(604l) Multi-Objective Optimization in Industrial Chemical Process Development by Navigation On Pareto Fronts

Process development in the chemical industry is a multi-objective optimization problem. There are many objectives, e.g. product quality, operation and capital cost, energy efficiency or reliability. For a given process configuration and given design specification, typical parameters of this optimization problem are, e.g. reactor volumes, overall and feed stage numbers of distillation columns, temperatures, pressures, mass flows and compositions of internal streams.  In industrial process design, the developer usually generates only a certain number of process variants by varying process parameters intuitively. The aim is to find a solution that is acceptable regarding all or at least most criteria, optimality can usually not be guaranteed.

In the present work a feasible alternative approach is presented. It is based on the concept of Pareto optimality. The Pareto front describes the set of solutions for which no improvement in one objective can be reached without a decline in at least one other objective. Hence the Pareto front describes all optimal trade-offs between objectives. The final choice among these alternatives is left to the engineer. In the present approach, tools were developed and implemented in BASF’s in-house process simulation environment that a) allow determining the Pareto front for a given design problem and b) support the user in the exploration of that front by navigation. The depth of the underlying process model can be varied from “short-cut” to the industrial standard of “equilibrium-stage based”. The navigation on the Pareto front enables the designer to efficiently find the best compromise between the objectives. Upon navigating, the designer gains insight in the solution space that allows him to rationally justify his final decision.

The Pareto front is approximated within a user defined accuracy by highly efficient optimization techniques; at the same time, a robust and accurate process simulation is used.

In the presentation the approach will be described and results of the application of the Pareto navigation for different processes will be presented. The examples include a comparison of a pressure swing distillation to an entrainer distillation as well as the optimization of an industrial reaction-distillation train system.