Data processing: generic control systems or specific application – Generic control system – apparatus or process – Optimization or adaptive control
Reexamination Certificate
1999-04-16
2002-10-22
Picard, Leo (Department: 2125)
Data processing: generic control systems or specific application
Generic control system, apparatus or process
Optimization or adaptive control
C700S188000
Reexamination Certificate
active
06470225
ABSTRACT:
TECHNICAL FIELD
The present invention relates to tuning feedforward parameters in a motion control system, and, more particularly, in one embodiment, to a method and apparatus which automatically tunes feedforward gains, such as acceleration feedforward gain, constant offset feedforward gain, Coulomb friction feedforward gain, and viscous friction feedforward gain, by providing a velocity command to a motor control system, measuring the corresponding torque command signals and velocity feedback signals, and using these values in an equation relating the sum of the torques acting on the system.
BACKGROUND OF THE INVENTION
Traditional machine tool position control systems are designed with feedback control in which “following error” (the difference between the actual and desired positions) produces a command which is delivered to a velocity control loop for driving an actuator connected to a movable member. The actuator and/or connected movable member are sometimes referred to as an “axis” of the machine tool or device having the movable member. In practice, it is difficult, if not impossible, to achieve zero following error during motion of the axis if only feedback control is utilized. Following errors generally do not cause a problem in the case of machining long linear spans, because the following error vector is along the desired straight line path, and, accordingly, the machine path follows the desired path. However, in the case of cut direction changes or in complex contouring, following errors can cause the machined path to deviate from the desired path resulting in a “contour error.” Typically, the magnitude of the contour error increases as the feedrate and/or path curvature increases. In addition to following errors, disturbance forces, such as Coulomb friction, may also cause the machine to deviate from the desired path.
Accordingly, “feedforward” control was developed to attempt to minimize following errors and errors caused by disturbance forces. Feedforward control attempts to minimize the following errors and to compensate “in advance” for the errors caused by known disturbance forces by anticipating the response of the machine tool system. The required information to allow feedforward to anticipate in advance is obtained by performing measurements of machine state variables while the machine undergoes various controlled movements, and/or a theoretical analysis from knowledge of system components, and geometries (e.g., system inertia). Feedforward compensation involves applying additional control signals, such as additional torque or velocity commands for example, based upon this knowledge of the system dynamics, rather than waiting for feedback errors to build up, as occurs in feedback control. Feedforward is used in addition to feedback, because feedback is still required to account for imperfections in the feedforward system model, and to counter unexpected forces, such as cutting forces.
It has been known that, for optimum performance, a number of feedforward compensations should be utilized. For example, a velocity feedforward compensation can be utilized to compensate for velocity feedback control dynamics, and an acceleration feedforward compensation can be utilized to compensate for inertial forces. In addition, a viscous friction feedforward compensation can be applied to compensate for viscous friction forces, a Coulomb friction feedforward compensation can be applied to compensate for Coulomb friction forces, and a constant offset feedforward compensation can be applied to compensate for gravitational forces.
In discrete time systems (i.e. sampled data systems), the feedforward command values can be generated at each update interval. The feedforward command values can then be added to the velocity command and/or torque commands, and the servo can hold these signals constant over the update interval.
Previously, many feedforward compensation levels had to be determined manually. For example, to determine a Coulomb friction feedforward gain, the axis could be moved in a positive direction at a constant velocity and then moved in the negative direction at a constant velocity. Motor current measurements could be taken during these movements, and the measurements used to manually calculate a Coulomb friction feedforward gain. To determine an acceleration feedforward gain, the inertia level of each pertinent mechanical element in the system could be calculated individually, and the sum of these values used to determine an appropriate acceleration feedforward gain.
As can be understood, such manual calculations had a number of disadvantages. For example, manual tuning took a considerable amount of time, and often provided less than accurate estimates. Moreover, the manual tuning of feedforward gains required an engineer or technician with sufficient knowledge of the control system and the required calculations. In addition, there was not a reliable way of determining whether the model developed from the manual calculations accurately described the actual system dynamics. Furthermore, when test movements and measurements were used, such as when deriving a Coulomb friction gain, there was no way of determining whether the test sufficiently “excited” the axis by moving it through enough modes of operation (e.g., enough different velocities, accelerations, and position changes.) In addition, such manual testing procedures did not provide a way of automatically determining an optimum excitation program (i.e., test motion) for the axis based upon user-enterable operating parameters.
U.S. Pat. No. 5,684,375, which issued to Chaffee, discloses a method for tuning gains in electronic motion control systems. According to the patent, the motor is commanded through a single, unidirectional motion event for tuning these parameters. As disclosed in the patent, the event comprises driving the motor at a predetermined tuning torque level until the motor reaches a predetermined maximum velocity, and then decelerating the motor back to zero at the tuning torque level. FIG.
5
and
FIG. 6
of Chaffee illustrate the control signals generated and the velocity profile of the motor during this tuning motion. During the motion, feedback signals from the motor are monitored and parameters, such as the electrical current applied to the motor, the velocity of the motor, and the acceleration and deceleration times are recorded. The patent discloses that a number of values can be calculated based upon the parameters measured and recorded, such as the effective inertia value J
m
, the position feedback error proportional gain K
p(pos)
, the velocity feedback error proportional gain K
p(vel)
, the position feedback error integral gain K
i(pos)
, and the velocity feedback error integral gain K
i(vel)
. The feedforward gain K
ff
, on the other hand, is apparently not tuned, as the patent discloses that it can be set either to unity or to zero, depending on whether feedforward is to be applied.
U.S. Pat. No. 5,159,254, which issued to Teshima, discloses a method of estimating mass, inertia, viscous friction coefficient, or sliding frictional force of a movable part in a machine, by using a spring mass system as a model for the dynamic characteristic of the movable part. According to the reference, the following equation is used to represent the relationship between x
1
, (which indicates a value obtained by converting an angular value of a motor shaft into a traveling direction of the movable part), and x
2
(which indicates a position of the movable part):
K
(
x
2
−x
1
)=
m {umlaut over (x)}
2
+c {dot over (x)}
2
+fr
(
if {dot over (x)}
2
>0)
K
(
x
2
−x
1
)=
m {umlaut over (x)}
2
+c {dot over (x)}
2
−fr
(
if {dot over (x)}
2
<0)
where x
2
is the velocity of the movable part, x
2
is the acceleration of the movable part, m is the mass, c is the viscous friction coefficient, and fr is the sliding frictional force of the movable part, respectively. These equations are then converted to an equation relating the motor current feedback va
Dinsmore & Shohl LLP
Picard Leo
Rapp Chad
Siemens Energy & Automation Inc.
LandOfFree
Method and apparatus for automatically tuning feedforward... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for automatically tuning feedforward..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for automatically tuning feedforward... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2993923