Sunteți pe pagina 1din 10

MECH TOC

Proceedings of DETC02 Proceedings of the 2002 ASME Design Engineering Technical Conference: ASME 2002 Design Engineering Conferences 27th Biennial MechanismsTechnical and Robotics Conference September 29October 2, 2002, Montreal, Canada and Computer and Information in Engineering Conference Montreal, Canada, September 29-October 2, 2002

DETC/MECH-34231 DETC2002/ MECH-34231


REDUCTION OF IMPACT AND VIBRATION IN AN INDUSTRIAL CAM-FOLLOWER SYSTEM USING THE SPLINEDYNE METHOD: A CASE STUDY

Robert P. Gordon
Engineering and Implementation Group The Gillette Company Boston, MA 02127

Robert L. Norton
Dept. of Mechanical Engineering Worcester Polytechnic Institute Worcester, MA 01609

ABSTRACT The splinedyne method of cam design is a variation on the polydyne method that uses B-spline functions to obtain superior control of follower motion and its derivatives in combination with a dynamic model of the follower train to minimize vibration in the follower at any one design speed. This paper presents the results of the application of this technique to an automated assembly machine cam-follower train whose vibratory behavior before redesign was limiting machine speed and causing increased product scrap rate due to impacts and vibrations. The improved splinedyne system contributed to a measured 14% increase in production rate in combination with a 1% reduction in scrap rate from that station. INTRODUCTION The splinedyne method of cam-follower design, a contraction of spline and dynamic, is a modern variation on the long established polydyne method of cam design. The term polydyne is a contraction of "polynomial" and "dynamic." It was coined by Thoren, Engemann, and Stoddart (Thoren 1952) to describe a cam design method first proposed and implemented by Dudley (Dudley 1948) who was the first to use a dynamic model of a cam-follower system to determine a cam profile that would, in effect, compensate for dynamic vibration of the follower train, at least at one particular cam speed. In essence it applies the desired motion function to the end effector of the cam-follower system and back-calculates the cam profile function needed to deliver that motion to the end effector considering the dynamic response of the follower train based on a lumped parameter model and uses B-splines to control motion. In the 1950-1970 period, polydyne cams were used in automotive pushrod, overhead valve-train systems to control vibrations but are seldom used in that application today. Modern valve train designs have stiffer and lighter follower trains, which reduces the need for polydyne methods. Also, a polydyne cam must

be designed for a particular camshaft speed, and in an engine application where speed is not constant, this proves a limitation. Industrial cam-follower systems can have relatively heavy and flexible follower trains, and are being required to run at ever higher speeds that make follower vibrations a significant problem in many instances. This makes them a candidate for the polydyne approach, especially since they tend to run at constant speed. Polydyne cams have another potential problem related to the mathematical limitations of polynomial functions. To control vibration with the polydyne method will usually require creating a polynomial function that uses a large number of boundary conditions. This drives the order of the polynomial up with concomitant problems of high peak values of acceleration and the potential for oscillation of the displacement function beyond desired limits. The automotive valve train polydyne functions of the 1950s typically used polynomials with terms as high as 50th power (Thoren 1952). B-spline functions (Schumaker 1969, Sanchez 1980, Schumaker 1981, MacCarthy 1985, MacCarthy 1988, Tsay 1988) eliminate the problems with high order polynomials and allow virtually unlimited application of boundary conditions needed for dynamic control while keeping the order of the function reasonable and thus well behaved. B-splines can be thought of as super polynomials that provide a set of knots that can be used to control the function shape while preserving its conformance to any set of desired boundary conditions. In fact, any polynomial function can be exactly reproduced with a B-spline, making the former a subset of the latter. This paper presents an approach to cam-follower system design that combines the principles of polydyne cam design with the advantages of B-spline functions, dubbed the Splinedyne approach, a combination of spline and dynamic. (Norton 2002a), This paper also describes the application of the splinedyne technique to a cam-follower system for one particular station of an automated assembly machine. The same technique has recently been applied to several other systems on the same and

Copyright 2002 by ASME

similar assembly machines with equally successful results. In general, residual vibration of the end effector tooling during dwell has been successfully minimized to levels approaching zero, and the motion functions have been customized with Bsplines to control velocities in order to limit impact levels while still keeping acceleration in good control. Significant increases in machine speed have been accomplished with simultaneous increases in machine efficiency and reduction of scrap rate. The machine in question was designed over 20 years ago and has undergone several speedups and modifications since that time. It now can be thought of as a 3rd-generation design whose speed is over twice its original design speed. The machine is typical of many automated production machines that have an indexing chassis to carry the product through the process in nests on a conveyor mechanism. As each nest stops at a workstation, camdriven tooling performs an operation on the product. A CAD solid model of the subject cam-follower mechanism is shown in Figure 1. The cams are on one or more camshafts (not shown) that run the length of the machine and are driven in time with the conveyor motion. Air cylinder springs are typically used to close the cam joint.

