THE PENNSYLVANIA STATE UNIVERSITY SCHREYER HONORS COLLEGE

DEPARTMENT OF MECHANICAL AND NUCLEAR ENGINEERING

BENCHMARKING THREE BALL BEARING SPINDLES WITH SPINDLE METROLOGY

JAMES A. KING, III Spring 2007

A thesis submitted in partial fulfillment of the requirements for a baccalaureate degree in Mechanical Engineering with honors in Mechanical Engineering

Reviewed and approved* by the following: Eric R. Marsh Associate Professor of Mechanical Engineering Thesis Supervisor Mary I. Frecker Associate Professor of Mechanical Engineering Honors Adviser

* Signatures are on file in the Schreyer Honors College.

We approve the thesis of James A. King, III: Date of Signature

_____________________________________ Eric R. Marsh Associate Professor of Mechanical Engineering Thesis Supervisor

______________________

_____________________________________ Mary I. Frecker Associate Professor of Mechanical Engineering Honors Adviser

______________________

9-5668-1777

Abstract
The purpose of this thesis is first and foremost to establish the feasibility of applying advanced spindle metrology techniques to conventional ball bearing spindles with relatively inexpensive equipment. To that end, the spindle performance of a surface grinder, CNC lathe, and CNC mill will be benchmarked. The results of these measurements will be analyzed for validity and usefulness for various applications in manufacturing. Spindle metrology on conventional spindles is useful for many purposes, ranging from better utilization of machinery to preventative maintenance. By benchmarking a

machine, its capabilities and limitations can be determined to a degree that will provide for properly informed decisions on which machines are used for which tasks. Furthermore, once the error motion of a machine is established, steps taken to improve said error motion can be evaluated for effectiveness. These and many more applications, such as measuring the error motion of a spindle over time to predict its failure, bode well for a promising future for spindle metrology in the machine shop.

iii

Table of Contents
List of Figures ................................................................................................................ v Acknowledgments .........................................................................................................vi Chapter 1: Introduction .................................................................................................. 1 1.1 Introduction................................................................................................................................ 1 1.2 Axis of Rotation Metrology...................................................................................................... 2 1.2.1 Spatial Measurement Classification ................................................................................. 2 1.2.2 Frequency Classification of Error Motion ..................................................................... 4 1.2.3 Sensitive Direction Classification of Error Motion ...................................................... 6 1.3 Previous Work............................................................................................................................ 8 1.4 Contribution .............................................................................................................. 8 Chapter 2: Surface Grinder........................................................................................... 10 2.1 Setup of Surface Grinder Experiment ..................................................................................10 2.2 Results of Surface Grinder Experiment ...............................................................................12 Chapter 3: CNC Lathe ................................................................................................. 16 3.1 Setup of CNC Lathe Experiment..........................................................................................16 3.2 Results of CNC Lathe Experiment .......................................................................................18 Chapter 4: CNC Mill .................................................................................................... 20 4.1 Setup of CNC Mill Experiment .............................................................................................20 4.2 Results of CNC Mill Experiment ..........................................................................................22 Chapter 5: Conclusions ................................................................................................ 24 Appendix A: PhotoWorks............................................................................................. 25 A.1 Surfaces and Modeling ...........................................................................................................25 A.2: Optics ......................................................................................................................................28 A.3 Example ...................................................................................................................................30 References .................................................................................................................... 40

iv

List of Figures
Figure 1.1: Axial and face error motions of an axis of rotation after B89.3.4M. ........................ 3 Figure 1.2: Radial and tilt error motions of an axis of rotation after B89.3.4M. ........................ 3 Figure 1.3 Synchronous and asynchronous components of error motion .................................. 5 Figure 1.4 Spindle error motion classifications and components ................................................. 7 Figure 2.1: Test setup of Surface Grinder ......................................................................................10 Figure 2.2: Measured displacement including both spindle and structural error motion ........12 Figure 2.3: Thermal drift over time .................................................................................................14 Figure 3.1: Test setup of CNC Lathe ..............................................................................................16 Figure 3.2: Measured displacement including both spindle and structural error motion ........18 Figure 4.1: Test setup of CNC Mill .................................................................................................20 Figure 4.2: Measured displacement including both spindle and structural error motion ........22 Figure A.1: Reflection on a satin finished surface .........................................................................26 Figure A.2: SolidWorks model (a) without perspective and (b) with perspective ....................28 Figure A.3: Caustics ...........................................................................................................................30 Figure A.4: Default Settings ..............................................................................................................31 Figure A.5: Lights Added ..................................................................................................................32 Figure A.6: Shadows Removed ........................................................................................................33 Figure A.7: Satin Surface ...................................................................................................................34 Figure A.8: No Reflectivity, Less Glossiness .................................................................................35 Figure A.9: Close up of Figure A.6 ..................................................................................................36 Figure A.10: Increased Sampling Rate ............................................................................................36 Figure A.11: Advanced Options Enabled.......................................................................................37 Figure A.12: Artifact Added to Model ............................................................................................38 Figure A.13: Adjusted Artifact .........................................................................................................39

Acknowledgments
First and foremost, thanks to Professor Eric Marsh, Alex Moerlein, and Ted Deakyne, for their advise, expertise, support, and camaraderie during my time in the Machine Dynamics Research Lab. The amount of effort that Professor Marsh especially has invested in me and my thesis is much appreciated and will not soon be forgotten.

Thanks are also due to Dave Arneson, Mel Liebers and Steve Sanner of Professional Instruments Company, for providing equipment, knowledge, and even some of the data needed for this research. Don Martin, Mark Kretschmar and Ray Herbst provided critical data and support as well.

And thanks to Professor Mary Frecker, for her guidance in my four short years at Penn State. Her counsel proved valuable time and again throughout my collegiate career.

Finally, thanks to my family for your advice, support, and inspiration. Your tireless efforts have enabled me set lofty goals and realize my dreams, and I am forever grateful.

vi

Chapter 1: Introduction
This chapter introduces the terminology and concepts of axis of rotation metrology that will be used throughout this thesis. The conventions used to characterize and quantify error motion of spindles will be explained, and the methodology utilized to analyze error will be discussed. The chapter also includes an overview of previous work in the field of spindle measurement, as well as a description of the contributions of this thesis.

1.1 Introduction
The capabilities of machine tools have increased rapidly and consistently over the past century. This trend was a direct result of the demands placed upon the state of the art first by mass production of automobiles and other large machinery, followed by nuclear weaponry and facilities [1], and most recently by microelectronics [2]. As new technologies were developed and applied to the many challenges facing the manufacturing industry, equally sophisticated measurement techniques were needed to monitor the progress of the state of the art. Today, air bearings are the leader in rotational accuracy in a variety of precision machining applications, with error motions at the nanometer level. In order to properly measure and characterize the performance of these incredibly precise spindles, advances in the field of spindle metrology were needed. These advances came in the form of both hardware, such as capacitance probes, and analysis techniques, including Donaldson and Estler reversals [3].

2 The progress in spindle metrology used by the most precise air bearing spindles can also be applied to conventional ball bearing spindles. Using these advanced spindle

technology processes, the performance of conventional spindles can be benchmarked to a degree that was not possible before without the most sophisticated and expensive equipment available. Now that the aforementioned advances in spindle metrology are established, and they can be applied with relatively inexpensive and accessible equipment, it is time to use them to improve the performance of conventional ball bearing spindles.

1.2 Axis of Rotation Metrology

First, it is important to introduce the nomenclature and definitions of axis of rotation metrology, which is used throughout this thesis. A spindle is the physical embodiment of an axis of rotation it restricts movement in all but one rotational direction of freedom. In its most basic form, a spindle is comprised of two parts: the non-rotating part, or stator, and the rotating part, or rotor. If any motion exists in addition to the desired rotation, it is

considered error motion. This includes any linear motion normal to or along the axis of rotation, as well as any rotation about axes other than the axis of rotation. These error motions can be considered in many different ways, but there are usually three classifications that are used to define error motion spatial measurement, frequency, and sensitive direction.

1.2.1 Spatial Measurement Classification

The spatial measurement classification of error motion is defined by the ANSI/ASME B89.3.4 standard as four different components axial, face, radial, and tilt. These four components of spatial measurement are depicted in Figures 1.1 and 1.2.

Figure 1.1: Axial and face error motions of an axis of rotation after B89.3.4M. [4]

Figure 1.2: Radial and tilt error motions of an axis of rotation after B89.3.4M. [5] The figures define each type of spatial measurement, as well as where to place a sensor to measure that motion. For example, in Figure 1.1, axial error motion is depicted as the motion in line with the axis of revolution, which in this case happens to be vertical. Also, a capacitance probe is placed directly atop the cylindrical artifact in order to record only axial error motion as the spindle rotates.

4 The four components of spatial measurement are very much interrelated, and not all need to be measured for analysis. For instance, face error motion is a function of axial error motion, tilt error motion, and the position of the sensor used to detect it. Because face error motion can be extrapolated for any position mounted to the spindle, it is often not measured directly, and instead simply derived from the other measurements. Similarly, radial error motion is needed from multiple positions to compute tilt error motion. However, in applications that do not require tilt error motion, a single radial error motion is measured and recorded.

1.2.2 Frequency Classification of Error Motion

The frequency classification of error motion can be considered for any or all of the types of spatial measurement defined in the previous section. Because the error motion of a spindle is a function of its orientation rather than its speed, measurements of spindles are often normalized by rotational velocity, in order to better depict the behavior of the spindle. These normalized measurements can be displayed on polar plots, as shown in Figure 1.3. This type of plot provides a very clear depiction of the actual behavior of the spindle being tested. The measurements can be analyzed to determine which error motions are occurring repeatedly with every revolution of the spindle, and which are uncorrelated to rotation. The time-dependent error is referred to as synchronous error motion, and appears as a result of flaws in the spindle. The remaining error is asynchronous error. The following figure illustrates the difference between synchronous and asynchronous error as it is separated from the raw probe data.

Figure 1.3 Synchronous and asynchronous components of error motion

Separating synchronous error motion from asynchronous error motion is surprisingly simple. First, the error is tabulated by undulations per revolution (upr) using a fast Fourier transform. Synchronous error motions occur repeatedly at integer multiples of 1 upr make up the synchronous error. Asynchronous error will not repeat with every revolution of the spindle. Therefore, asynchronous error can be identified as that which occurs at frequencies that are not integer multiples of 1 upr. As Figure 1.3 depicts, the synchronous error has one point per measured increment. The asynchronous error, by contrast, is made up of a cloud of motions that fill out the raw probe data. It is clear that identifying and separating synchronous and asynchronous data for each spatial measurement collected is very important in order to better understand the error motion of the spindle in question.

1.2.3 Sensitive Direction Classification of Error Motion

In addition to the spatial and frequency classifications of error motion, sensitive direction is an important aspect of spindle metrology. Depending on the application of the spindle, the sensitive direction can be either fixed or rotating. Fixed sensitive direction cutting consists of machining when the point of contact between the tool and the workpiece is stationary. Examples of fixed sensitive direction cutting include outer diameter turning and grinding. Even though both the tool and the grinding wheel are usually moving during outer diameter grinding, the point of contact between the two remains at a constant position. In both of these examples, the single fixed sensitive direction normal to where the tool contacts the workpiece is the only sensitive direction. Vertical or axial movement during outer diameter grinding has at worst a minimal effect on the workpiece, while horizontal movement towards or away from the tool will have a substantial effect on the part, hence the classification as a single, fixed sensitive direction. Operations classified as rotating sensitive direction have a rotating orientation of the tool at the point of contact with the workpiece. Conventional boring serves as a prime example of a rotating sensitive direction operation. In this case, the only sensitive direction is in line with the tool at its current orientation. Movement orthogonal to this direction will have a minimal effect on the workpiece, if any at all.

7 Figure 1.4 Spindle error motion classifications and components Figure 1.4 is a schematic that serves as a recap of the different classifications of error motion discussed in this section. Spatial measurement, consisting of axial, face, radial, and tilt error motions, is the error motion of the spindle as measured by the sensors. The frequency characteristic of error motion, consisting of synchronous and asynchronous error motions, is the error motion in terms of the fundamental artifact rotation frequency. And the sensitive direction characteristic of error motion, either fixed or rotating, determines how the error motion of a spindle will manifest itself onto a part. It is important to remember that movement of the spindle induced by changes in either the thermal or loading conditions of the spindle are not considered error motion. Thermal drift will occur as a machine heats during normal operation, as well as during the course of the day in work environments without climate control. And until the day that spindles can be built with infinite stiffness, they will continue to move as a result of loads applied by drive systems, payloads, and manufacturing processes. While both important aspects of spindle design and application, these movements are not a result of inherent flaws in the spindle, and therefore are not error motions. Also, vibrations outside of the spindle contribute additional error motion. Usually referred to as structural error motion, this error motion will contaminate the measurements of spindle and needs to be accounted for. Minimizing the structural loop between the spindle and the workpiece/measuring device is the best way to reduce structural error motion. Of course, structural error motion should be kept at a minimum in any machining operation, but strict attention must be paid to this issue more than ever when collecting measurements for spindle metrology.

1.3 Previous Work

Much work has been done in the testing of machine tools and their components over the last century. The major contribution to this testing began in the early 1900s, when the total indicated reading (TIR) of a spindle was established as the basis on which spindles would be compared for quality [6]. To this day, TIR is still sometimes used to measure spindles. However, it is not the most accurate of methods because spindles are often ground after assembly, a spindle with substantial error motion can be measured to have zero TIR. By the late 1980s, with the help of new capacitance sensors and recent advances in electronics, spindle analyzers were built that were capable of measuring the error motion of spindles to within 5 nanometers [7,8]. During the following decade, analyzers were

commercialized and made available to measure the error motion of spindles. Lion Precision of St. Paul, Minnesota was one of the first to market spindle analyzer software complete with capacitance gauging, probe nests, and dual sphere artifacts [9]. By the end of the 1990s, further advances in the techniques applied to analyze the data collected by the sensors were made, which allowed for measurement of the error motion of high performance spindles to the nanometer level. These advances allowed for techniques such as Donaldson and Estler reversals to be carried out in each degree of freedom with only a single fixed probe [10].

1.4 Contribution
The purpose of this thesis is to benchmark the performance of three common types of machines that use conventional ball bearing spindles. This analysis will be done using methods developed to observe and quantify the error of precise air bearing spindles. The

9 three machines examined a surface grinder, a CNC lathe, and a CNC mill each have distinct applications with their own inherent operating conditions. The applications of spindle metrology are many and varied. Analyses of machines such as those considered in this thesis can be used to better understand the characteristics of a machine, thereby providing a better understanding of both the capabilities and the limitations of that machine. Spindle metrology can also be applied to monitor error both in the spindle and in the structural loop of the machine, which can in turn be used to detect problems in either and lead to the proper course of action to solve the problem. Interestingly, the error motion of ball bearing spindles has been found to steadily increase with regular wear until the spindle reaches the point of catastrophic failure [11]. This effect can be recorded and monitored over time with spindle metrology techniques, which can in turn be used as a preventative maintenance measure by detecting and replacing worn spindles before they fail. These and many more applications can be performed using

measurements collected with the methods demonstrated in this thesis.

10

Chapter 2: Surface Grinder

2.1 Setup of Surface Grinder Experiment

The first machine to be examined is a manually operated 6 18 surface grinder. The surface grinder has a single speed, ball bearing spindle. Because the grinding wheel is held stationary while the planar workpiece is passed back and forth beneath it, the point of contact is stationary, which makes this a fixed sensitive direction operation. As such, the only error motion of the spindle that is relevant is the radial (in this case, vertical) displacement. Any error in the spindle will still be apparent through the single reading, while error motion in the axial direction is left unchecked.

Figure 2.1: Test setup of Surface Grinder In order to accurately measure the radial error motion of the spindle, a capacitance probe is secured directly beneath a spherical artifact, which has been mounted to the spindle.

11 Mounting the artifact to the spindle, however, is easier said than done. Because the surface grinder has only a standard wheel mount for a grinding wheel, most spherical artifacts cannot be attached directly to the grinder. One possible solution is to prepare and mount a special cylindrical artifact to the grinder in place of the lapped spherical artifact. However, this cylindrical artifact may become a source of error separate from the spindle, unless special metrology capabilities are available to compensate for these imperfections. The lapping process used to create conventional spherical artifacts is capable of producing a measurable surface with considerably less form error then a ground cylinder, thereby enhancing the reliability in precision spindle measurement. For these reasons, the solution is to mount a simple cylindrical plate to the grinding wheel, along with a tubular extension, which will provide a solid mount for the artifact. It is also worth noting that the majority of the fixturing used in this experiment consists of common shop hardware. Only the mounting plate for the artifact and the probe holder for the capacitance probe are specially made parts designed only for spindle metrology. In this example, the probe holder is held in position by a 246 block, and fastened with regular cap screws. The magnetic chuck of the surface grinder is particularly useful, as it allows for the stable mounting of a myriad of fixturing hardware. Of course, fixturing will often be unique to each application, but most will not require custom hardware. While the capacitance probe may be mounted using nothing more than common hardware, suitable attention has still been paid to the stiffness of the fixturing. As discussed in the previous chapter, minimizing error due to the structural loop of the measurement system can significantly improve the quality of the measurements taken and subsequent analysis. This issue is clearly addressed in Figure 2.1, where the probe holder is mounted with minimal overhang beyond the edge of the 246 block. It is important that these and

12 other steps be taken to reduce the chance that variations in the fixturing interfere with the measurements of the spindle. These steps include selecting fixturing that is relatively

insensitive to thermal gradients, as well as lightweight in order to lower the chance of resonant frequencies in the fixtures interfering with the measurements.

2.2 Results of Surface Grinder Experiment

The results of the analysis of the surface grinder are interesting.

(a)

(b)

Figure 2.2: Measured displacement including both spindle and structural error motion Figure 2.2(a) is an exaggerated polar plot of the recorded positions of the spindle. The black line represents the synchronous error motion of the spindle, which averages a remarkably low 0.3 m deviation from perfectly round. The asynchronous error motion, illustrated in blue, appears in a cloud around the synchronous error. Although the worst of the asynchronous error motion is nearly ten times larger than the synchronous error, the standard deviation of the asynchronous error motion is under 0.5 m, which results in

13 approximately 95% of the asynchronous error motion remaining within a range of 1.8 m. The Peak-to-Valley error (P-V), which is the total range of error motions in the data, is 2.5 m. This defines an absolute worst case scenario of the error motion of the surface grinder. However, it also makes clear that even the worst of the outliers in the measurements are within 0.7 m of the normal distribution. The bell curve positioned over the polar plot of the error motion in Figure 2.2(a) provides a graphical representation of the distribution of the error motion of the spindle. It is clear that the previously mentioned range of 1.8 m is a valid quantification of the distribution of the error motion. This plot also serves as a visualization of the distribution of the error motion of the spindle better than any list of quantities ever could. To the right of the polar plot of the error motion of the spindle is Figure 2.2(b), which is a graph of the amplitude and frequencies of the error motion detected in the spindle. Using a Fast Fourier Transform, the graph was developed by normalizing the frequencies of motions in the data by the speed at which the surface grinder was running when the data was recorded. The motion at the first-order or fundamental frequency was removed from this graph so as to not distract from the other trends in the data. Interestingly, the largest recurring error in the data is found at about 45% of the fundamental frequency. One possible source of this error could be the movement of a ball inside one of the ball bearings of the spindle. Because it occurs regularly, but at a speed only a fraction of that of the artifact, the ball bearing explanation is viable because the ball will travel around the inner race of the bearing at a speed slower than but proportional to the actual speed of the spindle. Another interesting trend in the data depicted in Figure 2.2(b) is the recurring spikes at each multiple of the fundamental frequency. This synchronous error could be the result

14 of flaws in the spindle, or possibly an imbalance in the fixturing mounting the artifact to the spindle or perhaps even a flaw in the artifact itself. The remaining smaller peaks in the chart are most likely error inherent in precise data collection, although the buildup of error around the fourth-order frequency does look particularly suspect. While this thesis will provide no speculation into the source of that error, it may be an interesting topic for another day.

Figure 2.3: Thermal drift over time In addition to the measurements taken to quantify the error motion of the surface grinder, data demonstrating the effects of thermal drift was collected during the experiment. This data is displayed in Figure 2.3, and it provides for an interesting glimpse into another aspect of spindle metrology. The wheel-work gap used as the vertical axis of the upper graph in Figure 2.3 is normally the space between the grinding wheel and the surface of the workpiece being ground. In this experiment, however, a spherical artifact is mounted to the spindle and a capacitance probe is in the place of the workpiece, so the wheel-work gap is actually the raw data taken from the capacitance probe over twenty hours of continuous operation. The most conspicuous trend in the data is the drop in the wheel-work gap over the first six hours of operation. This suggests that when operating around 21C, the surface

15 grinder will reach its steady-state temperature in about six hours. This is quite a long time, in fact the majority of most working days, so it the effects of thermal drift can often plague an operation the majority of every workday. Another interesting trend in the data is the slight increase in wheel-work gap in the first hour of operation. Since the surface grinder is a machine comprised of many

components that are each subjected to the heat generated by the friction of operation differently, it is likely that some components increase the wheel-work gap rather than decrease it like the overall response of the machine. It is also possible that some of these components that increase the wheel-work gap react to the temperature changes quicker than the rest of the machine, either because of material properties or proximity to the strongest heat sources. The case may be that the reaction of some components counter to the overall reaction of the machine may reach their maximum effect before the rest of the machine, thereby heavily influencing the early response of the entire machine. However, proof of this conjecture would require a good deal of meticulous observation of the machine in varying conditions. A final interesting trend in the data depicted in Figure 2.3 is the direct effect of changes in ambient temperature on the wheel-work gap. Despite the hypothesis for the initial increase in the wheel-work gap offered earlier, there are clearly at least some parts of the machine that react quickly and noticeably in a uniform overall direction when subjected to a change in temperature. This effect is obvious when the work environment was heated at approximately 15 and 18 hours into the experiment, but it can also be detected by the careful observer at 4 and 9.5 hours, among other places in the data. While the changes in both the temperature and the wheel-work gap at 4 and 9.5 hours may not be as pronounced as those at 15 and 18 hours, the relationship is apparent and too clear to be denied.

16

Chapter 3: CNC Lathe

3.1 Setup of CNC Lathe Experiment

The second spindle is part of a CNC Lathe with a 10 20 inch workpiece capacity that operates at up to 4000 RPM. Like the surface grinder, this machine is used in fixed sensitive direction operations, so vertical motion in the spindle will have only a minimal effect on the workpiece. Unlike the grinder, however, any tilt error motion in the spindle is unacceptable, as it will detract from the lathes accuracy. Therefore, the lathe measurements require no less than three probes, in order to get the complete depiction of the motion of the spindle.

Figure 3.1: Test setup of CNC Lathe Fortunately, stably mounting the multiple probes required is not difficult with a probe nest from the Professional Instruments Company. Just as in the previous experiment,

17 minimizing the structural loop is crucial in order to produce accurate measurement data. The probe nest can be mounted into the standard tool turret already present in most CNC lathes without much difficulty, and the artifact will fit in the jaws of any lathe that is large enough with no special modifications. By mounting the measurement equipment in this way, the structural loop is minimized to that which is present in regular turning operations.

18

3.2 Results of CNC Lathe Experiment

The test provided a clear picture of the error motion of this spindle. Figure 3.2 lays out the performance of the spindle, mapping its axial motion, as well as its radial motion at two points.

Figure 3.2: Measured displacement including both spindle and structural error motion

19 The plots in Figure 3.2 clearly reveal that the asynchronous axial error motion of the spindle is very, very low. In fact, at a mere fraction of a micron, the axial error motion is so concentrated that it is hardly a concern when considered along with the radial motion of the spindle. The asynchronous radial error motion of the spindle, while not quite as minute as the axial motion, still remains below 5 m throughout the rotation of the spindle. Perhaps a more interesting behavior revealed by the asynchronous error motion is the magnification of vibrations in the machine at points measured further from the spindle. Because the artifact is not particularly stiff in comparison to most of the other components in the structural loop of the lathe, it is most vulnerable to error due to vibration. The increased asynchronous error in the measurement taken further from the spindle indicates that the amplitude of vibrations passed into the artifact from the spindle is increasing. This serves as yet another reminder of the importance of considering the structural loop in spindle metrology, as even the artifact can be a source of error. Below the polar plots of the total error motion distribution in Figure 3.2 are polar plots of the first two revolutions of the spindle that were captured in the experiment. These plots serve as a terrific demonstration of how the cloud of asynchronous error motion is created. It is clear that in the first two revolutions, the error motion of the spindle oscillates around a general circular path. This general circular path is of course the synchronous error motion of the spindle, and the oscillations are the asynchronous variations that have been discussed earlier in this thesis.

20

Chapter 4: CNC Mill

4.1 Setup of CNC Mill Experiment

The third and final spindle is from a CNC mill with a 30 16 20 inch work envelope. The spindle is run at up to 7500 RPM while passing 20 hp from the motor to the cutting tool. Because any and all error motion of a spindle in a mill will have a detrimental effect on the final product, it is important that the test setup is capable of monitoring all types of error motion. Five probes are needed to capture all of the relevant motions of the machine, as radial and tilt error motion can vary at different points in the rotation of the spindle.

Figure 4.1: Test setup of CNC Mill In order to accurately capture all five measurements required to properly map radial, axial, face, and tilt error motion, it is best to use a probe nest. As depicted in Figure 4.1, the nest can be mounted directly to the table, with the spindle head lowered sufficiently to allow

21 the artifact to fit into the nest. The artifact can be mounted easily into the spindle with a specialized collet. Once again, minimizing the structural loop is of the utmost importance. Just like in the previous example of the CNC lathe, mounting the measurement equipment as demonstrated in this thesis produces results that are indicative of the actual performance of the entire structural loop of the machine, focusing particularly on the error motion of the spindle.

22

4.2 Results of CNC Mill Experiment

With five probes collecting data for the CNC mill experiment, there is plenty of information to consider. The results of the experiment are organized in Figure 4.2 below.

Figure 4.2: Measured displacement including both spindle and structural error motion

23 As expected, the spindle of the CNC mill exhibits at least some of the same characteristics as the spindle of the CNC lathe. The asynchronous axial error motion, for instance, is generally more concentrated and predictable than the radial error motions. This is demonstrated best by the polar plots of the first two revolutions of the spindles in Figure 4.2, which clearly show the first two revolutions following almost identical paths on a nearly perfect circle. Also, the radial probes mounted farther from the spindle recorded more asynchronous error motion than those mounted closer to the spindle. Once again, this is a result of vibrations originating in the spindle and/or machine being amplified by the distance from the spindle. Unlike the CNC lathe, the spindle of the CNC mill exhibits a tilt error motion that amplifies the synchronous error motion of the artifact at points further from the spindle. Also, the distribution of the asynchronous error motion falls in a nearly perfect bell curve, as expected. This distribution is even more impressive when considering its small range. Both the synchronous and asynchronous error motions of the CNC mill are smaller than those of the surface grinder and CNC lathe presented in this thesis. While the synchronous error motion of the spindle remained constant regardless of the orientation of the probes about the axis of rotation, the asynchronous error did not. This may be a result of vibration in the probe nest that was used to mount the capacitance probes in the experiment.

24

Chapter 5: Conclusions

The experiments described in this thesis were successful in demonstrating the feasibility of applying spindle metrology to conventional ball bearing spindles. The performance and characteristics of a surface grinder, CNC lathe, and CNC mill each with a conventional ball bearing spindle were each benchmarked. Some results were expected and easily explained, while others raised further questions about the behavior of spindles. From tilt error motion to thermal drift, many different aspects of spindle metrology were successfully employed in characterizing the three machines. Spindle metrology has a promising future in the machining industry thanks in part to its many possible applications, as discussed at length in this thesis. Management of

machinery will be performed more efficiently thanks to reliable, quantitative information on each spindle in the shop. Spindles will be run at full capacity without fear of failure thanks to new methods of preventative maintenance made possible by spindle metrology. And the performance of machines will be fine-tuned to a level never before possible thanks to precise input data using the same metrology methods applied in this thesis.

25

Appendix A: PhotoWorks
This appendix is intended to serve as an introduction to PhotoWorks, an add-in produced by Mental Ray for SolidWorks. PhotoWorks is a very useful add-in that provides designers with the power to produce lifelike images of their SolidWorks models without spending the time and resources required to take quality photographs of fabricated parts.

A.1 Surfaces and Modeling

No matter how powerful the rendering technique being applied, in order to produce a quality image, a model will need to have detailed descriptions of every surface of every object in the model. PhotoWorks uses over a dozen different illumination properties to define how each surface will react to different lighting scenarios. These properties are described at length, with sample images, in the PhotoWorks help files, but there are a few that bear mentioning in this introduction. Ambient, diffuse, glossy, and specular all control the amount of light that reflects off of the surface in question. Sampling Rate and Roughness are the main contributors to the quality of surfaces with satin finishes. Many other options are available, such as mapping 2-D images over surfaces or using transparent materials, but they will not be covered here.

26

Figure A.1: Reflection on a satin finished surface Satin finishes are of particular interest when modeling metallic surfaces, especially machined surfaces that reflect images just enough to be noticed with the eye. A satin finish takes into account how far away the source of a reflection is from the reflecting surface, and blurs and/or fades the reflection accordingly. For instance, if a dark rod is held

perpendicular to a sheet of aluminum, the reflection of the rod will be clearly defined near the point of contact between the two objects, as illustrated in Figure A.1 above. However, the reflection will be continuously less distinct at distances further from the point of contact, until eventually it cannot be seen at all. In PhotoWorks, satin finishes can be applied to any surface, and they can be tweaked using the sampling rate, roughness, and reflectivity properties of the surface to produce the exact reflection desired. While the number of properties may seem overwhelming at first, after enough tweaking and hands-on experience, anyone can become proficient at selecting realistic surfaces for their models. PhotoWorks also provides some assistance in the form of a library of preset combinations for many different materials commonly used in engineering applications. Using these presets as a starting point, any material can be represented in PhotoWorks with minimal difficulty.

