SQP SAND Strategies that Link to Existing Modeling Systems
We consider variants of Successive Quadratic Programming (SQP) applied to Simultaneous Analysis and Design (SAND)implementations for existing engineering modeling systems. Such programs include large scale PDE solvers, chemical process simulators and programs for electrical circuit analysis. Several aspects of the SQP algorithm will be considered. In particular, options need to be selected that are compatible with the information available from the modeling system. Here the following three areas are considered for SAND optimization algorithms: efficient and reliable strategies for global convergence based on a newly developed filter line search method, improved second order information for reduced Hessian methods, and a barrier approach for inequality constraints. The benefits of each of these options will be supported by numerical results and the relevance of these options to PDE models will be explored.
KeywordsLine Search Merit Function Exact Penalty Newton Step Exact Penalty Function
Unable to display preview. Download preview PDF.
- 4.Biegler, L. T., Claudia Schmid, and David Ternet, “A Multiplier-Free, Reduced Hessian Method For Process Optimization,” Large-Scale Optimization with Applications, Part II: Optimal Design and Control, L. T. Biegler, T. F. Coleman, A. R. Conn and F. N. Santosa (eds.), p. 101, IMA Volumes in Mathematics and Applications, Springer Verlag (1997)Google Scholar
- 5.Biegler, L. T., A. Cervantes and A. Wachter, “Advances in Simultaneous Strategies for Dynamic Process Optimization,” Chemical Engineering Science, to appear (2002)Google Scholar
- 6.A. S. Bondarenko, D. M. Bortz, and J. J. Moré, “COPS: Large-Scale Nonlinearly Constrained Optimization Problems,” Technical Memorandum ANL/MCS-TM237, September 1998, revised October 1999.Google Scholar
- 8.E. D. Dolan and J. J. More, “Benchmarking Optimization Software with COPS,” Technical Memorandum ANL/MCS-TM-246, November 2000.Google Scholar
- 11.Fletcher, R. and S. Leyffer, “Nonlinear programming without a penalty function,” Technical Report NA/l71, University of Dundee Numerical Analysis Report, Dundee, Scotland, UK (1998)Google Scholar
- 14.G. Itle, A. Salinger, R. Pawlowski, D. Alkaya and L. Biegler “Constraint Aggregation for Large-scale NLP Algorithms,” presented at Annual AIChE meeting, Los Angeles, CA, November, 2000Google Scholar
- 17.Schmid, C. and L. T. Biegler, “A Simultaneous Approach for Process Optimization with Existing Modeling Procedures,” Proc. European Symposium on Computer Aided Process Engineering, I Chemical Engineering Symposium Series No. 133, p. 231 (1994). (Also in Chemical Engineering Research and Design, Part A, 72, p. 382 (1994))Google Scholar
- 19.van Bloemen Waanders, B., A. Salinger, R. Pawlowski, L. Biegler and R. Bartlett, “SAND Optimization of Massively Parallel Simulation Codes Using and Object Oriented Framework,” presented at SIAM Meeting on Parallel Computing, Norfolk, VA, March (2001)Google Scholar
- 20.Wächter, A. and L. T. Biegler. “Global and local convergence of line search filter methods for nonlinear programming,” Technical Report CAPD B-01-09, Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA, USA (2001)Google Scholar