Abstract
In this work we present a new approach that we use to simulate and optimize multiple dividing wall columns at the same time. Instead of considering all model equations as constraints and all process variables as optimization variables in a large and highly nonlinear optimization problem we only incorporate a subset of the model equations as constraints and a subset of the process variables as optimization variables. The remaining process variables are calculated from this subset by a robust and fast calculation procedure. This calculation procedure also ensures that the remaining model equations are satisfied. A comparison with the commercial process simulator Aspen Plus shows that with the new approach multiple dividing wall columns can be optimized more stable and better solutions are found. Moreover the time needed to find an optimal design decreases significantly.