27 In addition to all of the complex properties of the surfaces in the model, a color must be selected for each object in the model in order to achieve a truly lifelike image. Some colors are included with the preset materials in the PhotoWorks library, but oftentimes it takes just a slight adjustment of this color to produce a stunningly realistic result. A few final adjustments to the model before rendering an image are actually some of the SolidWorks settings for the part file. The first adjustment is necessary because of a default setting in SolidWorks intended to conserve computing power. SolidWorks will often be set by default to approximate curves as rather large-sided polygons, which can be useful, especially on particularly large parts or assemblies. For the modeling process, this is usually not a problem, and often a welcome aid in reducing the lag experienced as the computer renders different angles of a model. However, for an image that is intended to look as lifelike as possible, this is clearly not an option. Fortunately, it is not difficult to switch this option to display curves as curves, as long as the user knows where to look. In SolidWorks 2006, the option is located in the Document Properties, which can be found by navigating to Options under the Tools menu. The Document Properties tab is located at the top of the Options window, and the option to improve the quality of the image is aptly titled Image Quality. All that remains is to move the HLR/HLV Resolution slider to the right, towards the High side. The example circle beside the slider will begin to resemble a circle again and not a polygon, as well as all of the curves in the model.

28

(a)

(b)

Figure A.2: SolidWorks model (a) without perspective and (b) with perspective A second adjustment to consider is perspective. By default, SolidWorks displays features of the same length equally, regardless of distance from the viewer. However, there is an option to take perspective into account, which will skew the model in order to make it appear more realistic. In SolidWorks 2006, this option, simply called Perspective, can be found by navigating to the View menu, then to Display submenu. Enabling the

perspective option will have an immediate effect on the appearance of the model, which is demonstrated above in Figure A.2. The effect of the option can be tweaked in the Modify submenu, listed directly below the Display submenu. By changing the observer position, the effect of the perspective option can be fine-tuned until the user is content. Note that not all models will benefit greatly from perspective option, while for others, it will be essential in order to achieve a lifelike image.

A.2: Optics
Some of the more advanced options in PhotoWorks make use of ray tracing to model the interactions between optical surfaces. As the name implies, rays of light are traced by the

29 model as they originate from light sources and interact with the optical surfaces in the model. These interactions can be as simple as the surfaces in the model being lit by the light source or as complicated as reflections of refracted images. In the most basic cases, a ray will be traced from a light source to a surface, where it reflects back at the viewer, resulting in the image of the model. In more complicated applications, a ray might pass through a glass of water, refracting through both the glass and the water, then reach a surface, then reflect to the viewer. While these models take longer to render, they result in more

impressive and realistic images. In PhotoWorks, the advanced illumination options allow the user to use the ray tracing capabilities to create effects with global illumination, indirect illumination, and caustics. Global illumination, as the name suggests, simply creates light everywhere throughout the model, without a specific source. This allows for complex parts with small enclosed features to be visible regardless of whether or not a directional light source can reach down inside the part. It also prevents outer surfaces from being washed out in too much light while not enough light is reaching into deep enclosures. Indirect illumination uses ray tracing to transfer light from a reflective surface to light up another surface. For instance, if a model contained a spot light shined at a mirror, when indirect illumination is enabled, the mirror will reflect that spot light onto another surface, assuming the model is lined up so that the reflection hits another surface. Finally, caustics are the concentrations of refracted light that result from a light shining through or reflection off of a curved surface. The familiar example of a light shining through a glass of water is pictured below. Note the particularly bright spot where the concentration of light after the glass is the greatest.

30

Figure A.3: Caustics [12] In PhotoWorks, when the caustics option is enabled, these concentrations will be found using ray tracing and displayed accordingly, allowing for some very impressive effects.

A.3 Example
Now that the concepts behind the optics and surface properties have been introduced, here is an example of the process of applying these concepts to a SolidWorks model in order to create a realistic image. The model used in this example consists of a basic T-slot table and the case of an air-bearing spindle made by Professional Instruments. The first image was generated using basic SolidWorks presets for steel for both the table and the machined surface of the spindle case, and a plastic for the painted surface of the case. Also, the default SolidWorks lighting scheme was used in the first image.

31

Figure A.4: Default Settings Clearly, these settings leave the model much to dark to bring out any of its detail. So, for the second iteration, five directional lights were added in order to make the model more visible.

32

Figure A.5: Lights Added While the lights provide a better view of the exterior of the model, a new problem arises. The shadows on the model are being developed by a feature of PhotoWorks that automatically applies shadows as a darker tint to any area that does not have a direct line of sight between the light source and said area. While this method is much quicker for the program to compute and useful for quick renderings, there are better ways to create shadows more naturally that will make the image appear more realistic, rather than simply blanketing most of the model. The next image has the shadows removed, in order to better illustrate the reflections that are already beginning to appear in the image.

