Accession Number:

ADA535693

Title:

Sensor Fusion, Prognostics, Diagnostics and Failure Mode Control for Complex Aerospace Systems

Descriptive Note:

Final rept. 27 Mar 2009-27 Sep 2010

Corporate Author:

SALFORD UNIV (UNITED KINGDOM)

Personal Author(s):

Report Date:

2010-10-01

Pagination or Media Count:

36.0

Abstract:

This report results from a contract tasking University of Salford as follows This project examined the speed-up of search for hardware-in-the-loop based searches. The details for carrying out parallel and distributed computations with the search system are conveyed and used to perform large searches. Two methods of parallelization are available due to the two loop feature of the search. In the first method, the master processes each iteration of the outer loop and subsequent inner loops serially. The slaves are used to evaluate the fitness values of the inner loop individuals in parallel. The result of this arrangement is that the processing is a form that is analogous to standard parallel genetic algorithms, and as such existing parallel methodologies can be used. The second method involves the master processing the outer loop only the inner loop for each outer individual is processed in its entirety by a slave. Hence, outer loop individuals are processed in parallel. The advantage of this second method is that communication overheads are reduced by a factor of Gi Xi. This is because communication between master and slaves only occurs at each iteration of the outer loop, rather than at each iteration of the inner loop, as in the first method. A methodology for minimizing the number of trials when using hardware-in- the-loop to direct the search is presented. An experiment is conducted demonstrating that control solutions developed using models under simulation do not always perform as expected in hardware. The demonstration shows a controller developed from a simplified model for an Electric Thrust Reversal Actuation System ETRAS, that when implemented in hardware does match the performance under simulation. This is because the controller is fundamentally based upon the accuracy of the model from which it has been derived. It is sometimes the case that very accurate models are not available and therefore it is desirable to use hardware-in-the-loop to direct

Subject Categories:

  • Aerodynamics
  • Cybernetics

Distribution Statement:

APPROVED FOR PUBLIC RELEASE