F.2 Initialisation for large problems

Distillation simulation
Solution of very large steady state simulation problems (more than 10000 equations and variables) may require very long computing times, even with the fastest computers. If steady state simulation is performed for process optimisation, even longer computing times may be necessary. One way of reducing the computing times is to provide better initial estimates so that a faster convergence to the solution can be obtained for the simulation/optimisation problem. For single and multiple column distillation problems, an efficient initialisation procedure has been developed and validated with several test examples. Large reduction of computational times have been achieved by first solving a reduced problem. A unique feature of the reduced problem is that it represents a linear set of algebraic equations and the key distillation model variables (for the large problem) is predicted with reasonable accuracy, thereby, providing very good initial estimates. The applicability of this approach for other problems will be investigated in future.