Optimisation de procédés chimiques par une méthode de gradient réduit. II: Exemples d’illustration. Comparaison avec d’autres méthodes. (Optimization of chemical processes by a reduced gradient method. II: Illustrative examples. Comparison with other methods). (French) Zbl 0581.90082

[For part I see ibid. 19, 247-274 (1985; Zbl 0571.90055).] - A reduced gradient algorithm, designed for solving several optimization problems in the field of Chemical Engineering has been presented in part I of this article. The algorithm is now compared on the one hand, with some unconstrained optimization methods (Hooke and Jeeves, Rosenbrock, conjugate gradient) needing some transformations of the problem statements, and on the other hand, with Rosen’s projection method. The resolution of three numerical problems and of three chemical engineering problems (chemical equilibrium, compressors sequence, and simplified cellosolve acetate plant optimization) is detailed.
Then, the reduced gradient algorithm is particularly illustrated by the means of seven examples including four numerical large-scale problems (up to 302 variables), and the optimization of an ethyl acetate production, of an industrial reactor and of a complete cellosolve acetate plant. The reduced gradient algorithm and Rosen’s method are nearly equivalent for solving problems involving few variables (\(\leq 5)\). Nevertheless, the advantage of the reduced gradient procedure is greater as the number of variables increases (\(\geq 20)\).


90C30 Nonlinear programming
90C90 Applications of mathematical programming
65K05 Numerical mathematical programming methods


Zbl 0571.90055
Full Text: DOI EuDML