MEMS-based Active Control
of Macro Scale Objects

Active Control of Structural Instability

Active structural enhancement consists of the use of active control to modify structural behavior. This enhancement can be used to actively stiffen, or strengthen (against Euler buckling) a given structure. Therefore, actively controlled structures can adaptively modify their stiffness properties to be either stiff or flexible as demanded. This research investigates optimal closed-loop control strategies to maximize the critical buckling load of a pinned-end column using MEMS based sensors and actuators. In this instance, active control is employed to actively stabilize a structural member to prevent it from collapsing. From a controls standpoint this problem consists of developing control strategies for a time-varying, inherently unstable system.

This project is a part of a larger research program sponsored by the Defense Advanced Research Projects Agency, and is a joint effort between Xerox PARC and the Sarcos Research Corp.

This page is organized as follows:

Experimental Setup

Control Architectures and Experiments

Notable Results

Future Work

Acknowledgments


Experimental Setup

This project integrates PZT (piezo-ceramic) actuators and MEMS based UAST strain sensors with a structural beam element to control linear buckling phenomena. Using optimal feedback control strategies we have been able to increase the buckling load of this column by 2.9 times. The picture shows the beam and fixture apparatus. The filament based PZT actuators and the UAST sensors are both technologies developed by Sarcos. The combination of smart material actuators and MEMS based UAST sensors with external computation emulates smart matter.

The beam is clamped vertically within the test fixture which allows us to apply a compressive axial load to the pinned-end beam. The load is applied mechanically using a load arm (shown in the foreground) and a moving weight. The weight is driven by a motor under computer control so that load may be applied dynamically.

The picture on the right shows UAST strain sensors (the little black boxes) and filament PZT patches between them. The beam material is G10 fiberglass composite. For reference, the beam is 18 inches long, 2 inches wide and 0.093 inches thick.

The UAST's sense strain with an accuracy of 3 micro-strain. Nine pairs are attached to this beam to determine bending strains. Structural deformation and can be determined from these measurements. The PZT patches are used as actuators for the control system. Voltages applied to the PZT enable it to expand or contract, creating surface strains on the beam. Eight pairs of PZT patches create bending strains which are applied as the control input. The control is implemented with a servo rate of 1000Hz using a PC and a TI DSP card.

The following animation shows the beam under control in both its straight position and the buckled position.

Control Architectures and Experiments

Several control strategies were developed and implemented for this beam. The general goal is to minimize deviation from the straight position at all times. The specific controllers developed include: Optimal buckling control uses semi-definite programming (SDP) to optimally specify the first critical buckling load and the structural damping. As such, multiple modes can be controlled. A finite element model was used to define the eigenvalue problems used in the optimization. This optimization problem is convex, and easily solved using SDPSOL. The Modal PID controller is a SISO controller based on controlling only the first buckling mode. This controller could be modified to control additional modes under the assumption that they are decoupled. The Local PD controller used only local sensor measurements to determine the associated local actuation. This controller's objective was to minimize local response in maintaining beam stability. This collection of controllers represents a wide range in the number of inputs and outputs, and in global/local control design.

All of these controllers were tested on the experimental apparatus. The maximum axial load achieved was 2.9 times the critical buckling load (PCR). These controllers supported all loads up to and including this value, where the load was steadily increased using the load arm mechanism. At loads lower than this value, the beam was able to support the load indefinitely (>30 minutes). The optimal buckling control was robust to sensor noise, adequately supporting the beam even when a sensor failed.

We were unable to determine which controller was best, because no controller outperformed the others. This result is due to problems with fixturing, beam imperfections, and external disturbances which caused the beam to buckle before the full actuator authority was used. As part of future work we will be examining new fixturing methods to eliminate this effect. 

Notable Results

This project has made several notable achievements :

Future Work

The future directions of this project will focus on two main areas. The first is controlling buckling for a greater number of physical dimensions and the second is greater control decentralization. The current project stabilized buckling for a single dimension. The next step is to generalize this work to multiple dimensions such as plate, shell, and compound structures. This generalization will require much larger numbers of sensors and actuators. Control decentralization is required for systems with greater numbers of sensors and actuators to avoid the communications and processing bottlenecks, and design intractability of a global controller.

Other topics include:

 

Acknowledgments

This research is the result of a collaboration between Sarcos Research Corporation and Xerox Palo Alto Research Center. The project or effort depicted is sponsored in part under contracts DABT63-95-C-0025 and DABT63-95-C-0033 by the Defense Advanced Research Projects Agency (DARPA). The content of the information does not necessarily reflect the position or the policy of the Government and no official endorsement should be inferred.

last modified June 26, 1998