The Effect of Input Shaping on Coordinate Measuring Machine Repeatability

(Proceedings of the 1995 IFToMM World Congress on the Theory of Machines and Mechanisms)

William E. Singhose and Warren P. Seering Department of Mechanical Engineering Massachusetts Institute of Technology Cambridge, MA Keywords: Vibration, Input Shaping, Measurement Abstract Coordinate Measuring Machines (CMMs) are a strategic element in the manufacture of high-precision parts because their measurements of part geometries are used for quality control and feedback on the manufacturing process. Proper CMM operation requires accurate knowledge of the position of the CMM's part sensor. The performance of a CMM is limited by background vibration levels, environmental conditions (temperature changes, cleanliness of the environment, etc.), accuracy in the measurement equipment (encoders and part sensor), and structural deflections between the encoders and the part sensor. The most important limitation depends on the type of CMM and the operating conditions; however, structural deflection is always an important limitation because it introduces an error in the indicated position of the part sensor. The effect of input shaping, a method of reducing residual vibration, on the quality of CMM measurements has been investigated. Tests were performed that verified the reduction of structural deflections when input shaping is used. Standard diagnostic tests were used to evaluate the repeatability of the CMM. Non-standard tests aimed at evaluating the CMM under adverse conditions were also performed. Input shaping improved measurement repeatability over a large range of operating parameters. Introduction Coordinate Measuring Machines (CMMs) measure manufactured parts to determine if tolerance specifications have been achieved. A CMM consists of a workspace in which parts are fixtured, a sensor for detecting the part surfaces, a mechanical assembly for moving the part sensor around the workspace, and a computer for calculating the part dimensions based on the sensor measurements. A sketch of a typical CMM is shown in Figure 1. The CMM sketched in Figure 1 is shown with a touchtrigger probe part sensor This sensor uses a ruby-tipped stylus to sense the part. When the stylus is brought into contact with a part surface, the deflection of the stylus triggers the computer to record the position indicated by the x, y, and z encoders. By probing the part on critical surfaces and recording their locations, the critical dimensions of the part can be calculated. A single cycle of a CMM measurement consists of four phases. First, the CMM performs a gross motion to move the part sensor to the vicinity of the part geometry that is to be measured. Second, the probe is allowed to come to rest. Third, the probe is reaccelerated to a small constant velocity in the direction of the part. This constant velocity portion of the measurement is called the pre-hit region because it immediately precedes the contact between the probe and the part. Finally, the stylus contacts the part and the computer records the location of the contact. The position of a touch-trigger probe during a measurement with a 2 mm pre-hit distance is shown in Figure 2.

Neil C. Singer Convolve, Inc. Armonk, NY

z y x y Encoder z Encoder
TouchTrigger Probe

x Encoder
Figure 1: Sketch of a Typical CMM. The constant velocity in the pre-hit region is necessary so that the time delay between actual sensor contact and the computer's recognition of the contact can be subtracted from the measurements. If a constant velocity approach is not used, it is difficult to determine how far the probe travels during the signal propagation delay. If a CMM is to provide valuable quality control, then its accuracy and repeatability must be greater than the tolerance specifications for the part. Many CMM designs strive for accuracies of 8-12 m and repeatabilities of 3-5 m. The measurement quality of a CMM is limited by background vibration levels, environmental conditions (temperature changes, cleanliness of the environment, etc.), accuracy in the measurement equipment (encoders and part sensor), and structural deflections between the encoders and the part sensor. The difficulty of obtaining micron level performance can be appreciated by examining the probe vibration when the CMM is subject to standard environmental disturbances. In order to detect the micron level probe vibration that is important in CMM measurements, a laser interferometer with a resolution of 3 nanometers was used to measure x-direction motions of the probe. The interferometer retro-reflector was mounted on the z-axis structural member in close proximity with the probe. Figure 3 shows the measured vibration when a 180 lb. person with a normal gait walks past the CMM. Virtually every foot strike causes a vibration with an initial amplitude of 3-4 m even though the machine weighs several thousand pounds and is stationed on a concrete floor. The most important limitation on CMM performance depends on the design of the CMM and the operating conditions, however, structural deflection between the encoders and touch-trigger probe is always an important limitation because it introduces an error in the measurement.

10

L
Pre-Hit Distance 3 Pre-Hit Phase (Constant Velocity) 2 Stop Before Pre-Hit 1 Approach Phase 4 Probe Touches Part and then Retracts

Position (mm)

6 4 2 0

Encoder Gives Position of Left Face CMM Structure

y Encoder z

Encoder Gives Position of Right Face

1.0

1.2

1.4

1.6

1.8

2.0

Time (sec)

Touch-Trigger Probe Part

Figure 2: Probe Position During a Measurement.