33

Figure A.6: Shadows Removed This image is a good example of the types of reflections that can be generated in PhotoWorks renderings. The rear of the case has a circular opening that is not quite as large as the prominently displayed opening in the front. Notice that the reflection of this opening in the rear is skewed by the curvature of the inner surface of the case. While this rendering clearly displays the features of the model and has at least the beginnings of some neat visual features, there are ways to make the image much more realistic. One method is to apply satin finishes to the metallic surfaces of the unpainted portions of the model. As described earlier, satin surfaces take into account how far the light reflecting off of a surface travels before reaching that surface, and blurs or fades the reflection accordingly. This effect can be seen on the metallic surfaces of the case in the following image.

34

Figure A.7: Satin Surface When compared to the previous image, this iteration of the rendering is clearly approaching a much more realistic finish. The reflection is no longer mirror-like, but somewhat blurred like true reflections in metallic surfaces. Also, the stripe of light seen inside the case is the reflection of one of the five lights that were applied earlier. However, the reflection is still much clearer than most machined surfaces, and the aforementioned stripe of light may not be desired if the intention was for the material to appear dull. In the following image, the unpainted material of the case has had the reflectivity property reduced to its minimum, in order to rid the image of any sharp reflections. Also, the glossiness of the material has been reduced in order to lessen the amount of light being reflected from the machined surfaces of the case.

35

Figure A.8: No Reflectivity, Less Glossiness The surface within the case now has the desired dull look of machined steel. However, upon closer inspection, the image appears rather coarse, as seen in the following close-up.

36

Figure A.9: Close up of Figure A.6 The lack of detail in this image can be fixed by increasing the sampling rate in the properties of the unpainted material of the case. The next image shows the same model with only the sampling rate increased.

Figure A.10: Increased Sampling Rate Now that the material for the surface in the case has been determined, the more advanced options can be used to create a very realistic image. It is worth noting that while most of the images thus far have taken a few minutes to render, once the advanced options are enabled, the render times will reach upwards of three hours, and this is a simple model.

37

Figure A.11: Advanced Options Enabled Now that the global illumination, indirect illumination, and caustics have been enabled, shadows begin to appear again in the less exposed areas of the case. However, these shadows are more natural than those in the initial iterations of the image. Rather than showing up as a dark swath applied uniformly to large areas, these shadows appear where fewer rays of light reflect from the surface to the viewer. Other changes in this image include the use of a satin finish on the T-slot table, which produces a nice reflection of the case. Also, the lights applied earlier no longer have an effect when global illumination is enabled, so they have been removed. In order to further illustrate the reflections of light and images, a very shiny artifact is added to the model.

38

Figure A.12: Artifact Added to Model The addition of the artifact has a profound effect on the entire model and how it is rendered. Its reflectivity is set so high that the amount of light coming back towards the viewer washes out the image. Even the case beside the artifact appears brighter thanks to all the light being reflected its way. With a few adjustments, however, the artifact will fit right into the rest of the model and make for a much more realistic image.

39

Figure A.13: Adjusted Artifact By reducing the reflectivity along with the ambient, diffuse, and specular properties of both the sphere and the base of the artifact, a much more lifelike image is produced. Also, the base of the artifact was switched to a satin finish, and the colors of both portions of the artifact were adjusted. Now, the sphere displays a magnificent reflection of both the T-slot table and the spindle case, which create the effect of a highly polished steel set atop a machined base. The process is now complete, and the result is an impressively realistic image.

40

References

[1] Taniguchi N. Current Status in, and Future Trends of, Ultraprecision Machining and Ultrafine Materials Processing. CIRP, 32 (2), 1983; 573-582. [2] McKeown PA. The Role of Precision Engineering in Manufacturing of The Future. CIRP, 36 (2), 1987; 495-501. [3] Grejda RD. Investigation of a Master Axis of Rotation for Use in Spindle Metrology. Penn State University MS Thesis, 1999. [4] ANSI/ASME B89.3.4M, Axes of Rotation; Methods for Specifying and Testing. American Society of Mechanical Engineers, 1985 (1992). [5] ANSI/ASME B89.3.4M, Axes of Rotation; Methods for Specifying and Testing. American Society of Mechanical Engineers, 1985 (1992). [6] Schlesinger G. Testing Machine Tools: for the use of Machine Tool Makers, Users, Inspectors, and Plant Engineers. 8th Edition. NewYork: Pergamon Press, 1978. [7] DeBra DB. Spindle Metrology A Student Project. Presentation at SME 7th International Precision Machining and Gaging Symposium, 1986. [8] Chapman PD. A Capacitance Based Ultra-precision Spindle Analyser. Engineering, 7 (3), 1985; 129-137. Precision

[9] Lion Precision. When Precision Matters. Technical and Sales Literature. Saint Paul, Minnesota. 2000. [10] Grejda RD. Use and Calibration of Ultraprecision Axes of Rotation with Nanometer Level Metrology. Penn State University PhD Thesis, 2002. [11] Levantsevich MA, Skorynin YV. Ways of Incresing the Life of High-Speed Electric Spindles. Soviet Engineering Research, 6, 1986; 62-64. [12] Image:Kaustik.jpg, [online document], 2006 Jan 22, [cited 2007 Mar 27], Available HTTP: http://en.wikipedia.org/wiki/Image:Kaustik.jpg