Generation of process alternatives using abstract models and case-based reasoning

An approach to manage the complexity involved in the redesign/retrofit of chemical processes is proposed in this paper. This approach uses model-based reasoning (MBR) to generate automatically alternative representations of an existing chemical process at several levels of abstraction. The hierarchical representation results in sets of equipments and sections organised according to their functions and purposes inside the overall process. Our approach is an extension of the multimodelling and the multilevel flow modelling methodologies. These, together with the methodology of Douglas and Turton, are used to represent a process hierarchically, thus improving the identification of analogous units (equipment) or meta-units (process sections) from different chemical processes. The hierarchical representation results in sets of units and meta-units organised according to their functional and teleological model. The implemented tool includes a CBR system to retrieve from a library of cases similar units and/or meta-units to the one selected by the user (similarity is measured according to the function and teleology of a unit and its neighbours). As an output of the CBR system we obtain a set of cases (units and meta-units) ranked according to their similarity. In the case of meta-units, the replacement of process sections may result in changes to the process topology while keeping the same functionality (e.g. a reactive-distillation unit replacing a reactor and separation system). The CBR operates at every level of the abstraction hierarchy both for the problem case and the library cases, opening for the first time the possibility to identify similar solutions at different levels of abstraction and thus widening the scope and the quality of search for similar solutions in the library of cases. (c) 2006 Elsevier Ltd. All rights reserved.