This product assembly process has very tight tolerance requirements for positioning of its component parts. Any vibration of the tooling end effectors during the assembly process can potentially compromise accurate placement and thus increase product scrap rate. It is desired to have the smoothest possible motions of the end effectors and to have minimal residual vibration during the dwell portion of the cycle when operations are performed on the product. The function of the mechanism in Figure 1 is simply to clamp the assembly of product parts while a fastening operation is done. If it is vibrating during the dwells clamp cycle, the assembly may be out of tolerance after fastening. Because of the need to clamp this assembly, the end effector must be provided with a small amount of compliance and the cam-follower motion with a small amount of overstroke, in order to guaranty contact with sufficient clamping force despite differences in product position from nest to nest due to conveyor and nest tolerances. This then requires that the tooling impact the product on the nest. Any impact will, of course, ring the system and cause structural dynamic vibrations. Thus it is desirable to minimize the impact velocity between tooling and workpiece. B-spline functions provide an easy way to do this. Including the dynamics of the follower train in the cam design serves to minimize residual vibrations in the follower at one design speed. (Dudley 1948) This splinedyne approach has been found to be extremely effective in cam design for automated machinery as we will show in this case study. DYNAMIC MODELS Figure 2 shows a schematic version of the mechanism of Figure 1. We wish to reduce this to a simple, single-mass single degreeof-freedom (SDOF) dynamic model. (Norton 2002b) Figure 3 shows the simplest SDOF lumped parameter model and the freebody-diagram (FBD) for the system of Figure 2. The spring k1 and damper c1 represent those characteristics, respectively, of the joint closure spring (which may be an air spring). The modeling of an air spring is described in (Norton 2002c). The spring k2 and damper c2 represent, respectively, the stiffness and damping within the links and joints of the follower train. The effective mass m represents all the moving mass in the follower system. Its system equation is:
F = mx x ) k 2 x c2 x Fi = mx k1 ( s x ) + c1 ( s x+ c1 + c2 k +k k c + 1 2 x = 1 s+ 1 s + Fi x m m m m (1a)

FIGURE 1 Cam, follower train, tooling, conveyor nests, and frame of one station of an automated assembly machine

If we assume that contact is always maintained between cam and follower by sufficient joint closure force, the system simplifies to that shown in Figure 4 with the equation

Copyright 2002 by ASME

bellcrank (4) I4, k4 and k5

r4

r5

m x sx k c

ground (1)

connecting rod (3) m3, k3 l3

tooling (5) m5

cam

s+

x+

roller follower

. . k(s x) c(s x)
FIGURE 4 SDOF model and free-body diagram (FBD) of a form-closed industrial cam-follower system

ground (1) r1 ra r2 air cylinder

follower arm (2) I2, k2

+ x

c k k c + x = s+ s x m m m m

(1b)

ground (1)

FIGURE 2 Schematic of the industrial cam-follower mechanism of Figure 1

The follower acceleration x has units of length/sec2. Convert it to an angle base (in degrees) and introduce the camshaft angular velocity N in rpm.
length length 360 2 deg 2 N 2 rev 2 = sec 2 deg 2 rev 2 60 2 sec 2
x c2

m Fi sx k2 s k1 c1

x=

d2x d2x = 36 N 2 2 2 dt d

(2)

Koster (1974) has shown, and we have confirmed by measurement in industrial cam-follower systems, that the range of values of damping ratios in these systems is about = 0.05 to 0.10. These are very underdamped systems, and so are potentially subject to significant vibration. Assuming then that damping is negligible, equation 1b is rewritten without the damping terms.
x+ k k x= s m m (3a)

s+

x+

k2(s x) k2(s x) Fi

. . c2(s x)
. . c2(s x) . c1(s)
Fc

Rearranging to solve for s:


s= m x+x k (3b)

k1(s)
FIGURE 3

Substitute equation 2 in 3b
s = 36 N 2 m d2x +x k d 2 (4a)

Simplest SDOF model and free-body diagrams (FBD) of a typical industrial cam-follower system

