You are on page 1of 85

The Pennsylvania State University The Graduate School

USE OF VARYING WORK-PIECE DISTANCE TO STUDY THE EFFECTS ON MACHINE STIFFNESS IN PLUNGE GRINDING

A Thesis in Mechanical Engineering by Theodore R.S. Deakyne

c 2007 Theodore R.S. Deakyne

Submitted in Partial Fulllment of the Requirements for the Degree of

Master of Science

August 2007

I grant The Pennsylvania State University the non-exclusive right to use this work for the Universitys own purposes and to make single copies of the work available to the public on a not-for-prot basis if copies are not otherwise available.

Theodore R.S. Deakyne

The thesis of Theodore R.S. Deakyne was reviewed and approved by the following:

Eric R. Marsh Professor of Mechanical Engineering Thesis Advisor

Eric Mockensturm Associate Professor of Mechanical Engineering Optional Title Here

Karen A. Thole Professor of Mechanical Engineering Head of the Department of Mechanical and Nuclear Engineering

Signatures are on le in the Graduate School.

Abstract

This work explores the role of grinding machine structural loop stiness on workpiece quality. Specically, the loop stiness and grinding parameters (e.g., feeds and speeds) are mapped to the measured forces, actual depth of cut, workpiece form and surface nish. The work is based on outer diameter grinding experiments run on a CNC Moore machine tted with two air bearing spindles. The work head is a special 4R spindle from Professional Instruments that has four Lion Precision capacitance probes imbedded directly into the stator, which allows for the calculation of the normal and tangential grinding forces. The workpiece used to vary the eective stiness of the grinding machine features three anges at set distances from the face of the spindle. Plunge grinding at the dierent anges provides changes in the machines eective stiness; the farther the ange from the spindle, the lower the stiness. The distances of the anges from the face of the spindle are determined by basic calculations and experimentally veried. There are many variables in the grinding process that can aect the quality of the workpiece. For this experiment only a small subset are considered, the main one being the loop stiness. The others are programmed depth of cut, infeed rates, and the rotational speed of the workpiece. The tangential and normal grinding forces are found to increase as a result of increasing stiness. The increase in programmed infeed rates also increase the grinding forces while lower stiness at the increased infeeds result is less material removal. Others have examined the eects of structural loop stiness on surface grinders and brittle material. The contribution of this work is the special instrumentation used in OD grinding. Most modeling work has been formulated for plunge grinding and this experimental eort will be used to explore the validity of these models. With this setup deeper insight into machine stiness eects on the grinding process is given with plots and calculations.

iii

Table of Contents

List of Figures List of Tables List of Symbols Acknowledgments Chapter 1 Background 1.1 Overview . . . . . . . . . . . . . . 1.2 Introduction to grinding . . . . . . 1.3 Literature Review . . . . . . . . . 1.3.1 Eects of Machine Stiness 1.3.2 Chatter Eects . . . . . . . 1.3.3 Grinding Parameters . . . .

vi ix x xi

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

1 1 1 3 3 5 7

Chapter 2 Testbed Design 2.1 Introduction . . . . . . . . . . . . . . . . . 2.2 Machine Design . . . . . . . . . . . . . . . 2.3 Instrumented Spindle . . . . . . . . . . . . 2.4 Theoretical Force Calculation . . . . . . . 2.5 Wheel . . . . . . . . . . . . . . . . . . . . 2.6 Pretests and Measured Force Calibrations 2.7 Conclusion . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

9 9 9 10 11 15 17 21

Chapter 3 Data Collection and Processing 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Software and Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22 22 22 25

iv

Chapter 4 Design of Variable Stiness 4.1 Introduction . . . . . . . . . . 4.2 Workpiece . . . . . . . . . . . 4.3 Probe Holder . . . . . . . . . 4.4 Theoretical Modeling . . . . . 4.5 Frequency Response Function 4.6 Conclusion . . . . . . . . . . Chapter 5 Tests and Results 5.1 Introduction . . . . 5.2 Testing Parameters 5.3 Testing Matrix . . 5.4 Results . . . . . . . Chapter 6 Conclusion

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

33 33 33 35 35 37 38

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

42 42 42 44 44

57

Appendix A Processing Program A.1 MATLAB PROCESSING CODE . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix B DOI MODELING B.1 MATLAB MODELING CODE . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bibliography

58 58

68 68 72

List of Figures
1.1 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 Shows the simplied model for the cylindrical plunge grinding process . . . . . . Testbed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Assembled Rathead setup on left, disassembled on right . . . . . . . . . . . . . . Left shows a standard 4R stator, center complete 4R spindle, Right 4R Rathead stator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shows multiple FRF of a 4R spindle set at dierent air pressure . . . . . . . . . . Simplied 4R (Left), Equivalent 4R (Right) . . . . . . . . . . . . . . . . . . . . . Three grinding wheels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . In descending order starting rst with the new softer wheel, new machined pilot and new machined wheel cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Showing the new pilot being dialed in with indictor . . . . . . . . . . . . . . . . . Bolting up the new wheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Trueing(concentric) and dressing(sharpening) the new wheel with diamond tool and coolant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stepped graph of load cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pushing the load cell into the negative y-direction of the second ange on workpiece Normal Force data collected by spindle in Blue compared to the Load Cell data collected in green for all three anges of the workpiece. All data is graphed to the same scale, it is just the push test was not done exactly the same. Green data lags the blue, which was expected since during the pretests with the load cell it was noticed the response was not instant . . . . . . . . . . . . . . . . . . . . . . . Capacitance probe used in collection of data. The sensing ability of the probe is limited to a tiny circle of area 0.125in. Care should be taken that the sensor is in the proper location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Modular system with seven slot cabinet. One of the drivers is pulled halfway out of cabinet blocking the three other cabinets behind it. . . . . . . . . . . . . . . . Two lters one stacked on top of the other . . . . . . . . . . . . . . . . . . . . . . Four channels sampling the same signal, and shows the eects of cuto frequency SCB-68 68-Pin shielded connector block . . . . . . . . . . . . . . . . . . . . . . . The larger plot shows one revolution of data that contains 1024 points of data. The Smaller plot shows a close up of only a section of the data to illustrate the DAQs resolution. As can be seen the bins marked by red lines are integer multiples of 0.004883 Volts. The data points shown in green round to the red lines. . . . . . . Graphical User Interface (GUI) created in LABVIEW CVI . . . . . . . . . . . . 2 10 11 13 14 15 16 16 17 17 18 18 19

20

3.1

22 23 24 25 26

3.2 3.3 3.4 3.5 3.6

3.7

27 28

vi

3.8

Raw data taken directly from output of LabWindows code before any MATLAB processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.9 The top graph again shows the raw data. The lower graph shows the synchronous error motion caused by the spindle. The red box is a close up of synchronous error motion and further illustrates the repeating error motion . . . . . . . . . . . . . . 3.10 Top plot show data set with repeating error motion removed. The bottom plot shows the before and after thermal drift is removed . . . . . . . . . . . . . . . . . 3.11 These two plots show the spots selected in order to line up and compare the workpiece material removal to the table motion . . . . . . . . . . . . . . . . . . . 3.12 Top plot show data after lining it up. The bottom plot shows the nished process data with a lter used in MATLAB, and the correct sensitives used. . . . . . . . 4.1 4.2 4.3 4.4 4.5 4.6 Variable Stiness Workpiece . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Probe holder machined from steel beam . . . . . . . . . . . . . . . . . . . . . . . User interface for SIGLAB unit. Currently shows coherence and input force . . . Amplitude vs Frequency, the most sti responds shown in blue and the least in . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Model of the combined workpiece, front/back adaptor plates, and the rotor of both the airbearing spindle and the motor. The stiness kr and kt are found from the table. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . FRF plots of measured compared to model . . . . . . . . . . . . . . . . . . . . . Basic test: actual depth of cut (material removed from workpiece) vs. the set depth of cut (Table motion) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Show the normal force data of twelve tests where the workpiece speed and length from spindle center line are varied. The feedrate is kept constant at 0.05mm/min Compares the rst three columns of the force data from the previous gure at 150, 200, and 300 RPM at a feedrate of 0.050mm/min . . . . . . . . . . . . . . . Show the normal force data of twelve tests where the workpiece speed and length from spindle center line are varied. The feedrate is kept constant at 0.025mm/min Force and part data during chattering at a feed rate of 0.025mm/min, work speed 600 RPM, and at Flange2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compares the rst three columns of the force data from the previous gure at 150, 200, and 300 RPM at a feed rate of 0.025mm/min . . . . . . . . . . . . . . . Show the normal force data of twelve tests where the workpiece speed and length from spindle center line are varied. The feedrate is kept constant at 0.0125mm/min Compares the rst three columns of the force data from the previous gure at 150, 200, and 300 RPM at a feedrate of 0.0125mm/min . . . . . . . . . . . . . . Steady state grinding forces vs. feed rate at a workpiece speed of 150 RPMs . . . Steady state grinding forces vs. feed rate at a workpiece speed of 200 RPMs . . . Steady state grinding forces vs. feed rate at a workpiece speed of 300 RPMs . . . Shows the part material removal in blue compared to the table in feed in green for twelve data sets at a feed rate of 0.050mm/min . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shows the part material removal in blue compared to the table in feed in green for twelve data sets at a feed rate of 0.025mm/min . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

29 30 31 32 34 35 38 39 39

4.7 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 5.15

40 41

45 46 46 47 48 49 50 50 51 52 53 54 54 55 55

vii

5.16 Shows the part material removal in blue compared to the table infeed in green for twelve data sets at a feed rate of 0.0125mm/min . . . . . . . . . . . . . . . . . . 5.17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56 56

viii

List of Tables
2.1 5.1 5.2 5.3 Specs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 42 43 44

System Parameters set by user . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Norton Company published in Grinding Technology . . . . . . . . . . . . . . FORCE/PART TESTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

List of Symbols
S yc ym kef f loop kp kc kw kef f R Workpiece wear, p. 2 Grinding wheel wear, p. 2 Deformation of the contact area between the workpiece and grinding wheel, p. 2 Lumped deformation of the machine structural loop, p. 2 Lumped eective stiness of grinding process, p. 2 Workpiece spindle stiness, p. 2 Stiness of the contact area between the workpiece and grinding wheel, p. 2 Stiness of the grinding wheel spindle, p. 2 Lumped eective stiness of the 4R airbearing spindle, p. 12 The radius from the axis of rotation to the probe sensor in the stator, p. 13

L The length from the center of the spindle to the location of the grind, p. 13 kt ka kr The displacement between the stator and thrust plate, p. 13 The tilt stiness coecient of the 4R airbearing spindle, p. 13 The axial stiness coecient of the 4R airbearing spindle, p. 12 The radial stiness coecient of the 4R airbearing spindle, p. 12

b Radial distance from 4R spindle centerline, p. 12 a Tspring n Kstatic E I J Length from the instrumented spindles centerline to the location of cut, p. 12 Torque, p. 13 Length from the instrumented spindles centerline to the location of cut, p. 14 Static stiness, p. 35 Modulus of elasticity, p. 35 Moment of inertia, p. 35 Mass moment of inertia, p. 36 x

Acknowledgments
First, I would like to thank Dr. Eric Marsh for all of his support and help. I have learned so much from him not only in precision machining, but also lessons for life. Thank you! Alex Moerlein, both my lab partner and friend, thank you for your help, advise, and camaraderie. I know I can be overbearing with all of the questions I ask, but you managed to survive them all and always answered them the best you could. Including some of the more bizarre questions. Special thanks goes to Jim King for both your friendship and making the lab experience more entertaining. Next, I would like to thank all of the people at Professional Instruments who made this thesis possible. Especially, Dave Arneson and Mel Liebers who were always willing to take the time and help me with precision machining questions, which saved weeks worth of work. Thank you both Phil Irwin and Larry Millinder or your patience and help with machining and supplies My parents Marianne and Roy you have always been there for me. Many of my values and my condence comes from your support and lessons, thank you. Julianna you are the best sister a guy could have, and thank you for your support. Finally, I would like to thank my ancee, Jennifer Cathcart. I know I can be a complicated man to live with, but you have never left my side. I love you so much.

xi

Chapter

Background
1.1 Overview

The goal of this work is to investigate the role of eective machine stiness and its eect on the grinding forces, workpiece form, and the actual depth of cut into the workpiece. The setup involved an instrumented air bearing spindle capable of measuring the grinding forces during a cylindrical plunge grind operation. The overall eective machine stiness was varied using a special workpiece that would increase the length from the targeted grind location to the center line of the instrumented spindle. Two capacitance probes embedded in the spindle measured real time grinding forces while two additional probes measured the material removal rate (MRR) of the part along with the grinding wheel position. These four measurements were used to investigate the grinding process

1.2

Introduction to grinding

Grinding can be dened as a machining process used to sharpen, shape, or remove material through abrasion. The process can be used to obtain precise dimensional accuracy and surface nish or remove material from parts too hard to machine with conventional methods. It sounds like a simple process yet, grinding has been called the least understood material removal process [1] The act of abrasion is not a new concept. The use of abrasives for shaping and nishing materials has occurred for thousands of years all over the world. Continuous developments in the grinding processes have occurred making it more ecient and predictable. One of the rst improvements in grinding and nishing were made with the introduction of aluminum oxide and silicon carbide abrasives. These new wheels replaced natural abrasives, because they contained no impurities and could be controlled and modied for dierent grinding processes, which made the process more controllable. Scientists since the early 50s have analyzed the dynamics of

u(t) w kw kc p kp

rp rw
Figure 1.1. Shows the simplied model for the cylindrical plunge grinding process

several grinding processes and published numerous papers on their ndings. Still modeling even the simplest grinding process is dicult due to the large number of parameters . Unlike conventional machining on lathes and mills in which the dynamics are well understood, and use a single point tool on the workpiece, the grinding wheel can be though of as a multi point tool. Each of the grits used in the process acts as a single cutting tool so that the point of contact becomes an area of contact. The experiments here measure the real-time process forces, MRR, and depth of cut for a cylindrical plunge grinding operation. A Plunge grinding operation is where the workpiece, shorter dimension in thickness than the wheel face, is ground by feeding the grinding wheel straight into the workpiece [2]. There are many models of the dynamics to model the cylindrical plunge grind. One of the most frequently sited models comes from Snoey and Brown [3], which the governing equation to the process is: (t) = u(t) yc (t) ym (t) s (t) (1.1)

In this equation (t) is the material removed from the workpiece, u(t) is the set infeed of the wheel into the workpiece set by the program,yc (t) is the contact deformation between the wheel and part, ym (t) is the deformation of the machine and s (t) is the material removed from the wheel. This equations comes from looking at the simplied model of the plunge grind shown in Figure 1.1 which is three linear springs, the workpiece spindle stiness kp , the contact stiness kc , and the wheel spindle stiness kw . Where the eective stiness to this model becomes: Kef f loop = 1/(1/kp + 1/kw + 1/kc ) (1.2)

3 Finally the forces to this model is composed as follows: Ftotal = fsliding + fploughing + fcutting (1.3)

These equations and model help explain typical plunge grinding curves seen during operation. [The force curves have an exponential look, where in the beginning, when the wheel rst comes into contact with the part, the wheel is more ploughing and sliding on the workpiece. After some time cutting dominates and the forces reach a steady state.] Also looking at the material removal and infeed rate equation 1.1 helps to explain why they are proportional, but the removal rate lags the infeed rate due to compliances in the structural loop yc (t), ym (t), and s (t). So when the machinist sets a depth of cut to 1in without sparkout he may only really remove 0.95in o the part leading to undersizing. The stiness of the machine is examined in this work to see the impact on ym (t), the machine stiness, and its eects on (t), Ftotal along with the form error of the part.

1.3

Literature Review

As stated earlier there are many parameters that eect the grinding process which make it dicult to predict. Machine stiness is one of the most important considerations for ultraprecision machines. However, there is much debate over which is better, higher structural loop stiness, or a lower structural loop stiness. The true answer is, it depends. It depends on what is the necessary output to the machine, whether it is high production volume, low quantity high quality precision parts, the type of materials used, etc. The only way to truly nd out is through experimentation. Many researchers have examined this and their knowledge is built upon in this work. Below is a brief synopsis of some of their work.

1.3.1

Eects of Machine Stiness

Byron R. Knapps thesis [4] stresses the need to account for machine stiness of brittle material grinding, and looked at the machine stiness eects on the equivalent chip thickness, grinding forces, surface nish and workpiece form on brittle materials. The experiments were conducted on a manual Moore No3 base with the same instrumented spindle used in this thesis. The grinding operations used a silicon wafers resin bond diamond grinding wheel to conduct a plunge surface grind. The machine stiness was varied using an annular exure. The increase in machine stiness was shown to have two consequences on the grinding force. First, the amplitude of both the average grinding force, as well as high frequency components, are greater. Secondly, the build up of the low stiness grinding force is less than that of the high stiness force. Also low infeed rates showed an abatement of both the low and high frequency components of the grinding force. The machine acts as a mechanical lter of form errors in the workpiece. The cutability is decreased as stiness increases. Cutability here is the relation between the cutting force to the instantaneous depth of cut. Also found was that surface roughness could be independent from

4 machine stiness if the grinding cycles included sparkout. The main point here is that the higher structural loop stiness resulted in higher forces and that the form of the workpiece could be maintained with sparkout. Lee and Furukawa [5] looked at stiness and its eects on the residual stock removal behavior and self-excited chatter vibration results which are usually caused by deections of the grinding wheel and work-support systems in an ordinary grinding machine. Their work presents a way to determine how much static stiness should be included to obtain proper sparkout time for a residual stock removal. Zhang [6] investigated the eects of machine stiness on grinding of silicon nitride, with specic concerns on grinding forces, actual depth of cut (ADOC) of a grinding wheel and workpiece strength. A grinding system with an adjustable stiness was created through the use of a compliant workholder added to a precision grinder. Single-pass and multi-pass grinding experiments were conducted to evaluate the eect of machine stiness. Cup-type diamond wheels of two dierent bond types and three grit sizes were used in the grinding experiments. Static and dynamic simulations were carried out to correlate grinding forces and actual depth of cut with machine stiness. Since the simulation used a time-domain model, it could accommodate non-linearity caused by the eect of machine stiness on grinding forces and actual wheel depth of cut, workpiece regeneration, wheel wear, as well as wheel bond type and grit size eects, etc. The general eects of machine stiness consists of a static and dynamic element. The static is commonly observable in grinding processes while the dynamic is not desired since it is often detrimental to workpiece quality and accuracies. Based on the simulation and experimental results, it was found that machine stiness gives more eect in single pass grinding on wheel ADOC, grinding forces and eventually on the exural strength of ground workpieces than in multi-pass grinding. One of the important ndings of this study is that the exural strength of ground workpieces is dependent more on the dynamic component than on the static component of a grinding force Horiuchi and Shibata [7] investigated the inuences of the workpiece stiness on grinding accuracy by computer simulations of the grinding process for cylindrical plunge grinding. The reliability of the simulation program was demonstrated by identication with experimental results. The eects of various errors on grinding accuracy were investigated for dierent work stiness. Accuracy of the ground surface was aected in dierent manners by dierent error causes. The results from this work showed that the grinding time constant is shorter and irregular on the pre-machined surface and can be removed quickly if work stiness is increased. The waviness due to the runout of the grinding wheel will increase and the depth of dent due to a keyway decrease if the work stiness is increased. These results show that the machine stiness depends on the type of machine process and that it is not always true that higher stiness results in higher accuracy. Jenkins and Kuyrfess [8] looked at operating a machine tool, such as a grinder, so that the spindle speed is at the systems natural frequency. The dynamic stiness was lower at the natural frequency and thus can be more tolerant to force variations, which will hopefully produce a

5 better quality surface. The point of this investigation was to determine if operating the grinding wheel at a natural frequency of the system results in improved surface nish due to the increased compliance in the system, which cushions the grinding wheel and lowers the normal force variations. The dynamic stiness of a three-axis grinding system was experimentally determined. Dynamic stiness is considered to be important in machine tool design because it eects the quality of the part being machined, machine tool life, and the repeatability of the part being manufactured. Grinding experiments were conducted to determine the eects of dynamic stiness on surface nish in terms of the surface prole characteristics. Displacements induced by grinding wheels typically exhibit a once-per-revolution eect. Thus, the rst experimental grinding wheel speed (in terms of revolutions per second) is selected by examining the empirical dynamic stiness data and choosing a resonant frequency (cycles per second) in the range of achievable grinding wheel speeds (revolutions per second). For comparison, a grinding run was performed using a higher grinding speed, where the system dynamic stiness at the corresponding frequency was larger. Simulated results support preliminary experimental data, which indicate smoother surfaces result when grinding with rotational speeds corresponding to an natural frequency. A current and logically accepted practice in designing a machine tool is to make the tool stiness as large as possible. By designing and implementing very high stiness in grinding, displacement variations from the cutting forces are reduced. However this has created machines that are large in size, weight and cost. Alternative is a compliant force-controlled grinding system. Current machine design philosophies target the reduction of tool vibration eects by increasing stiness and adding damping. From this research it appears that resonant mode grinding may be benecial in attaining smoother surface nishes in some applications. The ground material hardness was correlated to the eectiveness of resonant mode grinding. Chatter may have been pronounced in the harder material.

1.3.2

Chatter Eects

Another eect on the grinding process witnessed during experimentation for this thesis was chatter. Chatter is a condition of vibration that occurs in many machining processes such as milling, turning, and frequently occurs in precision grinding operations. It is an interaction between the setup with cutter whether it is an end mill or grinding wheel and the workpiece. Once this condition arises, it is often self-sustaining until the problem is corrected. Chatter can be identied when lines or grooves appear at regular intervals in the workpiece, shaking of the machine, and an excess of loud pulsing noise. The eects of chatter on precision parts is detrimental and interferes with the precision requirements of the part. Also these vibrations reduce the maximum permissible grinding rate and accordingly the rate of production [9]. At the start of testing for this thesis the chattering eects from the grinding process were so bad, that the rst tests had to be ended prematurely with no attainable results on the eects of stiness. Thanks to Dave Arneson from Professional Instruments, the chatter eects were eliminated from

6 the setup by replacing the grinding wheel. However, before talking to Dave a lot of research was done to learn more about this eect. One of the rst to study this phenomenon of regenerative chatter was Hahn [9]. In his work an analysis is made, based on the proportionality of the instantaneous wheel depth of cut to the instantaneous dynamic force existing between wheel and work, which yields two stability criteria from the Nyquist diagram. The grinding system will be unconditionally stable if one stability criterion is satised. If this criterion is not satised, a second criterion may yet be satised by proper adjustment of the cycles of vibration per revolution of the workpiece. The criterion is created from eective mass of the vibrating system, in contact spring rate, eective damping in the vibrating system, and a grinding-rate factor which he denes as the radial component of wheel force required to produce a radial advance of one unit, of the wheel into the work during a work revolution. This theory agreed with practical grinding, that a grinding wheel with a wide face is more apt to chatter than a narrow one because the grinding rate factor increases in proportion to the width of the wheel. Also, that as grinding wheels become dull it becomes more likely to chatter. From observations, chatter is more likely to occur at high work speeds. To improve the setup in this thesis the grinding wheel was dressed in such a way to reduce its width by half while making sure it was sharp, and the work speed decreased. Chatter conditions did not improve in the setup. Mohammed [10] looked at the eects of dynamic changes in the grinding force components due to changes in the grinding wheel wear at area and the workpiece material on the vibration behavior of the grinding spindle. The steady-state dynamics and vibration behavior of the grinding machine spindle is simulated by a ve degree of freedom model. The results indicate that when grinding dierent materials using a grinding wheel with xed wear at area, dierent vibration behavior is noticed. As the grinding wheel at area increases, the level of vibration increases for all the degrees of freedom, which indicates that there is an upper limit for the level of wear on the grinding wheel that can be tolerated, and after that level dressing operation must be conducted on the grinding wheel. Another point mentioned in this work is that vibrations cause major problems in the grinding process as the grinding wheel depth of cut is so small that even the slightest amplitude of vibration can have dramatically damaging eects on surface nish, wheel wear, and form-holding. The use of increased wheel exibility was examined to improve grinding performance by reducing chatter by Sexton [11].This paper describes the development of a technique for suppressing chatter which is particularly applicable to the use of diamond and CBN wheels. He states that it has been predicted theoretically and conrmed by experiment that chatter may be suppressed, or even eliminated by increasing the radial exibility of a grinding wheel, while at the same time maintaining high values of the wheels natural frequency and damping. It has been shown that the increased exibility may be achieved by mounting the abrasive rim of the wheel via a exible coupling or by manufacturing a wheel by conventional means but with a exible hub material. Basically, chatter often develops when grinding with diamond or cubic boron nitride wheels, whose chief characteristic is that they are very hard-wearing. Chatter is more likely to

7 occur for harder type grinding wheels. Originally the pretest for this thesis began with an O hardness grinding wheel. (The hardness scale of grinding wheels is set to the alphabet where A represents the softest wheel and Z the hardest.) The original wheel was later replaced by an I hardness wheel which did not improve the occurrence of chatter. Later it was learn that the decrease in hardness from O to I was not great enough. Baylis [12] also looked at the wheel exibility to improve the grinding process along with other parameters. He reviewed previous papers showing the benets of exible grinding wheels in demonstrating chatter suppression, and developed a mathematical solution which allows a more substantial investigation of all the parameters. A theoretical model of the process was developed and a digital simulation was used to show that the process either underwent an exponential growth or decay of vibrations. The use of a digital simulation resulted in time constraints on investigating all the relevant parameters. The results conrm that the model adequately predicts the very slow growth rates often encountered in practice. It also shows that the very slow precession rates of wheel waves around the circumference of the wheel may be predicted theoretically. Finally the advantages of a exible wheel described by Sexton, was conrmed by this model.

1.3.3

Grinding Parameters

As mentioned earlier in the introduction, there are many parameters that eect the grinding process which make it dicult to predict. Also with so many parameters it is dicult coming up with which parameter to vary and what is eecting something else in the system. To help with some of this complications Snoeys et al [3] paper was reviewed. Snoeys and Brown established a dynamic model for the external plunge grinding process based on the stability analysis with the classical control theory. Their work has become a cornerstone in machining dynamics. The model looks at stability requirements such as geometrical loci and the stability criterium formula. Important system parameters are studied,such as the cutting stiness, wheel wear resistance, contact stiness, and machine stiness. It also looks at geometrical limitations of chatter of transverse Feed Filtering and Filtering due to the nite length of contact. Finally mentioned are the grinding chatter results on high frequency grinding chatter and regeneration on workpiece and on grinding wheel. The grinding parameters, which are shown to be important from the theoretical analysis, are discussed in terms of their eect on grinding process stability. Some important cases of instability, which can occur in practice, are discussed in a theoretical manner. It is shown that the theoretical derivation of the formulated stability requirements is an extension and a generalization of the classical single tool cutting stability criterion. A set of chatter tests was performed in order to verify these theoretical predictions. The theoretical correlation between the machine characteristic, the work speed, the compliance of the grinding wheel-workpiece contact area and the appearance of work generation or wheel regeneration has been checked. Here some of these parameters are discussed in more details with the model introduced

8 earlier, which came from Snoeys paper. Equation 1.1 used is based upon the statement that at any instant of time the sum of the total wear of the wheel and part must be equal to the total infeed minus the deformation of the machine and the deection of the contact area. The instantaneous depth of cut or wear must be combined with the shape of the wheel and the form of the workpiece, which are determined by the previous depths of cut and the wear history of the grinding wheel. Another important parameters in the plunge grinding process is the workpiece speed. The inuence of the speed of workpiece and grinding wheel is conrmed by practical chatter experience. It is often recognized that an increase of work-speed leads to chatter problems and that an increase of the wheel-speed may contribute to a better dynamical behavior. The importance of speed of the workpiece is also demonstrated in the results section for the plots showing the twelve tests comparing the normal forces for dierent speeds and ange locations Lastly the nal parameters discussed which are shown in the model are the cutting stiness, wheel wear resistance, contact stiness and machine stiness. The cutting stiness is the ratio of the cutting force to the instantaneous depth of cut. The cutting stiness is a function of the wheel speed where the increase of the wheel speed decreases the grinding force. The wear stiness is about 100 to 10,000 times higher than the cutting stiness and consequently the inuence of the cutting stiness upon the limit of stability will be negligible with respect to the rst term. The value of the contact stiness is unique to grinding and is inuenced by a number of parameters such as the grinding wheel composition, the dressing conditions, the wheel and workpiece diameter. However, the pre-load of the contact area is the most important factor, which is why a softer wheel usually leads to smaller cutting forces. Preloads on hard wheels are higher in which case the contact area reacts stier and the stability is inuenced unfavorably. Lastly is the machine stiness. High static stiness of a grinding machine always contributes to an increase of the stability of the cutting operation with other factors being constant.

Chapter

Testbed Design
2.1 Introduction

Proper Design of the base, machine test bed, will ensure that the result measured will be due to the grinding process. Improper setups can measure erroneous data that can be mistaken for the grinding process. Loose probes, improper grinding wheels, noisy spindles (such as vibrations in ball bearing spindles with cages and deective balls) can ruin the results of the test. Also Vibrations outside of the spindle contribute additional error motion. Usually referred to as structural error motion, this error motion will contaminate the measurements from the spindle and need 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. Therefore all of these factors were examined before continuing with the experiments.

2.2

Machine Design

The testbed for this experiment was made up of a plain way No. 3 Moore machine base tted with a motorized Professional Instrument twin mount 4R air bearing spindle mounted with a 10in grinding wheel, and a second, instrumented spindle mounted with the Part. This setup is shown in Figure 2.1. The Moore base has been designed with accuracy and stiness in mind. The bed is large enough to support the table in both the x and y axis in full travel yet it is not so massive in size that it deects under its own weight. Since the table is fully supported and will never hang o the edge of the bed, the deection and error are decreased while the stiness is increased improving the measurement qualities. Also the tables axis run on double V ways that provides the best guides for straight line travel. These types of ways are the best for supporting loads, such as the workpiece weight, no matter where it is placed on the table. [13] Both of the air bearing spindles on the setup are very accurate and sti which complement

10

Figure 2.1. Testbed

the Moore Base. Grejda [14] has demonstrated that these bearings exhibit error motions as low as 3nm without motorization. As seen in Table 1 the basic 4R air bearing spindle has both a radial and tilt error motion of less then 1microinch. Combine this with the stiness, and one gets a spindle far more accurate and sti then conventional ball bearing spindles. With these two features the test bed will provide accurate and precise results and a good structural loop for plunge grinding.

2.3

Instrumented Spindle

The instrumented spindle is a 4R blockhead air bearing spindle. The complete setup for the instrumented spindle is shown in Figure 2.2. It is comprised of a special 4R air bearing spindle, motor with mount and encoder. This spindle is very unique from any other 4R blockhead in that it has four capacitance probes embedded directly into the stator of the spindle. These four probes target the bottom thrust plate of the spindle. The adaptor plates and workpiece are then bolted to this thrust plate.The deection of the grinding system is composed of various elastic deformations, e.g. deection of the machine structure, deection of the work and its supporting system, elastic deformation at the contact zone between work and grinding wheel as shown earlier [7]. As the workpiece and grinding wheel come together, the deection caused

11
Motor Housing and Encoder

4R Airbearing Spindle

Supporting Mount

Figure 2.2. Assembled Rathead setup on left, disassembled on right

between the stator and the thrust plate is measured by the capacitance probes in the stator. Table 1 shows the blockhead air bearing spindle specications for the 4R at 150psi.
Table 2.1. Specs

Load Capacity Radial Axial Tilt Stiness Radial Axial Tilt Error Motion Radial Axial Tilt

Ultimate 100lb 400lb 400lb-in 0.67lb/microinch 2.0lb/microinch 4.0lb-in/microradian <1microinch <1microinch <0.1microradian

Working 50lb 200lb 200lb-in

The values represented in Table 2.1 are associated with the spindles centerline. These values along with the voltage values recorded from the four capacitance probes are used to examine in-process normal grinding forces.

2.4

Theoretical Force Calculation

Grinding forces have been examined in the past on the Instrumented spindle in many dierent applications. Two previous students had setups very close to the current one used in this experiment. Both Jeremiah Coueys and Byron Knapps thesis are used extensively for help with force

12 equations and trouble shooting with the setup [1, 4]. However both of their experiments are very dierent from this one and are discussed in more detail here. Byron Knapp used a resin bond diamond grinding wheel and mounted the workpiece on an annual exure which used reeds of dierent thicknesses to vary the stiness. There are nine steps for the assembly process for the exure in order to swap out the reeds of dierent thicknesses. In this experiment there is not an assembly process. Once the workpiece is dialed in on the instrumented spindle all of the experiments can be completed. The diamond wheel was then plunged into the face of the workpiece instead of the outer diameter. As a result the sensors embedded in the instrument spindle measured both axial and tilt displacement from the thrust plate of the instrumented spindle. This led the equation for eective stiness to become: Kef f = [ ab 1 + ]1 kt ka (2.1)

The variables kt and ka are found in Table 2.1 for tilt and axial stiness of the spindle. The variable a is the length from the instrumented spindles centerline to the location of cut and b is the radial distance. From this equation the stiness is behaving as two springs in parallel. Jeremiah Couey also used the same basic setup but with an alumina oxide wheel, and instead of using the instrumented spindle with the embedded sensors he tried to develop a sensor rig that could be bolted to any of the air-bearing spindles in the lab. This rig mounted two capacitance probes in the x and y-directions tracking the displacement between the rig and outer diameter of the spinning thrust plate of a Professional instruments twin mounted spindle. The type of cut was a outer diameter plunge grind into a workpiece similar to the cuts taken and explained in this work. As a result the equation had to adapt to accommodate both radial and tilt in which case the equation for the eective stiness becomes: Kef f = [ l2 1 + ]1 kr kt (2.2)

Both of these examples are similar in terms of the setup used in this thesis. However, the equations to the eective stiness are very dierent. These results were based on the type of cut into the workpiece along with the placement of the measurement sensors. This knowledge is used here to development the new eective stiness. The development of the equation for eective spindle stiness came directly from looking at the output from the embedded sensors in the instrumented spindle and the type of cut on the workpiece. Shown above in Figure 2.3 is a standard 4R block head spindle with the two thrust plates and stator labeled. On the left is a standard stator to the 4R and to the right is the instrumented stator. As can be seen in the gure, the four circles within the instrumented stator are the sensors which target the back of the top thrust plate. Looking at the stator from the side as Figure 2.5, the sensors can not see the displacement from the radial motion. Instead they can only detect the tilt of the thrust plate. Therefore the 4R becomes a torsional spring. From

13
Standard Stator Top Thrust Plate Rathead Stator Capacatance Probes

Stator

Bottom Thrust Plate

Figure 2.3. Left shows a standard 4R stator, center complete 4R spindle, Right 4R Rathead stator

the Figure and assuming small angle approximation the theorectical deection measured at the sensor becomes: = R Also using the Figure the torsional spring force becomes: Tspring = Fwheel L = kt (2.4) (2.3)

In the above equations L is the length from the centerline of the spindle to the target of grinding forces Fwheel , and R is the distance from the axis of rotation to the sensor. Combining these two equations by solving of theta in equation 2.3, substituting it into equation 2.4 and nally solving for the force of the wheel, Fwheel we get: Fwheel = kt = Kef f x RL (2.5)

Therefore the eective stiness base on the output location of the sensor is: Kef f = kt LR (2.6)

The kt in this equation is the spindle tilt stiness acquired from Table 1. As mentioned earlier the values in Table 1 are for the 4R at 150psi. During all the tests for this thesis the average pressure is measured at 110psi. The stiness of an air bearing spindle is eected by the supplied air pressure. In the air bearing the stator and rotor never come into contact during operation. Contact between these elements would be very bad and would eventually lead to the destruction of the bearing. Instead the rotor is supported by an air lm between the stator. This lm is what gives the bearing

14
RMS vibration (pm) 100 50 0
Magnitude of spindle frequency response at different pressures

10

atm 6

kHz

Figure 2.4. Shows multiple FRF of a 4R spindle set at dierent air pressure

its stiness. The stiness of the air lm increases at higher air supply while inertia remains the same. This eect is illustrated in Figure 2.4. The gure clearly shows the increasing natural frequencies that result from increasing air pressure through the FRF. Equation 2.7 in general is used to calculate the natural frequency of a single degree of freedom system. For this case since the mass is not changing, the stiness is proportional to the natural frequency k 2 [15]. =
n

n =

kspindle mspindle

(2.7)

However, there is a limiting return to increasing the pressure. As the pressure increases the gap between the stator, thrust plates, and rotor increases until more air is lost and results in diminishing returns. The calculated force needs a factor so that the eective stiness equation becomes: Kef f = () kt LR (2.8)

Where the factor is a factor caused by the supplied air pressure. This factor is determined experimentally in this work. Finally the normal force of the cylindrical plunge grind will be: Fnorm = Kef f x Where x is the measured displacement between the thrust plate and stator. (2.9)

15

Wheel

Tilt Wheel

Radial L

} Measure Gap
Sensor R

Kt

Figure 2.5. Simplied 4R (Left), Equivalent 4R (Right)

2.5

Wheel

Vibrations can be induced due to a number of reasons, such as insucient machine stiness, wheel wear, poor dressing and truing of the grinding wheel, and wheel unbalance [6]. Here the vibrations were caused by the hardness of the wheel There are many dierent types of wheels used for grinding that vary in size, grit number, hardness, material etc. For the experiments conducted in this work, the wheel used is a 10in diameter by 1in thick with a 3in hole, 57A60FVBE wheel (Need Figure). Choosing the correct wheel for the grinding application can avoid problems such as chatter, loading of the wheel, loss of surface nish, high grinding forces, machine failure etc. Figure 2.6 shows some of the grinding wheels tested in this thesis. In the pretests there were some initial diculties with the plunge grinding process. The wheels stopped cutting the workpiece and began to plow into the surface causing small pieces of the workpiece to tear o and embed themselves into the wheel. This made the whole process much worse as the metal on metal rubbing could be heard. This became louder as time went on causing severe surface damage along with the force measurements to increase to very large magnitudes, and the test bed itself started to vibrate. This is a form of extreme excited chatter. Many dierent theories for this phenomenon were talked about but to correct this problem we simply used a softer wheel and more coolant. Initially both the I and O hardness wheels were used which lead to chatter. The softest wheel that could be found was the F. As can be seen by the center hole of this wheel in Figure 2.6 it would not t the original setup. As a result a whole new wheel pilot and cap were designed to hold the new wheel. These new addition to the test setup are shown in Figure 2.7. As can be seen in the Figure the new wheel pilot can sit ush on a at surface indicating it does not have a lapped carbide hub as did the old pilot. This is important since this hub is used to ensure the old pilot concentric with the rotor of the spindle. The wheel will be spun up to about 1906 RPM. The last thing needed is a large 10in

Sensor

16

Figure 2.6. Three grinding wheels

Figure 2.7. In descending order starting rst with the new softer wheel, new machined pilot and new machined wheel cap

wheel spinning o axis creating unbalanced forces, chatter or even failure. Therefore this new pilot is carefully dialed in as shown in Figure 2.8. Another important thing to remember about a grinding wheel is to not over tighten. Over tightening can lead to cracks or some other kind of damage. Again a wheel with a crack spinning at 1906 RPM leads to disaster. For this setup the six screws of the wheel cap were tightened up to 45lb-in in increments of 5lb-in as shown in Figure 2.9. Tightening the bolts instantly to 45lb-in one at a time, again could lead to disaster. Finally the wheel is made concentric to the spindle by dressing as shown in Figure 2.10. Further information about the dressing parameters is given in chapter 5.

17

Figure 2.8. Showing the new pilot being dialed in with indictor

Figure 2.9. Bolting up the new wheel

2.6

Pretests and Measured Force Calibrations

With the theoretical eective stiness calculated and the new wheel in place, initial pretest of the system could begin. The theoretical calculation is evaluated against the measured values from force pretests. The force pretests are completed with an Interface SM Series Load cell connected to a DP25-S Strain Gage Panel Meter on the test setup with the instrumented spindle and workpiece. The meter was then connected via BCN cable directly into our data acquisition system where voltage signals coming through the meter could be recorded directly onto the PC and processed with the program MATLAB. Before data could be collected the load cell with meter had to be calibrated. This was done with a known weight. The known weight was acquired on an Acculab3 digital scale. Once the

18

Figure 2.10. Trueing(concentric) and dressing(sharpening) the new wheel with diamond tool and coolant

Calibrate Load Cell

0.08

0.06

Voltage

0.04

0.02

0 0 100 300 Points


Figure 2.11. Stepped graph of load cell

500

700

19

Figure 2.12. Pushing the load cell into the negative y-direction of the second ange on workpiece

weight was known, it was placed on top of the load cell and a voltage measurement was recorded and graphed in MATLAB as shown in Figure 2.11. As can be seen by the gure the resolution of the meter was 0.016V. The range of the meter goes from 0 to 4V. with 250 divisions or bins going from 0 to 4V. Each division represented about 46.3grams so the meter can measure up to about 11575grams before reaching saturation. With this knowledge we could then calculate the sensitivity of the load cell which is about 2894 grams per volt or about 28 Newtons per volt. Once the load cell was calibrated it is used to test the forces of the spindle that were calculated from Equation 2.10. This was done with the setup that included the Moore machine, spindle, and special workpiece bolted to the spindle. The spindle was then spun at 300rpm while the load cell was lightly pushed into each of the three anges of the workpiece, in the negative y-direction as shown in Figure 2.12. As can be seen in the picture a square block of wood was bolted to the load cell in order to avoid directly wearing the load cell (friction between metal spinning disk and metal C bar). The load cell is calibrated with the piece of wood. Also baby oil was used to lubricate the contact between the block of wood and the workpiece. Special attention was given to hold the blue housing of the load cell and not the metal C bar. This C bar contained the strain gage and would corrupt the data if the bar was pushed on this instead of the housing.

20

Flange 1

Normal (Spindle) Load Cell

Flange 2

Flange 3

Force (Newtons)

10

15

20

20

40

60 Time (sec)

80

100

120

Figure 2.13. Normal Force data collected by spindle in Blue compared to the Load Cell data collected in green for all three anges of the workpiece. All data is graphed to the same scale, it is just the push test was not done exactly the same. Green data lags the blue, which was expected since during the pretests with the load cell it was noticed the response was not instant

21 The test occurred over durations of 2minutes for a total of 300 revolutions of data collected. Each revolution contains 1024 pts of digital data stored. Figure 2.13 shows the results from three tests taken at each of the anges. The factor was multiplied to the spindle data so that it had the same magnitude as the load cell data. The hand calculations are lower from the actual force applied to the spindle as expected since it was tested at 110psi where the value of kt taken from Table 1 is assuming a pressure of 150psi.

2.7

Conclusion

This section talked about the design and thought process of the test bed and showed that the bed design is accurate for use in determining the eects of eective workpiece stiness on the grinding process. Also proven is the ability of the instrumented spindle to actually measure real time in process grinding forces.

Chapter

Data Collection and Processing


3.1 Introduction

The dierence between converting an analog signal into digital data that can be used to describe the grinding process or into garbage relies on the ability and knowledge of the user with both the hardware and software. In this section both the hardware used to collect and convert the analog signal into a digital signal and the software used to process this digital signal in order to interprets the eects to the grinding process is discussed.

3.2

Hardware

The data collection setup was comprised of four main components. These include the capacitance probes, modular system with drivers, mechanical lters, and a PC with a data acquisition (DAQ) card installed. The data collection begins with the four capacitance probes. Non-contact sensors developed by Lion Precision are shown in Figure 3.1. Two of these, as mentioned earlier, are imbedded in

Figure 3.1. Capacitance probe used in collection of data. The sensing ability of the probe is limited to a tiny circle of area 0.125in. Care should be taken that the sensor is in the proper location

23

Figure 3.2. Modular system with seven slot cabinet. One of the drivers is pulled halfway out of cabinet blocking the three other cabinets behind it.

the stator of the instrumented spindle and are factory calibrated with the proper driver. The output sensitivity of these two probes is 1V /m. The other two sensors were added to collect data on the part and table movement of the machine. Each of the added sensors are calibrated with their specic driver. The driver and corresponding sensor serial numbers must match those on the calibration report provided by Lion Precision. The part uses sensor with serial number 99212-08 and the table measurements are done with sensor 99212-06, both having a sensitivity of 10mV /m. Further information on the two probes are found in the appendix. Next the analog signal is passed to the corresponding drivers housed in the modular system shown in Figure 3.2. The modular system can contain up to seven drivers. The purpose of this housing is to allow multiple drivers to be connected together without having to worry about interference between these dierent drivers. The green board that can be seen in Figure 3.2 is the actual driver for the x-direction force probe. The drivers used for this setup are two DMT12, dual sensitivity probe driver modules, used for the force measurements and two DMT10, single sensitivity probe driver modules, used for the part and table. They create, control, and monitor an electric eld (A.K.A the signal) at the sensor on the tip of the probe. The driver then detects changes in the eld and produces an output voltage that is linearly proportional to the change in the gap between the sensor and target. This is how the displacements, table movement, and outer diameter of the part to the system are measured. More information about the drivers can be found in the Lion Precisions manual [16].

24

Figure 3.3. Two lters one stacked on top of the other

From the drivers the analog signal is then carried to the lters by a BNS connector through coaxial cables. The mechanical lters with digital displays are shown in Figure 3.3, The X, Y, RD, P tape labels in the picture stand for the X-force, Yforce, Machine table, and part signals. The lters are Krohn-Hite Model 3362 fourth order set to a 1000Hz cuto, Low-Pass, Butterworth lter. In general a lter is roughly analogous to a spring mass damper. A damper has the possibility of creating a time lag between what goes into the system and what comes out. The collection setup is not an exception from this as can be seen in Figure 3.3. For this gure the analog signal from the Y-Force sensor was split into four signals using a BNS splitter. Each of the four signals was rst sampled under a dierent cuto frequency producing the graph on the left, and then each of the signals was sampled under the same 1Khz frequency producing the graph on the right. This shows how the cuto frequencies eect the signal. In order to avoid the impact of time lag the signals, during data collection, were all sampled at the same cuto frequency. It should be noted that the output data still has the time lag but since they all have the same time lag it is not worried about. More information about the lters can be found in the user manual [?]. After ltering, the analog signal is distributed to the Input/Output (I/O) connector block shown in Figure 3.5. This is the last step before the analog signal is converted into a digital signal. The input to the I/O block is made up of two components from each of four signals, a DC voltage and ground (GND). These eight inputs make up the four channels. The output is sent from a 68-pin cable to the DAQ card installed on the PC. More information on the I/O block can be found in the SCB-68 68-Pin shielded connector block user manual.[]

25
Different Filter Cutoffs 1 1 Filter 1KHz Cutoff

Volts

Volts

1 11.89

300Hz(Ch1) 1KHz(Ch2) 3KHz(Ch3) no filter(Ch4) 11.895 11.9 11.905 time (sec) 11.91 11.915

ch0 ch1 ch2 ch3 1 14.798 14.8 14.802 14.804 14.806 14.808 time (sec) 14.81 14.812 14.814

Figure 3.4. Four channels sampling the same signal, and shows the eects of cuto frequency

Finally the four analog signals are converted to digital signals with the National Instruments DAQ PCI-6110E board [17]. The DAQ board features a 12-bit analog to digital converter with four simultaneously sampling analog inputs. The range for each of these channel inputs is from -10 V to +10 V. For a 12-bit converter the range is broken into 4096 bins. In other words the resolution becomes 20 Volts divide by 4096 which equals 0.004883 Volts. The sample points are all rounded into integer multiple bins.This is shown in Figure 3.6. Another important feature of the DAQ is the maximum sample rate the board can handle. As stated in DAQ boards manual, the maximum sampling rate is 5 million samples per second. The minimum sample rate for this board is 1000 samples per second. The experiments for this thesis were sampling 1024 samples per revolution of the spindle. The spindle speed was limited to a minimum of 150 RPMs and a maximum of 600 RPMs. This translates to a minimum\maximum sampling rate of 2560\10240 samples per second. Below the 150 RPM, the spindle with its low torque had trouble maintaining consistent speeds for testing. Above the 600 RPM and the spindle was going too fast for comfort.

3.3

Software and Processing

A LabWindows code original developed by J. Couey [1] is re-used as the graphical user interface (GUI) between the PC with DAQ card installed and user. Figure 3.7 shows the GUI, using this interface. The raw digital voltage signal ltered at 1KHZ was collected and transferred over to MATLAB for processing and analysis. Figure 3.8 shows an example of one of the data sets extracted from this setup. This is one data set out of forty eight,and will be used as an example to show the step by step processing with the software developed in MATLAB. Initially the data is delivered to MATLAB as an array sized to the multiplied value of the time required per test, the number of samples, and the speed of the workpiece. For this case the array is 768000 X 1. The units of this array are in terms of voltage from the capacitance probe. The next step is to remove the synchronous error motion from the

26

Normal Force

Tangential Force Machine

NI PCI-6110E Part

Figure 3.5. SCB-68 68-Pin shielded connector block

data. This synchronous error motion is because of the fact that the bottom of the thrust plate the probe is targeting, is not perfectly at. As the spindle rotates, the gap is changing due to the imperfection, however since the error motion is built into the thrust and does not change from revolution to revolution it can be easily pulled out by reshaping the array. For this example the 768000 X 1 array is reshaped in MATLAB to a 1024 X 750 matrix where the rows have become the data points for the column which represents the particular revolution of the spindle. At the beginning of each test the rst twenty ve revolutions of data are taken without any grinding.

27
Showing DAQ Resolution 0.4

0.2

Volts 0 0.2 0 points

1000

0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 900 1000

Figure 3.6. The larger plot shows one revolution of data that contains 1024 points of data. The Smaller plot shows a close up of only a section of the data to illustrate the DAQs resolution. As can be seen the bins marked by red lines are integer multiples of 0.004883 Volts. The data points shown in green round to the red lines.

Using twenty revolutions from this set (1024 X 20) the mean of each column is taken to produce a 1 X 20 matrix. This 1 X 20 matrix is then subtracted from each of the 750 revolutions of the raw data set. The resulting graphs are shown in Figure 3.9. With the synchronous error motion removed the data is now ready to have its drift removed. As seen in Figure 3.10 the data drifts away from the zero x-axis. This is further amplied in the lower graph of this gure. The drifting of the data set is believed to be a result of the temperature increasing due to grinding. Increasing coolant along with allowing the spindle to rotate awhile before testing has improve this condition but does not totally remove it. The only way to totally remove the drift is mathematically with

28

Figure 3.7. Graphical User Interface (GUI) created in LABVIEW CVI


0.5 Y(sensor2) X(sensor1)

Voltage

1 Table(sensor3) Part(sensor4)

10

Voltage

10

100 Time (sec)

200

300

Figure 3.8. Raw data taken directly from output of LabWindows code before any MATLAB processing

the program. A curve is t to the drift in the data which is then removed. This begins by taking the reshaped matrix from before and averaging all of the data points in each revolution to create a 1 X 750 array which is the lower plot of Figure 3.10. Using the ginput command in MATLAB, seven points (x,y) are chosen, shown as red circles in the lower plot. At these points no grinding has taken place and therefore the forces should be zero. Using MATLAB the seven points are created into a curve that is subtracted from the data. The results to this may be seen in Figure 3.12 Since the data sets will be compared to understand the change

29
Y(sensor2) X(sensor1)

Voltage

0.5

0.5

0 Table(sensor3) Part(sensor4)

10

Voltage

10

100 Time (sec)

200

300
0.6

Voltage

0.4

0.2

8.2

8.6

9 Time (sec)

9.4

9.8

0.6

Voltage

0.4

0.2

10 Time (sec)

15

20

Figure 3.9. The top graph again shows the raw data. The lower graph shows the synchronous error motion caused by the spindle. The red box is a close up of synchronous error motion and further illustrates the repeating error motion

in forces and part removable as a result of length from ange to spindle center line and spindle speed it would help if the data sets all started at the same time. This is achieve by once again using the ginput command. Figure 3.11 shows the two points (x,y) to be chosen. The upper plot is that of the table motion. The red dot on this plot is chosen to line the data on the x-axis. On the graph before the red dot is a force spike that results from an initial touch o between the grinding wheel and workpiece. This is done to keep the experiments constant. The nal results from the processing of the data is shown in the bottom of Figure 3.12. Now the data is ready to be interrupted and discussed in the results section.

30

0.1 0.05 Voltage 0 0.05 0.1 0.15

Y(sensor2) X(sensor1)

10 5 Voltage 0 5 10

Table(sensor3) Part(sensor4)

0.02

0.02 Voltage

0.04

0.06 Before Curve After

0.08

0.1

50

100

150 Time (sec)

200

250

300

Figure 3.10. Top plot show data set with repeating error motion removed. The bottom plot shows the before and after thermal drift is removed

31

10 8 6 4 2 0 2 4 6 8 10 10 8 6 4 2 0 2 4 6

100

200

300

Figure 3.11. These two plots show the spots selected in order to line up and compare the workpiece material removal to the table motion

32

0.05 0 Voltage 0.05 0.1

Y(sensor2) X(sensor1)

20 10 Voltage 0 10 0 50 100 150 Time (sec) 200

Table(sensor3) Part(sensor4)

250

YForce(sensor2) XForce(sensor1)

Newtons

10 Table(sensor3) Part(sensor4) Displacement ( m) 40 20 0 20 40 0 50 100 150 Time (sec) 200 250

Figure 3.12. Top plot show data after lining it up. The bottom plot shows the nished process data with a lter used in MATLAB, and the correct sensitives used.

Chapter

Design of Variable Stiness


4.1 Introduction

This section discusses the workpiece and model with corresponding Frequency Response Functions (FRF) of the system which is composed of the workpiece, adaptor plates, spindle and motor rotor. The FRFs are used to show the eective spindle stiness decrease as an eect of increase in length from one ange to the next. Initially a model is created and theoretical FRFs are produced which aid in the design of the workpiece. Once the workpiece is machined it is tested via modal analysis and the measured FRFs compared to the theoretical ones. Precision machining performance, whether it is grinding, milling or turning, is dependent upon the weakest link in the structural loop. Here the grinding performance is dependent upon the rigidity of the Moore base, wheel with spindle and the workpiece with spindle. If the grinding structural loop is too exible chatter will develop ruining the surface nish; or the grinding wheel will wear too rapidly causing frequent dressing/truing and lead to more down time during grinding operations; or long sparkouts will be required to hold nish and size tolerances; or all of these can occur at the same time [18]. Even the most rigid, well supported base like the Moore No. 3 machine used here and the stiest air bearing spindles like the 4R can not protect against the eects of a imsy workpiece. This knowledge is taken advantage of, to develop a workpiece so that the stiness of the grinding machine and its eect on the grinding forces, workpiece form, and the actual depth of cut into the workpiece can be examined. [Development and testing of this workpiece similar to the one found in Colding [18] but dierent is discussed in this section.]

4.2

Workpiece

The deection of a grinding structural loop is composed of various elastic deformations such as deection of the machine structure, deection of the work and its supporting system and elastic

34
7.255

3.000

Tap 10-32 Equally Spaced Twelve total 0.5in deep

2.500

3.500

0.500 0.250 0.250 0.250

0.500

Tap 10-32 Equally Spaced Twelve total 0.5in deep

0.375 2.675 5.875

Figure 4.1. Variable Stiness Workpiece

deformation at the contacting zone between work and grinding wheel and so on. [7] By increasing or decreasing the stiness of one of these components the deections are eected which overall eect the performance of the grinding cycle in terms of forces, precession,etc. The workpiece was key in the eects of eective machine stiness on plunge grinding for this experiment. Any changes to the grinding structural loop can vary the eective machine stiness more than expected. In order to control the eective stiness by a theoretical calculated amount a single workpiece is used that could vary the length from the face of the spindle centerline to the location of the targeted grinding forces. This workpiece can be seen in the Figure 4.1. In this gure the workpiece looks like a cylinder with two bolt circles on either end and three anges on the shaft. This cylindrical workpiece is made out of a solid piece of 4inch round stock steel. The center was bored out and the anges turned on the piece of stock via automated lathe. The reason the workpiece is made out of a piece of solid round stock instead of tubing stock is to insure a homogenous material workpiece. Most tubed round stock is welded which creates a seam. The material at the seam may have dierent properties due to the method the seam is made. Also the workpiece wall thickness is more uniform when it is turned and bored on the same automated lathe, than from original tubed stock. The three 0.25in by 3.5 diameter anges are the three locations of the grinding tests. Their length from the face of the spindle to the location of grinding varies from 0.375in, 2.675in, and 5.875in. Also added to these lengths is 2in from the center line to the face of the spindle along with 0.75in from each of the two adaptor plates. The static stiness for the end of the workpiece calculated by treating it as a cantilever beam using equation 4.1, is 497,000lb/in. These values were all determined theoretically as the next section discusses.

35

Figure 4.2. Probe holder machined from steel beam

Kstatic = 3EI/L3

(4.1)

4.3

Probe Holder

With such a long workpiece a new probe holder had to be machined in order to collect data on the second and third ange. This new holder is shown in Figure 4.2. This holder is machined from a steel channel C6 X 10.5 beam. The specications for this beam were found and used to evaluate the stiness to ensure its natural frequencies would not eect the results from the part data. (Information about air piston cap probe)

4.4

Theoretical Modeling

Using analytical methods, a theoretical model is made of the workpiece, adaptor plates, spindle and motor rotor as shown in Figure 4.6, to examine the eective stiness as a result of the change in length of the location of the input force. This model has two independent coordinate outputs that are used to determine the position of the center of mass (COM) at any instant in time. The two outputs to this system are the tilt, and the bounce, x.

36 This model is built and assembled using the program SOLIDWORKS. All the material in the
lb model is either a 416 or 1018 steel which has relatively the same density properties of 0.28 in3 .

Using this information and the program, the eective mass (m), inertia (J), and center of mass to which they are applied are found. The stiness of the system comes from the air lm of the bearing between the rotor and stator. It is estimated to behave as two springs, a linear and torsional, in series. The values for these springs are taken out of table 2.1 from the previous chapter. Now the two equations that describes the motion of the model shown in Figure 4.6 are: m + kr x kr l1 = f x and
2 J kr l1 x + (kr l1 + kt ) = f l3

(4.2)

(4.3)

In these equations l1 is from the center of mass to spring connection and l3 is from the center of mass to the input force located at ange 1. Two addition lengths l4 and l5 can also be dened and replace l3 as the locations of the force move to ange 2 and 3. The damping of the system is unknown but assumed to be proportional to the mass and stiness matrix through equation: [C] = [M ] + [K] (4.4)

where the constants and are found by making the damping constant equation to 0.05. This is just a rst guess and adjusted later so that the theoretical model better ts the actually results. Now that the equations of motion are complete with damping, the Laplace transform is taken with all initial conditions set to zero, meaning the system starts at rest with no initial osets. From here the equations in terms of outputs equal to the impedance matrix times the inputs are acquired. The current outputs to the system are located at the COM, but the output will be measured from the end of the workpiece where the accelerometer will be located for the modal analysis tests. Therefore using equation: Y (s) = X(s) + l3 (s) (4.5)

The outputs X and from the center of mass are related to the output Y at the end of the workpiece. With this, the transform function of the output Y divided by the input force F can be found as shown in equation: (J + l2 l3 )2 s2 + (l1 + l2 )(l1 + l3 )kr + kt Y (s) = 2 F (s) mJs4 + (kr J + (l1 kr + kt )m)s2 + kr kt (4.6)

Finally the theoretical FRFs for the model are found from the transform function by replacing s = i. The results are plotted in MATLAB and used to determine the dimension of the workpiece

37 shown in Figure 4.1.

4.5

Frequency Response Function

After the theoretical model is created and the workpiece is machined, the actual model could be constructed on the the testbed and tested. This part of the text gives a brief discussion of FRF along with details about the modal analysis test setup. A FRF is a three-dimensional quantity that is made up of amplitude vs. phase vs. frequency. Here this quantity is represented as a Bode plot which plots the information into two graphs of amplitude vs. frequency and phase vs. frequency. These plots show the characteristics of the systems measured response resulting from a known applied input. The systems response is measured by a Kistler accelerometer with a range from 50g and sensitivity of 101mV/g while the input is applied by a Kistler impact hammer with a force range from 0 to 500N and a sensitivity of 10mV/N. Both the input and output data to the system are recorded by a SIGLAB unit and processed in MATLAB. Figure 4.3 shows the user interface used with the SIGLAB unit. In this window the bandwidth is set to 2000Hz and a soft hammer tip is used for the best resolution with in this range. Five averages are taken per test. There are a variety of graphs that can be viewed in this window. The three most important graphs viewed are the coherence, input, and FRF graphs. The rst two are more important in determinating if the data is corrupt while the latter contain the information about the system. The coherence here is a frequency domain function computed to show the degree of linear, noise-free relationship between the systems input and output. Values of coherence satisfy the relationship: 0.0 COH 1.0 (4.7)

where a value of 0.0 indicates no causal relationship between an input and the output, and 1.0 indicates the existence of linear noise-free frequency response function between input and output. Basically it is used here to access FRFs. If the coherence graph was not close to a straight horizontal line at one as shown in Figure 4.3 the data was discarded. The second graph is the time domain input into the system, or the impulse. The impulse into the system is somewhat an art. The impulse needs to be a single hit with the hammer. This helps with a snap of the wrist. Sometimes double hits occurred which distorts the FRF and tell wrong information about the system. Also a pre trigger was used in order to obtain the entire impulse. With the rst two graphs veried, the FRFs were collected and exported to MATLAB. The results to the system for amplitude vs frequency are shown in Figure 4.4 and phase vs frequency are shown in Figure ??. For these graphs the accelerometer measuring the output is located at the very end of the workpiece for all three tests, which matches the output for the theoretical

38

Figure 4.3. User interface for SIGLAB unit. Currently shows coherence and input force

result as stated earlier. The input from the hammer moves from ange 1 for test 1 to ange 3 for test 3. The nal discussion for the FRF deals with main peak in the plots. This peak is the location of where resonance would occur. (Resonances are where for the same input the largest output occurs). For example, a person holding a rubber band with a weight attached to the end. At resonance the person is putting in a minimum vibration with their hand where the weight is traveling large distances. It is also known to occur at the natural frequency or frequencies of the system. For the case here it was estimated in MATLAB with the theoretical model that the system will have two natural frequencies which occur at about 275Hz and 550Hz. Looking at Figure 4.4 the rst peak is very noticable at about 275Hz, The second peak is not noticable. Figure 4.7 compares the theoretical to the actual measured FRF. As can be see they are a close t. The second peak is there but severely attenuated due to damping. Therefore the dynamics of the system are mostly controlled by the rst mode to the system which is a combination of both tilt and bounce.

4.6

Conclusion

In this section a theoretical model was developed in MATLAB starting with the dierential equations of motion. Once the model was completed the workpiece was machined based on the model. The important part here is that the workpiece was designed to only change the eective stiness, and the natural frequencies of the system do not change as shown in the previous gures. Also using the model and comparing it to actual measured data showed a close likeness. The machine eective stiness inuence on the plunge grinding process can now be examined with condence.

39

1.2

x 10

Measured Frequency Response of the Workpiece flange 1 flange 2 flange 3

0.8 Magnitude [m/N]

0.6

0.4

0.2

100

200

300

400 500 Frequency (Hz)

600

700

800

Figure 4.4. Amplitude vs Frequency, the most sti responds shown in blue and the least in

20

60 Phase (Degrees)

100

140

180

200

400 Frequency (Hz)

600

800

Figure 4.5.

