A method to circumvent the combinatorial difficulties in the optimal design of distillation columns

•A single integer column model for structural optimization of distillation units.•MINLP formulation in combination of derivative-free pattern search with SQP.•Systematic enumeration of combinatorial alternatives using pattern search.•-Reliability of the implemented algorithm for use in industrial pr...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	Computers & chemical engineering Ročník 143; s. 107076
Hlavní autoři:	Ishii, Yoshikazu, Otto, Fred D.
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Elsevier Ltd 05.12.2020
Témata:	Combinatorial optimization of distillation columns Derivative-free optimization MINLP (mixed-integer nonlinear programming) Pattern search methods SQP (sequential quadratic programming) Combinatorial optimization of distillation columns MINLP (mixed-integer nonlinear programming) Pattern search methods SQP (sequential quadratic programming) Derivative-free optimization
ISSN:	0098-1354, 1873-4375
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	•A single integer column model for structural optimization of distillation units.•MINLP formulation in combination of derivative-free pattern search with SQP.•Systematic enumeration of combinatorial alternatives using pattern search.•-Reliability of the implemented algorithm for use in industrial practice. A new method is presented for the optimal design of distillation columns with structural and operational decisions. This has been a challenging issue because of difficulties such as complexity of problem formulation, numerical instabilities and lengthy computational times. The new approach presented here solves combinatorial optimization problems by a pattern search method. In the method, the best combination of integer values, each representing the number of stages placed in a column section of a distillation unit, is identified by comparing the objective values evaluated by the embedded NLP solver, while derivatives are not used explicitly. The reliability of the implemented algorithm is numerically confirmed by solving DWC (dividing wall column) problems. This reliability is due to the use of globally convergent pattern search in combination with a full-space Hessian SQP algorithm where global convergence in large and highly nonlinear flowsheeting problems is ensured (Ishii & Otto, 2018).
ISSN:	0098-1354 1873-4375
DOI:	10.1016/j.compchemeng.2020.107076