Failure Detection in Actively Controlled Structures

University Of Houston, Houston TX

Investigators

Abstract

Project Abstract The overall objective of this research program is the development of analytical and numerical techniques to autonomously monitor the integrity of actively controlled structures. For the control system, the failure may be the loss of a sensor or actuator, or a more compensatible loss such as a change in the gain of a sensor. For the structural system, failure may be the partial or total loss of structural stiffness or mass in a region of the structure. The basis for all proposed methodologies is to utilize changes in the measured dynamics before and after a failure has occurred to determine the current "state" of the actively controlled structure. Specifically, we will develop methodologies to determine a statistically meaningful methodology to establish the existence of a failure using system realization theory, time-domain signal processing and outlier analysis. The proposed methodology will be able to distinguish between changes in operational environment and actual failures. With existence established, system realization redundancy will be used to isolate the failure to either the control system or the structure and once isolated to determine the individual component that has failed. Once the failure extent has been estimated, the final step is to compensate for the failure to maximize the performance of the "new" system using a Genetic Algorithm learning control concept and a Neural Network adaptive controller. All methods developed will be evaluating using both numerical and actual experimental tests. There are several systems which will benefit from this research, including actively controlled civil structures (buildings/bridges), aircraft, offshore platforms and deep-water risers, re-useable launch vehicles, the International Space Station, and long term exploratory space probes.