ....................................................................... 12
Figure 1.2 – Geometric Spindle Errors [3] ........................................................................................... 13
Figure 2.1 – Typical setup for ETVE or Spindle Heating Test [2] .......................................................... 17
Figure 2.2 – Five-sensor system for measurement of rotating sensitive direction spindle error
motions [3] ......................................................................................................................................... 18
Figure 2.3 – Error motion polar plot displaying synchronous and asynchronous error motion [3] ..... 18
Figure 2.4 – Principle of capacitance sensor [5] .................................................................................. 20
Figure 2.5 – Laser triangulation principle [38] .................................................................................... 21
Figure 2.6 – Principle of eddy current sensor [7] ................................................................................ 22
Figure 3.1 – Thermal Warm up Test Setup.......................................................................................... 32
Figure 3.2 – 15 min Warm Test with Aluminium Fixture ..................................................................... 33
Figure 3.3 – 15 min Warm Up Test with Carbon Fibre Fixture ............................................................ 33
Figure 3.4 – Thermal Image Showing Spot Positions .......................................................................... 34
Figure 3.5 – Plot of Spot Temperatures .............................................................................................. 35
Figure 3.6 – Displacement during 2 hour stability test ....................................................................... 35
Figure 3.7 – Static resolution test setup ............................................................................................. 36
Figure 3.8 – Static resolution with 0 x averaging displacement data .................................................. 36
Figure 3.9 – Static resolution with 16 x averaging displacement data ................................................ 37
Figure 3.10 – Dynamic resolution test setup ...................................................................................... 38
Figure 3.11 – Dynamic resolution with 0 x averaging displacement data ........................................... 38
Figure 3.12 – Kaman eddy current sensor and signal conditioning unit ............................................. 40
Figure 3.13 – Thermal stability test setup ........................................................................................... 41
Figure 3.14 – Comparison of displacement and temperature during thermal stability test ............... 41
Figure 3.15 – Comparison of displacement and temperature during second thermal stability test ... 42
Figure 3.16 – Sensor 2 temperature against displacement ................................................................. 43
Figure 3.17 – Sensor 3 temperature against displacement ................................................................. 43
Figure 3.18 – 50 hour thermal stability test ........................................................................................ 44
Figure 3.19 – Non-linear output from eddy current sensor ................................................................ 45
Figure 3.20 – Repeatability results for eddy current sensor output ................................................... 46
Figure 3.21 – Repeatability results for eddy current sensor output ................................................... 46
7
Figure 3.22 – Eddy current static resolution test setup ...................................................................... 47
Figure 3.23 – Eddy current static resolution ....................................................................................... 48
Figure 3.24 – High resolution piezo platform dynamic resolution test setup ..................................... 49
Figure 3.25 – Renishaw XL80 laser output on vibration rig ................................................................. 49
Figure 3.26 – Output from eddy current sensor on vibration rig ........................................................ 50
Figure 4.1 – Manufacture of short aluminium test-bar ....................................................................... 58
Figure 4.2 – Short Test-bar measuring radial error motion................................................................. 58
Figure 4.3 – Long Test-bar measuring tilt error motion ...................................................................... 59
Figure 4.4 – Adjustable fixture of radial error motion testing ............................................................. 60
Figure 4.5 – Invar fixture for thermal testing ...................................................................................... 61
Figure 4.6 – National Instruments data acquisition device ................................................................. 62
Figure 4.7 – Raw thermal data plot converted to microns .................................................................. 63
Figure 4.8 – Post possessing averaged data plot................................................................................. 64
Figure 4.9 – Surface mountable temperature sensor ......................................................................... 64
Figure 4.10 – Temperature data ......................................................................................................... 65
Figure 4.11 – Donaldson reversal technique [20] ............................................................................... 66
Figure 4.12 – Peak finding data ........................................................................................................... 67
Figure 4.13 – Peak finding data ........................................................................................................... 68
Figure 4.14 – FFT plot of spindle vibration in the Z-Axis direction when rotating at 1000rpm ........... 69
Figure 4.15 – Zoomed in FFT plot of spindle vibration in the Z-Axis direction when rotating at
1000rpm ............................................................................................................................................. 69
Figure 5.1 – Spindle displacement during a 6 hour spindle heating test ............................................. 72
Figure 5.2 – Spindle temperature during a 6 hour spindle heating test .............................................. 73
Figure 5.3 – Comparison of temperature and displacement during a 6 hour spindle heating test ..... 73
Figure 5.4 – Thermal imaging at various times of a heating and cooling test ..................................... 74
Figure 5.5 – All run plotted from run 1 ............................................................................................... 75
Figure 5.6 – All runs plotted from run 2 .............................................................................................. 76
Figure 5.7 – All runs plotted in a polar plot ......................................................................................... 76
Figure 5.8 – Averaged data run-out plot ............................................................................................. 77
Figure 5.9 – Vibration data from an air bearing spindle running at 250rpm ....................................... 78
Figure 5.10 – Vibration data from an air bearing spindle running at 500rpm ..................................... 78
Figure 7.1 – Adjustable tool holder to 0.000µm run-out [39] ............................................................. 80
Figure 7.2 – Ultrafine adjustment screws [40] .................................................................................... 81
Figure 9.1 – Linearity of Micro-Epsilon eddy current sensor............................................................... 87
8
Figure 9.2 – Stability of Micro-Epsilon eddy current sensor ............................................................... 88
Figure 9.3 – Thermal step heating test displacement measurement .................................................. 89
Figure 9.4 – Thermal step heating test temperature measurement ................................................... 90
Figure 9.5 – Thermal step heating test thermal imaging .................................................................... 90
Figure 9.6 – Thermal step heating test temperature and displacement comparison ......................... 91
Figure 9.7 – Radial run-out of ball bearing spindle at 1000rpm .......................................................... 92
Figure 9.8 – Tilt error of ball bearing spindle at 1000rpm ................................................................... 92
Figure 9.9 – Vibration of ball bearing spindle at 1000rpm .................................................................. 93
9
Table of Tables
Table 1 - Minimum Required Specification ......................................................................................... 30
Table 2 - Keyence Laser Specification ................................................................................................. 31
Table 3 – Comparison of Non-Contact Measurement Sensors Part 1 ................................................. 53
Table 4 - Comparison of Non-Contact Measurement Sensors Part 2 .................................................. 54
Table 5 – Advantages and Disadvantages of the Potential Systems ................................................... 55
Table 6 – Spindle Analysis B.O.M ........................................................................................................ 70
Table 7 – Potential Resolutions at various positions in the sensor range ........................................... 88
1.0
Introduction
The precision engineering industries demand for high volume automated production has led to the
continuing development of CNC machine tools. Over the years, machine tool and spindle
manufacturers have been faced with the challenge of producing faster and higher accuracy
equipment to cope with the industries continuous improvement ethos.
It is the primary goal of machine tool manufacturers and maintenance teams to bring their machine
tools in to a state where they can consistently produce parts to within a specified tolerance. Within
manufacturing, especially the aerospace industry, part tolerances have tightened considerably over
recent years. These tight tolerances require high levels of repeatability and accuracy from the
machine tools and spindles used to produce these parts.
The ability to produce accurate components has many advantages, with some of the most important
being:




A reduction in the costs incurred by remanufacturing or scrapping out of tolerance
components
Tolerances can be reduced allowing the production of more accurate assemblies
An increased possibility of a part being both roughed and finished on the same machine,
resulting in reduced setup time
More accurate parts being produced, leading to improved efficiency e.g. in engines and
compressors
While understanding machine tool errors is a large and complex subject and should not be
understated, spindle error measurement is becoming increasingly important to machine tool users
in high precision manufacturing, as they look to characterise their spindle performance capabilities.
Historically, machine tool spindles were driven by gear or belt systems, with the speed only variable
by a transmission. The requirement from industry for increased productivity led to the development
of new bearings, power electronics, inverter systems and low cost 3 phase motors, which in turn led
to the production of high speed motorised spindles.
Today, the majority of machine tools are equipped with these motorised spindles [1]. This reduces
the number of moving parts and the potential for unwanted vibration from mechanical transmission
elements like gears and couplings.
11
Figure 1.1 – Sectional View of a motorised spindle [1]
Figure 1.1 shows a typical configuration of a motorised machine tool spindle, with all the individual
elements identified.
The spindles have at a minimum of two sets of bearings; predominantly ball bearings are used
although air bearings are becoming more popular in very high precision operations. The bearing
system is the component with the greatest influence on the lifetime of a spindle, so periodic
monitoring of bearing condition is advisable.
Due to high power outputs of the motor and high rotational speeds, active cooling is often required.
This is generally implemented through water based cooling. The coolant flows through a cooling
sleeve around the stator of the motor and often the outer bearing rings.
Seals at the tool end of the spindle prevent the intrusion of swarf and coolant; this is often achieved
with purge air and a labyrinth seal. A standardized tool interface such as HSK or SK is utilised at the
spindles front end with a clamping system for fast automatic tool changes.
Spindle errors typically account for up to 30% of overall machine tool errors, and catastrophic
spindle failures bring production to a standstill, until a new spindle can be fitted. At worst case this
can result in several weeks downtime waiting for a new spindle, or companies holding expensive
spares (replacement spindles typically cost between £20k and £50k). Based on the past records of a
well known British manufacturer, spindle failure is a significant factor affecting machine availability,
it is important to understand their errors and point of failure, to be able to predict maintenance
requirements.
12
1.1. Spindle Error Sources
As previously mentioned, high volume, tight tolerance components are now the norm in industry.
These tight tolerances require high levels of repeatability and accuracy from the machine tools and
spindles used to produce them. The sources of error responsible for affecting these spindle
performance parameters can be split into three main areas:



Geometric errors
Non-rigid errors
Thermal errors
1.1.1.
Geometric Spindle Errors
As the spindle rotates about its axis, any deviation from the true axis of rotation will result in a
geometric spindle error. This could be a radial error in the X and Y directions, an axial error in the Z
direction or a tilt error in the X and Y directions (see figure 1.2).
Figure 1.2 – Geometric Spindle Errors [3]
These errors occur as a result of small mechanical imperfections and misalignment of the spindle
assembles. The magnitude of these errors is dependent on the quality of the manufacturing process.
Despite the majority of spindle manufactures being able to minimise these errors through good
13
design and careful assembly, some level of inherent error will remain. In addition, continued use of
the spindle will result in bearing wear and degradation of the overall bearing condition over time.
This will increase the magnitude of the error and will eventually lead to spindle failure.
1.1.2.
Non-Rigid Spindle Errors
Non-rigid errors also known as load induced errors result from factors such as spindle inertia and
cutting forces. It could be induced from any moving part within the spindle, i.e. bearings, races,
shaft. These errors vary with speed and dependant on magnitude can cause a poor quality surface
finish on manufactured parts.
Sources of these errors resulting in poor surface finish could include:



An out of balance shaft, which would produce a once per revolution error
A damaged ball bearing, which would produce an asynchronous error at the ball rotation
frequency
Damaged bearing races, which again would produce an asynchronous error at the ball pass
frequency
When cutting in a sensitive direction, these errors will result in a one for one error on the part at the
magnitude of the induced vibration.
1.1.3.
Spindle Thermal Errors
Thermally induced errors occur due to changes in temperature of the spindle and surrounding
structure. These can be categorised into internally and externally generated.
Internally generated heat is produced from the moving elements within the spindle such as,
bearings, motors and gears where high friction causes a temperature increase of the components.
This generated heat will flow through the spindle and machine structure causing thermal expansion
and resulting in distortion between tool and work-piece. This distortion could vary dependant on
processes being carried out, for example processes requiring a high spindle speed will induce more
heat and therefore a larger error.
14
Externally generated heat is produced from changes in temperature around the machine. This could
be from an environmental temperature fluctuation throughout the course of the day, the opening
and closing of workshop doors, a workshop heating system or heat generated by other machines in
close vicinity.
1.2. Chapter Conclusions
A demand for tighter tolerances and faster process times in the manufacturing industry has resulted
in the requirement for extremely accurate and high speed spindles. This has led to machine tool
users having a requirement to quantify spindle errors so that they can assess their capability for
producing components to specification. The sources of error affecting spindle performance are:



Geometric errors
Non-rigid errors
Thermal errors
This has led to the investigative work described in this report, into the most significant static and
dynamic errors associated with a variety of common machine tool spindles. Also, the development
of robust non-contact measurement hardware and analysis methods for efficient characterisation of
these errors.
Surprisingly, there are but a few commercial systems available on the market that are capable of
analysing spindle errors on precision machine tools. However, the systems use capacitance sensing
technology, which may not always be suitable for use in a manufacturing environment. The systems
are also very expensive and so may not be a viable option or at least difficult to budget for the
ajo it of s all to
ediu
size e te p ises “ME s . Another potential problem is that a single
system may not have sufficient sensitivity or range to measure both geometric and thermal errors
on a range of spindles potentially present in such factories. Section 2 includes a review of the
specifications of some existing systems.
The desired solution is a spindle analysis system that is capable of being taken to customer sites by
service organisations or utilised by maintenance teams, to perform spindle analysis testing on their
machine tool spindles in all types of manufacturing environments.
15
2.0
Literature Review
As mentioned in the introduction, a demand for tighter tolerances and faster processing times in the
manufacturing industry has resulted in the need for extremely accurate and high speed spindles. The
requirement of the machine tool manufactures and maintenance teams to measure and quantify the
inherent errors present in these spindles has resulted in the production of a series of international
standards.
2.1. ISO Standards
These standards have been prepared by a technical committee with the aim to standardise test
methods for determining spindle errors. The standards relevant to spindle analysis Include:



ISO 230 part 3 - Determination of thermal affects
ISO 230 part 7 - Geometric accuracy of axes of rotation
ISO 230 part 8 - Determination of vibration levels
2.1.1.
I“O
pa t
Determination of Thermal Affects
Dete
i atio of The
al Affe ts is the standard for assessing internal and external
thermal influences on machine tool and spindle behaviour. The standard describes three main tests:



Environmental Temperature Variation Error (ETVE) test
Thermal Distortion Caused by Rotating Spindle test (spindle heating test)
Thermal Distortion Caused by Linear Motion of Components test (axis heating test)
An ETVE test can be performed to assess the effect of environmental temperature changes on the
machines positional accuracy. A spindle heating test can be conducted to identify the effects of the
internally heat generated by rotation of the spindle(s) and the resultant temperature gradient along
the structure and the distortion of the structure observed between the work-piece and the tool. An
axis heating test can be carried out to identify the effects of internal heat generated by the driving
system components, such as the ball-screw, nut and support bearings, on the distortion of the
machine tool structure observed between the work-piece and the tool. [2]
16
The standard suggests using non contact measurement sensors to measure the linear and angular
displacement of the spindle relative to the work-piece as it rotates, while monitoring the
temperature of the machine structure near the spindle nose, the machine ambient temperature and
the spindle ambient temperature. An example setup used in the ISO standard can be seen in figure
2.1.
Figure 2.1 – Typical setup for ETVE or Spindle Heating Test [2]
Non-contact measurement sensors are recommended for thermal testing however, many machine
tool use s a d se i e tea s do t ha e a ess to this t pe of te h olog a d so
a use i di to
clocks or other contact probe method. This involves a simple routine that measures the
displacement of the spindle relative to the work piece, usually performed every 15 minutes, and
requires the spindle to be stopped in order to take the measurement. This method in not very
accurate and can result in poor thermal profiling of a machine.
2.1.2.
I“O
pa t
Geometric Accuracy of Axes of Rotation
Geo et i A u a
of A es of ‘otatio
is the standard for assessing the accuracy of
machine tool spindles, rotary heads and rotary / swivelling tables. The standard describes tests for
two machine configurations, rotating sensitive direction and fixed sensitive direction. For each of
these there are three tests to measure the geometric error:



Radial Error Motion
Axial Error Motion
Tilt Error Motion
17
For all three of these tests, the standard suggests using non contact measurement sensors to
measure the displacement of the spindle relative to the work-piece as it rotates. An example setup
used in the ISO standard can be seen in figure 2.2.
Figure 2.2 – Five-sensor system for measurement of rotating sensitive direction spindle error motions [3]
The output from radial and axial error motion tests are two error motion values, calculated from the
error motion polar plot. These are the synchronous error motion value, errors that occur at the same
frequency as the rotational speed of the spindle (once per revolution errors) and the asynchronous
error motion value, errors that occur at frequencies other than the rotational speed of the spindle.
Figure 2.3 shows a typical polar plot and displays the points of maximum synchronous and
asynchronous error motion.
Figure 2.3 – Error motion polar plot displaying synchronous and asynchronous error motion [3]
18
2.1.3.
ISO 230 part
Determination of Vibration Levels
Dete
i atio of Vi atio Le els is the standard for assessing the affect of vibration
on work-piece accuracy and surface finish quality. The requirement for vibration control is
recognised through the manufacturing industry so that vibration causing undesirable effects can be
mitigated. These effects are identified principally as: [4]





unacceptable cutting performance with regard to surface finish and accuracy
premature wear or damage of machine components
reduced tool life
unacceptable noise level
physiological harm to operators
Only vibration errors that occur as a result of unacceptable cutting performance are covered in the
scope of ISO 230. It is these errors that are investigated during spindle analysis.
This part of the ISO 230 standards is in the form of a technical report and as such does not describe
standard methods of test for identifying unwanted vibration magnitudes.
2.1.4. ISO Standards Conclusions
The tests described above from the ISO standards all require the machine to be freed up to allow
testing to take place. When faced with the types of production targets talked about previously,
production teams are apprehensive about taking a machine off line to perform capability testing. A
system with a standard setup that could be used to measure all spindle error would be extremely
advantageous in terms of reduced setup time of equipment and testing times. The sta da d does t
specify a type or make of instrumentation to be used to carry out testing, so it is left to the user to
find a solution. It is not likely that this solution will comprise of a single piece of equipment and
based on the limited number available, capable of dynamic measurement, it will most likely be
expensive and sometimes prohibitively so. The alternative is to use a service provider but similar
issues may occur if lots of measurements are required, for example for lots of machines and/or for
statistical process control (SPC) or preventative maintenance.
19
2.2. Non-Contact Sensing Technology
The aforementioned standards suggest the use of non contact measurement to analyse the spindle
behaviour. The ability to monitor a target without physical contact offers several advantages over
contact measurement, including the ability to achieve higher measurement resolution and increased
dynamic response to moving targets. They are also virtually free of hysteresis and there is limited
risk of damaging fragile targets because of contact with a measurement probe. Three typical non
contact measurement sensing technologies (as suggested by ISO) are [2] [3]:



Capacitance
Eddy Current
Laser triangulation
Each of these technologies has its advantages and disadvantages for use in spindle analysis. The
following sections look at these sensors in detail and describes there suitability for use in spindle
analysis measurement.
2.2.1.
Capacitance Sensors
Capacitive sensors work by measuring changes in an electrical property called capacitance.
Capacitance describes how two conductive objects with a space between them respond to a voltage
difference applied to them. When a voltage is applied to the conductors, an electric field is created
between them causing positive and negative charges to collect on each object (see figure 2.4). If the
polarity of the voltage is reversed, the charges will also reverse.
Figure 2.4 – Principle of capacitance sensor [5]
Capacitive sensors use an alternating voltage which causes the charges to continually reverse their
positions. The moving of the charges creates an alternating electric current which is detected by the
sensor. The amount of current flow is determined by the capacitance, and the capacitance is
20
determined by the area and proximity of the conductive objects. Larger and closer objects cause
greater current than smaller and more distant objects. The capacitance is also affected by the type
of nonconductive material in the gap between the objects.
Capacitance sensors can achieve nanometre resolutions with high accuracy measurement and are
not sensitive to material changes. However, they are very sensitive to changes in the interfacial fluid
i.e. they are affected by dirty environments where dust, oil and coolant may be present between the
sensor and the measuring surface.
2.2.1.
Laser Triangulation Sensors
Laser triangulation sensors work by emitting a laser beam onto a target and reflecting it back onto a
detector in a triangular configuration. The laser beam is projected through a lens to the target and is
then reflected from a target surface to a collection lens. This lens is typically located adjacent to the
laser emitter. The lens focuses an image of the spot on a linear array camera. The camera views the
measurement range from an angle that varies from 45 to 65 degrees at the centre of the
measurement range. The position of the spot image on the pixels of the camera is then processed to
determine the distance to the target as in figure 2.5.
Figure 2.5 – Laser triangulation principle [38]
21
The signal from the detector is used to determine the relative distance to the target. This
information is then typically available through an analog output, a digital interface or a digital display
for processing.
Laser triangulation sensors have a large standoff so there is a reduced risk of damage during setup
and also offer very high sampling frequencies. However, they are susceptible to self heating,
environmental influences such as humidity and material in the gap between the sensor and the
target. They are also quite bulky and expensive.
2.2.2.
Eddy Current Sensors
Eddy current sensors operate on a principle based on Le z s la [6]. Most eddy current sensors are
constructed with a sensing coil, which is a coil of wire in the head of the probe. When an alternating
current is passed through the coil it creates an alternating magnetic field. When a metallic target is
present in this magnetic field the electromagnetic induction causes an eddy current in the target
material in a perpendicular plane to the magnetic field of direction. This induced eddy current
generates an opposing magnetic field which resists the field generated by the sensing coil. The
interaction of the two magnetic fields is sensed using electronics and converted into an output
voltage that is directly proportional to the distance between the sensor and the target.
Figure 2.6 – Principle of eddy current sensor [7]
Compared to the other noncontact sensing technologies, high-performance eddy-current sensors
have some distinct advantages.
22




