Optimization of VARICOL SMB processes using hybrid modeling and nonlinear programming

Abstract

The Varicol process exhibits continuous and discrete-time dynamics and can be represented by a hybrid model, where the continuous-time dynamics is described by mass-balance partial differential equations, whereas the discrete-time dynamics is described by a timed transition Petri net (TTPN). This hybrid model allows to simplify the solution of a normally difficult Mixed-Integer Non-Linear Programming problem for maximizing the productivity of a Varicol process subject to constraints on the desired purities of the extract and the raffinate components. Indeed, the coverability tree of the TTPN can be constructed to generate a set of candidate integer parameters, and the corresponding nonlinear problems are solved for a subset of candidates selected on the basis of an approximate evaluation using the (true moving bed) triangle theory. The optimal solution is the Varicol configuration (column sequence) for which the productivity is maximal. A repeatability and a robustness analysis show the feasibility of this approach.