Copyright 2002 by ASME

m4 eff

r4

r5 O4 m5

m4 eff

r4 C link 4 k5

r5 D O4 m5

link 4

D
k4

link 3
m3

m3 k3 link 3

(a)

cam

(a) lumped masses

cam
m1 O2

m2 eff

B A
r1 r2

A
m2 eff

link 2 m1 r1 r2 B Fcyl cam k2

link 2
Fcyl

O2

cam

meff
(b) lumped spring model

keff
(b) lumped mass model

r1

Ffol
r1 Ffol
FIGURE 6 Combining lumped springs for the system of Figure 2

FIGURE 5 Combining lumped masses for the system of Figure 2

Differentiate with respect to to get the splinedyne cam velocity and acceleration functions.
s = 36 N 2 m d x dx + k d 3 d m d4x d2x + k d 4 d 2
3

allow control of the functions shape and peak value independent of the number of boundary conditions required. Equation 4a shows that the cam profile will be different for any value of N selected. Thus the dynamic behavior of the system can be optimized (i.e., vibration minimized) only for one operating speed. This is not a severe limitation in this case as assembly machines are typically operated at one speed for long calendar periods. MODELING THE SYSTEM To apply the above models to a system such as that in Figures 1 and 2, the distributed mass and stiffness properties must be defined and reduced to properly configured lumped parameters as shown in Norton 2002a. Figure 5a shows lumped masses of the various elements of Figure 2. The rotating links are modeled as point masses at one of their connecting pins based on an equiva4

(4b)

s = 36 N 2

(4c)

Note that the cam acceleration function involves the fourth derivative of the selected follower displacement function, which we call "ping." This means that any function selected for the follower displacement should be continuous through at least four derivatives, velocity, acceleration, jerk, and ping. Since spline functions allow control of continuity at the ends of the segment to any derivative, they provide a useful means to this end and also

Copyright 2002 by ASME

(a) Original modified trapezoidal cam function

FIGURE 8 Velocity and acceleration of a modified trapezoidal cam (b) Improved asymmetric 3-4-5-6 polynomial cam function

(FEA) software is available, perhaps the best way to obtain an estimate of the individual parts spring rates is to mesh the solids model in an FEA package and calculate the deflection for a load applied at the point where the link joins its neighbor, being careful to replicate the boundary conditions correctly. Absent the availability of these CAD tools, the spring rates can be calculated using classical beam and column theory as was done for this design. If parts exist, then a simple experiment can be done by supporting the part as it is held in the machine, applying a known load to the pin joint, and measuring the deflection with a dial indicator. The data for the system of Figure 1 was calculated to be: lent mass moment of inertia. This simplification is acceptable for small angular rotation of the link, which is typical in these mechanisms. The lumped masses are combined taking all link ratios into account until an effective mass is defined to be acting at the cam follower as shown in Figure 5b. The existence of a CAD solids model of the assembly (Figure 1) makes it easy to determine the mass properties of the individual links needed to calculate the effective mass for the dynamic model of Figure 3 or Figure 4 as was done here. The effective spring rate of the follower linkage is found in a similar fashion as shown in Figure 6. The spring rate of each elastic element such as the follower arm, connecting rod, and bellcrank must be calculated, measured, or estimated and then combined properly to obtain an effective spring rate. When CAD solid models of the parts exist and when finite element analysis
meff = 15.04 kg keff = 5.410 E 6 N/m (5)

(c) Latest B-spline function timing diagram FIGURE 7 Timing diagrams of three successive cam designs for Fig 1

The effective spring rate and preload of the air cylinder were:
kspring = 3.793E 4 N/m Fi = 603 N (6)

CAM TIMING DIAGRAMS Over the history of this particular machine, there have been three different cam designs applied to this follower train. The timing diagrams for them are shown in Figure 7. The original design was a conventional (for the time) modified trapezoidal acceleration (MT) double dwell motion (Figure 7a). During an earlier

Copyright 2002 by ASME

FIGURE 9 Velocity and acceleration of a 3-4-5-6 polynomial cam

speedup program, an asymmetric polynomial design was developed that incorporated an increase of stroke coupled with changes to the rise and fall durations to control acceleration. This polynomial design (Figure 7b) is essentially a modified 3-4-5 polynomial double-dwell motion with an added interior boundary condition that sets acceleration to zero at other than the midpoint of the motion, making it an asymmetric double-dwell 3-4-5-6 polynomial. This redesign, along with similar changes to many of the other stations of the machine, allowed a 25% increase in speed. The most recent cam design (Figure 7c) uses a quintic B-spline function to control the velocity and acceleration functions in order to reduce impact and dynamic forces and has allowed an additional 14% increase in speed. The follower train design has remained essentially unchanged throughout all the cam design changes. These three cam designs will now be described in more detail with respect to their dynamic performance, all defined at the same machine speed. MODIFIED TRAPEZOIDAL CAM PERFORMANCE Figure 8 shows (from top to bottom) the kinematic velocity (v), simulated follower velocity ( x ), kinematic acceleration (a), and simulated follower acceleration ( x ) for the modified trapezoidal (MT) design, simulated at current machine speed. Note the oscillation in the simulation during the dwell in both velocity and acceleration curves as compared to the theoretical kinematic functions. There is vibration in acceleration through more than half of the nominal dwell period.

