Selection of separation sequences by case-based reasoning
Pajula, E., Seuranen, T. and Hurme, M.
Helsinki University of Technology, Laboratory of Chemical Engineering and Plant Design,
P.O. Box 6100, FIN-02015 HUT, Finland
The objective of this paper is to introduce a method for finding feasible separation process
sequences and separation process structures utilising case-based reasoning (CBR). This means
finding the most similar existing separation processes and applying the knowledge of their
concept and separation sequencing for solving new problems in the early phases of process
design.
1. INTRODUCTION
Typical task in process design is to determine the configuration of a separation sequence.
When dealing with multicomponent mixtures, the number of possible separation methods,
their combinations and process structures to be screened is huge as well as the work involved.
The method used, case-based reasoning (CBR), uses existing design cases stored in the
database for solving new separation problems. The synthesis method studies the physical and
chemical properties of the species present in the mixture and uses the properties presenting
most favourable possibilities for successful separation for retrieving the nearest cases to the
current problem. The presented method has the advantage of not losing any information
because no generalisations are used.
2. METHODOLOGY
It is a well-known fact that in the majority of cases distillation is the most feasible way to
separate components (Barnicki and Fair 1990). Therefore the distillation related properties are
studied first in the methodology. The strategy is to find first a feasible distillation sequence
for the separations where ordinary distillation is possible, and then to solve the remaining
separation problems with further reasoning which apply other separation methods than
ordinary distillation (steps 2-4 in Table 1). The main steps of this algorithm are presented in
Table 1.
Step 1: For all components α’s are calculated and the presence of reactive components is also
considered. The most similar cases to the current problem are searched from database based
on these parameters. When it is possible (and no known cases for better procedures are found)
ordinary distillation is applied using the same separation strategy as defined by the sequence
of the nearest case found in database. The separation strategy is described in the cases as a set
of heuristic rules or as textual description.
The separations are classified in the database based on relative volatility (α29 values as easy
(α>=1.2), possible, where mass separating agent (MSA) could be useful (1.1< α <1.2) and
difficult (α =<1.1) separations. In a simple situation a search using component names would
© 2001 Elsevier Science. Reprinted with permission from Computer-Aided Chemical Engineering, Vol. 9, 469-474, 2001.