|
In this dissertation, we develop a real-time adaptive control
technique, which enables the high-performance control of an ac
induction spindle motor drive for CNC machine tools, and develop
a transputer-based parallel processing technique for the
realization of the complicated and real-time adaptive control
algorithm. The induction spindle drive can adaptively regulate
the speed performance in contending with varying load, torque
disturbance, and motor parameter variations. The control system
consists of adaptive speed control, adaptive torque control, and
adaptive current control. The adaptive torque control comprises
rotor-time constant adaptation for field-oriented vector
control, load-torque estimation and feedforward compensation,
and field-weakening flux control. The adaptive speed controller,
which precedes the field-oriented control loop, consists of a
two-degree-of-freedom controller and a speed-controlled plant
model estimator. The two-degree-of-freedom controller is
designed by a pole-placement technique with polynomial
manipulations. Its parameters are adjusted adaptively in terms
of estimated model parameters. Estimating the model parameters
entails a second-order least-squares estimator with constant
trace to avoid estimator windup. The adaptive current
controller, operating in the stationary two-axis frame for
providing the persistently exciting condition for parameter
estimation convergence, consists of an one-step-ahead predictive
controller and a model estimator. The predictive controller*s
parameters are adjusted adaptively in terms of estimated model
parameters. Estimating the model parameters entails a first-
order four-dimensional least-squares estimator with variable
forgetting factor to detect the variations of the motor
parameters and back emf. In the adaptive torque controller,
which precedes the current control loop, the design of the
feedforward load torque compensator is based on an estimated
load-torque model. Estimating the load torque entails a first-
order three-dimensional least-squares estimator with variable
forgetting factor and covariance resetting, whose purposes are
to detect any slow or sudden changes of torque disturbance,
respectively. A simple implementation scheme for the PWM
waveform generation, which modulates the current control outputs
as three-phase PWM pulses to drive the motor, based on space
vector concept is also presented. The computation of the
resulting adaptive speed, torque, and current controller is very
complex. However, the system exhibits some implicit parallel
characteristics because of the nested control loops. So, it has
been implemented in parallel by IMS T800-20 transputers. For
the realization of the parallel adaptive control algorithm, a
unified controller architecture comprising transputer-based
parallel computing boards and input/output boards suitable for
the real-time control of various types of motor drives is also
presented. The system can increase its computing and input/
output processing capability by paralleling these boards. A host
server based on a personal computer for user interface is also
developed. The control functions can easily be implemented in
parallel by using the high-level programming language Occam. A
comparison with two existing parallel controllers shows the
performance and architecture features of the system.
Experimental results show that the adaptive control system
maintains the desired torque-producing current and speed
performance in the presence of varying load and disturbance. The
work can be the basis of the research for high-speed and high-
power ac induction spindle drive for CNC machine tools in the
future.
|