FIGURE 10 B-spline functions for rise and fall

POLYNOMIAL CAM PERFORMANCE Figure 9 shows (from top to bottom) the kinematic velocity (v), simulated follower velocity ( x ), kinematic acceleration (a), and simulated follower acceleration ( x ) for the asymmetric 3-4-5-6 polynomial design. Compared to the MT design in Figure 7, the

Copyright 2002 by ASME

ence is striking. The polynomial shows ringing during the dwell and the splinedyne does not. Figure 12b shows the velocity functions for the polynomial and splinedyne functions superposed. The splinedyne solution shows no oscillation during the dwell, but the polynomial function does. Figure 12c shows the displacement functions for the two designs. Note that, by design, the splinedyne function (which has much less dynamic error) also has substantially greater displacement then the prior polynomial design. There are good engineering reasons for this having to do with other desired changes to the process specifications at this station, but they are not germane to the issues addressed here. The point, however, is that, despite the additional stroke that increases peak acceleration significantly, its dynamics are superior, due to the splinedyne approach. Figure 12d shows the error function between the cam displacement command signal (defined at the follower end effector) and the actual displacement of the end effector (according to the simulation). The significant difference is seen in the dwell. The polynomial design shows ringing that lasts for nearly all of the dwell period. The splinedyne solution shows essentially no follower oscillation during the dwell. Since the primary goal of the cam-follower mechanism is to move the end effector to its final position and then hold it stationary and quiet during the dwell so that the desired operation can be accomplished, any oscillation during the dwell defeats its purpose. Thus the splinedyne solution shows a clear advantage over the others described herein. EXPERIMENTAL DATA SPLINEDYNE CAM DESIGN AND PERFORMANCE Figure 10 shows the design of the B-spline functions for rise and fall of the splinedyne cam using program DYNACAM (Norton 2002a). Note the knots (vertical lines) that control the shape of the curve. They are arranged to bias the curves asymmetrically in order to reduce the velocity as the impact point is approached. The fall motion is the one that impacts the nest. Toward the end of the fall the function is shaped so as to approach the impact point slowly with low velocity. Despite a designed increase in the total stroke, the peak acceleration is nevertheless controlled to reasonable values. Figure 11 shows (from top to bottom) the kinematic velocity (v), simulated follower velocity ( x ), kinematic acceleration (a), and simulated follower acceleration ( x ) for the splinedyne cam design. Note that this is the result of both the superior spline function shape and the inclusion of the dynamic response of the follower train in the cam design. Compared to both of the earlier designs, note that the simulated dynamic responses of the follower velocity and acceleration are essentially identical to the theoretical kinematic velocity and accelerations. Figure 12a shows the splinedyne solution follower acceleration superimposed on that of the polynomial design. The differThe splinedyne cam was installed on several machines and will eventually be fitted to all machines of this type. The acceleration at the end effector was measured on one machine with the polynomial cam design and on one to which the splinedyne cam had been fitted. Figure 13 shows the results. The acceleration was measured with a Dytran model 3110a piezoelectric accelerometer and recorded in a Hewlett Packard 35670a Dynamic Signal Analyzer. There are 1176 data points in each trace over one revolution of the cam. The raw data is averaged over 50 cam revolutions, and is also shown as a 10-point moving average trend line fitted to the raw data The traces of both cams are in time phase. The polynomial data shows more oscillation during the dwells than does the splinedyne data, which is essentially quiet after the ringout of the impact in the first third of the dwell. The rise and fall motion events show more oscillation than was predicted in the simulation probably due to several factors. The simulation is only an approximation of the true behavior and does not include the effects of impact nor of other systems that are running simultaneously in the machine. There are more than 20 cam-follower trains in this machine and vibrations from any one train can affect the others. The most significant difference between the two traces shown in Figure 13 is the quiescence of the second dwell following the fall event and impact ringout of the splinedyne sys-

FIGURE 11 Velocity and acceleration of a quintic B-spline, splinedyne cam function over entire interval including dwells

follower velocity is better controlled and is close to the kinematic velocity. The simulated follower acceleration shows oscillation to a lesser degree than the MT design, but still has an undesirable level of distortion and residual vibration during the dwell.