Position (m)

4 2 0 -2 -4 0 2

Measured Part Width = L 4(a) L+2

z
4 6 8 10 12

Time (sec) Figure 3: Probe Vibration Caused by a Person Walking Past the CMM.

y Part Measured Part Width = L+2 4(b) Figure 4: Deflections Cause Measurement Errors. modeling errors. Several papers give detailed explanations of input shaping[3, 5, 7-10, 12, 15, 18, 19]. Methods for increasing the robustness of input shaping to modeling errors have been developed[14, 15]. Input shaping reduced residual vibration and maximum deflections during the slewing of a large nonlinear spacebased antenna[1, 2]. Two-mode input shapers were used to increase the throughput of a silicon wafer handling robot[10]. Input shaping was shown to be beneficial for long-reach manipulators[6, 8] and trajectory following applications[4, 13, 16]. This paper will investigate the effect of input shaping on the accuracy and repeatability of CMM measurements. The operation of a CMM is not the typical point-to-point application for which input shaping was designed. Rather, the important phase of the operation (the part detection) occurs while the machine is in motion and its timing occurs with some uncertainty because there is no way to know a priori where the part is located or how much the structure will be deflected at the time of contact. The remainder of this paper is organized as follows. First, important parameters of a CMM measurement cycle are discussed. Experimental results showing the effect of input shaping on the structural deflections during the measurement will then be presented. The effect of input shaping on the measurement quality will be demonstrated with the use of repeatability studies. Finally, conclusions will be drawn from the experimental results. Important Parameters of the Measurement Cycle There are several important parameters that determine the vibration during the critical pre-hit region. During the approach to pre-hit, the acceleration and approach distance are significant parameters. As the acceleration is increased,

Figure 4 graphically demonstrates how deflection in a CMM structure can adversely effect measurement accuracy. In Figure 4a the part width is determined from two measurements. First, the probe is moved into contact with the left side of the part. At the moment the contact is made, the position indicated by the encoders is recorded. Next, the probe is moved to the opposite side of the part and brought into contact with the right face. The two encoder positions are then subtracted to obtain the part dimension. In Figure 4a the part is measured accurately because the encoders indicate the true position of the probe. In Figure 4b the measurement is inaccurate because the structural deflections make the encoder readings differ from the true location of the contact points. If the structure is vibrating with an amplitude of during the pre-hit region, then the calculated dimension can have an error of 2 . Several methods exist for limiting the structural deflection of a CMM. First, the acceleration and velocity of the machine can be limited. This solution is effective at reducing deflections, but it causes a decrease in throughput which makes it an unattractive solution. Second, the mechanical structure can be modified with additional stiffening members, damping materials, or configuration changes. This solution involves many of the classic design tradeoffs such as cost versus performance and solutions may not apply across a product line containing CMMs of many sizes and accuracy levels. Alternatively, the hardware can be left unaltered and the command signals sent to the motors can be shaped so that deflections during the measurement phase are decreased[7, 11, 12]. Input shaping is a procedure that eliminates residual vibration by convolving a sequence of impulses with the desired command signal. The result of the convolution is then used to command the system. The impulse sequence used to shape the input is derived by solving a set of constraint equations which limit the residual vibration of the system. Input shaping requires that the residual vibration remain at a low level even in the presence of