Tolerance of dirty environments
Not sensitive to the interfacial material between the probe and target
Often less expensive and much smaller than laser triangulation devices
Often less expensive than capacitive sensors
Eddy current sensors are not without their disadvantages. The main challenge is that the output
changes with the use of different target materials. This leads to a requirement for careful calibration
of the sensors.
2.2.3.
Eddy Current Sensor Calibration
It is normal practice that the eddy current sensors are calibrated by the manufacturer before they
are delivered to the customer [8]. Each sensor is calibrated to its own individual signal conditioning
unit; these cannot be interchanged between sensors without affecting the output. The length of
cable used to connect the sensor to the signal conditioning unit also has an effect on the output, so
once calibrated cannot be exchanged for a cable of differing length. The main challenge for high-end
specification eddy current sensors with linear outputs in the region of 0.2% is that they must be
calibrated to a specific target material. There is a significant difference in the output of the sensors
when used with a ferrous target compared to a non-ferrous target and as such the sensors are
usually calibrated to either one or the other.
This leaves room for a certain amount of measurement uncertainty between materials of a similar
nature. For example the output from an aluminium alloy target would be different to that of pure
aluminium. Some sensor manufactures provide a method for self calibration for fine adjustment [8].
However, if it is necessary to use the sensors on both ferrous and non-ferrous target then either a
new set of sensors will be required or they will continuously need to be sent back to the
manufacturer for re-calibration to a new target material.
Alternatively low cost eddy current sensors are available that are not calibrated to a specific target
material. However, the output from these sensors can be very non-linear and as such requires
careful calibration (see section 3.3.2 for details on calibration methods).
23
2.2.4.
Electrical Run-out
A further concern when using eddy current sensors is the occurrence of a phenomenon known as
electrical run-out. As described by Yating [9], the electrical run-out problem is caused by the change
or uneven distribution of sample conductivity and relative permeability of the target material. As
eddy current sensors work by penetrating the surface of a target when creating a magnetic field, any
defects that appear under the surface will affect the reciprocated magnetic field and hence the
voltage, which is used to determine the distance between the sensor and the target. The
penetration depth of the magnetic field depends on the permeability of the target material. Teruhiro
& Ma [10] [11] look at the permeability of a number of metal composite materials. The affects of this
problem and methods used to mitigate it are explored in later chapters.
2.3. Related Work
In order to gain an understanding spindle error it is important to understand spindle design and the
latest technologies being employed. The CIRP Annals 2010 [1] investigates the latest in machine tool
spindle units, providing detailed information about spindle components and there thermal and
dynamic behaviour. The majority of machine tools in use today are equipped with motorised
spindles to enable higher speeds to be achieved and reduce the number of mechanical parts with
the potential to fail.
These spindles have a minimum of two bearing sets, predominately using a ball bearing design and it
is this bearing system that has the greatest influence on the lifetime of the spindle. Angular contact
ball bearings are most commonly used in high speed spindles due to their low friction properties and
ability to withstand external loads in both axial and radial directions. A method of predicting bearing
life and as a result spindle life would be extremely beneficial to machine tool users when planning
maintenance schedules.
Some of the highest precision spindles are found in grinding machines where high accuracy is
necessary during finishing operations. The present internal cylindrical grinding spindles have a runout requirement of less than 1µm. A submicron resolution would therefore be required from the
chosen sensor.
24
In the past an increase in reliable rotational speed was desired to enable greater metal removal
rates. The focus has now shifted to spindles with higher achievable torque up to 15,000rpm. When
performing spindle analysis this would result in a requirement for a sensor capable of logging data at
a minimum of 15 kHz, to capture 60 points per revolution. The sta da d does t spe if a e ui ed
number of points per revolution, but from experience, anything less than 60 points and
measurement accuracy starts to be affected.
This reduction in spindle speed also reduces the level of friction and therefore the heat generated
from the bearings and motor. According to Bryan [12], machine tool thermal distortion can account
for 70% of the total machining error. Methods for reducing thermally induced error through
machine and spindle design can help to minimise this error. Postlethwaite [13] discusses thermal
error reduction techniques through design and compensation. The majority of error reduction
methods involve large amounts of data being capture and used to apply compensation to a machine.
This works well for the machine on which the data was captured however, transferring the
technique to other machines can be difficult and time consuming. Methods of compensation and
thermal error prediction have been well explored [14] [15] [16].
This is not the only method of error reduction, many solutions exist for the machine tool builder to
reduce thermal errors that can be applied at the design stage including; symmetric machine tool
structures to allow for a uniformed thermal growth and therefore better prediction of thermal error.
Liquid circulation cooling systems, which pump chilled oil, water or other coolant around the moving
elements to extract some of the heat generated. Low thermal expansion coefficient materials may
be applied if they maintain other key material properties. Where this is not a viable option, the
machine can be designed so that the direction of thermal growth is away from the thermal datum so
as to limit the affect on tool part accuracy.
According to the ISO standards, geometric spindle measurement can be conducted using a number
of types of non-contact measurement sensors. Jywe [18] presents a measurement system using laser
diode and a quadrant sensor for measuring the radial error motion of a wide range of machine tool
spindles. Due to the high sampling frequency of laser measurement, very high speed spindles can
still be measured using this method. However, the achievable resolution is only 0.7µm which would
not be sufficient for the majority of spindles in production today.
Castro [19] also presents a method of evaluating spindle errors using laser measurement. However,
this method implements a laser interferometer, which enables a high resolution of 1nm to be
achieved, while still having the benefits of the high sampling frequency. However, using a laser
25
interferometer would mean that only one measurement at a time would be possible. This would
require longer test periods, which would result in longer machine downtime periods.
Eric Marsh [20] has produced a book on precision spindle metrology, describing concepts and
techniques for measuring spindle performance.
When running, the spindle speed will fluctuate by some amount. It is therefore important to
synchronise the data with the rotation of the spindle. Many spindles have rotary encoders, but
access to the output may be difficult. A method for determining the spindle rotation will therefore
be required. Marsh suggests using a slightly eccentric target to provide a once per revolution
component to the output data. Using this method will require careful control so as not to introduce
vibration from imbalance. Xiaodong [21] [22] presents a new method for characterising radial error
motion and discusses the limitations of current spindle motion analysis techniques.
Spindle data will contain a variety of components, occurring at different frequencies. Data
acquisition requires a fixed sampling frequency, but there is no guarantee that the spectral content
of the measurement will be limited to a range within the sampling frequency. Any frequency content
above one half of the sampling rate will be incorrectly sampled as a lower frequency during analog
to digital conversion.
In high precision spindles that offer nanometre level error motion, it is important to distinguish
between spindle error and the out of roundness of a test-bar. As long as the artefact form and
synchronous spindle error are repeatable then reversal methods proposed by Donaldson and Estler
or a multi probe method will result in complete separation of test bar form errors and spindle error
motion. Ma sh e ie s the ad a tages of ea h i his pape , A o pa iso of e e sal a d
ulti-
probe error separation [23].
As previously mentioned, asynchronous spindle error caused by bearing defects and out of balance
shafts can provide unwanted vibration that results in poor surface finishes and form error of
machined parts. Martin [24] discusses the need for precision spindle bearing analysis and proposes
methods for analysing these errors.
A major source of vibration in precision high speed machine tool spindles is induced from bearing
sets. This can be caused by inherent geometrical errors in the assembly, including out of round
bearings or out of balance shaft and stator. It may also be caused due to defects in bearing races and
other contact surfaces. The magnitude of potential causes often makes vibration diagnostics a
difficult task and when it may result in expensive remedial action may be undertaken, it is important
to isolate and identify specific vibration causes. Vafaei [25] investigates the use of spectral analysis
26
techniques to monitor vibration in high speed machine tool spindles. He proposes the use of an
auto-regression moving average (ARMA) as opposed to the traditional Fourier analysis. The ARMA
method offers higher frequency resolution due to the averaging method employed and better
vibration magnitude estimation. However, this method requires a large quantity of data processing
to be carried out.
Felten [26] has produced a guide to understanding and identifying bearing vibration frequencies. It
provides details on the defect frequencies, including ball pass, ball spin and fundamental frequencies
and how to calculate them based on specific spindle speeds and bearing specifications. Once these
frequencies have been calculated it is easier to identify them on the vibration spectra and
monitoring the magnitude over time, predict bearing failure. ISO standards also exist for calculating
load ratings and predicting bearing life [27] [28].
When performing dimensional metrology it is important to employ robust methods of test to enable
confidence of test results. Bell and Flack [29] [30] have produced good practise guides to tackle key
issues such as traceability and uncertainty of measurement.
2.4. Conclusions
Spindle analysis is a relatively new concept and as such there has been a limited amount of research
conducted around the subject. There are solutions for the measurement procedure and
interpretation of the errors that have been standardised but the availability of measuring systems is
very limited and in most cases not suitable for measuring all the spindle related errors or have
limitations when considering their application as a tool or maintenance engineers working in dirty
workshop environments.
2.4.1.
Aims and Objectives
The aim of the research described in this work is:


To research and review the differing types of sensing technologies for their applicability to
machine tool spindle error measurement
To investigate data capture and processing methods with consideration of application in
workshop conditions
27
The main objectives of the research are:





To investigate differing non-contact sensing technologies for their suitability for spindle
analysis
To design suitable test-bars and fixtures for use in a spindle analysis system
To develop data processing methods
To provide a more affordable solution for service organisations and maintenance teams
To be able to perform spindle analysis in manufacturing environments
28
3.0
Sensor Testing & Selection
As stated in the literature review, there are three different types of non-contact sensing
technologies recommended in the ISO standards. Testing was carried out on some of these sensors
to determine the most suitable solution for the desired application, as defined in section 1.2.
3.1. Sensor Specification Requirements
The selection of the most appropriate sensor for the specified job is key to the success of the design.
From the point of view of this project the most important factors were:







Resolution
Range
Standoff Distance
Bandwidth or Sampling Frequency
Linearity
Accuracy and Repeatability
Thermal Stability
Resolution – This is the smallest increment of displacement that the sensor can measure, for
example when measuring the run-out of a machine tool spindle a resolution of 0.05µm would be
sufficient for the vast majority of spindles in production. Sometimes the theoretical resolution is not
achievable due to noise, typically electrical noise, which may vary depending on how or where the
sensor is used, so this must be considered as well.
Range – This is simply the distance over which the sensor produces an output, for example in volts
that can then be converted into a displacement measurement. The maximum displacements are
seen during spindle thermal testing, and experience suggests that anything over 500 µm is
uncommon. Therefore, this would be the minimum range requirement of the sensor.
Standoff – This is the distance between the sensor and the target before an output is produced. A
sensor with a very small standoff will increase the likelihood of it being damaged during test setups.
29
Sampling Frequency – This is the defined by the number of samples logged over a certain time,
usually 1 second and is measured in hertz. Toda s high to ue spi dles ha e a
a i u
spi dle
speed of around 15,000rpm. If a minimum of 60 samples per revolution are required, so that
accuracy of the error measurement is not compromised, then this would require a sampling
frequency of 15 kHz.
Linearity - In an ideal world the output from any sensor would be perfectly linear and not deviate
from a straight line at any point. However, in reality there will be slight deviations from this line
which define the system linearity. Typically, linearity is specified as a percentage of the Full Scale
Measurement Range (FSR). While linearity is a factor, sensors can be calibrated to minimise the
error.
Accuracy and Repeatability – the accuracy of a sensor is its ability to provide an output as close to
that of the true value as possible and the repeatability is its ability to repeat that output for the
same measurement over a number of separate tests.
Thermal Stability – is the stability of the sensor output over a period of time when subjected to
internal or external sources of heat.
Other factors
The size of the required target area will vary between sensors. For example, laser sensors have a
very small spot size (in the order of ø25µm). This allows small test bars to be used which would allow
high rotational speeds when measuring high speed spindles. Conversely, eddy current sensors have
a larger target area and therefore require large diameter test-bars to be used.
The cost of the sensors is also a significant factor when choosing a solution, as a minimum of 5
sensors are required. Cost is also an issue when replacing broken or damaged sensors; this is
especially relevant when dealing with sensors with very small standoff distances.
Table 1 - Minimum Required Specification
Minimum Specification
Resolution
0.05 µm
Range
500 µm
Sampling Frequency
15 kHz
30
Table 1 shows the required specification for the most important criteria of the chosen sensor.
Although the other factors mentioned in this section are not present in the table, they are still
important. However, it is the criteria in the table that will have the largest influence on the sensor
selection.
3.2. Laser Triangulation Testing
A number of laser triangulation sensors were researched and investigated, to find the most
appropriate solution, with the laser that best matched the required specification being the Keyence
LK-H022. The sensor had many features that stood it apart from others available including being
approximately 60% of the conventional size, with an impressive specification:
Table 2 - Keyence Laser Specification
Specification
Range
6 mm
Sampling Frequency
392 kHz
Spot Size
Ø25µm
Linearity
±0.02% of range
Repeatability
0.02 µm
The specification did not include the sensor resolution; this needed to be determined during testing.
There was a concern that the sensor would generate a substantial amount of heat, which may cause
problems when performing thermal tests, this also needed to be quantified during testing. The
following tests were conducted to help determine whether or not the sensors were capable of being
used to perform spindle analysis tests:




Warm up period test
Thermal stability test
Static resolution test
Dynamic resolution test
31
3.2.1. Warm Up Period Test
The aim of this test was to assess the affect of internally generated heat from the sensor head and
the time taken to stabilise.
The test was carried out by setting up the laser in an aluminium fixture against a thermally stable
granite table target (see figure 3.1). The sensor was switched on and left to log the displacement of
the sensor, while the temperature of the sensor was monitored using a thermal imaging camera.
Thermal camera
Laser Sensor
Aluminium Fixture
Granite Table
Figure 3.1 – Thermal Warm up Test Setup
The sensor was left logging for 15 minutes, as can be seen in figure 3.2, during which time it
displaced by 7 µm before stabilising after approximately 8 minutes.
32
1
0
Displacement (microns)
-1
-2
-3
-4
-5
-6
-7
-8
0
5
10
15
Time (minutes)
Figure 3.2 – 15 min Warm Test with Aluminium Fixture
The test was then performed again, this time with a carbon fibre fixture, to limit the expansion. As
can be seen from figure 3.3, the sensor displaced by 1.5µm before stabilising after approximately 5
minutes.
0.5
Displacement (microns)
0
-0.5
-1
-1.5
-2
0
5
10
15
Time (minutes)
Figure 3.3 – 15 min Warm Up Test with Carbon Fibre Fixture
33
3.2.2. Thermal Stability Test
The aim of this test was to assess the stability of the sensor over a longer period of time once the
warm up cycle had been carried out.
The test was carried out using the same setup as in the previous test (see figure 3.1) but again using
the carbon fibre fixture to limit the thermal growth. The sensor was switched on and left to warm up
for the required 5 minutes, the displacement of the sensor was then logged for a two hour duration,
while the temperature of the sensor was monitored using a thermal imaging camera. The test was
run for 2 hours to give the sensor enough time to be exposed to external sources of heat which
could affect the stability of the measurement.
Figure 3.4 shows a thermal image of the sensor and fixture setup with 5 spot positions identified.
The temperature at these spots was monitored for the two hour test duration to monitor the flow of
heat through the sensor and fixture.
38.4°C
SP05
SP01
35
30
SP02
SP04
SP03
25
21.8°C
Figure 3.4 – Thermal Image Showing Spot Positions
Figure 3.5 shows the spot temperatures of the sensor and fixture during the two hour thermal
stability test. As can be seen from the plot the temperature at each position remained stable to
within 0.5 ˚C for the duration of the test. The sensor itself is very hot, approximately 36.5˚C, which
may influence tests such as a long term ETVE tests.
34
SPO1
SPO2
SPO3
SPO4
SPO5
Figure 3.5 – Plot of Spot Temperatures
Figure 3.6 shows the displacement of the sensor during the two hour thermal stability test. As can be
seen the displacement is less than ±0.3µm, which is very good for a test of this nature. A
displacement of such a small magnitude can be attributed to environmental influence.
0.5
0.4
Displacement (microns)
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
0
20
40
60
80
100
120
Time (minutes)
Figure 3.6 – Displacement during 2 hour stability test
35
3.2.3. Static Resolution Test
The aim of this test was to assess the static resolution of the sensor when measuring against a
cylindrical test bar, as would be the case during spindle analysis.
The test was carried out by positioning a test-bar against to granite square to limit the influence of
any external vibration, as shown in figure 3.7. The sensor was used to log data at the full sampling
capability of 392 kHz with no averaging in place.
Figure 3.7 – Static resolution test setup
Figure 3.8 shows the displacement data of the sensor when logging at 392 kHz with no averaging.
When logging at such high frequencies there was a lot of noise on the signal, which gives the
appearance of the static resolution to be 1 µm.
Displacement (microns)
1.5
1
0.5
0
-0.5
-1
0
5000
10000
15000
20000
25000
30000
Number of Samples
Figure 3.8 – Static resolution with 0 x averaging displacement data
36
Using standard deviation on the data set to identify the minimum possible signal, measurable with
the sensor gives 0.22 µm. This is quite good; however it exceeds the minimum requirement of 0.05
µm.
Figure 3.9 shows the displacement data of the sensor when logging at 392 kHz with 16 times
averaging. Even with 16 times averaging, this still provides a sampling frequency of 15 kHz, which is
in line with other available systems. As can be seen, this reduces the noise on the signal, which gives
the appearance of the static resolution to be 0.3 µm.
0.3
Displacement (microns)
0.2
0.1
0
-0.1
-0.2
-0.3
0
5000
10000
15000
20000
25000
30000
Number of Samples
Figure 3.9 – Static resolution with 16 x averaging displacement data
Using standard deviation on the data set to identify the potential resolution achievable with the
sensor, gives 0.06 µm. This is comparable with the required resolution.
3.2.4. Dynamic Resolution Test
The aim of this test was to assess the dynamic resolution of the sensor when measuring against a
cylindrical test bar, as would be the case during spindle analysis.
37
The test was carried out by positioning a test-bar in a typical machine tool spindle and rotating it at
1000rpm, as can be seen figure 3.10. The sensor was used to log data at the full sampling capability
of 392 kHz with no averaging in place.
Figure 3.10 – Dynamic resolution test setup
Figure 3.11 shows the displacement data of the sensor when logging at 392 kHz with no averaging.
As can be seen from the plot, when logging during a dynamic measurement, the noise on the signal
increases significantly. This gives the appearance of the dynamic resolution to be approximately 10
µm.
30
25
Displacement (microns)
20
15
10
5
0
-5
-10
-15
0
5000
10000
15000
20000
25000
30000
Number of Samples
Figure 3.11 – Dynamic resolution with 0 x averaging displacement data
38
Using standard deviation on the data set to identify the potential dynamic resolution achievable with
the sensor, gives 6.63 µm. This is far outside the required resolution of 0.05 µm.
3.2.5. Conclusions of Laser Testing
The initial concerns about the sensors self heating were dispelled following good results from the
warm up and thermal stability tests. Once the sensor had been allowed to warm up for 5 minutes
the displacement output then remained stable to within 0.6 µm for the two hour test duration.
There still remained a slight concern about the affects the sensor would have on an ETVE test by
adding to any externally induced heat.
The laser performed well under static resolution testing providing a possible resolution of 0.06 µm
when the data was averaged at 16 times (15 kHz sampling rate). However, when the laser was tested
under dynamic conditions the resolution increased significantly to 6.63 µm at 0 averaging. While this
data could be averaged 16 times to bring it to a sample rate of 15 kHz, it would still be far outside
the required resolution of 0.1 µm.
This significant increase in noise was due to light scatter of the laser while trying to measure on a
rotating surface. Keyence recommend setting the averaging to 16384 to stabilise the sensor reading,
bringing the sampling rate down to 24 Hz. This is considerably lower than our required sampling
frequency.
Therefore, while this sensor seemed to offer everything required on its specification, testing
revealed that it was not suitable for the application of dynamic spindle analysis.
39
3.3. Eddy Current Testing
A number of eddy current sensors were investigated, a summary of which are listed in table 3, to
find the most appropriate solution. Two sets of sensors were tested with the sensor that performed
best during testing being described in this section. The results of testing from the other sensor can
be found in the appendix.
Figure 3.12 – Kaman eddy current sensor and signal conditioning unit
Figure 3.12 shows a CAD model of the Kaman eddy current sensor and the signal conditioning unit.
When initial testing was conducted there were some concerns about the robustness of the system.
When pressure was applied to the signal conditioning unit or it was moved during measurement, the
sensor readout was affected. However, this was overcome by enclosing the signal conditioning units
in a small robust case that ensured stability of the output (see figure 3.13).
3.3.1. Thermal Stability
The aim of this test was to assess the thermal stability of the Kaman eddy current sensors and their
signal amplifiers, firstly at room temperature and then with external heat source.
Two sensors were setup measuring against a steel surface to measure the change in temperature
with change in ambient temperature, to assess the thermal stability of the sensors (see figure 3.12).
Two temperature sensors were also setup, one to measure the ambient temperature near the
sensors and the other to measure the ambient temperature of the signal conditioning unit.
40
Logging was initially for 1 hour to measure the stability of the sensors. After 1 hour the ambient
temperature around the amplifier was heated up to 42˚C and then left to cool, to assess the effect
this had on the sensor output. The ambient temperature around the sensors was then heated to
approximately 32˚C to assess the effe t this had o the se so output a d
as the agai left to
cool.
Figure 3.13 – Thermal stability test setup
Figure 3.14 below shows the temperature and displacement of the eddy current sensors and signal
amplifier over the 100 minute test duration.
7
45
6
40
5
Sensor Ambient
Amplifier Ambient
Temperature (C)
Sensor 3
35
3
2
30
Displacement (microns)
4
Sensor 2
1
0
25
-1
20
-2
0
10
20
30
40
50
Time (minutes)
60
70
80
90
100
Figure 3.14 – Comparison of displacement and temperature during thermal stability test
41
During the first hour of the stability test the ambient temperature of both the sensors and the
amplifier remained stable to within 0.5˚C. This resulted in a displacement of sensor 2 and sensor 3 of
0.9 µm and 0.6 µm respectively.
The amplifier was then heated to 42˚C which also heated the ambient of the sensors to 29˚C, which
resulted in a change in displacement of approximately 2 µm in both sensors.
A second test was then performed by heating the ambient temperature around the sensor to 36˚C
logging the temperature and displacement every 2 seconds, as can be seen in figure 3.15.
10
40
9
38
Sensor 2
Sensor 3
Sensor Ambient
Sig Con Ambient
7
6
36
34
32
5
30
4
28
3
26
2
24
1
22
0
20
0
2
4
6
8
10
12
14
16
18
20
Time (minutes)
Figure 3.15 – Comparison of displacement and temperature during second thermal stability test
The above plot shows a displacement of sensors 2 and 3 of 9 µm and 6µm respectively. There is also
a good correlation between the ambient temperature around the sensors and the displacement
data.
42
Temperature (C)
Displacement (microns)
8
Figure 3.16 shows the temperature plotted against the displacement of sensor 2 as the setup cools.
40
Temperature (C)
35
30
y = 1.445x + 21.811
25
20
15
1
2
3
4
5
6
7
8
9
10
Displacement (microns)
Figure 3.16 – Sensor 2 temperature against displacement
As can be seen there is a good linear relationship between the temperature and displacement
indicating good thermal stability of the sensor. There is a change of 0.69µm / ˚C.
Figure3.17 shows the temperature plotted against the displacement of sensor 3 as the setup cools.
40
Temperature (C)
35
30
y = 2.226x + 19.739
25
20
15
1
2
3
4
5
6
7
Displacement (microns)
Figure 3.17 – Sensor 3 temperature against displacement
Again, there is a good linear relationship between the temperature and displacement indicating
good thermal stability of the sensor. There is a change of 1.08µm / ˚C.
43
The sensors were then left logging for a 50 hour period. Figure 3.18 shows the temperature and
displacement data.
0.6
25
S1
S2
Sensor
Sig Con
24.8
24.6
24.4
0.2
24.2
0
24
23.8
-0.2
23.6
Temperature (C)
Displacement (microns)
0.4
23.4
-0.4
23.2
-0.6
23
0
5
10
15
20
25
30
35
40
45
50
Time (hours)
Figure 3.18 – 50 hour thermal stability test
The sensors displaced by approximately ±0.3 µm, showing very good thermal stability. There is also
an excellent correlation between the ambient temperature data and the displacement.
44
3.3.2. Linearity and Calibration
The measurement range of the eddy current sensor is 500µm. However, only the first 100µm was
required for the purpose of testing. This was due to the most sensitive section of the sensor
appearing in this part of the range. The blue trace in figure 3.19 shows the output over the first
100µm and as can be seen the output is very non-linear.
8
7
Output (volts)
6
5
4
3
2
1
0
0
20
40
60
Displacement (microns)
80
100
Figure 3.19 – Non-linear output from eddy current sensor
The output data file from this test was then input into Matlab and a curve fitting tool used to fit to
the data. A low order polynomial was used to ensure robustness of the fit and to enable the fit to be
performed to within 0.1µm the curve was chopped into sections. The end result is a calibration file
for each individual part of the curve fit to within 0.1µm accuracy.
The data in figure 3.19 was captured using a 16-bit National Instruments (NI) data acquisition device
and a bespoke computer application which uses standard NI APIs. The low order polynomial
calibration file can then be loaded into the software whenever the sensor is being utilised with the
target material for which it has been calibrated.
With the calibration file input into the windows application the repeatability of the calibrated sensor
can be tested. The axis was moved in steps of 10 µm in the same direction from the same starting
position for five separate runs and the deviation from linearity plotted in figure 3.20.
45
0.15
Run
Run
Run
Run
Run
Repeatability (microns)
0.1
0.05
1
2
3
4
5
0
-0.05
-0.1
-0.15
-0.2
0
10
20
30
40
50
60
Displacement (microns)
70
80
90
Figure 3.20 – Repeatability results for eddy current sensor output
As can be seen the sensor measured repeatably to within 0.2 µm over the 100µm range, validating
the method.
3.3.3. Sensor Sensitivity Test
The sensitivity of the sensor was calculated and plotted over the first 40 µm of the range, as this was
the area where the non-linear section occurred. Figure 3.21 shows the peak sensitivity occurring at
approximately 15 µm from the beginning of the range.
500
450
400
Sensitivity (mV)
350
300
250
200
150
100
50
0
0
5
10
15
20
25
30
35
40
45
Displacement (microns)
Figure 3.21 – Repeatability results for eddy current sensor output
46
3.3.4. Static Resolution Test
The aim of this test was to assess the static resolution of the sensor when measuring in the most
sensitive part of the measurement range.
The test was carried out by positioning the sensor to measure off the same aluminium target as used
during its calibration. The sensor was moved to the most sensitive part of the range determine the
sensors highest static resolution. Data was logged at 10 kHz using a 16 bit National Instruments (NI)
data acquisition device.
Figure 3.22 – Eddy current static resolution test setup
Figure 3.23 shows the displacement data of the sensor over 3 seconds when logging at 10 kHz. As
can be seen from the plot, when logging during a static measurement, the displacement is
approximately 0.06 µm, which is comparable with the required resolution of 0.05 µm.
47
0.05
0.04
0.03
Displacement (microns)
0.02
0.01
0
-0.01
-0.02
-0.03
-0.04
-0.05
-0.06
0
5000
10000
15000
20000
25000
30000
Number of Samples
Figure 3.23 – Eddy current static resolution
Using standard deviation on the data set to identify the potential static resolution achievable with
the sensor, gives 0.011 µm, which is much better than the required resolution.
3.3.5. Dynamic Resolution Test
The aim of this test was to assess the dynamic resolution of the sensor when measuring in the most
sensitive part of the measurement range.
Using the calibration file that was previously created, a dynamic measurement was taken using a
high resolution piezo actuated platform flexure rig. The eddy current sensor was set up to measure
against the same target and the software loaded with the calibration file captured on a standard
milling machine using the proposed method. A Renishaw XL80 laser interferometer was used as a
traceable reference device (see figure 3.24).
48
Laser Optics
Eddy Current
Sensor
Movement
Figure 3.24 – High resolution piezo platform dynamic resolution test setup
The rig was set to vibrate at 100Hz and the magnitude of the oscillation adjusted to be
approximately 0.008 µm (this was the minimum magnitude before the signal became distorted).
Figure 3.25 shows the output from the Renishaw XL80 laser interferometer. There are some
imperfections of the sine wave but this is to be expected at this level of resolution and can be
attributed to the stability of the setup.
Figure 3.25 – Renishaw XL80 laser output on vibration rig
Figure 3.26 shows the output from the eddy current sensor sampling at 10 kHz. As is expected at this
resolution there is some noise on the output signal but a sine wave can still clearly be seen at a
magnitude of between 0.008µm and 0.01µm.
49
8
6
Displacement (nm)
4
2
0
-2
-4
-6
-8
-10
0
1000
2000
3000
4000
5000 6000
Samples (Hz)
7000
8000
9000
10000
Figure 3.26 – Output from eddy current sensor on vibration rig
The standard deviation of the data in figure 3.26 shows the noise level to be at approximately 0.0037
µm. Since the data was sampled at 10 kHz averaging could be applied to clean up the signal, if this
does not impinge on the required sampling frequency of the application.
3.3.6. Conclusions of Eddy Current Sensor Testing
There were some initial concerns about the stability of the output signal due to the quality of the
signal amplifiers. However, once these had been mounted in a rigid container the signal seemed a
lot more stable and they performed extremely well during the thermal stability testing.
Due to the non-linearity of the sensors they required careful calibration to the intended target
material. However, a method was successfully identified and implemented.
Static and dynamic resolution tests also showed the sensor to exceed the specifications with a very
small 0.008µm sinusoidal signal, correctly measured by the sensor. This indicates a theoretical
sensitivity based on the standard deviation of less than 0.004µm (4nm), which far surpasses the
resolution requirement and would allow for testing of extremely high precision spindles.
The testing described in this section was on low cost un-calibrated eddy current sensors but by
exploiting the non-linear output to get high resolution and calibrated to get accuracy, their suitability
has been confirmed. Testing was also conducted on other eddy current sensors, which offer good
linearity but the resolution is not quite as good, the results of this testing can be found in the
appendix.
50
3.4. Capacitance Sensor Testing
Testing was not carried out on any capacitance sensors as they have already been proven for use in
spindle analysis. While this system offers a good solution, there are some limitations when it comes
to practical spindle analysis.
Firstly, and most importantly, as previously mentioned the output of a capacitance sensor is affected
by any particles present in the area between the sensor and the target. As a result, they may not be
suitable for use in typical manufacturing environments, where swarf, dust and coolant may be
present in testing areas and where the uncertainty of such effects cannot be eliminated through
cleaning, because of the nature of application, a quick check by maintenance departments or even
by operators.
The system only utilises 5 channels, for measuring error in X, Y and Z and a further two sensors
measuring in X and Y for tilt. It is sometimes beneficial to use a 6 th channel during thermal testing to
offer a differential between the spindle nose and the end of the test-bar.
The commercial systems available at the time of writing this report are very expensive and therefore
may not be a ia le optio fo the
ajo it of s all to
ediu
size e te p ises “ME s looking to
implement spindle analysis processes. Tables 3 and 4 examine typical costs per channel.
51
3.5. Non-Contact Sensor Solution
There are many factors to consider when choosing the most suitable non-contact sensor, the most
important of which were identified at the beginning of this section. The testing described has
assessed the prospective sensing technologies against these criteria to establish their suitability.
Testing and research was carried out on further sensors not mentioned in this section. Results from
this testing can be found in the appendix, with the rest of the information gathered taken from the
manufacturer specifications.
The following pages display all investigated non-contact sensors in table form to allow for
comparison.
52
Table 3 – Comparison of Non-Contact Measurement Sensors Part 1
COMPARISON OF NON-CONTACT MEASUREMENT SENSORS
SENSOR TYPE
EDDY CURRENT (OLD
MICRO EPSILON)
LION PRECISION
SYSTEM
Spot Size
22mm2
-
 Ф 30 µm
Range
1mm
0.2mm
250µm
2mm
125 µm
10mm
0.2mm
0.2mm
9kHz
15kHz
8kHz
10kHz
10kHz
0.1µm
15nm
0.6 µm
0.1 µm
0.04 µm, Varies
Output Voltage
Range
0 – 10 VDC
±10V
0 – 10 VDC
0 – 10 VDC
0 – 10 VDC
Cost / channel
Circa £2,000
N/A
£785.00
£785.00
£610.00
Temperature Range
-
4°C – 50°C
-10°C – 60°C
-10°C – 60°C
-55°C - 105°C
Environmental
Resistant
YES
NO
NO
YES
YES
Sensitivity of input
Voltage
N/A
N/A
Negligible
None (12-24v)
None (12-24v)
Offset
Max Sampling
Frequency
Resolution
RIFTEK LASER
SMALL EDDY CURRENT
(4U ON KD-2446)
SMALL EDDY CURRENT
(2U ON KD-2446)
11.3mm2
3.14mm2
1.3mm
0.5mm
NOTES
Values in red are taken from manufacturer specifications
Values in blue are results of testing conducted on the sensors
53
Table 4 - Comparison of Non-Contact Measurement Sensors Part 2
Table 4 – Comparison of Non-Contact Measurement Sensors Part 2
SENSOR TYPE
MICRO EPSILON
LASERS
KEYENCE LASERS
(LK-G-32)
KEYENCE LASERS
(LK-H022)
CAPACITEC
CAPACITANCE SENSORS
LION CPL 190
DRIVER SYSTEM
Ф 30 µm
Ф 25 µm
-
-
Spot Size
 Ф 30 µm
Range
2 mm
10 mm
6 mm
2 mm
500 µm
Offset
10 mm
30 mm
20 mm
0.5 mm
250 µm
Max Sampling
Frequency
10 kHz
50 kHz
392 kHz
15 kHz
15 kHz
Resolution
0.03 µm
0.05 µm
0.02 µm, varies
TBC
10nm
Output Voltage
Range
-
-
-
0 – 10 VDC
±10V
Cost / channel
TBC ( £4k)
£2423.00
£3486.97
£2884.28
£4047.00
Temperature Range
-
0°C – 50°C
0°C – 50°C
-
-
Environmental
Resistant
No
No
No
No
No
Sensitivity of input
Voltage
-
-
-
-
-
NOTES
Values in red are taken from manufacturer specifications
Values in blue are results of testing conducted on the sensors
54
Through experimentation it has been found that in some cases the
a ufa tu e s spe ifi atio s is
not always what is achievable in the required applications.
Table 5 details the advantages and disadvantages of all 3 of the best performing potential noncontact sensing technologies.
Table 5 – Advantages and Disadvantages of the Potential Systems
SYSTEM
ADVANTAGES


