CBET Award Achievements
Notable Accomplishments from CBET Awards
 

CAREER: Mixed-integer Nonlinear Programming Approaches
for Hydrogen Production Systems


Edward P. Gatzke  -  University of South Carolina


Dr. Gatzke and his research group continue to develop deterministic global optimization methods, which guarantee that the best available point in a search space is located.  Even in cases where the solution cannot be guaranteed to be the global solution, deterministic methods provide bounds on the worst-case value for the solution.  These methods have been applied successfully to online process control applications, where the control system must find the optimal value to regulate a dynamic process.  Traditional local methods provide no guarantees and may provide suboptimal solutions.  Local solutions may allow a chemical plant to experience dangerous transients, while a better solution found by global methods may lead to safer operation and recovery from dangerous events.  These methods are to be used in the future for application to the design of fuel cell systems.

In order to find global solutions efficiently, new methods for the generation of tight bounds on a nonlinear problem have been developed.  Typically, a nonlinear problem that has multiple local minima is bounded from below by a convex nonlinear or convex linear relaxation.  The bounding method uses a piecewise linear relaxation of the nonlinear function, resulting in a tight relaxation and a Mixed-Integer Linear Programming (MILP) problem to be solved.  This is significant in that well-established MILP methods can be used for rapid solution.

In terms of broader impact, a Matlab-based software package for global optimization is nearing completion and will be released as part of the Computational Infrastructure for Operations Research (COIN-OR) repository, free for public use.  Dr. Gatzke has served as head referee for the local middle school FIRST Robotics Lego League competition.  He taught an intermediate level process control elective covering discrete time systems and applied numerical optimization methods.  New experimental systems for demonstration of process dynamics have been developed and details of these experiments will be published in the Journal of Chemical Engineering Education.

The project is notable because chemical process safety is important to industry and this work is not only developing fundamental applied mathematical methodology, but it also has potential applicability in the chemical process industries.

This work involves multidisciplinary research.  Automatic control of chemical processes involves knowledge of the chemical manufacturing process as well as new and novel methods in applied mathematics needed to develop more efficient control algorithms.  The disciplines involved are chemical engineering and applied mathematics.




Edward P. Gatzke Figure

MILP-based Piecewise Relaxation Technique

Credit:  Edward P. Gatzke - University of South Carolina



     
Program Officer:   Maria Burka
     
NSF Award Number:   0238663
     
Award Title:   CAREER: Mixed-Integer Nonlinear Programming Approaches for Hydrogen Production Systems
     
PI Name:   Edward P. Gatzke
     
Institution Name:   University South Carolina
     
Program Element:   1403
     
CBET Nugget:   FY 2006
     

Top of Page


This Nugget was Updated on 24 September 2008.