Combining KARL and CRLM for designing vertical transportation systems

); match (); refine () Par ∈ KnownParam ← Par ∈ Param ∧ ¬ dependencies(p: Par). A parameter is inferred as a known parameter if it does not depend on an. ... Figure 1. The different description levels of MIKE parts: Section four describes the model of the domain knowledge, while section five focuses on the problem-solving method which is a variant of propose-and-revise Marcus (1988a). Section six describes the implemented system and section seven supplies a sample trace. The paper concludes with an evaluation and discussion of the achieved solution and, especially, the advantages of the combination of both approaches. 2 Knowledge Modeling Approaches In the following, we will shortly sketch the two different approaches we have combined in this experiment. Both approaches are model-based in the sense that they explicitly distinguish different types of knowledge and use generic problem-solving methods as the behavior model of an expert system. In spite of their similarities, the underlying principles and points of interest differ significantly in the two approaches. A detailed comparison of both approaches can be found in Fensel and Poeck (1994). MIKE is strongly influenced by the results of the KADS-I and CommonKADS projects, c.f. Wielinga, Schreiber and Breuker (1992) and Schreiber, Wielinga, Akkermans, van de Velde and de Hoog (1994), and by work in software engineering and information system design, cf. Angele, Fensel and Studer (1990). It is based on the distinction of different phases in the software development process such as, e.g., analysis, design, and implementation. An important means of MIKE is the formal and operational knowledge specification language KARL, cf. Angele at al. (1994) and Fensel (1995a), which allows a precise description of a model of expertise resulting from the analysis phase. MIKE assumes that, during modeling expertise, a large gap has to be bridged between informal descriptions of the expertise gained from the expert using knowledge acquisition methods and the final realization of the expert system. Decomposing this gap into smaller ones reduces the complexity of the whole modeling process, since in every step particular aspects may be considered independently of other aspects. MIKE provides five different description levels of a task and the required knowledge (see figure 1). First, knowledge and task are described in natural language documents. These documents may result from interviews or observations or can already exist as manuals, books, etc. Second, these