PROVEN
CAPACITANCE
SYSTEM



It is a proven system
It has full traceability of
measurement
Excellent resolution
Can be used on any target
material
Can be used on small targets
DISADVANTAGES








KAMAN EDDY
CURRENT



KEYENCE LASER
LK-H022


Very inexpensive
Can be used in dirty
manufacturing environments
High sensitivity over short range
enables high resolution to be
achieved
Good measurement range to
resolution ratio
Extremely high sampling
frequency, allowing for high speed
testing without reduced accuracy
Large offset and range for ease of
setup and reduced risk of damage
Very small spot size, so extremely
small targets can be used allowing
for higher rpm testing






Very Expensive. May be exacerbated
by requiring dual range system or
multiple systems to measure error
motion and thermal on a range of
spindles
Long lead time on kit and possible
spares
Limited to a 5 channel system
Low maximum sampling frequency
compared to laser systems
Must be used in clean environment
Very small target range
Sensors need calibrating to the
target material
Require a larger target area,
although the miniature sensor head
is just 2mm in diameter, requiring
only a 6mm target
Relatively expensive
Resolution during dynamic
measurement is very poor due to
light scatter
Sensors are large, making fixture
design difficult
Sensor heads get very warm so
fixtures need to be thermally stable
55
After comprehensive testing and careful consideration of the key elements and requirements of the
type of sensor to be utilised, the Kaman eddy current sensor was chosen as the most suitable for the
required application.
This sensor fulfilled all the key requirements of resolution, sampling frequency and range, whilst also
offering the added benefit of being able to be used in manufacturing environments. They are also
very inexpensive in comparison to the other sensor options, enabling a more affordable overall
system to be produced, which must be considered as a key motivating factor for broader application
of this important measurement.
56
4.0
System Design
The selection of an appropriate displacement sensor completes one of the main parts of the project.
To enable spindle analysis testing to be carried out to ISO standards, further system hardware and
software needed to be designed and manufactured.
4.1. Equipment Design
The main elements of hardware to be identified or design are:



Precision Test-bars
Fixtures
Data Acquisition Device
The design of suitable test-bars and fixtures is very important to enable high accuracy measurement
to take place.
4.1.1. Test-bar Design
Although electrical run-out (see section 2.2.4) can be removed using one of the reversal methods, it
is a good idea to limit it so that it does not dominate the signal. Aluminium has one of the lowest
permeability values available, so for error motion measurement, aluminium test-bars were
developed.
Figure 4.1 shows the manufacture of a short test-bar on a diamond turning machine with a radial
and axial run-out of <50nm. The test-bar was initially machined on a standard turning machine and a
target area (20mm in length) machined at a very low material removal rate with a diamond tipped
cutter to produce a high precision mirror finished target area.
57
Figure 4.1 – Manufacture of short aluminium test-bar
Figure 4.2 shows the short aluminium test bar being used on a high precision air bearing spindle
measuring the radial error motion.
Eddy Current Sensor
Figure 4.2 – Short Test-bar measuring radial error motion
Figure 4.3 shows a long aluminium test-bar being used to measure the tilt error motion of a spindle.
This was manufactured using the same method as the short test-bar, only this one required two
precision surfaces at a set distance of 100mm, apart to facilitate the measurement of tilt.
58
Figure 4.3 – Long Test-bar measuring tilt error motion
The test-bars have a diameter of 20mm, as the eddy current sensors selected are a 2mm diameter
unshielded model. The unshielded eddy current sensors require a target area 3 times that of the
sensor diameter, therefore a minimum target area of 6mm diameter is required. As two sensors are
required to measure in two directions, it is important their magnetic fields do not interfere with
each other. To ensure sufficient distance between sensors a 20mm test-bar was chosen.
When using the flanged test-bars in figures 4.2 and 4.3 on a spindle with a face plate interface, there
is no fine adjustment for centring them. The adjustment comes from the clearance of the bolt holes
and requires the test-bar to be tapped into position. Using this method it is difficult to achieve a testbar run-out of less than 1µm. It can also be extremely time consuming, especially as the setup needs
to be done twice per error measurement.
When performing spindle analysis on a milling spindle with a tool holder interface, an un-flanged
test-bar can be used with precision collets that typically provide a run-out of <3 µm. This is an
acceptable level of run-out for a ball bearing spindle, but when measuring a high precision air
bearing spindle a run-out of < 1 µm is required to achieve the required level of resolution. This is due
to the high sensitivity area of the sensor only being over a small range.
59
Manufacturing the test-bars out of aluminium means that they are not ideal for thermal testing due
to its high expansion coefficient. The thermal expansion of the test-bar could be calculated and
removed from the data, if the temperature of the test-bar could be monitored to a sufficient level of
accuracy.
Alternatively, a test-bar manufactured from a thermally stable material such as Invar may be
preferable.
4.1.2.
Fixture Design
A set of fixtures has been developed for mounting the eddy current sensors during spindle analysis.
The fixture allows for adjustment in two planes so that the sensors can be positioned within the
required measuring range and the point where the sensitivity of the sensor is at its highest.
The fixture is manufactured from 10mm thick steel for structural stability. When measuring at a sub
micron level, it is important to reduce the external influences of vibration as much as possible. By
having a thick steel fixture it increases the stability of the setup and reduced the amount of vibration
induced into the sensors, which may affect the measurement.
Figure 4.4 shows the adjustable fixture being used to measure the radial error motion in the X and Y
directions.
Figure 4.4 – Adjustable fixture of radial error motion testing
60
When measuring thermal displacement, the steel fixtures are not appropriate, as the thermal
expansion properties of steel would influence the accuracy of the measurement. Figure 4.5 shows
the use of Invar bar and fixtures during a thermal test. Invar has a very low thermal coefficient of
expansion and so is a perfect material for manufacturing the fixtures.
Invar Bar
Invar Fixture
Figure 4.5 – Invar fixture for thermal testing
Further development of the fixtures would be beneficial, to allow for sub micron level adjustment,
which would reduce setup times. A five sensor nest would allow simultaneous measurements to
take place and enable thermal measurements to be taken using the same setup if manufactured
from a thermally stable material such as Invar.
4.1.3. Data Acquisition
A data acquisition device and logging software are also required, to capture and log the data from
the sensors. The required device must have the following:




8 channels – to allow scope for additional sensors to be used if necessary
16-bit resolution – this allows potential measurement to a resolution of 1nm
Capable of voltage logging
Simultaneous sampling – to enable accurate measurement at faster speeds
61
Figure 4.6 – National Instruments data acquisition device
Figure 4.6 shows a suitable data acquisition device from National Instruments. For ease of use, this
device has a USB connection to the laptop or PC to be used for data capture. It works with NI signal
express software, which is a basic NI software, for which a low cost software licence can be
obtained.
62
4.2. Data Processing
It is necessary to convert the raw voltage recorded by the NI software to microns, following data
capture. This data will then require processing to put it into a usable format for data interpretation.
The method of processing the data varies depending on the type of spindle testing; the following
section describes how this is done for each type of spindle analysis test.
4.2.1.
Thermal Data Processing
The amount of post processing required with thermal data depends on the type of test bar used.
Previous methods of thermal testing implemented the use of a steel test-bar which had a large form
error of approximately 20 µm. If a test-bar such as this, with a large run-out, is used then the data
needs to be averaged out so the magnitude of thermal growth can be determined. If a test-bar with
a small run-out (in the order of 1µm) is used then it may be possible to use the raw data without
averaging, assuming that the thermal error is of sufficient magnitude.
380
250 rpm
500 rpm
750 rpm
1000 rpm
370
Displacement (microns)
360
350
340
330
320
310
300
290
280
0
60
120
180
240
Time (minutes)
Figure 4.7 – Raw thermal data plot converted to microns
Figure 4.7 shows the Z-axis displacement of a spindle during a 4 hour step heating test, with the raw
voltage data converted into microns. As can be seen, the test-bar run-out is approximately 30µm
and it is only possible to obtain a rough order of magnitude of the spindle displacement. The data
63
was averaged and inverted to produce the plot in figure 4.8, which displays more clearly the
displacement of the spindle in the Z-axis direction to be approximately 20µm over the 4 hour run
time.
25
250 rpm
500 rpm
750 rpm
1000 rpm
Displacement (microns)
20
15
10
5
0
Z Displacement
-5
0
60
120
180
240
Time (Minutes)
Figure 4.8 – Post possessing averaged data plot
Thermal displacement data on its own is not that meaningful. It is important to monitor temperature
at various points around the machine as well, so that the cause of thermal expansion can be
identified. This can be done by using cheap surface mountable temperature sensors to measure
areas of thermal interest, as seen in figure 4.9.
Figure 4.9 – Surface mountable temperature sensor
64
These sensors offer an accuracy of ±0.5˚C from -10˚C +
˚C
ith a u i ue -wire interface which
requires only one port pin for communication. They can be easily assembled in series for positioning
around a machine tool, with extension wires for use on larger machines. Each device has a unique 64
bit serial code in an onboard ROM so each sensor can be easily identified and labelled appropriately
during data logging. The price per unit is approximately £2, so a full temperature measurement
system can be assembled for a very affordable price.
Figure 4.10 shows the associated temperature data from the displacement data in figure 4.8. The
temperature is logged from the temperature sensors using specialist software and plotted against
time to show the change in temperature during the test period.
24
250 rpm
500 rpm
1000 rpm
750 rpm
23
Temperature (C)
22
21
20
Spindle
Machine
Spindle
Spindle
Spindle
Machine
19
18
12:30:00
13:30:00
14:30:00
15:30:00
Front Left
Base
Ambient
Back
Front Right
Ambient
16:30:00
Time (hh:mm)
Figure 4.10 – Temperature data
4.2.2.
Geometric Data Processing
Data obtained from geometric testing requires the most amount of processing following capture.
When measuring high accuracy spindles with radial run-outs in the order of 1µm or less, then test65
bar run-out and form error cannot be ignored. Also, as mentioned in the literature review, when
using eddy current sensors electrical run-out error is also present in the signal. All of these errors can
be removed through the use of the following reversal technique.
4.2.2.1.
Donaldson reversal method
This method requires the fewest measurements (two), although setup time maybe increased. The
method works by taking two measurements, one in the forward direction and one in the reverse
direction. In the forward direction setup, the angular position of the spindle rotor, the spindle stator
and the test-bar and the eddy current sensor are all aligned. In the reverse direction setup, the eddy
current sensor and the test-bar are rotated through 180˚
ith espe t to the spi dle oto a d
spindle stator, as can be seen in figure 4.11.
Setup 1
Setup 2
Figure 4.11 – Donaldson reversal technique [20]
The t o
easu e e ts a e a o
i atio of the spi dle s adial e o
otio a d the test-bar run-
out, test-bar form error and electrical run-out. Using the following mathematical equations, the two
can be separated.
66
Other methods of are available for the removal of test bar form error and electrical run-out.
As mentioned in the literature review, Marsh [20] suggests using a slightly eccentric target to
provide a once per revolution component to the output data. This can introduce unwanted vibration
from an out of balance test-bar, when measuring at high precision levels of accuracy.
To overcome this, another method for measuring the angular position of the test-bar with respect to
the spindle was devised that uses a small mark in the test bar and an additional sensor to detect the
mark (this sensor is available because not all are used for dynamic measurement). This can then be
correlated with the mark from the second set of data to ensure the two data sets match up, even if
small variations in the rotational speed of the spindle occur. Figure 4.12 shows the run-out data from
the sensor with a sharp peak where the mark on the test-bar is located. Using a combination of low
pass filtering and a peak finding function in Matlab the individual peaks can be identified.
6.1
6
5.9
Displacement (Volts)
5.8
5.7
5.6
5.5
5.4
5.3
5.2
5.1
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
Number of Samples
Figure 4.12 – Peak finding data
This is important when measuring spindles where the rotation speed may not be constant and
therefore there will be a differing number of samples per revolution. By identifying the peaks the
67
number of samples per revolution can be found and the data interpolated to match up exactly with
the data from the reverse direction setup.
Figure 4.13 shows a zoomed in peak, and as can be seen the profile of the mark is not a smooth
curve. It is therefore import not to choose the highest peak as this could vary, but to use a low pass
filter so that the once per revolution peak is more accurate. As can be seen from figure 4.13, the
mark for the peak is central despite the rough extents of the mark.
6.14
Displacement (Volts)
6.12
6.1
6.08
6.06
6.04
6.02
6
1.5
1.502
1.504
1.506
1.508
1.51
1.512
1.514
1.516
1.518
4
Number of Samples
x 10
Figure 4.13 – Peak finding data
4.2.3.
Vibration Data Processing
Due to the high sensitivity of the sensors over a small range, it is possible to take vibration
measurements to identify the magnitude of spindle vibration frequencies. This can be either a
measurement taken on a static structural part of the spindle or a dynamic measurement taken
against the test-bar.
The data is the same raw voltage output file from the eddy current sensors as used in the other test.
This can be input into Matlab and used in a Fast Fourier Transfer (FFT) function to produce an FFT
plot, which can identify resonant frequencies in the spindle. An example of this can be seen in figure
68
4.14, which shows data taken from a high precision air bearing spindle. As can be seen, the vibration
is dominated by the rotation frequency of the spindle, as would be expected with a spindle of this
quality. At 1000rpm the spindle rotation frequency is 16.67 Hz, which is the frequency of the first
peak in the plot below; the repeated peaks are harmonics of the rotational speed.
Power spectral density (Log Y)
5
0
-5
-10
-15
-20
0
100
200
300
400
500
600
Frequency (Hz)
700
800
900
1000
Figure 4.14 – FFT plot of spindle vibration in the Z-Axis direction when rotating at 1000rpm
Figure 4.15 displays a zoomed in section of the frequency plot to show the clear domination of the
rotational frequency of 16.67 Hz.
Power spectral density (Log Y)
4
2
0
-2
-4
-6
-8
-10
-12
-14
-16
0
10
20
30
40
50
60
Frequency (Hz)
70
80
90
100
Figure 4.15 – Zoomed in FFT plot of spindle vibration in the Z-Axis direction when rotating at 1000rpm
69
4.3. Chapter Conclusions
The sensor selection is only part of the problem, the rest of the hardware and software is just as
important. The use of suitable test-bars and fixtures is essential to enable high accuracy
measurement to take place.
Once the hardware is capable of logging data, data processing is still required to convert the
captured data to a format that can be understood and presented in accordance with ISO standards.
This section described the processes required for processing data from thermal, geometric and
vibration analysis testing.
A full list of the required hardware and software to enable spindle analysis to be performed can be
found in table 6:
Table 6 – Spindle Analysis B.O.M
Hardware
Item
Non-Contact Displacement Sensors
Signal Amplifier
Data Acquisition Device
Power Supply
Sensor Fixture
20mm Test Bar (short)
20mm Test Bar (long)
Tool Holders
HSK 100
HSK 63
ISO 50
SK 40
SK 50
Capto
Temperature Sensors
Temperature Sensor USB Adapter
Temperature Sensor Extension Cables
Adapter Power Cable
Thermal Imaging Camera
Quantity
6
6
1
1
1
1
1
1
1
1
1
1
1
12
1
4
1
1
Software
WinTcal (temperature logger)
NI Signal Express
Compiled Matlab functions
1
1
1
70
5.0
Practical Spindle Analysis
This section looks at real case studies of spindle analysis taken in typical manufacturing
environments. It identifies potential drawbacks of the newly employed system and how they can be
overcome.
5.1. In situ Calibration
From practical experience, it is not always possible to use the same target material when performing
spindle measurements. In many cases different machines will have differing test bar material. In
certain situations it may be necessary to perform a measurement against a component mounted in a
spindle. With this in mind a simple in-situ method for calibrating the sensor to the necessary target
material is extremely beneficial
The linear scale of a machine tool is specified with linear errors on a typical modern machine in the
region of 5 µm/m, so when calibrating over 0.1 mm, any error can be considered negligible. There is
a possibility of stiction error when moving the machine in small increments but experience shows
that this is typically negligible but may be the case on some older machines and as such, calibration,
where possible, should take place on newer machine tools or where responsiveness has been
ascertained.
To ensure the stability and accuracy of the measurement the following is taken into consideration:


