Computation of an extractive distillation column with affine arithmetic

Abstract

The need of reliably solving systems of nonlinear equations often arises in the everyday practice of chemical engineering. In general, standard methods cannot provide theoretical guarantee for convergence to a solution, cannot reliably find multiple solutions, and cannot prove non-existence of solutions. Interval methods provide tools to overcome these problems, thus achieving reliability. To the authors' best knowledge, computation of distillation columns with interval methods have not yet been considered in the literature. This paper presents significant enhancements compared to a previously published interval method of the authors. The proposed branch-and-prune algorithm is guaranteed to converge, and is fairly general at the same time. If no solution exists then this information is provided by the method as a result. Power of the suggested method is demonstrated by solving, with guaranteed convergence, even the MESH equations of a 22 stage extractive distillation column with a ternary mixture. Topical heading: Separations

Computing multiple steady states in homogeneous azeotropic and ideal two‐product distillation

“…The sequence in which the subproblems in Eq. 4 are processed has a great impact on the required computational efforts 45. Achterberg's heuristic45 for subproblem selection is followed.…”

Adaptive filtering strategy for numerical constraint satisfaction problems

“…In some cases, ACID only calls once the method enforcing the weak consistency. The Achterberg' heuristic proposed in Baharev, Achterberg, and Rév (2009) reduces the number of calls of the Simplex algorithm in linear-relaxation-based contractors. These contractors generally reduce each bound of the variable domains by applying the Simplex algorithm to minimize or maximize the variable subject to a linear relaxation of the nonlinear system.…”

Finding All DC Solutions of Diode and BJT Circuits by Interval Method and Modified LP‐Narrowing Technique

“…It can be seen that one needs to solve 4m LP subproblems for narrowing one box. But this number can be reduced by using Achterberg's heuristic [10]. In this implementation, for the current box to be narrowed, we prepare two lists {L} and {U }.…”