Title:
Adaptive Control for a Microgravity Vibration Isolation System
Adaptive Control for a Microgravity Vibration Isolation System
Author(s)
Yang, Bong-Jun
Calise, Anthony J.
Craig, James I.
Whorton, Mark S.
Calise, Anthony J.
Craig, James I.
Whorton, Mark S.
Advisor(s)
Editor(s)
Collections
Supplementary to
Permanent Link
Abstract
Most active vibration isolation systems that try to a provide quiescent acceleration environment for space-science
experiments have utilized linear design methods. In this report, we address adaptive control augmentation
of an existing classical controller that combines a high-gain acceleration inner-loop feedback together with a
low-gain position outer-loop feedback to regulate the platform about its center position. The control design
considers both parametric and dynamic uncertainties because the isolation system must accommodate a variety
of payloads having different inertial and dynamic characteristics. We show how adaptive control is beneficial in
three important aspects in design of a controller for uncertain systems: performance, robustness, and transient
responses. First, performance is treated in the setting that an accelerometer and an actuator is located at the
same location, as is the current hardware configuration for g-LIMIT. Second, robustness for the control system
becomes more of an issue when the sensor is non-collocated with the actuator. We illustrate that adaptive
control can stabilize otherwise unstable dynamics due to the presence of unmodelled dynamics. Third, transient
responses of the position of the isolation system are significantly influenced by a high-gain acceleration controller
when it includes integral action. An important aspect of the g-LIMIT is the accelerometer bias and the deviation
of the platform it causes as a result of integral control. By employing adaptive neural networks for both the
inner-loop and outer-loop controllers, we illustrate that adaptive control can improve both steady-state responses
and transient responses in position. A feature in the design is that high-band pass and low pass filters are applied
to the error signal used to adapt the weights in the neural network and the adaptive signals, so that the adaptive
processes operate over targeted ranges of frequency. This prevents the inner and outer loop adaptive processes
from interfering with each other.
Sponsor
Research Supported by:
NASA Marshall Space Flight Center, Huntsville, AL
Grant No. NAG8-1292.
Date Issued
2005
Extent
Resource Type
Text
Resource Subtype
Technical Report