The accuracy and responsiveness of the axis over the region to be used must first be
established by standard methods.
Before any measurement is taken, it is important to move the axis of the machine over the
region in which the measurement is to be taken to lubrication of the guide ways to counter

the stiction.
All calibration measurement is taken when approaching the target from the same direction
so that no reversal error is introduced; an axis over-run [7] before the test ensures
unidirectional calibration.
71
5.2. Thermal Analysis
Thermal analysis of machine tool spindles is a good way of setting up a thermal profile of the
machine to allow for the errors, or for compensation to be applied during production.
40
Spindle
Stopped
Spindle Running
X Far
Y Far
X Near
Y Near
Z
35
Displacement (microns)
30
25
20
15
10
5
0
-5
0
60
120
180
240
300
360
Time (minutes)
Figure 5.1 – Spindle displacement during a 6 hour spindle heating test
Figure 5.1 shows the displacement of a milling spindle in relation to the table when running at a
constant 5000rpm for a 4 hour period. The largest displacement is in the Z direction where the
spindle displaces by 25 µm over the first 80 minutes, it then remains stable to within 4 µm for the
remaining run time.
In addition, once the spindle is stopped there is a rapid increase in displacement, up to 38 µm,
before retracting back. This is most likely due to residual heat from the test flowing into and
expanding the test-bar, therefore increasing the displacement.
While monitoring the displacement of the spindle with the non-contact displacement sensors, it is
also important to measure the temperature of the spindle at various locations. The ISO standard
suggests monitoring the surface temperature at the spindle nose as well as the ambient
temperature of the spindle and machine. This is considered to be the minimum data, the more
temperature sensors used the better the thermal profile that can be obtained.
72
25.5
25
Column Top
Spindle Nose
Machine Bed
Spindle Ambient
Column Middle
Machine Ambient
24.5
Temperature (C)
24
23.5
23
22.5
22
21.5
21
20.5
0
60
120
180
240
Time (minutes)
Figure 5.2 – Spindle temperature during a 6 hour spindle heating test
Figure 5.2 shows the temperature of the spindle and surrounding area during the 6 hour heating
test. When compared to the displacement data, there is a direct correlation between the
temperature at the spindle nose and displacement in the Z direction, as can be seen in figure 5.3.
The te pe atu e at the spi dle ose ui kl heats up to its
a i u
of a ou d
˚C. It took slightl
longer for the Z axis to achieve its maximum displacement but this could be due to the flow of heat
into the test bar.
30
26
25.5
25
20
24.5
24
15
23.5
10
23
5
22.5
Z
Spindle Nose
0
Temperature (C)
Displacement (microns)
25
22
21.5
-5
21
0
60
120
180
240
Time (minutes)
Figure 5.3 – Comparison of temperature and displacement during a 6 hour spindle heating test
73
5.2.1.
Thermal Imaging
The use of thermal imaging during testing can also provide comprehensive data about the flow of
heat through the structure of the machine tool by way of the high spatial resolution of temperature
information. One drawback of thermal imaging is that the accuracy is dependent on the knowledge
of the emissivity of the surfaces being measured. Also, a typical specification for a camera is in the
order of ±2ºC. Methods exist to improve this accuracy, including applying masking tape (with a
known emissivity of 0.95) to areas of thermal interest and averaging the images to reduce noise.
Accurate surface mountable temperature sensors in the field of view of the camera can also be used
to adjust emissivity to improve accuracy. Figure 5.4 shows thermal images taken at different periods
during a spindle heating test on a large vertical turning machine.
Figure 5.4 – Thermal imaging at various times of a heating and cooling test
Using the test methods described in ISO2
pa t
dete
i atio of the
al effe ts , o
i ed
with thermal imaging sequences, data can be obtained to establish a thermal profile of a machine
tool in it operational environment. This data can then be used to establish production capability,
corrective action or when creating thermal compensation files to compensate out any thermal error
present in the machine.
74
5.3. Geometric Analysis
Geometric analysis of a machine tool spindle involves measuring the error motion of the spindle (see
section 4.2.2).
The example in this section is measuring the radial error motion of a high precision air bearing
spindle. Using the Donaldson reversal technique described in the previous section, two sets of data
were captured.
Figure 5.5 shows 20 revolutions of a test-bar from setup 1(see section 4.2.2.1) of the reversal
measurement plotted together.
Channel '1' all revs
1
Displacement (microns)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
0
500
1000
1500
2000
2500
Number of Points
Figure 5.5 – All run plotted from run 1
The data includes test-bar run-out, form error, electrical run-out error and spindle error. The largest
error can be seen as the test-bar run-out, which is approximately 1µm. The rest of the errors require
separation using the reversal technique.
Figure 5.6 shows 20 revolutions of a test-bar from setup 2 of the reversal measurement plotted
together. The data includes test-bar run-out, form error, electrical run-out error and spindle error.
75
Channel '1' all revs
1.4
1.2
Displacement (microns)
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
0
500
1000
1500
2000
2500
Number of Points
Figure 5.6 – All runs plotted from run 2
These two sets of data are then input in the Donaldson reversal formulas (see page 60) to separate
the errors so that the remaining spindle error can be plotted on a polar plot and the asynchronous
and synchronous error motion values can be calculated.
Asynchronous error motion is the non-repeating change in position of the spindle on successive
rotations.
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
Figure 5.7 – All runs plotted in a polar plot
76
The asynchronous error value is the difference between the maximum and minimum values at a
particular angle in the above plot. A description of this can be seen in figure 2.3 on page 17. In this
case it is 0.13 µm.
Synchronous error is the once per revolution error that occurs as a result of the spindles deviation
from its true centre line. It characterises the spindle ability to cut a round hole. Figure 5.8 shows the
synchronous error plot, which is an average of the 20 revolutions.
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
Figure 5.8 – Averaged data run-out plot
The synchronous error value is the difference between the maximum and minimum values in the
above plot. In this case it is 0.53 µm.
77
5.4. Vibration Analysis
Vibration analysis can be linked to the asynchronous error motion values to identify the frequency at
which vibration may be occurring.
The vibration data in figure 5.9 is from an air bearing spindle being run at 250rpm. The first big peak
at 4.16Hz is due the rotational frequency of the spindle.
Power spectral density (Log Y)
0
-5
-10
-15
-20
0
50
100
150
Frequency (Hz)
200
250
300
Figure 5.9 – Vibration data from an air bearing spindle running at 250rpm
The 5.10 shows the same air bearing spindle being run at 500rpm. As can be seen, the first big peak
is at 8.3 Hz; again this is due to the rotational speed of the spindle. Some of the vibration peaks
occur at the same frequency as the previous plot, such as at 67 Hz and 200Hz. This suggests that this
is vibration due to the resonant frequencies of the spindle and surrounding structure.
Power spectral density (Log Y)
0
-5
-10
-15
-20
0
50
100
150
200
Frequency (Hz)
250
300
350
400
Figure 5.10 – Vibration data from an air bearing spindle running at 500rpm
Further examples of practical spindle analysis can be found in the appendix.
78
6.0
Conclusions
Research was conducted into relevant work in the area of spindle metrology. With it being a
relatively new concept there was a limited amount of directly related work available to assess.
However, there was a substantial amount of work in related fields such as structural vibration, this
coupled with the appropriate ISO standards aided with the research process.
The non-contact sensing technologies, suggested in the ISO standards as suitable for use in spindle
analysis, were investigated to find the most appropriate solution for an affordable system useable
on a regular basis by maintenance engineers in a typical workshop environment. This investigation
involved extensive research and testing to assess the sensors against their manufacture
specifications.
This resulted in eddy current and capacitance sensors being confirmed as appropriate for spindle
analysis testing, but with the laser triangulation sensors not meeting the required dynamic
resolution. In the end eddy current sensors were chosen as the most appropriate solution, for the
following reasons:


They can be used in manufacturing environments where swarf, dust and coolant may be
present.
They offered a much cheaper solution, which was important as the sensors make up a large
percentage of the overall system cost.
The overall outcome was the design and development of an affordable spindle analysis system that
can be taken into a manufacturing environment to perform comprehensive spindle analysis testing
and post processing of data. A critical contribution from this solution is the ability to measure very
small errors from, for example an air bearing or high quality rolling element bearing spindle to large
thermal errors using a single system.
79
7.0
Future Work
As described and demonstrated in this work, the designed system is capable of performing robust
spindle analysis. However, further development of both system hardware and software would be
beneficial to improve the efficiency of measurement and data processing.
7.1. Test-bar Design and Holding
Manufacturing a set of invar test-bars would improve the thermal properties and enable more
accurate thermal error measurement, without the influence of thermally expanding test-bars.
However, while this is an improvement for the thermal error measurement, further testing will be
required to assess the permeability of invar and whether or not this would have an effect on
accuracy during the geometric testing.
A set of tool holders would be a good addition to the system, to hold the test-bars when measuring
milling spindles. Figure 7.1 shows a commercially available tool holder that can be adjusted to
remove the run-out of the test-bar.
Figure 7.1 – Adjustable tool holder to 0.000µm run-out [39]
Being able to centre the test bar would reduce vibration when running at high speeds and allow for
more accurate measurement of high precision spindles.
80
When measuring turning spindles with face plate mounting, an interface holder can be developed
with the same idea of being able to adjust the test-bar to remove the run-out. This would also result
in a much quicker test setup than the current method, therefore improving testing efficiency and
minimising machine downtime.
7.2. Fixtures
The manufacture of one Invar fixture that could be used for all the testing would reduce setup time
between tests as the same fixture would be used and also provide thermal stability during thermal
testing.
Figure 7.2 – Ultrafine adjustment screws [40]
The fixture would need to be fully adjustable, allowing for very fine movements during precision test
setups. Figure 7.2 shows a set of ultrafine adjustment thumb screws, which allow for an adjustment
sensitivity of 0.5µm.
81
7.3. Data Processing Software
The current data processing method is run as a series of script files. This is easy to use but can
sometimes be quite time consuming. An ideal solution would be to compile an executable
application that can run all the scripts from one place easily.
The inclusion of a sensitivity displays would also be beneficial, so that the most sensitive part of the
individual sensors range can easily be identified and positioned.
7.4. Test Methodology
Further investigation in to the multi probe method would be beneficial to establish if the additional
cost of purchasing additional sensors required can be offset by the time saved from reduced setup
times of tests.
82
8.0
[1]
References
A ele, E. Alti tas, Y. B e he , C. Ma hi e Tool “pi dle U its , CI‘P A
als – Macnufacturing
Technology, 2010.
[2]
ISO 230-3, Test code for machine tools – Part 3: Determination of thermal effects. 2007.
[3]
ISO 230-7, Test code for machine tools – Part 7: Geometric accuracy of axes of rotation.2006.
[4]
ISO 230-8, Test code for machine tools – Part 8: Determination of vibration levels. 2009.
[5]
Lio P e isio , Capacitive sensor operation and optimisation , Te hNote LT03-0020, 2011.
[6]
Na a i, M‘ A o el i te fa e fo edd
u e t displa e e t se so s , IEEE t a sa tio s o
instrumentation and measurement, vol. 58, no.5, May 2009.
[7]
Lio P e isio , Diffe e e et ee
apa iti e a d edd
u e t se so s , Te hNote LT
-
0011, 2009.
[8]
Lio P e isio , Cali atio of edd
[9]
Yati g, Y. I estigatio o
p o le
of edd
u e t se so s , Edd
o t i utio of o du ti it a d pe
u e t displa e e t se so ,
[10] Te uhi o, K. High f ue
u e t Te hNote LT
pe
-0013, 2007.
ea ilit o ele t i al u out
.
ea ilit of fe o ag eti
etal o posite
ate ials , Jou al
of magnetism and magnetic materials 310, pp 2566 – 2568, 2007.
[11] Ma, X.
Edd
u e t
easu e e ts of ele t i al o du ti it a d
porus metals , NDT & E I te atio al
, pp
ag eti pe
ea ilit of
– 568, 2006.
[12] Bryan, J. International status of thermal error research. Annals of the CIRP, Vol. 39, Part 2,
645-656, 1990.
[13] Postleth aite, “. Ma hi e tool the
al e o edui tio – an appraisal, Journal of Engineering
Manufacture, 1999.
[14] Ramesh,R. Error compensationin machine tools – a review: Part II: thermal errors.
International Journal of Machine Tools and Manufacture, Vol 40, Issue 9, 1257-1284, 2000.
[15] Mia N“, Flet he “F, Lo gstaff AP, M e s A,
Effi ie t thermal error prediction in a
a hi e tool usi g fi ite ele e t a al sis , Measurement Science and Technology, 22(8): p.
085-107.
[16] Flet he “, Lo gstaff AP a d M e s A,
assessment and error- o pe satio
Measu e e t
ethods fo effi ie t the
al
Proceedings of the Topical Meeting: Thermal Effects in
Precision Systems – Maastricht.
[17] Lo gstaff AP, Flet he “, a d Fo d DG,
efe e e to I“O
P a ti al e pe ie e of the
al testi g
ith
Pa t , Laser Metrology and Machine Performance VI: p. 473-483.
83
[18] J
e, WY. The de elop e t of high speed spi dle
a d a uad a ts se so , I te atio al jou al of
easu e e t s ste
a hi e tools a d
usi g lase diode
a ufa tu e
,
:
p1162-1170.
[19] Cast o, HFF. A
ethod fo e aluati g spi dle otatio e o s of
a hi e tools usi g a lase
i te fe o ete , Measurement 41, 2008: p526 - 537.
[20] Marsh, E. Precision spindle metrology, second edition, DEStech Publications Inc, 2010.
[21] Lu, X. A e
di e sio
[22] Lu, X.
ethod fo characterising axis of rotation radial error motion: Part 1 two-
adial e o
A
e
otio theo
ethod fo
, P e isio E gi ee i g,
.
ha a te isi g a is of otatio
adial e o
otio : Pa t
E pe i e tal esults , P e isio E gi eering, 2011.
[23] Ma sh, E. A o pa iso of e e sal a d
ultip o e e o sepe atio , P e isio E gi ee i g
34, pp 85 – 91, 2010.
[24] Ma ti , D.L. P e isio spi dle ea i g e o a al sis , I te atio al jou al of
a hi e tools
& manufacture vol 35, pp187 – 193, 1995.
[25] Vafaei, “. Vi atio
o ito i g of high speed spi dles usi g spe t al a al sis te h i ues ,
[26] Felte , D. U de sta di g ea i g i atio f e ue ies
.
[27] ISO Standard 76-1987, Load ratings and fatigue life, 1987.
[28]
I“O
:
ea i g life sta dard – a d the a s e is? , t i olog
a d lu i atio
technology, 2010.
[29] Bell, “. A egi
e s guide to u e tai t of
easu e e t , Measu e e t good p a ti e
guide No. 11 Issue 2, 2001.
[30] Fla k, D. Ha
afo d, J. Fu da e tal good p a ti e i di e sio al
et olog , Measu e e t
good practice guide No.80, 2005.
[31] Wilso , J“. “e so te h olog ha d ook , Else ie ,
[32] “hieh, J. The sele tio of se so s , P ogess i
[33] Lai. Y,
Edd
u e t displa e e t se so
.
ate ials s ie e
, pp
– 504, 2001.
ith LTCC te h olog ,
Ph.D. disse tatio ,
Universität Freiburg, Switzerland, 2005.
[34] Tia . G. Y, )hao. ). X a d Bai es. ‘.W, The esea h of i ho oge eit i edd
u e t
se so s , Sens. Actuators A, Phys., vol. 69, no. 2, p148– 151, Aug. 1998.
[35] )ha g, H. A app oa h of eddy current sensor calibration in state estimation for maglev
s ste
, Ele t i al
a hi es a d s ste s, p
[36] ‘ao, BPC. P a ti al edd
-1958, 2007.
u e t testi g , O fo d: Alpha “ ie e,
[37] Ka g, Y. I teg ated CAE st ategies fo the desig of
.
a hi e tool spi dle ea i g s ste s ,
Finite Element in Analysis and Design 37, 2001.
[38] International journal of machine tools & manufacture vol 42, pp1223 – 1234, 2002.
84
[39] http://www.mtiinstruments.com/products/lasertriangulation.aspx
[40] http://www.gb.schunk.com/schunk/schunk_websites/news/press_release_detail.html?article
_id=15440&submenu=3700&submenu2=244&archive=2010&country=GBR&lngCode=EN&lng
Code2=EN
[41] http://www.standa.lt/products/catalog/fine_adjustment?item=179
85
9.0
Appendix
Page
Micro Epsilon Eddy Current Testing
86
Further Examples of Practical Spindle Analysis
88
86
9.1. Micro Epsilon Eddy Current Testing
Linearity Test
Figure 9.1 shows the output of the micro epsilon eddy current sensor, with the sensor output is
plotted in blue and a linear fit is plotted in red. The linear fit is very close to that of the sensor
output, showing that the sensor has a very good linear output over its full range.
5
4.5
data 1
linear
y = 0.0095*x + 0.01
4
3.5
3
2.5
2
1.5
1
0.5
0
0
50
100
150
200
250
300
350
400
450
500
Figure 9.1 – Linearity of Micro-Epsilon eddy current sensor
87
Static Resolution Test
The static resolution test was carried out using the same methodology as with the Kaman eddy
current sensors.
0.04
0.02
Displacement (microns)
0
-0.02
-0.04
-0.06
-0.08
-0.1
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
Number of Samples
Figure 9.2 – Stability of Micro-Epsilon eddy current sensor
Figure 9.2 shows the static displacement to be approximately 50nm which is what was suggested by
the manufacturer specification. This would be a good enough resolution to measure the majority of
machine tool spindles. When standard deviation is applied to this data it identifies a potential
resolution of 14.3nm.
This test was repeated at various positions in the sensor range with the following results:
Table 7 – Potential Resolutions at various positions in the sensor range
Distance from start of range (µm)
170
300
500
800
Potential resolution when standard
deviation is applied to the data (nm)
12.5
15.9
24
47.2
88
9.2. Further Examples of Practical Spindle Analysis
Thermal Step Heating Test
Figure 9.5 shows the displacement data from a thermal step heating test carried out on a ball
bearing spindle. The spindle was run at 500, 1000, 2000 and 3000 rpm for 1 hour periods and then
left to cool.
40
X Near
Y Near
X Far
Y Far
Z Near
Z Far
30
Displacement (microns)
20
10
0
-10
-20
-30
-40
0
60
120
180
240
300
360
Time (minutes)
Figure 9.3 – Thermal step heating test displacement measurement
Displacement is positive for increasing distance between the sensor head and the spindle nose.
The maximum spindle displacement during the 6-hour test was 34 µm in the Z-axis direction, 5 µm in
the X-axis direction and 34 µm in the Y-axis direction.
The above plot also shows the displacement of the spindle stabilises during each of the 1 hour run
times at the different speeds.
This sort of data provides good information to users about potential warm up cycles and thermal
compensation to be applied.
89
Figure 9.6 shows the temperature data from the same thermal step heating test carried out on a ball
bearing spindle.
35
Spindle Ambient
Test Rig Ambient
Spindle Middle
Spindle Front
Spindle Back
33
31
Temperature (C)
29
27
25
23
21
19
17
15
0
60
120
180
240
300
360
Time (minutes)
Figure 9.4 – Thermal step heating test temperature measurement
Figure 9.7 shows the thermal images for the same test which can be used for verification of the
temperature data.
Figure 9.5 – Thermal step heating test thermal imaging
90
Figure 9.8 shows a correlation between the spindle displacement and the spindle ambient
temperature.
40
1000 rpm
500 rpm
2000 rpm
3000 rpm
Spindle Stopped
34
X Near
Y Near
30
Z Near
32
Z Far
20
Spindle Front
10
28
0
26
-10
Temperature (C)
Displacement (microns)
30
24
-20
22
-30
20
-40
18
0
60
120
180
Time (minutes)
240
300
360
Figure 9.6 – Thermal step heating test temperature and displacement comparison
91
Error Motion Tests
A radial error motion test of carried out on a mechanical ball bearing spindle while running at
1000rpm.
0.5µm / division
Figure 9.7 – Radial run-out of ball bearing spindle at 1000rpm
Figure 9.9 shows a polar plot of the averaged data from 20 runs, as per the ISO standard, to show
the synchronous error. In this case it was 0.64µm, which is very good for a mechanical spindle and
shows the requirement for very resolution sensors to be used.
8
0.5
0.4
6
0.3
4
0.1
2
0
0
Tilt (µradians)
Displacement (microns)
0.2
-0.1
-0.2
-2
-0.3
Near
-4
Far
-0.4
Tilt
-6
-0.5
0
60
120
180
240
300
360
Angle (degrees)
Figure 9.8 – Tilt error of ball bearing spindle at 1000rpm
92
Figure 9.10 shows the tilt error measurement from the same spindle when running at 1000rpm. This
identifies a further element of required further work; to develop polar plots to display the tilt error is
a clearer way.
Vibration Analysis
Vibration data was also logged from the same ball bearing spindle when it was running at 1000rpm.
Power spectral density (Log Y)
-8
-10
Fundamental
Ball Spin
Ball Pass Inner
Ball Pass Outer
-12
-14
-16
-18
-20
-22
0
50
100
150
200
250
300
Frequency (Hz)
350
400
450
500
Figure 9.9 – Vibration of ball bearing spindle at 1000rpm
Figure 9.11 shows the FFT plot for the vibration data with the frequencies of interested identified,
that are related to the ball bearing sets used in the spindle.
93
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement