Design and Experimentation of A Roundness Tester
Design and Experimentation of A Roundness Tester
A thesis submitted in partial fulfillment of the requirements for a baccalaureate degree in Mechanical Engineering with honors in Mechanical Engineering
ABSTRACT The purpose of this thesis is to explain the issues and describe the mechanics involved in the development of a roundness tester. After gaining background knowledge of precision mechanism design and mechanical accuracy, a roundness tester was assembled using a Professional Instruments BLOCK-HEAD 4R high-precision air-bearing spindle as the axis of rotation. The roundness profiles are measured to sub-microinch repeatability using a Lion Precision DMT 10 capacitance gage and driver module. A Hewlett-Packard 35670A Dynamic Signal Analyzer captures the capacitance gage readings, and a Hewlett-Packard HEDS 9140-I00 3-channel optical encoder attached to the base of the BLOCK-HEAD 4R relays angular position. Furthermore, a MATLAB program converts the capacitance gage and encoder output into roundness data which can be plotted as a polar diagram and summarized in computed values of out-of-roundness. A typical polar plot is show in Figure 1, which is the circular profile of a diamond-turned aluminum cylinder.
Figure 1 Polar diagram of a diamond-turned aluminum log, values in microinches. Note that the trace shows roundness with respect to a true circle, but is not an absolute measurement of diameter. The repeatability, precision, and accuracy of the machine design was considered under strict discretion since the test parts are round to 2 millionths of an inch. For comparative purposes, commercial roundness tester specifications were researched, and a demonstration was requested of a Federal Formscan 6100 roundness tester. Many issues and properties affecting mechanical accuracy must be considered when manufacturing a commercial roundness tester. This is an advantage for the MDRL, with knowledge of precision mechanics and dynamics, a roundness tester could be built with repeatability and accuracy that exceeds its commercial
counterparts. In contrast, the efficiency and user-friendliness that is designed into commercial products is considerably lacking in the MDRL roundness tester. However, this project focused on the development of a roundness tester leaving automation as a future consideration.
Chapter 1: INTRODUCTION Design Process The design and implementation of the MDRL roundness tester includes many factors of precision mechanism design, structural mechanics and mechanical accuracy. Since the out-ofroundness value should be accurate and repeatable to less than one millionth of an inch, the design process requires that every issue contributing to the error budget be analyzed and addressed. These issues include: thermal effects, structural integrity, capacitance gage and optical encoder implementation [1], proper data acquisition, and processing in the computer program. Noting these effects and minimizing error by design or compensation contributes to the accuracy of the final form trace and out-of-roundness value. As will be discussed further, the most important part of the roundness tester is the axis of rotation [1]. The high accuracy and stiffness associated with the BLOCK-HEAD 4R air-bearing spindle shown in Figure 1.1 provides accurate rotation with minimal synchronous and asynchronous error motion. The error motions are less than one microinch in both the radial and axial directions and less than 0.1 microradians tilt error [3]. The stiffness of the BLOCK-HEAD 4R air-bearing spindle is: 0.67 lbs/microinch radial, 2.0 lbs/microinch axial, and 4.0 lb-ins/microradian tilt [3].
Figure 1.1 A BLOCK-HEAD 4R air-bearing spindle [3]. The other critical component of the roundness tester is the measuring instrument, in this case a Lion Precision DMT 10 capacitance gage and driver module. The capacitance gage is extremely sensitive and measures displacement with resolution finer than a millionth of an inch [4]. The orientation of the capacitance gage must be in line with the center of the part being measured and perpendicular to the parts vertical axis of rotation [5]. Lion Precision also offers a Contact Adapter Probe (the CAP) which uses an air-bearing plunger with a capacitance probe to obtain an extremely high resolution stylus measurement [4]. This would be a future consideration to further increase the accuracy and repeatability of the roundness measurement.
Design Parameters and Issues Certain design parameters were driven by the availability of both a BLOCK-HEAD 4R airbearing spindle and a Lion Precision DMT 10 capacitance gage for use with the roundness tester. The following is a brief list of additional design considerations. The measuring instrument fixture holding the capacitance gage would be rigidly mounted to the base. The instrument fixture would be adjustable to measure parts of varying diameter and height. The optical encoder and capacitance gage output would be captured on a signal analyzer to be further manipulated off-line on a computer running a single program. The computer program would calculate an out-of-roundness value for the measurement and a create a polar plot of the workpiece roundness. Initially the roundness tester was designed to measure a high-precision, diamond-turned aluminum cylinder, but is adjustable to measure additional parts. These parts include a lapped steel ball and a precision-ground steel disc, which are considered typical components requiring a roundness measurement. The roundness tester described in this report is considered the basic setup needed for an accurate, repeatable measurement. It allows for future components to be designed and replace the current setup to automate the measurement process and increase efficiency [2]. A future consideration would be motorization, perhaps in the form of a belt or friction-drive. Another possible option to automate the process would be to implement data acquisition in the computer so the roundness tester could perform nearly real-time measurements. Expanding on those ideas, a program could be outlined to start the test from a user input, run the test using the drive motor at a specified RPM, and output the polar diagram and roundness values on the monitor following the test. Such is the case for all commercial roundness testers, with variations in the system components and user interface capabilities between manufacturers. In summary, this project sacrifices the user-interface for extreme accuracy and reliability.
Chapter 2: ROUNDNESS TESTING BACKGROUND Roundness Measuring Techniques This chapter provides a brief background of roundness as a function and measurement. Roundness of a part is important for a wide variety of applications whether it be precision ball bearings conforming to a race or a cylindrical shaft being fitted into a bore. Roundness of a part depends on the way it was manufactured, or the method of generation. The most common error associated with roundness of parts is lobing [1]. Lobes can be difficult to detect using common metrology tools such as micrometers, as shown in Figure 2.1. In the figure, an odd number of lobes is completely masked in the diameter measurement and will appear to be within given roundness tolerances. Contrarily, even-lobed parts will be exposed by a diameter check, as displayed in Figure 2.2. However, since the number of lobes of a part is unknown, the diametral roundness measurement is considered unreliable. Further roundness measuring techniques have been developed to discover the nature of lobing.
The following techniques for measuring roundness are discussed in The Foundations of Mechanical Accuracy. Each has advantages and disadvantages, but the most accurate roundness test given is use of an accurate spindle, the method that is associated with the MDRL roundness tester. The first method explained is the diametral method, previously mentioned as unreliable for detecting out-of-roundness of parts with odd-numbered lobes. A diametral method consists of measuring multiple diameters of a part through a common center point. Physically the measurement can be thought of as rotating a part through opposed confining gages. When a quick roundness check needs to be done on a mass-produced part, the circumferential confining gage may be used. This is a highly accurate master, shown in Figure
2.3, which can be placed around the part and rotated. Any stickiness or failure to rotate would imply a part that has out-of-roundness. Roundness of the master is crucial to this measurement because the clearance between the master and the part determines whether or not the part is acceptable. The drawback to this approach is that only a single size of manufactured part can be measured with a given master. Again, this is mostly used for a quick roundness check on the production floor.
Figure 2.4 Potential sources of error with rotating on centers [after Moore, 1].
The next roundness measuring technique discussed in Foundations of Mechanical Accuracy is rotating on centers. Rotating parts on centers contains a large opportunity for error and is not considered a universal method. The listed potential sources of inaccuracy are: alignment of centers, surface condition of center holes, angle of centers, runout or total indicated reading (TIR) of the piece, and roundness of centers. Figure 2.4 shows potential error sources of rotating parts on centers. A V-block measurement is only recommended for revealing an odd number of lobes. This measurement consists of a gage set in opposition to a highly accurate V-block in which a part is rotated. Certain angles of the V-block will exaggerate or mask the number of lobes of a part. If the number of lobes is unknown, a V-block may be used with an adjustable angle, shown in Figure 2.5. Using any angle has limitations for roundness measurements, though. For instance, the 60 V-block shown in Figure 2.6 will give accurate measurements of parts with 5 or 7 lobes, but magnify the error of a 3-lobed part and contain partial roundness error when measuring a part with randomly spaced lobes. Other angles used for V-blocks are shown in Figures 2.7 and 2.8 on the following page, with the number of lobes they most accurately measure respectively. Vblocks have difficulty with roundness measurements of any even number of lobes and give a smaller than actual value for out-of-roundness. Therefore, a V-block measurement would be
considered accurate for specified odd numbers of lobes using the corresponding angle of the Vblock, but is still not a universal method for measuring a variety of parts.
For a more accurate roundness measurement of a part in which lobing and other characteristics are unknown, a three-point probe can be used. Each probe is spaced 120 apart on a level surface. A 3-jaw inside micrometer is used for the measurement of roundness. Parts with three-lobes can be accurately measured, but limitations exist similar to the 60 V-block. For a production measurement, though, the three-point probe is considered acceptable. Finally, The Foundations of Mechanical Accuracy recognizes a type of measurement that is more universal and accurate than the previous techniques, the use of an accurate spindle. The two configurations for using accurate spindles are: holding the part fixed while rotating the indicator, Figure 2.9, or holding the indicator fixed while the part is rotated on a spindle, Figure 2.10. The latter is used for the MDRL roundness tester utilizing the high-precision BLOCK-HEAD
4R air-bearing as the accurate spindle. An adjustable stand provides the flexibility for different measuring functions such as: concentricity, tapering, and camming (circular flatness) in addition to roundness. The rotating table has a slight advantage to the overhead spindle in that it can measure squareness of the part relative to the datum of the table [1]. This is the method used in the American National Standard Measurement of Out-Of-Roundness B89.3.1, 1997 for obtaining an out-of-roundness value.
Figure 2.9 Accurate spindle, rotating table [after Moore, 1]. Measuring Out-Of-Roundness
Now that the technique of using the accurate spindle with rotating base has been recognized as the most accurate and reliable for measuring roundness, the actual method for attainment of an out-of-roundness value is discussed. Basic Concept The American National Standard Measurement of Out-Of-Roundness B89.3.1, 1997 states, The objective of roundness measurement is to evaluate form errors of components as opposed to feature size. Whenever one or more surfaces of a cylinder, cone, or sphere are required to have roundness of a high order, cross-sections of the feature must be measured to assure that the profile falls within the required form tolerance. Roundness is how much a part deviates from a true circle.
In order to compare roundness, five methods have been found in two publications, the American National Standard and American Machinist, How to Measure Roundness, December 1993, and are considered standard measurements: 1. Minimum Radial Separations (MRS) or Minimum Zone Circles (MZC) 2. Maximum Inscribed Circle (MIC) 3. Minimum Circumscribed Circle (MCC) 4. Least Squares Circle (LSC) 5. Departure From True Circle (DFTC) These methods have slight variations between one another, but all rely on the principle that out-of-roundness is the deviation from a true circle. The first three methods do not provide identical results, but each may be used without the benefit of a computer data acquisition system. In fact, older systems allowed the user to quickly compute roundness using a compass and a thermal paper plot of the profile. However, with increasing technology came the LSC and DFTC method which are more adaptable for roundness measurements involving a computer [6]. Assessment Methods 1. Minimum Radial Separation (MRS)
OUT-OFROUNDNESS
Figure 2.1 Minimum Radial Separation method. The roundness obtained by this method consists of the minimum radial separation of two concentric circles that still contain the entire polar trace [5]. The MRS method will result in the lowest possible roundness value for the part. The Minimum Radial Separation circles are acquired by tracing an inside circle using two maximum inward lying points and then tracing a concentric but circumscribing circle using a maximum outward lying point. The method could also be performed using two outward lying points and a single inward lying point. Both methods will result in similar values [5]. For this criterion to be valid, the eccentricity of the part must be either minimized prior to testing or removed after testing [5].
10
OUT-OFROUNDNESS
Figure 2.2 Least Squares Circle method. A reference circle is drawn such that the sum of the squares of the radial deviations between the circle and the trace is a minimum. The corresponding roundness value is the sum of maximum outward and maximum inward deviations from this reference circle divided by a chart amplification factor, if needed [5]. This is the method used for the Federal Formscan 6100 tests described in Chapter 3. 3. Maximum Inscribed Circle
OUT-OFROUNDNESS
Figure 2.3 Maximum Inscribed Circle method. The largest circle that can fit inside the profile determines the center of the profile. The circle is described by three points. The maximum outward deviation from this circle denotes the out-ofroundness value for this method [5]. 4. Minimum Circumscribed Circle
11
OUT-OFROUNDNESS
Figure 2.4 Minimum Circumscribe Circle method. Similar to the MIC method, the center of the cross-section is described by three points which define the smallest circle that contains the entire data set. The maximum inward deviation from this circle would be considered the out-of-roundness for the part [5]. 5. Departure From True Circle
OUT-OFROUNDNESS
Figure 2.5 Departure From True Circle method. The DFTC method also easily formulated in a computer algorithm. In American Machinist, How to Measure Roundness, December 1993, a window of specified angular width rotates around an entire revolution of data. The window that contains the largest difference of peak-tovalley becomes the roundness measurement for the part. When displaying the roundness measurement, the window width would be displayed with the DFTC notation. The window method is most often used for ball bearing roundness measurements where the window width is determined by the bearing contact area [6]. This value is displayed with the DFTC for the entire plot.
12
Method Used For Roundness Testing The technique used for this project is the Least Squares Circle to calculate a value of out-ofroundness. The capacitance readings are converted to values of microinches and the out-ofroundness value is the sum of the maximum and minimum departure from zero. Roundness Measuring Rules to Abide By Measuring roundness is considered to be one of the most demanding and constraining measurements in the area of form metrology. A tester must adhere to many subtleties and skills that would support the accurateness of the measurement. A collection of rules for high quality roundness testing has been formulated using excerpts from Quality Articles, The Dos and Donts of Roundness Measurement, November 1996 and abided by for tests performed on this roundness tester. Centering Prior to a test, the part must be centered to within 10 times the expected out-of roundness tolerance for the part. For this project, an experiment was performed exploring the relationship between roundness values and total indicated reading (TIR). Filtering The smooth rotation of the BLOCK-HEAD 4R eliminated any vibration that would require a high-pass or band-pass filter. However, once-around centering errors must be eliminated using software algorithms. Clamping of Part to Spindle When possible, centering the part with slight clamping pressure using three bolts and the threaded holes in the BLOCK-HEAD 4R is utilized. Then the bolts are tightened to hold the part in the centered position. This procedure mimics the part fixturing during manufacturing and usage. If a part is heavy enough, slippage between it and the spindle would not occur and clamping was unnecessary. Also since a non-contact indicator was used, radial load was insignificant. Reference Circle A reference circle must be calculated using one of the five techniques mentioned previously to give a complete roundness value. Instrument Alignment Abbe error occurs when datum axes are not aligned, affecting the instrument reading. Measuring Points A sampling frequency must be high enough to avoid aliasing when measuring the part. This corresponds to the number of data points taken during a revolution of the workpiece. The more data points taken, the less aliasing will occur. Rotational Speed Angular speed of the spindle is related to the sampling frequency in that if the speed is too great, not enough data points will be taken during a single revolution to accurately define the form of the cross-section and therefore the out-ofroundness value.
13
Environmental Influences Environmental influences are more significant on a production floor than in a metrology lab. Temperature fluctuations can affect measurements drastically on the sub-microinch level. The sensitivity to temperature fluctuations of the steel structure requires that any influence of a possible heat source be minimized. This includes movements of the user performing the tests.
Vibration The low error tolerance required for accurate measurements means that any slight vibrations will ruin the test. Vibrations such as outside noise, operating machines, closing doors, and even user movement in the vicinity of the tester may all have an effect on the measurement.
TIR Can be compared to a roundness value, but not substituted for it. TIR is the superposition of the out-of-roundness and the part eccentricity on the spindle. Concentricity versus Coaxiality Concentricity is the measurement of two circles on a common plane. An example would be measuring concentricity between an inside and outside diameter. Coaxiality concerns two circular measurements that lie on a common axis. For instance, a cylindrical part might require a coaxiality measurement between a circle near the base and one at the top.
Correlation Between Devices This rule is related to roundness testing in industry and describes the understanding that similar devices performing the same test will not always agree perfectly. For proper testing using different devices, standards should be imposed on certain variables such as filters, reference circles, sampling frequency, and stylus dimension to minimize variations.
Deviation From True Circle The measurement of out-of-roundness is a deviation from a true circle, not a measurement involving the workpiece diameter. The polar diagram of the circular form is not related to the diameter dimensions of the part.
Surface Quality and Cleanliness The surface condition of the part, not just the surface finish, is vital to an accurate roundness measurement. Even microscopic granules of dust can bias a measurement. Wiping the surface of fragile, high-precision parts with a clean rag is usually unacceptable. The surface may be dripped dry with an evaporating cleanser such as acetone and must not be touched prior to testing.
14
Chapter 3: COMMERCIAL ROUNDNESS TESTER COMPARISON Federal Formscan Model 6100 Demonstration
Figure 3.1 The Federal Formscan 6100 [after Federal, 8]. Background The functions and capabilities of a Formscan 6100 roundness tester were demonstrated by Lori Massarone of Federal Products to provide a more thorough knowledge of commercially available products. Ms. Massarone showed how to perform basic roundness measurements on the same aluminum log used in the development of the MDRL roundness tester including both a roundness test and a coaxiality test on the cylinder. The Formscan 6100 demonstrated retails for $13,850 [8] and is considered an entry-level roundness tester. Since the part under consideration was diamond turned, and known to be round to a few microinches, the part centering and table tilt were expected to have a significant impact on the roundness measurement. To study the effect of part centering, a series of tests was carried out to compare the total indicated reading (TIR) to roundness measured by the Formscan. During Ms. Massarone's demonstration, she first centered the air bearing table to 0.002 TIR. She explained that the table should be centered to within 0.005. Clearly 0.002 TIR would not give the most precise reading of the analyzed part, but for a quick roundness demonstration it sufficed. Tilt in the air bearing table was directly affected by the tables centering, in that table is first centered and then the tilt is removed. Any tilt left in the table would later affect coaxiality measurements. To measure the roundness of the part the aluminum log was placed on the table and then centered using the table's built-in adjustment. Because the table tilt was previously removed, the table could then be moved out of center while maintaining levelness. Once the part is centered, it could be measured. Any defects or burrs on the table would inherently cause leveling problems, however.
15
Test Procedure After the initial demonstration, the Formscan was used by MDRL researchers to more accurately center and level the table and to center the part. This was done to maximize the Formscan 6100s ability to measure the roundness of a high-precision (5in) part. This procedure included minimizing table centering and leveling error using the finest scale available (0.0001 per div). Once the table was centered and leveled to the desired precision, it was cleaned and stoned for burrs. This allowed for improved surface to surface contact between the table and the part. The base of the aluminum log was also stoned. The part was centered by tapping it into position, instead of adjusting the tables centering knobs. Using the diamond turned log, printouts were made of the instrument's computed roundness versus the measured TIR, which include the effect of centering errors. The tests were initially performed on a perfectly-centered test part and then rerun as the part was progressively tapped further and further away from true center. The results of the tests are polar diagrams showing how the computed roundness varies with TIR (part centering). The results also hint at the repeatability of the instrument. The following pages include results of tests performed on the aluminum log (expected roundness 5 to 10 in) as well as a lapped steel ball (expected roundness = 8in). Results The results of the tests performed on the diamond-turned aluminum log and lapped steel ball are shown in Table 3.1 and Table 3.2, respectively, and in Figure 3.2 and Figure 3.3.
Roundness in 6 5 5 8 3 11 20 20
16
Roundness in 11 11 10 11 13 13
For the repeatability analysis, Figure 3.4 shows a scanned image of some of the printouts obtained from the Formscan 6100. These represent the tests performed on the lapped steel ball at several values of TIR. The printouts are useful for determining out-of-roundness, but the results do not inspire much confidence shown from the varying roundness measurements.
17
Figure 3.4 Scanned image of Federal Formscan 6100 printouts. Note the lack of repeatability between measurements. TIR is in microinches. Discussion The measurement of the Formscan air bearing table roundness does not reflect on the accuracy of the instrument, but it does provide insight into the quality of the parts used in the instrument. The computed roundness was 150 in. The measurements taken on the aluminum log clearly show that the computed roundness gets worse as the TIR is increased (part centering gets worse). The results suggest that if a user did not tap the part in to within a thousandth of an inch, the computed roundness would be overestimated.
18
The measurements taken on the lapped steel ball show a similar trend and the roundness measurements suggest that the part should be tapped into center within perhaps 0.0005 inches. The repeatability of the steel ball measurements was also assessed over the several measurements and found to be fairly poor for an artifact of this roundness. Commercial Roundness Testers: Features and Specifications Information was gathered on three commercially produced roundness testers: Precitech FMS 8150 or Taylor/Hobson TalyRond 30 TS Mitutoyo Roundtest TA-300/400 Federal Formscan 6100.
Each model was considered a moderate-level roundness tester in terms of measuring characteristics and features. These models should provide the accuracy and tolerance values that would be expected from the MDRL roundness tester. The following sections will include a brief description of each of the models, list certain applicable features, and provide specifications and tolerances.
19
Precitech FMS 8150 The FMS 8150 and TalyRond 30 TS have the exact same features and specifications. The FMS 8150 features and specifications were provided by Precitech Precision: FMS 8150 Circular Form Measurement System, 1999. They are described as ideal for the production floor; heavyduty and easy to use. Control of the roundness tester is performed on a touchscreen which is provided with the spindle and indicator stand. A 6 or 8 air-bearing spindle is available equipment for this model. Also included is a thermal printer that will plot polar diagrams and tabulate specified values. The servo drive is capable of two speeds (6 and 10 rpm) and enables exact measurements of out-of-roundness, eccentricity, flatness, perpendicularity, and more [9]. Specifications denote that the spindle is capable of measuring a part that is 16 diameter by 17 height and up to 200 lbs. The FMS 8150 is shown in Figure 3.5, below.
Figure 3.5 The Precitech FMS 8150 [after Precitech, 9]. The measurement range is a minimum of 50 in and a maximum of 0.010 in. The corresponding chart scales per division are 5 in and 0.001 in. The results of measurements can be based on the following reference circles: LSC, MRS, MIC, and MCC (see Chapter 2). The operating pressure for the air-bearing spindle is 30 to 60 psig, and consumes air at a rate of 1.2 scfm. Price for this model could not be obtained from the information gathered.
20
Mitutoyo Roundtest RA-114/116 Mitutoyo manufactures several models of roundness testers but the features and tolerances of the RA-114/116 were chosen because they describe an entry-level tester that is durable enough to be used on the production floor. Information on the RA-114/116 was gathered from Mitutoyo Measuring Instruments Catalog No. 302, Roundtest RA-114/116, 1997. Similar to the FMS 8150 and TalyRond 30 TS, the RA-114/116 uses a high precision air-bearing spindle as the rotating table. However, the computer provided is a personal computer with no touchscreen controls. The roundness tester also provides a printer. The table can accommodate a part of 17.3 diameter, 11.0 height, and weighing up to 44 lbs. The rotating accuracy of the table is given to be 2.8 0.6H in., where H is the height of the part being measured. The table rotates at a single speed (6 RPM), and it has manual leveling and centering capabilities. The reference circles available for testing are: LSC, MZC (MRS), MIC, and MCC. The air-pressure needed is approximately 57 psi and consumes air at 1.59 cfm (normal state) [10]. Figure 5.6 shows the RA-114 model. The RA-116 model has a separate computer control with a monitor, keyboard, and mouse. The specifications remain the same for both, however.
Figure 3.6 Mitutoyo Roundtest RA-114 [after Mitutoyo, 10]. Federal Formscan 6100 The Formscan 6100 was the model used in the demonstration previously discussed in this chapter. All the information provided was obtained from The Federal Dimensional Gaging Products Catalog, 1998. An air-bearing spindle of 4-inch diameter is used as the table and has the ability to be manually or automatically centered and leveled. The roundness tester has a touchscreen computer control unit and thermal printer. The Formscan 6100 is capable of measuring a part of a maximum diameter of 7, a maximum height of 9, and a maximum workpiece weight of 50 lbs. The stated accuracy of the roundness tester is given to be 1 in., which is the most aggressive claim of the commercial testers researched. The table has the ability to rotate at either 6 RPM or 5 RPM. Reference circles for measurements include LSC, MRS, MIC, and MCC,
21
as seen for the other models. Finally, the air-bearing source requirements are 80-125 psig and the air is consumed at 2 scfm. The Federal Formscan 6100 is base priced at $13,850.
22
Chapter 4: DEVELOPMENT OF THE ROUNDNESS TESTER The roundness measurement requires analysis of the capacitance gage and optical encoder signals, transfer of these signals into readable data, and manipulation of this data to output a cross-section polar diagram and roundness values. Discussion of the development of these procedures is sectioned into three parts: Data Acquisition/Signal Analysis, Computer Program Components, and Testing Procedure. Data Acquisition/Signal Analysis Three separate channels are needed for the signal analysis of the roundness tester: optical encoder index pulse, optical encoder square-wave signal, and capacitance gage reading. The DC offset of each signal is removed to maximize the resolution of the analog to digital conversation used to sample the data. The TTL index pulse and square wave are viewed within a range of 5 volts. The voltage range on the A/D board must also be set so that the full 16 bit resolution of the converters get the best possible sampling of the capacitance gage signal. This voltage range is adjusted depending on the total indicated reading (TIR) of the part being measured and varies between 40 and 500 millivolts for the experiments performed. For the capacitance gage signal, a 1 volt/mil Lion Precision driver module is used. An LED display on the driver also shows where the capacitance gage location is in terms of its displacement range. The voltage range of the analyzer has to be adjusted so the signals are not clipped, but close enough so acceptable measurements are taken. The Hewlett Packard 35670A Dynamic Signal Analyzer has auto-ranging capabilities as well as an Overload LED which indicates when the signal is being clipped. Voltage clipping occurs when a signals voltage amplitude exceeds the voltage range being measured by the signal analyzer. The clipped voltage is instead displayed as an end value of the voltage range. This is a special issue when considering automation of the roundness tester since the voltage range will have to be automatically set at the lowest possible value to get maximum resolution of analog to digital conversion. Computer Program Components A MATLAB program was written to retrieve the data and ultimately output a polar diagram and corresponding roundness value. However, certain analysis is applied to process the data and extract the needed information. Once the data is downloaded from a time-capture file, it is in matrix form and each column vector is assigned as a variable. The index pulse signal, square wave, and capacitance gage reading are columns 2, 3, and 4 respectively. Column 1 is a time vector assigned to any trace recorded on the signal analyzer. This vector is not needed since the capacitance gage readings are coupled with the square wave, eliminating the effect of the spindle speed variation during the
23
test. Any remaining DC offset is removed by subtracting the mean of both the square wave and capacitance signal and subtracting the median of the pulse train from the entire vectors for the respective data. The median is subtracted from the pulse wave because the majority of the pulse wave is approximately zero volts and only deviates at index pulse markings. Therefore, subtracting the median more precisely zeroes the pulse wave. The spindle is not motorized and spins at slowly decreasing speed throughout the tests. The effect of the decreasing speed is easily removed and the motorless drive is extremely smooth. A nominal angular spindle speed of 0.3 rev/s (20 RPM) is used in most tests. Since the data is acquired via a time capture, an incomplete revolution occurs before the first pulse wave marker and following the final pulse wave marker. Therefore, the program analyzes each increment of the index pulse to find the complete revolutions and eliminate the initial and final incomplete revolutions. In other words, the number of useful revolutions is always one less than the number of index pulse markers found. The capacitance gage and square wave vectors are then systematically cropped at the incremented values of the first and last pulses. For the square wave, an algorithm was formulated that extracts the two data points above and below a specified voltage level on the upside of a single transition (one of the 512 counts). The level in this case is zero since the square waves DC offset has been removed and the mean of the data is now known to be zero. These two data increments are paired with corresponding increments on the capacitance gage reading, shown on Figure 4.1 with arrows. The values of the capacitance gage reading are then extracted. Finally, an interpolation is performed using both sets of data so the capacitance gage value precisely corresponds to the zero point of the square wave, marked on Figure 4.1 as dots. The result of extracting these points is a 512 point-perrevolution displacement vector.
24
Figure 5.1 This diagram shows how interpolation is used to extract a capacitance gage reading using the zero level of the square wave. First the points above and below the zero level on the square wave are matched up with those on the capacitance gage reading (arrows). Then the final capacitance gage value (circle) is found by interpolation using the zero point on the square wave (square). For this roundness tester, the part is first centered using a dial indicator accurate to 50 millionths, then centered using capacitance gage readings. Without removal of the eccentricity, an improperly centered part will have a polar plot in the form of a limacon. This eccentricity shows up in the Fourier Transform of the signal as the fundamental. After removing the eccentricity, synchronous and asynchronous error is calculated for agreement between revolutions. The synchronous error is the average value of the capacitance gage over duration of the test. This error is plotted on the same polar diagram as the test data to show agreement and repeatability between revolutions. If minimal error exists between each revolution, then future tests only have to measure a single revolution for a reliable roundness value and polar diagram. Next, asynchronous error, the amount the capacitance gage values deviate from the mean of the entire data, is calculated. Again, this error reflects on the reliability of the measurement.
25
Testing Procedure Proper testing procedure will result in the most accurate and reliable measurement attainable. The rules of measurement procedure listed in Chapter 2 are recognizable in the testing procedure for this project. When performing measurements accurate to millionths of an inch, thermal effects become more important. Slight changes in temperature cause all materials to change dimension as a function of their size and thermal conductivity. Steel is used in the MDRL roundness tester and is widely recognized to have a moderate coefficient of thermal expansion; however, any measurements performed to the microinch level will be very sensitive to temperature fluctuations. A human operator is roughly equivalent to a 100 Watt light bulb and great care must be taken to minimize this disturbance. Stiffness of the BLOCK-HEAD 4R is very high relative to the loads imparted from the weight of the part and manually rotating the spindle. Even though the BLOCK-HEAD 4R air-bearing spindle rides on a thin film of air and spins with error motions less than 1 microinch [3], the measurement is taken after the spindle has completed several revolutions and settled into pseudo-steady-state. Vibration effects are minimized since testing only occurs when no other machine is in use and after outside noise is determined to be insignificant. No movement in the lab is allowed once the test has started since vibrations from the floor can impart error in the measurement. The procedure for a single test was fairly simple and has the capability for further automation. In terms of setting the part up for measurement, key aspects should be mentioned that greatly increase the precision of leveling and centering the part. Mating contact and therefore levelness between the part and the air-bearing spindle is considered nearly perfect using a concept known as flat stoning. Flat stoning uses the concept of self-checking by using three separate surfaces to achieve nearly perfect contact [1]. Stoning is a standard procedure used to remove asperities between mating surfaces of high-precision components. The objective for the roundness tester is good contact between the part being measured and the BLOCK-HEAD 4R air-bearing spindle. Once the part is stoned and centered on the spindle, a time capture is taken over approximately 5 revolutions (nominally 20 RPM). This angular speed is greater than what should be used (5 to 6 RPM), but because repeatability between revolutions is being tested, 5 revolutions in 14 seconds is about the maximum amount for the size of the data files that occur (approx. 70 MB). The tests utilize the maximum sampling frequency of the Hewlett Packard 35670A Dynamic Signal Analyzer, which is 65 kHz. The time capture is downloaded on to a computer for post-processing in MATLAB. Once a series of tests is complete, the program runs a loop that loads each file, makes a plot, and outputs an out-of-roundness value using the method of Least Squares Circle.
26
Chapter 5: COMPONENTS OF THE ROUNDNESS TESTER Function and not necessarily form is stressed in the design and assembly of the MDRL roundness tester. The goal of this project is to manufacture a repeatable and reliable roundness tester using knowledge and background of precision mechanics coupled with the resources presently available. This particular roundness tester is an initial, functional prototype that will be expanded on and improved in the future. The roundness tester complete with dial indicator, used for an initial TIR reading simply for placement, is shown in Figure 5.1.
Figure 5.1 The roundness tester set up to analyze an aluminum log. The dial indicator is used to center the part initially, then the capacitance gage reading is used for a value of TIR. The location of the dial indicator is not assumed to be coaxial to the diamond-turned cylinder. Parts To Be Measured Three parts were used as test specimens for the roundness tester: a diamond-turned aluminum log, a lapped steel ball, and a precision ground steel disc, shown in Figures 5.2, 5.3, and 5.4. Each part was cleaned with acetone without handling the critical surface prior to measurement. The diamond-turned aluminum cylinder or log was known to be round within 5 in. The part was diamond turned on a Twin-Mount BLOCK-HEAD air-bearing spindle with an AC motorized drive. For roundness testing, the part is bolted down to facilitate the process of tapping into center to minimize eccentricity. The lapped steel ball is provided courtesy of Professional Instruments, manufacturers of the BLOCK-HEAD air-bearing spindles. Lapping the steel ball results in the measured roundness of
27
two micro-inches. The part was measured on a Federal Formscan 6100 during the design phase of the MDRL roundness tester with a roundness value of approximately 8 in; the difference is attributed to shop wear. A cylindrically ground discs with large amplitude chatter marks was also used in the testing. The expected roundness of the disc was not known, but estimated with a C. E. Johansson Mikrokator and found to be 250 to 300 microinches.
Figures 5.2, 5.3, 5.4 Parts measured by the roundness tester: Aluminum Log (left), Steel Ball (center), Steel Disc (right). The BLOCK-HEAD 4R Air-Bearing Spindle
28
A high-degree of rotational accuracy and stiffness comparable to similar size ball bearing spindles are some of the advantages of using the BLOCK-HEAD 4R air-bearing as the roundness tester spindle [3]. The BLOCK-HEAD 4R is operated on 100 120 psig, about the midrange of the manufacturers air pressure recommendation. Another advantage to the air-bearing is the extremely low rotational friction that allows very smooth movement even at high RPMs. The specified maximum rotational speed for the model 4R is 10,000 RPM [3]. For the roundness tester, the rotational speed will unlikely exceed 20 RPM. The working load capacity of the BLOCK-HEAD 4R at 150 psig is 50 lbs. in the radial direction and 200 lbs. in the axial direction [3]. Loads imparted during roundness testing will be well within the operating limits. Indicator Stand The stand is made from 123 and 246 blocks configured to hold the Lion Precision DMT 10 capacitance gage in a position to measure the part. The 123 and 246 blocks are rigid, rectangular blocks that have a high degree of squareness and stiffness. The capacitance gage holder was attached in a way that vertical position could be easily adjusted. The vertical range was about 4 inches, which would suffice for measurements of all three parts. Lion Precision DMT 10 Capacitance Gage and Driver Module The capacitance gage is used with a driver module, in this case a Lion Precision DMT 10, Single Range Driver, shown in Figure 4.5. The DMT 10 driver provides excellent repeatability and extremely high resolution. The features mentioned in Lion Precision: Modular Systems and Drivers include:
29
Precise linear output A low signal to noise ratio (1 part in 10,000 at 100 Hz bandwidth) High bandwidth (response flat to 10 kHz) LED bar graph range indicator Extended range and extremely high resolution calibration available Coarse and fine zero adjust and BNC output connector.
Hewlett Packard HEDS 9140 I00 Optical Encoder A Hewlett Packard HEDS 9140 I00 Optical Encoder is used for precise angular position measurements. The encoder model provides 512 counts per revolution (CPR) and three-channel output containing two channels of square waves in quadrature and an index pulse. The module contains a highly collimated LED source and a unique photodetector integrated circuit for high quality measurement characteristics [11]. These optical encoders are very practical for their low cost and accurate, repeatable output. The encoder was located on the underside of the base aligned with the codewheel attached to a shaft that rotated with the air-bearing, Figure 4.6.
Figure 4.6 Location of the HP encoder and codewheel (underside of air-bearing and base).
30
Chapter 6: RESULTS OF TIR VERSUS ROUNDNESS A series of experiments was carried out on the completed MDRL roundness tester to explore its capabilities. The first experiment was to compare TIR to out-of-roundness and observe repeatability, similar to the Federal Formscan 6100 tests. The data were organized such that the individual revolutions were considered in addition to the synchronous motion of the entire test. As expected, the out-of-roundness values computed for individual revolutions were very close to the out-of-roundness of the synchronous motion. The results from the MDRL roundness tester match the overall trends observed with the Federal Formscan, but with higher resolution and repeatability. Diamond Turned Aluminum Cylinder (expected roundness: 5 in.) Eight tests were performed on the diamond-turned aluminum log, Figure 4.2, at increasing total indicated readings. The TIR ranges from 30 to 4950 in. Figures 6.1 through 6.7 represent the plots that correspond to the increasing values of TIR found in Table 1. All values are in microinches.
Test 1 2 3 4 5 6 7
31
. Figure 6.5 TIR = 925 in. Figure 6.6 TIR = 2300 in.
Figure 6.7 TIR = 4950 in. Lapped Steel Ball (expected roundness: 8 in.) Five tests of TIR versus roundness were performed on the lapped steel ball, and the TIR ranged from 15 to 860 in. The LSC method [6] was again used for the out-of-roundness value. Figures 6.8 through 6.12 represent the plots that correspond to the increasing values of TIR found in Table 2. All values are in microinches.
32
Test 1 2 3 4 5
33
The following tests were performed separately from Figures 6.8 and 6.9,
Figure 6.12 TIR = 860 in. Precision-Ground Steel Disc (expected roundness: 250 in.) The expected roundness of the steel disc was approximately 25 to 50 times greater than the lapped steel ball and diamond-turned aluminum log. Five tests were performed for total indicated
34
readings between 50 and 1200 microinches. Figures 6.13 through 6.17 show the polar plots that correspond to the increasing values of TIR found in Table 3. All values are in microinches.
Test 1 2 3 4 5 TIR in 50 100 500 800 1200 Roundness (mean) in 280 280 280 270 270 Roundness (max-min) in 280 280 280 270 270
Table 6.3 TIR versus roundness test for ground steel disc.
35
36
4000
5000
6000
Figure 6.18 Graph showing TIR versus out-of-roundness for diamond-turned aluminum log.
Figure 6.19 Graph showing TIR versus out-of-roundness for lapped steel ball.
Figure 6.20 Graph showing TIR versus roundness for ground steel disc. Note y-axis scaling.
37
Repeatability Analysis Artifact Test Repeatability in log 1 2 3 4 5 6 7 ball 1 2 3 4 5 disc 1 2 3 4 5 0.2 0.7 0.1 0.3 0.4 0.2 0.6 0.1 0.2 0.1 0.2 0.2 0.20 0.20 0.20 0.00 0.00
Artifact
Specifics
log ball
(excludes Test 7) (Tests 2 and 3) (Tests 3, 4, and 5) (Tests 1, 2, and 3) (Tests 4 and 5)
disc
0.30 0.20
38
Table 6.5 Displays repeatability between different tests of same artifact. Certain tests are grouped to show capability of roundness tester.
39
Discussion of Results The most significant aspects of roundness testing are the repeatability and accuracy of the measurements. Over the span of tests for the three parts, the roundness measuring process was becoming more precise because of improvements in technique and skill. The repeatability of the roundness tester is based on the components as much as the process of taking measurements and obeying the rules of roundness testing. Extreme care must be taken when performing microinch-accuracy tests and any discrepancy between tests would most likely be attributed to not meeting the standards of the roundness testing process. For the diamond-turned aluminum log, the mean out-of-roundness values between tests show a repeatability within 0.5 microinches. The repeatability is expected to further improve by running the tests at a slower spindle speed and by reducing thermal fluctuations by movement. For tests where all the rules of roundness testing are adhered to and strict tolerances of minimal vibration and noise are kept, repeatability of this roundness tester could be within 0.1 microinches for a TIR range of at least 0 to 2300 in. For the lapped steel ball, the last three tests were taken on a different day. This resulted in dissimilar out-of-roundness values and polar plots comparing plots from the two days. More precise measurements were taken on the lapped steel ball than the aluminum log which included minimal vibration and thermal effects and resulted in higher repeatability. The polar diagrams of the lapped steel balls circular form all show a comparatively large imperfection across approximately 10 degrees. The location changes between tests performed on different days because the part was removed from the roundness tester following Test 2. However, the imperfection lies in the exact same position for tests performed on the same day. The two peaks that are noticeable in the Federal Formscan 6100 tests might also be at a different vertical location than the tests performed for Figures 6.8 through 6.12. Agreement between tests performed on similar days is very high. Results for the ground disc showed 24 lobes and out-of-roundness values that, overall, were within a range of 5.0 microinches (mean) and 5.0 microinches (max-min). Tests 1 through 3 are repeatable within 0.2 microinches. Since the out-of-roundness drops significantly at a TIR of 800 in., centering a part being measured becomes more significant as the expected out-ofroundness value increases. This is supported by the out-of-roundness agreement between revolutions for each test which are within at least 0.2 microinches for each test and perhaps closer, but the values were assumed accurate to only a tenth of a microinch due to averaging affects and the size of the values. These tests were performed with strict adherence to the rules of roundness testing [5] with vibrations, noise and thermal effects minimized.
40
Chapter 7: CONCLUSIONS Based on the comparison between results obtained by this roundness tester and those obtained from the Federal Formscan, repeatability of measurements is a function of maintaining an environment that would allow the most accurate measurements possible. By carefully minimizing vibrations and thermal fluctuations between tests and adhering to a strict procedure, the out-of-roundness value for a part can be repeated. This roundness tester meets and exceeds tolerance standards set by commercially manufactured roundness testers [28]. The superior components of this roundness tester including the BLOCK-HEAD 4R air-bearing spindle and Lion Precision DMT 10 capacitance gage and driver module increase the accuracy of out-ofroundness attained for various parts. Further improvement can be expected using a Lion Precision Contact Adapter Probe (CAP) [4]. Future modifications will lead towards increasing the user-interface capabilities and further instituting automation via motorization and signal acquisition. The stylus indicator of the Lion Precision CAP could be implemented on a fully adjustable stand with 2 or 3 degrees of mobility and fine adjustment near the indicator. For post-processing and computer analysis, programs could be written that calculated out-of-roundness using the alternative methods discussed in Chapter 2. The data processing could further be expanded to include tests such as concentricity and coaxility of parts. Repeatability of this roundness tester has been proven with understanding of the highly disciplined art of out-of-roundness measurement and testing on how the centering and TIR of a part effects the out-of-roundness value. Further modifications could increase the precision of roundness testing to within 0.1 in with exceptional repeatability.
41
REFERENCES
1. Moore, Wayne R. Foundations of Mechanical Accuracy. Connecticut: The Moore Special Tool Company, 1989. 2. Smith, S. T. and Chetwynd, D. G. Ultraprecision Mechanism Design. Netherlands: Gordon and Breach Science Publishers, 1997. 3. Professional Instruments Company. BLOCK-HEAD Operators Manual. Minnesota: Professional Instruments Company, 1999. http://www.thomasregister.com/olc/pico/ 4. Lion Precision, Modular Systems and Drivers Minnesota: Lion Precision, 1998. http://www.lionprecision.com/guide/modular.html 5. American Society of Mechanical Engineers. American National Standard Measurement of Out-Of-Roundness ANSI B89.3.1 New York: The American Society of Mechanical Engineers, 1997. 6. Castle, Philip How to Measure Roundness. American Machinist (December 1993): 41 43. 7. Litsikas, Mary The Dos & Donts of Roundness Measurement. Quality Articles (November 1996). http://qualitymag.com/1196fl.html 8. Federal Products Co. Dimensional Gaging Products, Formscan Model 6100 Hunt Valley, MD: Federal Products Co., 1998, http://www.federalproducts.com. 9. Precitech Precision. FMS 8150 Circular Form Measurement System. Precitech Precision, 1999. http://www.precitech.com/fms8150.htm 10. Mitutoyo, Mitutoyo Measuring Instruments Catalog No. 302, Roundtest RA-114/116. Mitutoyo, 1997. M-20. http://www.mitutoyo.com 11. Hewlett-Packard. Three Channel Optical Encoder Modules Technical Data Hewlett Packard, 1999. http://www.hp.com
42
%***********ROUNDNESS TESTER PROGRAM**************** %***********Get vectors from data file************** %******************FOR LOG FILES******************** clear all close all
for n = 1:7, eval(['load log' num2str(n) '.dat']); data = eval(['log' num2str(n)]); eval(['clear log' num2str(n)]); tach = data(:,2) - median(data(:,2)); square = data(:,3); gage cap = data(:,4); square wave off1 = mean(square); off2 = mean(cap); square = square - off1; cap = cap - off2; clear data off1 off2; % Find pulse points level = 2; j = 0; for inc = 1:length(tach)-1, if (tach(inc)<level)&(tach(inc+1)>level), j = j+1; start(j) = inc; stop(j) = inc+1; end %squarewave DC offset %capgage DC offset %to be compared to %to find where revs are %to be compared to cap
43
% Number of revolutions rev = length(start); rev = rev-1; disp(['Number of Revolutions = ' num2str(rev)]) % Crop start, stop, cap, square square = square(start(1):start(length(start))); cap = cap(start(1):start(length(start))); start = start - start(1); stop = stop - stop(1); % Squarewave marker loop for each revolution level = 0; j = 0; for cnt = 1:length(square)-1, if (square(cnt)< level)&(square(cnt+1)> level), j = j+1; start_sq(j,1) = cnt; stop_sq(j,1) = cnt+1; end end clear cnt j % Filter N = 100; upr = 120; cutoff = 2*upr*(length(start)-1)/length(cap); cap = filtfilt(fir1(N,cutoff), 1, cap); clear N upr cutoff
44
y2 = square(stop_sq); y1 = square(start_sq); tcross = (level + (y2.*t1 - y1.*t2))./(y2 - y1); c2 = cap(stop_sq); c1 = cap(start_sq); cap1 = (c2 - c1).*tcross + c1.*t2 - c2.*t1; clear tcross c2 c1 t2 t1 y2 y1 square start_sq stop_sq cap level % Remove once around points = length(cap1); theta = 2*pi*rev*[0: points-1]'/points; a1 = 2/points * sum(cap1.*cos(theta)); b1 = 2/points * sum(cap1.*sin(theta)); cap1 = cap1 - a1*cos(theta) - b1*sin(theta); clear a1 b1 points theta % Synchronous error cap1 = 1000*(cap1 - mean(cap1)); for inc = 1: rev, cap1matrix(:,inc) = cap1([512*(inc-1)+1:512*inc]); end if rev > 1, cap1mean = mean(cap1matrix')'; clear inc end if rev == 1, cap1mean = cap1; clear inc end % Asynchronous error
45
for inc = 1: rev, asynch(:,inc) = cap1([512*(inc-1)+1:512*inc]) - cap1mean; end clear inc % OOR value, Minimum Radial Separation method, (mean of all revs) and (min-max) for inc = 1:rev, OORv(1,inc) = max(cap1matrix(:,inc)) - min(cap1matrix(:,inc)); end OOR = (mean(OORv)); maxrnd = max(cap1)- min(cap1); clear inc rev OORv logrnds(n,1:2) = [OOR maxrnd]; clear OOR maxrnd rnds FFT analysis after filter, once around removed trans = fft(cap1); transig = abs(trans); clear trans % Plotting bradius = ceil(2.5*max(abs(cap1))); oradius = bradius + ceil(max(abs(cap1))) + 0*cap1; figure polar(2*pi*[1:length(cap1)]'/512, oradius, 'k') hold on polar(2*pi*[1:length(cap1)]'/512, bradius + cap1) polar(2*pi*[1:512]'/512, bradius + cap1mean, 'r') eval(['print log' num2str(n) ' -dbitmap']) close %figure %plot(mary) %eval(['print fft' num2str(n) ' -dbitmap']) %close
46
clear asynch bradius oradius cap1 cap1mean start stop cap1matrix mary disp(['End of Test ' num2str(n)]) close all end save logrnds logrnds;
47