40

7.72 4.02 Impact Force c 2.22


V K Kt Z V x
V

A Accelerometer

V y

Kr

2 2.19

8.35

Figure 4.6. Model of the combined workpiece, front/back adaptor plates, and the rotor of both the airbearing spindle and the motor. The stiness kr and kt are found from the table.

41

Frequency Response of the Workpiece 6 x 10


7

50

4 Magnitude [m/N] Phase (Degrees) 100

150

2 200

250

(a)
8 7 6 60 Magnitude [m/N] Phase (Degrees) 0 5 4 3 2 160 1 0 180 200 80 100 120 140 x 10
7

0 20 40

(b)
1.4 x 10
6

0 20 40 Theoretical Measured

1.2

1 60 Magnitude [m/N] 0.8 Phase (Degrees) 0 100 200 300 400 500 Frequency (Hz) 600 700 800 80 100 120 0.4 140 0.2 160 180

0.6

(c)

Figure 4.7. FRF plots of measured compared to model

Chapter

Tests and Results


5.1 Introduction

It is important to ensure that the results acquired from the testing are not only repeatable but also reproducible. Also it is important to identify the parameters to the system especially for the grinding process which has numerous variables which complicate the process. In this part the reader is rst taken through the testing parameters followed by the results from these parameters.

5.2

Testing Parameters

Table 5.1 list the parameters used during testing. Workpiece velocity, commanded plunge in feed rate, and the Length from the face to location of grind were varied in order to examine the eects they had on the grinding forces, workpiece form, and the actual depth of cut into the workpiece. The remaining parameters were set in order to make the test repeatable and reproducible. Each of the parameters eects on the grinding process are now briey discussed.
Table 5.1. System Parameters set by user

Parameter Workpiece Velocity Wheel Velocity Coolant Dressing Depth Dressing Feed rate Programmed Depth of Cut Spark out Commanded Plunge In feed Rate Length from face to grind

Value 150, 200, 300, & 600 rpm 1,910 rpm (5000 sf for 10 in wheel) Heavy mist of Blasocut BC40SW 0.01 mm 177.8 mm/min constant 0.01 mm 0s 0.0125, 0.025, & 0.05 mm/min 0.375, 2.675, & 5.875in

The wheel speed plays a very important part in the grinding process. Higher grinding wheel

43 speeds can result in higher production, ner surface nishes, longer wheel life, and lower grinding costs. The faster speeds can also cause the wheel to act harder and reduce the the grain grain depth of cut, where the slower speeds will cause the opposite eects. The most ecient speed for a grinding wheel is one that allows the abrasive grain to do the maximum amount of work without breaking away or dulling too rapidly [2].
Table 5.2. The Norton Company published in Grinding Technology

Peripheral Speed in Diameter of Wheel (in) 1 2 7 2728 3001 8 2387 2626 10 1910 2101

Revolutions per Minute 3 4 5 6 3274 3547 3820 4092 2865 3103 3342 3580 2292 2483 2674 2865

7 4366 3820 3056

8 4538 4058 3247

There are published tables of recommended wheel speeds that one may go to for suggestions on this parameter. Table 5.2 shows a segment of one of these tables used in this thesis. The nal wheel chosen for the tests was a 10in that could not exceed speeds of 2500rpm. The nal speed for the wheel chosen was based on pretest and observations which lead to a speed of 1910rpm. Thermal damage is one of the major process limitations in grinding where a workpiece experiences heating within a small grinding zone. At the grinding zone the peak temperatures generated by chip formation at individual cutting points are close to the melting temperature of the metal being ground. The overall grinding zone temperature if high enough can lead to typical thermal damages such as burning, possible rehardening,unfavorable residual tensile stresses, cracks, and distortions [19, 20]. During the initial pretest, burning was noticed on the anges of the workpiece, a brown with blue discoloration on the outer diameter of the anges during some of the experiments. This was eliminated through the use of the coolant, which lowers the grinding zone temperature due to lubrication. The coolant apparatus can be seen in Figure(need gure). It was angled to blast the grinding wheel which is believed to help pull the uid into the grinding zone. No optimization on the uid was used, it was simply run at its max ow rate. Besides lubrication, the coolant also serves two other important functions. It is used for bulk cooling of the workpiece which helps reduce thermal drift and ushes away swarf and dislodges wheel grits to prevent wheel loading and chatter. Dressing the wheel is another huge factor in the grinding process. Numerous papers have been written on the eects of dressing and trueing grinding wheels. Trueing restores the wheel geometry while dressing fractures the abrasive and opens the wheel surface by exposing grits from surrounding bond which can result in better chip ow and coolant ow into the grinding zone [21].Truing and dressing the wheel was done with a single point diamond dresser. Figure ?? shows the placement of the dressing tool. Base on the discussion by Steve Krar, the dressing for the setup was set to a depth of 0.0005in (0.01mm) for each pass and a feed rate of 7in/min (177.8mm/min) across the the grinding surface face. This will achieve a coarser dress of the wheel and will be completed in ve cycles [2]. This was programmed into the Bosch controller

44 using G-code. This program was run after each group of four consecutive tests at the same speed. This was done to ensure the grinding process was not eected by the conditions of the cutting wheel. Last but not least is the spark out parameter. This is the grinding of a workpiece at the end of a grind cycle without engaging any further down feed. The grinding forces are allowed to subside with time, ensuring a precision surface. The spark out has a signicant eect on the nal size, for this experiment the spark out was set to zero seconds which means the wheel was almost instantly backed away from the workpiece after each cut. (This was in order to show the eective of the part solely on structurally loop stiness.)

5.3

Testing Matrix

Table 5.3 shows the remaining parameters in a test matrix where the length is from the center of the spindle to the location of the grind and the numbers in each of the cells represent the test number. Each of the numbers is a dierent feed rate, e.g 1,2,3 is equal to a feed rate of 0.05, 0.025 ,0.0125 mm/min. To test the repeatability of each, tests were repeated four times consecutively and compared for a total of 144 tests.
Table 5.3. FORCE/PART TESTS

Workpiece Speed (RPM) Length(in) 150 200 300 600 2.22 1,2,3 ... ... ... 4.02 ... ... ... ... 7.72 ... ... ... 34,35,36

The basic test combined each cell into one test reducing the number of tests to 48. Final only the third test out of a set of four with the same parameters for in feed, work speed, and length were compared which brings the number of tests presented in this thesis to 36. Each of the 36 test contain two sets of data, the rst is the force and the second is for the part. One of the basic tests for part data is shown in Figure 5.1, here the table motion in red along with the material removal of the part is shown in blue. Attached to the table was the spindle with wheel. The table is programmed to make three cuts into the workpiece varying the feed rate into the part. After each cut the table moves away from the part until the wheel is no longer in contact and follows by pausing 20sec before continuing with another cut as can be seen by the at parts in the red line. After this is completed, three of the same tests are repeated before varying the workpiece speed to ensure repeatability as mentioned early.

5.4

Results

Figure 5.2 compares the normal grinding forces for the third test out of a data set of four, for a feed rate of 0.05mm/min. This shows no eects on the normal force magnitude as a result of

45
0.015 Part Table 150RPM

0.01

0.005

0.005 Depth of Cut (mm)

FEED RATE: 0.05 mm/min


0.01

0.015

FEED RATE: 0.025 mm/min


0.02

0.025

0.03

FEED RATE: 0.0125 mm/min

0.035 0 50 100 Time (sec) 150 200 250 300

Figure 5.1. Basic test: actual depth of cut (material removed from workpiece) vs. the set depth of cut (Table motion)

decreasing the eective spindle stiness. However, in terms of the eects of workpiece speeds, the grinding process becomes unstable at both 300 RPM and 600 RPM at the second and third ange as noted by the chattering eect. The forces increase exponentially, well above the average 6.3 Newton recorded from the other eight tests. This is in agreement with Hahn [9], discussed in the Chapter 1, which talks about chatter being more likely to occur at high work speeds. Figure 5.3 shows a close up of the rst three columns from the previous set of twelve normal forces. The fourth column was excluded due to the chattering eects making the graphs hard to read. As can be seen here between ange 1 and 2, the magnitude is unaected by the eective stiness. At ange 3 the eects of the eective stiness can be noticed. Although the force magnitude is not much less than the previous two it has a very dierent growth rate to steady state magnitude as seen by the slope. From this graph it can be shown there is a critical eective stiness, that until this value is reached will not have any eect on the grinding process. Figure 5.4 shows the next twelve data sets at a feed rate of 0.025 mm/min labeled on the same scale as the previous tests at 0.050mm/min. When comparing each of the twelve plots at a feed rate of 0.025mm/min to each other there appears to be no distinct increase in force as the stiness is increased. There is still chatter development in the right quadrant but not as severe as in the previous plots obtained at a feed rate of 0.050mm/min. In fact, the reduction in chatter

46

FEED RATE: 0.05mm/min Flange 1 150RPM 200RPM 300RPM 600RPM

FORCE (N)

Flange 2 40 Flange 3 0 40 0 25 TIME(sec)

Figure 5.2. Show the normal force data of twelve tests where the workpiece speed and length from spindle center line are varied. The feedrate is kept constant at 0.05mm/min
150RPM 8 6 4 2 0

200RPM 8 6 FORCE (N) 4 2 0 Flange1 Flange2 Flange3

300RPM 30 20 10 0 10 20

10 TIME (sec)

15

20

25

Figure 5.3. Compares the rst three columns of the force data from the previous gure at 150, 200, and 300 RPM at a feedrate of 0.050mm/min

47
FEED RATE: 0.025 mm/min Flange 1 150RPM 200RPM 300RPM 600RPM

FORCE (N)

Flange 2 40 Flange 3 0 40 0 50 TIME (sec)

Figure 5.4. Show the normal force data of twelve tests where the workpiece speed and length from spindle center line are varied. The feedrate is kept constant at 0.025mm/min

between the two data sets is the main eect of cutting the feed rate in half from 0.050mm/min to 0.025mm/min. Also for this group of tests at a feed rate of 0.025mm/min, a longer grinding session of 1.2 minutes was taken on the second ange at a work speed of 600 RPM. This data is shown in Figure 5.5. As can be seen in the rst plot the force pulses though out the entire cycle. This pulsing is also transferred to the workpiece as seen in the second plot which leads to visible lines or grooves appearing at regular intervals on the workpieces outer diameter. These groove interfere with the precision requirements of the part. Taking a closer look at these plots in Figure 5.6, we see that the maximum value of the normal force at 150 RPM has decreased by about 25percent for ange 1, 29percent for ange 2 and 33percent for ange 3. These are about the same percentage decreases for the other two speeds at 200 and 300 RPM for all three anges. In terms of decreasing stiness, the eects on the steady state normal forces are the same as in the earlier plots in having little eect until a critical stiness is reached. This is shown by how the forces to the rst and second anges are the same where the force data collected at the third ange has similar characteristics to the early plot in terms of the slope representing the growth of forces to stable conditions. However at higher speeds all three slopes grow at dierent rates. Finally the last data set of twelve force data sets are shown in Figure 5.7. These are the same in everything except that the feed rate for these is cut in half to 0.0125mm/min. These graphs show a substantial reduction in chatter formation by about sixty percent for the worst case scenario from a feed rate of 0.050mm/min to 0.0125mm/min. It is clear from the thirty six plots of force data that a reduction of the feed rate of the wheel into the workpiece will result

48

Flange 2 (600 RPM, 0.025mm/min) 40 30 20 10 0 10 20 30 40

Force (Newtons) Depth of Cut (mm)

0.015 0.01 0.005 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0 Part Table 0.5 1 Time (min) 1.5 2

Figure 5.5. Force and part data during chattering at a feed rate of 0.025mm/min, work speed 600 RPM, and at Flange2

49

150RPM 6 5 4 3 2 1 0

200RPM 6 5 FORCE (N) 4 3 2 1 0 Flange1 Flange2 Flange3

300RPM 20 10 0 10 20 0 5 10 15 20 25 TIME (sec) 30 35 40 45 50

Figure 5.6. Compares the rst three columns of the force data from the previous gure at 150, 200, and 300 RPM at a feed rate of 0.025mm/min

in a reduction of chatter. This is in agreement with Sexton [11] which states, The liability of chatter may be reduced by reducing the stock removal rates, but this is not normally an acceptable solution. This is one of the reason this work is important in order to nd other means of suppressing chatter, such as softer wheels or increasing the stiness. The increased stiness at ange 1 showed no signs of chatter for any of the work speeds as seen in all of the previous plots. Again the close up plots for a feed rate of 0.0125mm/min are shown in Figure 5.7. At the lower feed rate the maximum normal grinding force at 150 RPM has decreased by 47percent for ange 1, 51percent for ange 2 and 57percent at ange 3. These are about the same percentage decreases for the other two speeds at 200 and 300 RPM. Also to be noticed is the slope for all three dierent ange locations. The slopes or removal rates are starting to overlap and level o at the same maximum stable condition, which states the normal grind force will become independent of the eective stiness at a low enough feed rate. Figure 5.9 shows the steady state forces reached during the grinding process versus feed rate at a work speed of 150 RPMs. As can be seen in the Figure there is a linear increase of steady state forces to the feed rate of the plunge grind into the part. Figure 5.10 and 5.11 show the

50
FEED RATE: 0.0125 mm/min Flange 1 150RPM 200RPM 300RPM 600RPM

FORCE (N)

Flange 2 40 Flange 3 0 40 0 50 100 150 TIME (sec)

Figure 5.7. Show the normal force data of twelve tests where the workpiece speed and length from spindle center line are varied. The feedrate is kept constant at 0.0125mm/min
150RPM 4 3 2 1 0

200RPM 4 3 FORCE (N) 2 1 0 Flange1 Flange2 Flange3

300RPM 15 10 5 0 5 10 15 0 20 40 60 TIME (sec) 80 100 120

Figure 5.8. Compares the rst three columns of the force data from the previous gure at 150, 200, and 300 RPM at a feedrate of 0.0125mm/min

51

150RPM 7

5 Force (Newtons)

3 Flange1 Flange2 Flange3

1 0.01

0.015

0.02

0.025 0.03 0.035 Rate of Cut (mm/min)

0.04

0.045

0.05

Figure 5.9. Steady state grinding forces vs. feed rate at a workpiece speed of 150 RPMs

steady state forces reached during the grinding process versus feed rate at a work speed of 200 and 300 RPMs. These graphs show as the rate of cut increases the forces also increase in a linear fashion. Figure 5.12 shows the twelve data sets for the part outer diameter reduction in blue overlaid on the table motion in green. For this set the feed rate of the wheel into the part is 0.050 mm/min. The part and table are lined up using the initial tap in the beginning of the grinding cycle. As can be seen in the data the parts slope in mm/sec which can be thought of as the material removal rate is proportional to the feed rate of the table with the wheel into the part. The two data set are lined up and were taken at the same time with the same time scale. Looking at the plots we see that the removal rate lags the in feed. Figure 5.13 shows a close up of the rst three columns in the previous graph for the part. As can be seen here the lag of the removal rate increases with a decrease in stiness. Figure 5.14 shows the twelve data sets for a feed rate of 0.025 mm/min. Comparing these twelve to the previous the removal rate does not lag the in feed as great as before. Also for the

52

200RPM 7

5 Force (Newtons)

Flange1 Flange2 Flange3 0.015 0.02 0.025 0.03 0.035 Rate of Cut (mm/min) 0.04 0.045 0.05

1 0.01

Figure 5.10. Steady state grinding forces vs. feed rate at a workpiece speed of 200 RPMs

highest stiness there is no lag between the removal rate and in feed. In the plot at 300 RPM for ange1 the in feed lags the removal rate, which is just an error in the data, and should be ignored. Looking at the close up to the rst two columns for this part data in Figure 5.15 the removal rate is starting to over lap one on top of another regardless of the value for eective stiness. Finally Figure 5.14 shows the last twelve data sets for a feed rate of 0.0125 mm/min. Comparing these twelve to the previous the removal rate no longer lags the in feed regardless of eective stiness. This states that a longer spark out is no longer necessary in order to achieve proper part tolerances. Again, the plot at 300 RPM for ange1 the in feed lags the removal rate, and should be ignored. The close up plots to the rst two columns of data shown in Figure 5.17 again shows that longer spark out is un necessary since the removal rate no longer lags the table in feed as can be seen in that all the data overlaps one on top of another.

53

300RPM 7

5 Force (Newtons)

3 Flange1 Flange2 Flange3 0.015 0.02 0.025 0.03 0.035 Rate of Cut (mm/min) 0.04 0.045 0.05

1 0.01

Figure 5.11. Steady state grinding forces vs. feed rate at a workpiece speed of 300 RPMs

54

FEED RATE: 0.05 mm/min Flange 1 150RPM 200RPM 300RPM 600RPM

Depth of Cut (mm)

Flange 2

0.02 0.01
Flange 3

Part Table

0 0.01 0.01 0.02 0.02

10

20

30

40

TIME (sec)

Figure 5.12. Shows the part material removal in blue compared to the table in feed in green for twelve data sets at a feed rate of 0.050mm/min

0.015

FEED RATE: 0.05 mm/min


150RPM

0.01

0.005

0.005

Depth of Cut (mm)

0.01

0.015 0.015

Flange 1 Flange 2
200RPM

0.01

Flange 3

0.005

0.005

0.01

0.015 0 5 10 15 20 25 30 35

TIME (sec)

Figure 5.13.

55

FEED RATE: 0.025 mm/min Flange 1 150RPM 200RPM 300RPM 600RPM

Depth of Cut (mm)

Flange 2

0.02 0.01
Flange 3

Part Table

0 0.01 0.02 0 10 20 30 40
TIME (sec)

Figure 5.14. Shows the part material removal in blue compared to the table in feed in green for twelve data sets at a feed rate of 0.025mm/min

0.015

FEED RATE: 0.025 mm/min

0.01

150RPM

0.005

0.005

Depth of Cut (mm)

0.01

0.015 0.015

Flange 1 Flange 2 Flange 3


200RPM

0.01

0.005

0.005

0.01

0.015 0 5 10 15 20 25 30 35 40 45 50

TIME (sec)

Figure 5.15.

56

FEED RATE: 0.0125 mm/min Flange 1 150RPM 200RPM 300RPM 600RPM

Depth of Cut (mm)

Flange 2

0.02 0.01
Flange 3

Part Table

0 0.01 0.02 0 50 100


TIME (sec)

Figure 5.16. Shows the part material removal in blue compared to the table infeed in green for twelve data sets at a feed rate of 0.0125mm/min

0.015

FEED RATE: 0.0125 mm/min


0.01 150RPM

0.005

0.005

0.01

Depth of Cut (mm)

0.015

0.015

Flange 1 Flange 2
0.01

Flange 3

200RPM

0.005

0.005

0.01

0.015 0 20 40 60 80 100 120

TIME (sec)

Figure 5.17.

Chapter

Conclusion
In this work the relationship of eective machine stiness on the grinding forces, workpiece form, and the actual depth of cut into the workpiece for a cylindrical plunge grind process is investigated. An instrumented spindle with embedded capacitance probes measured the grinding forces while a special workpiece is used to vary the eective stiness of the structural grinding loop. A mathematical model was used to develop the spindle and workpiece setup which is compared to modal analysis experiments using MATLAB. The true beauties of the spindle and workpiece is shown as the ability to measure in-process grinding forces without adding other components such as a dynomometer. Added components such as a dynomometer is not innitely sti nor petite in size which would add unwanted side eects in the force measurements. The second beauty is varying the stiness with the workpiece but without changing the natural frequencies of the system. This allows the true dynamics of the grinding process to be test, while simplifying the testing procedures. It was shown that the steady state normal force measurements varied linearly with the feedrate. Also the cylindrical plunge grinding process is susceptible to chatter at combinations of work speed and eective spindle stiness. Chatter leads to an unstable process which lead to deterioration of the tolerances of a part as seen by the large displacements in the part data. However, at the spindle speed of 150 RPM no matter what the eective spindle stiness value did not result in chatter. Also at the stiest eective stiness no matter the velocity of the workpiece chatter did not occur. Using this information one can design a grinding machine to increase production time and save cost, but if the machine is already built one may be able to save defective parts due to chatter by slowing down the workpiece speed.

Appendix

Processing Program
A.1 MATLAB PROCESSING CODE

%====================================================================% % DATE CREATED:5/3/07 % % % %PROCESSING GRINDING DATA TED DEAKYNE DATE MODIFIED:5/29/07% % % % %

%====================================================================% % rawdata(:, 1) = capraw (Normal Force)

% rawdata(:, 2) = capraw2 (Tangential Force) % rawdata(:, 3) = capraw3 (Table) % rawdata(:, 4) = capraw4 (Part) %% clc clear all close all count150=0; count200=0; count300=0; count600=0; Path = T:\001_Thesis_Grinding_Data\001_RAW_DATA_GRINDING; folder = {001_Grinding_1flange, 002_Grinding_2flange, 003_Grinding_3flange}; for outer = 1:length(folder), ddirs = dir([Path \ folder{outer} \*.raw]) for inner = 1:length(ddirs);

59 load([Path \ folder{outer}, \, ddirs(inner).name], -mat) %======================= KNOWNS ================================== %WORK SPINDLE SPEED IN REVS/SEC work = work;%/60; %POINTS PER REVOLUTION after decimation ppr = 1024/8; %TOTAL NUMBER OF REVOLUTIONS PER TEST revs = length(rawdata)/ppr; %TOTAL NUMBER OF POINTS 1 to ppr %Npoints=1:ppr; %TOTAL NUMBER OF DATA POINTS PER TEST after decimation N=length(rawdata); %Time of experiment t = revs/work; %Total Time time = t*[0: N-1]/(N-1); %TOTAL NUMBER OF POINTS 1 to revs %Nrevs=1:revs; %CONVERSIONS FOR STIFFNESS EQUATION CFD=39.37; %in/m CFF=0.2248; %lb/N %radial distance from axis of rotation to cap probe in meters R = 1.75*0.0254; %m %================= AVERAGE ERROR FOR FIRST 20 REV ================ cap1err = mean(reshape(rawdata((1:20*ppr), 1),ppr,20)); cap2err = mean(reshape(rawdata((1:20*ppr), 2),ppr,20)); %Machine_err = mean(reshape(rawdata((1:20*ppr),3),ppr,20)); %part_err = mean(reshape(rawdata((1:20*ppr),4),ppr,20)); %====================== REMOVE SPINDLE FORM ERROR================= Xvoltraw=rawdata(:, 1)-reshape(cap1err * ones(1,revs), N,1); Yvoltraw=rawdata(:, 2)-reshape(cap2err * ones(1,revs), N,1); Machinevolts=rawdata(:, 3); Partvolts=rawdata(:, 4); clear cap1err cap2err rawdata %================= Calculate per-rev mean ======================== avg_disp_norm = mean(reshape(Xvoltraw,ppr,revs));

60 avg_disp_tan = mean(reshape(Yvoltraw,ppr,revs)); avg_partdata = mean(reshape(Partvolts,ppr,revs)); avg_ref_diam = mean(reshape(Machinevolts,ppr,revs)); ave_time = [0:revs-1]*time(ppr); %% ==================== CREATES CURVE ============================ %USED TO REMOVE THE DRIFT FROM THE FORCE DATA figure (2) plot (ave_time,avg_disp_norm) title Choose Seven Points on the Graph [x1,y1] = ginput(7);

curve1 = polyval(polyfit(x1,y1,2),time); curve3 = polyval(polyfit(x1,y1,2),ave_time);%testing %plot(ave_time,avg_disp_norm,ave_time,curve3,-.r,... ave_time,(avg_disp_norm-curve3)) %xlabel Time (sec) %ylabel Voltage %legend Before Curve After %pause %figure (2) %plot (ave_time,avg_disp_tan) % %[x2,y2] = ginput(7); % %curve2 = polyval(polyfit(x2,y2,2),time); %plot(time,curve2) figure (3) plot (ave_time,avg_disp_norm) title Choose Starting Location [x3,y3] = ginput(1); time=time-x3; %Machinevolts=Machinevolts-y3; %plot(time,curve2)

61 figure (3) plot (ave_time,avg_partdata) [x4,y4] = ginput(1); Partvolts=Partvolts-y4; %%%%%%%%%%%%%%%%%%%%%%%%%%%%% figure (3) plot (ave_time,avg_ref_diam) [x5,y5] = ginput(1); Machinevolts=Machinevolts-y5; %plot(time,curve2) %================================================================= Xvoltraw=Xvoltraw-curve1; % Yvoltraw=Yvoltraw-curve2; %Forces to different flanges if outer == 1 %Flange 1 L = 4.804 * 0.0254; num = 0; elseif outer ==2 %Flange 2 L=6.679 * 0.0254; num = 1; elseif outer ==3 %Flange 3 L=10.286 * 0.0254; num = 2; end %=============== Effective Spindle Stiffness ==================== %Radial Stiffness of Spindle kr=0.67*(1*10^6); kr=(kr/CFF)*CFD; %Tangential Stiffness of Spindle kt=4.00*(1*10^6); kt=kt/(CFD*CFF); Keq=(((110/150)^.4)*kt/(L*R))/10^6 %lb-in/radian %N-m/radian %lb/inch %N/m