the deflections during the approach phase become larger and the residual vibrations from these deflections in the pre-hit region increases. For short approaches, the deflection during the approach is smaller than for long approaches and, therefore, contributes less to the vibration during the pre-hit region. Certain approach distances lead to a natural cancellation of the approach phase vibration, i.e., the vibration induced by the acceleration is canceled by the vibration from the deceleration of the approach. This cancellation can be thought of as a naturally occurring posicast control[17]. It is difficult to utilize this natural cancellation on a consistent basis because its effectiveness is a strong function of approach distance, maximum acceleration, maximum velocity, probe location, and knowledge of the system dynamics. The period of time that is spent waiting between the approach phase and the pre-hit phase is also an important parameter because it allows the vibrations induced by the gross motion to settle out. As the waiting period increases, the residual structural vibration decreases, however, the throughput also decreases. The two important parameters effecting the vibration caused by the pre-hit phase are the pre-hit velocity and prehit distance. The pre-hit velocity is usually a small percentage of the maximum velocity used during the approach phase. (The pre-hit velocities used in our experiments was 1% of the maximum velocity.) Increasing the pre-hit velocity would be desirable for throughput, but would lead to larger deflections because the acceleration to pre-hit velocity would last longer. On the other hand, a larger pre-hit distance yields better repeatability because vibrations from both the approach phase and the acceleration to pre-hit velocity have a longer period of time to damp out before contact is made with the part. While increasing the pre-hit distance will improve repeatability, it will also degrade throughput considerably because the CMM is moving at a very low velocity during the pre-hit. Reducing Structural Deflections In order to measure the structural deflections of the CMM, the laser interferometer was used to measure the position of the touch-trigger probe. Then, the position measured by the encoders was subtracted from the laser measurements to obtain the deflection between the encoders and the touch-trigger probe. Figure 5 compares the deflections with the standard CMM controller to the deflections when input shaping is used. The two curves are both from a measurement cycle with a 25 mm approach distance. The curves have been shifted in time so that their pre-hit regions align. The large deflections during the first part of the move are caused by the accelerations during the approach to prehit. The amplitudes of these deflections during the approach are relatively unimportant, however, the amplitudes become important during the pre-hit phase, when the probe contacts the part. If the deflection is not zero when the part is encountered, then the deflection leads to an inaccuracy in the measurement, as was demonstrated in Figure 4. The deflection amplitude in the pre-hit region is decreased by a factor of about 3 when input shaping is used. Repeatability Tests Laser interferometers are not practical for evaluating the performance of every CMM, so the performance of a CMM is generally measured with a repeatability test. A

Deflection (Laser-Encoder) ( m)

60 40

Shaped Deflection Unshaped Deflection

Pre-Hit Region
20 0.0 -20 -40 -60 0.40 0.60 0.80 1.00 1.20

Time(sec)

Figure 5: Comparison of Deflections During Shaped and Unshaped Measurement Cycles. repeatability test is conducted by repeatedly measuring the same part. A standard diagnostic repeatability test consists of 10-50 measurements of a fixtured part. After each measurement cycle, the probe returns to its starting position. For each of the measurements, the approach distance, pre-hit distance, and pre-hit velocity are kept constant. The minimum measured value is subtracted from the maximum measurement to obtain the repeatability of the CMM. Most CMMs have a repeatability of just a few microns. The repeatability of a CMM depends on several parameters, including the parameters listed above for the measurement cycle. Figure 6 shows the 50 individual measurements of a repeatability test using a 1 mm pre-hit distance and a 20 mm approach distance. The repeatability for the unshaped measurements is: 625.2735 mm 625.2687 mm which equals 4.8 m. For the shaped measurements: 625.2683 mm - 625.2650 mm gives a repeat of 3.3 m. For this set of measurement parameters input shaping improves repeatability by 1.5 m. To ensure that the benefit of input shaping is not limited to this particular set of parameters, repeatability tests were conducted with several pre-hit and approach distances. Figure 7 shows the repeatability as a function of pre-hit distance for measurements with a 20 mm approach distance. Each data point in Figure 7 represents the range of measurements obtained from a 50 point repeatability test with the given pre-hit distance and a 20 mm approach distance. In general, repeatability improves with increasing pre-hit distance. Figure 7 also demonstrates that input shaping improves repeatability over a wide range of pre-hit distances. Figure 8 shows that input shaping also improves repeatability when a 10 or 15 mm approach distance is used. Note that as the pre-hit distance increases, the benefit of input shaping over the standard controller decreases. Variable Pre-Hit Distance Repeatability Tests In the repeatability tests shown in Figures 7 and 8, the pre-hit distance and approach distance were held constant throughout the 50 measurements. The standard repeatability tests are very valuable for evaluating CMM performance when parts are fixtured in the workspace consistently and when part geometries do not very greatly. However, if consistent part fixturing is not used, then prehit distances could vary significantly from part to part. To evaluate the CMM performance under adverse fixturing conditions, tests were performed with a pre-hit distance that changed for each measurement in the repeatability test. For the first measurement of the repeatability test the pre-hit distance was set at 1 mm. During each subsequent measure, the pre-hit distance was increased by 0.01 mm until the pre-hit distance reached 1.5

Measured Part Location (mm)

Measured Part Location (mm)

625.276 625.274 625.272

Shaped, 1mm Pre-Hit, 20 mm Approach Unshaped, 1mm Pre-Hit, 20 mm Approach

625.285 625.280 625.275

Shaped 15 mm Approach Unshaped 15 mm Approach

4.8 m
625.270 625.268 625.266 625.264 0.00

10.4 m
625.270 625.265 625.260 625.255 1.00

3.3 m

4.2 m

10.0

20.0

30.0

40.0

50.0

1.10

1.20

1.30

1.40

1.50

Measurement Number

Pre-Hit Distance (mm)

Figure 6: Measurements During a Standard 50 Point Repeatability Test.

Figure 9: Measurements During a Variable Pre-Hit Distance Repeatability Test (15 mm Approach).
25

Repeatability ( m)

Repeatability (m)

Shaped 20mm Approach Unshaped 20mm Approach

Shaped (1-1.5 mm Pre-Hit) Unshaped (1-1.5 mm Pre-Hit)

20 15 10 5 0 5

Data from Figure 6

Data from Figure 9

Pre-Hit Distance (mm)

Approach Distance (mm)

Figure 7: Comparison of Shaped and Unshaped Repeatability for Several Pre-Hit Distances.
10

Figure 10: Shaped and Unshaped Variable Pre-Hit Distance Repeatability (1-1.5 mm Pre-Hit Distance).
25 Shaped(1-1.5PH) Shaped(1.5-2PH) Shaped(2-2.5PH) Unshaped(1-1.5PH) Unshaped(1.5-2PH) Unshaped(2-2.5PH)

Repeatability (m)

8 7 6 5 4 3 2 1.0 1.5

Repeatability (m)

Shaped, 10 mm Approach Shaped, 15 mm Approach Shaped, 20 mm Approach Unshaped, 10 mm Approach Unshaped, 15 mm Approach Unshaped, 20 mm Approach

20 15 10 5 0 5 10

2.0

2.5

3.0

15

20

Pre-Hit Distance (mm)

Approach Distance (mm)

Figure 8: Shaped and Unshaped Repeatability with 10, 15, and 20 mm Approach Distances. mm. Each of the measurements composing this 51 point repeatability test are shown in Figure 9 for an approach distance of 15 mm. By subtracting the minimum measurement value from the maximum value we obtain a variable pre-hit distance repeatability of 4.2 m with shaping and 10.4 m without shaping. This improvement with input shaping is much larger than the improvement revealed by the standard repeatability tests shown in Figures 7 and 8. The large improvement with shaping is not restricted to an approach distance on 15 mm; Figure 10 shows significant improvement for approach distances ranging from 5 to 20 mm. Furthermore, Figure 11 shows that shaping improves repeatability when the variable pre-hit test covers the range of pre-hit distances of 1.5-2.0 mm and 2.0-2.5 mm. Figure 11 also shows a clear trend in repeatability with input shaping; as approach distance increases, repeatability slowly degrades. The repeatability without shaping is not only larger, it is more unpredictable. Variable Approach Distance Repeatability Tests In another variation on the standard repeatability test, the measurement approach distance can be varied instead of

Figure 11: Variable Pre-Hit Distance Repeatability for Several Ranges of Pre-Hit Distances. the pre-hit distance. Figure 12 shows the 51 measurements of a repeatability test when the approach distance was varied from 10-12 mm at a step of 0.04 mm and the pre-hit distance was held constant at 2 mm. For this variable approach distance repeatability test the input shaping improved the repeat from 7.2 m to 3.5 m. Figure 13 shows that shaping improves the 10-12 mm variable approach distance repeatability over a range of prehit distances. Figure 14 demonstrates the improvement in repeatability also occurs for approach distances varying from 5-7 mm and from 15-17 mm. Once again, the repeatability with input shaping is consistent over a wide range of parameters, while the unshaped measurement repeatability is considerably larger and unpredictable. Conclusions Deflections in the structural components of a coordinate measuring machine introduce error into the measurements because the CMM encoders do not indicate the true position of the part sensor. Input shaping decreases the deflections in a CMM structure during the critical phase when the sensor is brought into contact with the part. The decrease in deflection translates directly into improved

625.292

Shaped 2 mm Pre-Hit Unshaped 2 mm Pre-Hit 3.5 m

625.287

625.282

625.277

7.2 m
10.5 11.0 11.5 12.0

625.272 10.0

Approach Distance (mm)

Figure 12: Measurements During a Variable Approach Distance Repeatability Test (2 mm Pre-hit Distance).
15

Repeatability (m)

Shaped, 10-12 mm Approach Unshaped, 10-12 mm Approach Data from Figure 12

10

0 1.0

1.2

1.4

1.6

1.8

2.0

Pre-Hit Distance (mm)

Figure 13: Comparison of Shaped and Unshaped Variable Approach Distance Repeatability (10-12 mm Approach Distance).
14

Repeatability (m)

12 10 8 6 4 2 1.0 1.2

Shaped(5-7) Shaped(10-12) Shaped(15-17)

Unshaped(5-7) Unshaped(10-12) Unshaped(15-17)

1.4

1.6

1.8

2.0

Pre-Hit Distance (mm)

Figure 14: Variable Approach Distance Repeatability for Several Ranges of Approach Distances. measurement repeatability as measured by standard diagnostic tests. Non-standard tests designed to evaluate the CMM performance under adverse conditions that might be found in some industrial settings indicate that input shaping improves measurement quality over a wide range of measurement parameters. Acknowledgments Support for this work was provided by Convolve, Inc. under NSF contract ISI-9101441 and the Office of Naval Research Fellowship Program. We would like to thank the Management and Engineering staff of Brown & Sharpe whose assistance was invaluable in this project. References [1] Banerjee, A. and W. Singhose, "Slewing and Vibration Control of a Nonlinearly Elastic Shuttle Antenna," AIAA Guidance, Navigation, and Control Conference, Scottsdale, AZ, 1994. [2] Banerjee, A.K., "Dynamics and Control of the WISP Shuttle-Antennae System," The Journal of Astronautical Sciences, 41(1): p. 73-90, 1993.

[3] Bhat, S.P. and D.K. Miu, "Precise Point-to-Point Positioning Control of Flexible Structures," ASME Journal of Dynamic Systems, Measurement, and Control , 112(4): p. 667-674, 1990. [4] Drapeau, V. and D. Wang, "Verification of a Closedloop Shaped-input Controller for a Five-bar-linkage Manipulator," IEEE International Conference on Robotics and Automation, Atlanta, GA, Vol. 3, pp. 216-221, 1993. [5] Hyde, J.M. and W.P. Seering, "Inhibiting Multiple Mode Vibration in Controlled Flexible Systems," Proceedings of the American Control Conference, Boston, MA, 1991. [6] Jansen, J.F., Control and Analysis of a Single-Link Flexible Beam with Experimental Verification, ORNL/TM-12198 , Oak Ridge National Laboratory, 1992. [7] Jones, S.D., Quantification and Reduction of Dynamically Induced Errors in Coordinate Measuring Machines, Ph.D. Thesis, University of Michigan, Ann Arbor, MI, 1993. [8] Magee, D.P. and W.J. Book, "Filtering Schilling Manipulator Commands to Prevent Flexible Structure Vibration," American Control Conference, Baltimore, MD, 1994. [9] Murphy, B.R. and I. Watanabe, "Digital Shaping Filters for Reducing Machine Vibration," IEEE Transactions on Robotics and Automation, 8(April): p. 285289, 1992. [10] Rappole, B.W., N.C. Singer, and W.P. Seering, "Multiple-Mode Impulse Shaping Sequences for Reducing Residual Vibrations," ASME Mechanisms Conference, Minneapolis, MN, 1994. [11] Seth, N., K.S. Rattan, and R.W. Brandstetter, "Vibration Control of a Coordinate Measuring Machine," IEEE Conference on Control Applications , Dayton, OH, 1993. [12] Singer, N.C. and W.P. Seering, "Preshaping Command Inputs to Reduce System Vibration," ASME Journal of Dynamic Systems, Measurement, and Control, 112(March): p. 76-82, 1990. [13] Singhose, W., T. Chuang, and N. Singer, "Reducing Deviations From Trajectory Components with Input Shaping," American Control Conference, Seattle, WA, 1995. [14] Singhose, W., L. Porter, and N. Singer, "Vibration Reduction Using Multi-Hump Extra-Insensitive Input Shapers," American Control Conference , Seattle, WA, 1995. [15] Singhose, W., W. Seering, and N. Singer, "Residual Vibration Reduction Using Vector Diagrams to Generate Shaped Inputs," ASME Journal of Mechanical Design, 116(June): p. 654-659, 1994. [16] Singhose, W. and N. Singer, "Initial Investigations into the Effects of Input Shaping on Trajectory Following," American Control Conference, Baltimore, MD, 1994. [17] Smith, O.J.M., Feedback Control Systems. 1958, New York: McGraw-Hill Book Company, Inc. pp. 331-345. [18] Tuttle, T.D. and W.P. Seering, "A Zero-Placement Technique for Designing Shaped Inputs to Suppress Multiple-Mode Vibration," American Controls Conference, Baltimore, MD, 1994. [19] Zuo, K. and D. Wang, "Closed Loop Shaped-Input Control of a Class of Manipulators with a Single Flexible Link," Proceedings of the IEEE International Conference on Robotics and Automation, Nice, France, Vol. 1, pp. 782787, 1992.

Measured Part Location (mm)