Copyright 2002 by ASME

(a)

(b)

FIGURE 12 Part 1 Comparison of simulated polynomial and splinedyne system responses - acceleration and velocity

tem in comparison to that of the polynomial design. This is the period of clamping of the product that is required to be as vibration free as possible. High speed video of both designs confirmed that the splinedyne system had significantly less residual vibration during the dwell.

CONCLUSIONS The splinedyne approach to cam design has been shown in this case study to provide significant advantages in the control of vibrations and impact forces. B-splines offer significant design flexibility in terms of controlling follower motion by allowing large numbers of boundary conditions to be applied in combination with a practical limitation on function order that prevents unwanted oscillations. Inclusion of the dynamics of the follower system in order to alter the cam profile design so as to control the

Copyright 2002 by ASME

(a)

(b)

FIGURE 12 Part 1 Comparison of simulated polynomial and splinedyne system responses - acceleration and velocity

motion of the end effector provides an additional and powerful tool that can virtually eliminate unwanted vibrations in the output motion at any one machine speed. Application of these cam design principles has been shown to increase productivity by allowing operation at higher speeds than were possible before these techniques were introduced, and to simultaneously provide significant reductions in product scrap rates. In addition, acoustic noise levels from stations to which these splinedyne techniques were applied have been reduced by as much as 5 dB.

REFERENCES
Dudley, W. M. (1948). New Methods in Valve Cam Design. SAE Quarterly Transactions, 2(1), pp. 19-33. Fawcett, G. F., and J. N. Fawcett, eds. (1978). Comparison of Polydyne and NonPolydyne Cams. Cams and Cam Mechanisms, Jones, J. R., ed., Institution of Mechanical Engineers: London. Koster, M. P. (1974). Vibrations of Cam Mechanisms. Phillips Technical Library Series, Macmillan Press Ltd.: London.

Copyright 2002 by ASME

FIGURE 13 Experimentally measured accelerations in two machines, one with the polynomial cam and one with the splinedyne cam

MacCarthy, B. L., and N. D. Burns. (1985). An Evaluation of Spline Functions for Use in Cam Design. IMechE, 199(C3), pp. 239-248. MacCarthy, B. L. (1988). Quintic Splines for Kinematic Design. ComputerAided Design, 20(7), pp. 406-415. Norton, R. L. (2002a). Cam Design and Manufacturing Handbook. Industrial Press: New York. Norton (2002b): Norton, R. L., C. A. Gillis, C N. Maynard. Dynamic Modeling of the Typical Industrial Cam-Follower System: Part 1 Single-Degreeof-Freedom Models. Paper #DETC2002/MECH-34232, Proc. of ASME International Design Engineering Conference, Montreal, Canada. Norton (2002c): Norton, R. L., C. OBrien, P. Duperry, J. R. Hall, Determining the Effective Spring Rate of Air Cylinders Used to Close CamFollower Joints. Paper #DETC2002/MECH-34280, Proc. of ASME Intl Des. Eng Conf., Montreal, Canada. Sanchez, M. N., and J. G. deJalon. (1980). Application of B-Spline Functions to the Motion Specifications of Cams. Proc. of ASME Design Engineering Technical Conference, Beverly Hills, CA.

Schumaker, L. L., ed. (1969). Some Algorithms for the Computation and Approximating Spline Functions. Theory and Application of Spline Functions, Greville, T. N. E., ed., Academic Press: New York, pp. 87102. Schumaker, L. L. (1981). Spline Functions: Basic Theory. John Wiley & Sons: New York. Stoddart, D. A. (1953). Polydyne Cam Design-I. Machine Design, January, 1953, pp. 121-135. Stoddart, D. A. (1953). Polydyne Cam Design-II. Machine Design, February, 1953, pp. 146-154. Stoddart, D. A. (1953). Polydyne Cam Design-III. Machine Design, March, 1953, pp. 149-164. Thoren, T. R., et al. (1952). Cam Design as Related to Valve Train Dynamics. SAE Quarterly Transactions, 6(1), pp. 1-14. Tsay, D. M., and C. O. Huey. (1988). Cam Motion Synthesis Using Spline Functions. Journal of Mechanisms, Transmissions, and Automation in Design, 110, pp. 161-165.

10

10

Copyright 2002 by ASME

S-ar putea să vă placă și