Optimal Hypergraph-Based Decomposition of Design Problems

Abstracts ARCSIM CAPC CAMBAS DES Hypergraph Decomp. ILP Decomp. Nonlinear

Optimal Hypergraph-Based Decomposition of Design Problems

The online software section has been disabled indefinitely due to server usage policies. Please check back here at a later date.

Design of large engineering systems remains a challenge because increased size reduces the reliability and speed of design methods and algorithms. Decomposition of the original design model into smaller coordinated submodels is desirable or even necessary. This software implements a methodology for optimal model-based decomposition (OMBD) of design problems. The overall model is represented by a hypergraph, which is then optimally partitioned into weakly-connected subgraphs that can be identified with design subproblems. Design variables are represented by the hypergraph's edges, whereas the design relations interrelating these variables are represented by the nodes. Design relations may be algebraic or differential equations, simulation or analysis "black boxes", response surfaces, look-up tables, and so on. Design relations' dependencies on variables are given in a Functional Dependence Table (FDT). The formulation is robust enough to account for computational demands and resources, by means of design relation weighting coefficients and partitions sizes, respectively, and for the strength of interdependencies between the simulation modules contained in the model, by means of design variable weighting coefficients. More details about the methodology for OMBD of design problems can be found in the references listed here.

Input Data
Browse or specify the path of the file containing the Functional Dependence Table (FDT). The format of this file is available here.

Path of file containing Functional Dependence Table : or
Choose Functional Dependence Table example:

Number of design relations :

Number of design variables :

Browse or specify paths of files containing the design relation weights and the design variable weights OR choose one of the default choices. A description of the file formats is available here.

Path of file containing design relation weights : or
Design relation weights equal to one or
Design relation weights equal to the number of variables in relation or
Design relation weights equal to the sum of weights of variables in relation

Path of file containing design variable weights : or
Variable weights equal to one

Enter the number of subproblems you want the design problem to be divided into.

Number of subproblems:

Browse or specify the path of file containing the subproblem sizes OR choose equal-size subproblems. A description of the file format is available here. Also enter the allowed imbalance between subproblem sizes, as a percentage of the average size.

Path of file containing the subproblem sizes (% of problem size) : or
Equal-size subproblems

Allowed deviation of subproblem sizes (%) :

Acknowledgements

This site is supported by the Automotive Research Center at the University of Michigan, a US Army RDECOM Center of Excellence for Modeling and Simulation of Ground Vehicles, under Contract No. DAAE07-94-R094. We also thank Dr. Terrance Wagner for providing the powertrain model and for his suggestions on the design of this site.

Web Development Team

Steve Meckl, Web Specialist

Dr. Nestor Michelena, Project Leader