62 clear L R kr kt CFD CFF curve1 curve2 curve3 x1 x2 x3 x4 y1 y2... y3 y4 avg_* ave_time %========================= FORCES =============================== %Factor found from calibrations factor=1.12; capsense=1; %microns/Volt Xvoltraw = Xvoltraw*capsense*factor*Keq; Yvoltraw = Yvoltraw*capsense*factor*Keq; Partvolts=Partvolts/10000; Machinevolts=Machinevolts/10000; %hold on %plot(Xvoltraw) %Combined Speeds of the same values if work <=170 count150=count150+1; answer=1; Xforces150(:,count150)=Xvoltraw(:,1); part150(:,count150)=Partvolts(:,1); machine150(:,count150)=Machinevolts(:,1); time150(:,count150)=time(:,1)*60; if count150 == (num*4)+3 %Forces figure(5) subplot(3,4,(4*num+1)) plot(time150(:,count150),Xforces150(:,count150)) xlim([-20 5]) ylim([-30 10]) figure(6) subplot(3,4,(4*num+1)) plot(time150(:,count150),Xforces150(:,count150)) xlim([25 75]) ylim([-30 10]) figure(7) subplot(3,4,(4*num+1)) plot(time150(:,count150),Xforces150(:,count150)) xlim([80 200]) ylim([-20 10]) %in %in %um*N/um=N %um*N/um=N

63 %Part vs Machine figure(8) subplot(3,4,(4*num+1)) plot(time150(:,count150),-part150(:,count150),... time150(:,count150),machine150(:,count150)) xlim([-27 5]) ylim([0 12E-4]) figure(9) subplot(3,4,(4*num+1)) plot(time150(:,count150),-part150(:,count150),... time150(:,count150),machine150(:,count150)) xlim([20 70]) ylim([3E-4 14E-4]) figure(10) subplot(3,4,(4*num+1)) plot(time150(:,count150),-part150(:,count150),... time150(:,count150),machine150(:,count150)) xlim([80 200]) ylim([0.8E-3 2E-3]) end elseif work > 170 && work <=250 count200=count200+1; answer=2; Xforces200(:,count200)=Xvoltraw(:,1); part200(:,count200)=Partvolts(:,1); machine200(:,count200)=Machinevolts(:,1); time200(:,count200)=time(:,1)*60; if count200 == (num*4)+3 figure(5) subplot(3,4,(4*num+2)) plot(time200(:,count200),Xforces200(:,count200)) xlim([-20 5]) ylim([-30 10]) figure(6) subplot(3,4,(4*num+2)) plot(time200(:,count200),Xforces200(:,count200)) xlim([25 75]) ylim([-30 10])

64 figure(7) subplot(3,4,(4*num+2)) plot(time200(:,count200),Xforces200(:,count200)) xlim([80 200]) ylim([-20 10]) %Part vs Machine figure(8) subplot(3,4,(4*num+2)) plot(time200(:,count200),-part200(:,count200),... time200(:,count200),machine200(:,count200)) xlim([-27 5]) ylim([0 12E-4]) figure(9) subplot(3,4,(4*num+2)) plot(time200(:,count200),-part200(:,count200),... time200(:,count200),machine200(:,count200)) xlim([20 70]) ylim([3E-4 14E-4]) figure(10) subplot(3,4,(4*num+2)) plot(time200(:,count200),-part200(:,count200),... time200(:,count200),machine200(:,count200)) xlim([80 200]) ylim([0.8E-3 2E-3]) end elseif work > 250 && work <=450 count300=count300+1; answer=3; Xforces300(:,count300)=Xvoltraw(:,1); part300(:,count300)=Partvolts(:,1); machine300(:,count300)=Machinevolts(:,1); time300(:,count300)=time(:,1)*60; if count300 == (num*4)+3 figure(5) subplot(3,4,(4*num+3)) plot(time300(:,count300),Xforces300(:,count300)) xlim([-20 5]) ylim([-30 10]) figure(6)

65 subplot(3,4,(4*num+3)) plot(time300(:,count300),Xforces300(:,count300)) xlim([25 75]) ylim([-30 10]) figure(7) subplot(3,4,(4*num+3)) plot(time300(:,count300),Xforces300(:,count300)) xlim([80 200]) ylim([-20 10]) %Part vs Machine figure(8) subplot(3,4,(4*num+3)) plot(time300(:,count300),-part300(:,count300),... time300(:,count300),machine300(:,count300)) xlim([-27 5]) ylim([0 12E-4]) figure(9) subplot(3,4,(4*num+3)) plot(time300(:,count300),-part300(:,count300),... time300(:,count300),machine300(:,count300)) xlim([20 70]) ylim([3E-4 14E-4]) figure(10) subplot(3,4,(4*num+3)) plot(time300(:,count300),-part300(:,count300),... time300(:,count300),machine300(:,count300)) xlim([80 200]) ylim([0.8E-3 2E-3]) end elseif work > 450 && work <=700 count600=count600+1; answer=4; Xforces600(:,count600)=Xvoltraw(:,1); part600(:,count600)=Partvolts(:,1); machine600(:,count600)=Machinevolts(:,1); time600(:,count600)=time(:,1)*60; if count600 == (num*4)+3 figure(5) subplot(3,4,(4*num+4))

66 plot(time600(:,count600),Xforces600(:,count600)) xlim([-20 5]) ylim([-10 5]) figure(6) subplot(3,4,(4*num+4)) plot(time600(:,count600),Xforces600(:,count600)) xlim([25 75]) ylim([-10 5]) figure(7) subplot(3,4,(4*num+4)) plot(time600(:,count600),Xforces600(:,count600)) xlim([80 200]) ylim([-20 10]) %Part vs Machine figure(8) subplot(3,4,(4*num+4)) plot(time600(:,count600),-part600(:,count600),... time600(:,count600),machine600(:,count600)) xlim([-27 5]) ylim([0 12E-4]) figure(9) subplot(3,4,(4*num+4)) plot(time600(:,count600),-part600(:,count600),... time600(:,count600),machine600(:,count600)) xlim([20 70]) ylim([3E-4 14E-4]) figure(10) subplot(3,4,(4*num+4)) plot(time600(:,count600),-part600(:,count600),... time600(:,count600),machine600(:,count600)) xlim([80 200]) ylim([0.8E-3 2E-3]) end end clear Xvoltraw L time

end end

67

save ThesisResults

Appendix

DOI MODELING
B.1
% % % % % % Date Created:5/23/07 % =================================================================% clc clear all close all %clf %in l1 = 2.19; l2 = [2.22,4.02,7.72]; l3 = 8.35; %lengths to center of mass of parts (in) lm=1.5; lr=5; lab=7.375; laf=8.075; lw=12.03; %weight of parts (lb) mm=2.07; mr=8.03; mab=2.68; maf=2.36; mw=5.68; %total weight (lb) W=mm+mr+mab+maf+mw; %Location to the center of mass (in) Ted Deakyne MODEL

MATLAB MODELING CODE

69 cm=(1/W)*(mm*lm+mr*lr+mab*lab+maf*laf+mw*lw) %Acceleration acc=386; %Mass Matrix m=(W)/acc; J=(318.45)/acc; M=1.2*[m 0;0 J] for j=1:3 %Stiffness Matrix kr=670000; kt=4000000; K=.75*[kr -kr*l1;-kr*l1 (kr*l1^2+kt)] %Natural frequencies and mode shape [phi,wn]=eig(K,M); wn=sqrt(diag(wn)); fn=wn/(2*pi) %Damping Matrix %Assuming zeta zeta1=0.087; zeta2=0.087; %Assuming Beta a = [1 wn(1)^2; 1 wn(2)^2]\[2*zeta1*wn(1); 2*zeta2*wn(2)] C=a(1)*M+a(2)*K; %Frequency array f = 0: 800; w = 2*pi*f; s = i*w; y = zeros(size(f)); for inc = 1: length(f), B = s(inc)^2 * M + s(inc) * C + K; X = inv(B) * [1; l2(j)]; y(inc) = X(1) + l3 * X(2); end %lb-sec^2/in %lb-sec^2-in %in/sec^2

70

%Convert to m/N y=y*(.2248/39.37); %shg figure(j) hold on plot(f,abs(y)) hold on figure(j+4) pha(:,j)=angle(y)*(180/pi); pha((469:677),1)=pha((469:677),1)-(357.8948+1.9359); if j==2 pha((514:566),2)=pha((514:566),2)-(359.064); end plot(f,pha(:,j)) hold on % Experimental data from Ted load FRF_combined.txt -ascii %===================================================================% %Sensitvity of Accelerometer SA = 0.010306; %V/(m/sec^2) %Sensitvity of Impact hammer SH = 0.01; %V/N % ratio is acceleration over force in terms of V/V factor = SH/SA %Converts to (m/sec^2)/N %===================================================================% %Convert Magnitude freq1(:,j) = FRF_combined(:, 3*j-2); frf1(:,j) = factor * (FRF_combined(:,3*j-1) + i*FRF_combined(:,3*j)); frf1(:,j)=frf1(:,j)./(-(2*pi*freq1(:,j)).^2); phase(:,j)=FRF_combined(:, j+9); figure(j) plot(freq1(17:801,j),abs(frf1(17:801,j)),r) title Frequency Response of the Workpiece xlabel Frequency (Hz) ylabel Magnitude [m/N] xlim ([0 800]) legend Theoretical Measured

71

figure(j+4) plot(freq1(17:801,j),phase(17:801,j),r) title Frequency Response of the Workpiece xlabel Frequency (Hz) ylabel Phase (Degrees) xlim ([0 800]) legend Theoretical Measured end figure(4) plot(freq1(17:801,:),abs(frf1(17:801,:))) legend flange 1 flange 2 flange 3 title Measured Frequency Response of the Workpiece xlabel Frequency (Hz) ylabel Magnitude [m/N] xlim ([50 800])

Bibliography

[1] Couey, J. A. (2004) Precision Grinding with Real-Time Force Feedback from an Instrumented Spindle, Masters thesis, The Pennsylvania State University. [2] Krar, F. S. (1995) Grinding Technology, second ed., Delmar Publishers Inc. [3] Snoeys, R. and D. Brown (1969) Dominating Parameters in Grinding Wheel and Workpiece Regenerative Chatter, 10th M.T.D.R., (5), pp. 325 348. [4] Knapp, B. R. (2002) On the Use of and Instrumented Spindle to Determine the Eects of Machine Stiness in Grinding Brittle Materials, Ph.D. thesis, The Pennsylvania State University. [5] LEE Hwa, F. (1990) Stiness Design Method of Grinding Machine, Bulletin of JSPE, 24(5), pp. 136 141. [6] Zhang, J., F. Wang, Z. Yang, and Zhu (1998) The eect of machine stiness on grinding of silicon nitride, International Journal of Machine Tools and Manufacture, 39(1-2), pp. 12631283. [7] Horiuchi, O. and T. Shibata (2007) Computer Simulations of Cylinderical Plunge Grinding: Inuence of work stiness on grinding accuracy, Key Engineering Materials, 329(5), pp. 5156. [8] Jenkins, H. and K. T. (1997) Dynamic Stiness Implications for Multiaxis Grinding System, Journal of Vibration and Control, 3(5), pp. 297313. [9] Hahn, R. (1954) On the Theory of Regenerative Chatter in Precision Grinding Operations, Transactions of the ASME, 76, pp. 593597. [10] Mohammed Alfares, A. E. (2000) Eect of grinding forces on the vibration of grinding machine spindle system, International Journal of Machine Tools and Manufacture, 40(5), pp. 2003 2030. [11] Sexton, J. S., T. D. Howes, and B. J. Stone (1982) Use of increased wheel exibility to improve chatter performance in grinding. Proceedings of the Institution of Mechanical Engineers (London), 196, pp. 291 300. [12] Baylis, R. J. and B. J. Stone (1983) The Build up and decay of vibration during grinding. CIRP Annals, 32(1), pp. 265 268.

73 [13] Moore, R. W. (1970) Foundations of Mechanical Accuracy, rst ed., The Moore Special Tool Company. [14] Grejda, R. D. (2002) Use and Cailbration of Ultraprecision Axes of Rotation with Nanometer Level Metrology, Ph.D. thesis, The Pennsylvania State University. [15] Olson, M. W., D. A. Arneson, M. J. Liebers, and E. R. Marsh (2007) Eects of Gas Composition on Asynchronous Error Motion in Externally Pressurized Spindles, . [16] Lion Precision (2004) RD20,RD22 Driver Module, Lion Precision. [17] National Instruments (1998) PCI-6110E/6111E User Manual, National Instruments Corporation. [18] Colding, B. and L. P. Tarasov (1970) How Stiness Aects Grinding Performance, Machinery, (5), pp. 3 8. [19] Nguygen, T. and L. Zhang (2004) Modelling of the mist formation in a segmented grinding wheel system, International Journal of Machine Tools and Manufacture, 45(5), pp. 21 28. [20] Malkin, S. (1984) Grinding of Metals: Theory and Application, American Society for Metals, 3(5), pp. 94 109. [21] Prusak, Z., J. A. Webster, and I. Marinescu (1997) Inuence of dressing parameters on grinding performance of CBN/Seeded Gel hybrid wheels in cylindrical grinding, INT. J. PROD. RES., 35(10), pp. 2899 2915.

You might also like