ECE: Electrical & Computer Engineering
ECE News

CESCA Seminar Talk (10/25): “Optimal Architecture Synthesis for Aircraft Electrical Power Systems”

12:00 AM - 2:30 PM on Friday, October 25, 2013
Location: Lavery Hall 350

by Safa Messaoud, Ph.D. student in the ECE Department at Virginia Tech.


Aircraft Electrical Power Systems (EPS) are becoming complex
cyber-physical systems, consisting of a physical plant (generators,
AC-Buses, rectifiers, DC-Buses, switches and loads) and a cyber
component, namely the Bus Power Control Units (BPCU). To deal with the
increasing complexity of EPS while guaranteeing the satisfaction of
tight reliability and safety constraints, automated design tools for
the synthesis of the topology (interconnection of components) are
developed. These tools can significantly reduce the design time aiming
to generate correct-by-construction systems. In this presentation, we
discuss two optimization oriented design methodologies to synthesize a
reliable-by-construction EPS topology with minimal cost and
complexity. The two methodologies differently deal with the non linear
reliability constraints, that arise in the mathematical formulation of
the synthesis problem. The first methodology uses a Mixed Integer
Linear Programming Modulo Reliability (MILPMR) approach to generate
minimal cost topologies, given a set of a connectivity and power tow
constraints. The optimizer is placed in a loop with a reliability
analysis algorithm, which evaluates the failure probability of the
critical loads and implements several strategies that provide
suggestions to the optimizer to improve the reliability of the
topology, in case the requirements are not met. In the second
methodology, we automatically generate a symbolic expression of the
reliability analysis, linearize it and use it directly in a plain MILP
optimizer. We successfully implement the two methodologies and compare
them with respect to their runtime, scalability with the number of
components and accuracy. The obtained results show that the MILPMR can
result in a less efficient implementation because of the number of
iterations needed to converge to the final solution, whereas the plain
MILP normally converges in one iterations, but it relies on an
approximate reliability algebra to improve on scalability.


Safa Messaoud is currently a Ph.D. student in the Bradley Department
of Electrical and Computer Engineering at Virginia Tech. She received
a B.S. degree and a M.S. degree from the Technical University of
Munich in 2012 and 2013, respectively. From 2012 to 2013, she
concurrently enrolled in the M.S. program in the EECS Department at
University of California, Berkeley. This presentation is based on her
Master Thesis, which she performed as an integral part of the research
stay at the UC Berkeley.

(Previous CESCA Seminar talks are available at