Knowledge engineering in CAPE systems -Can chemical engineers finally get around programming ?
Systematic methods and tools for managing the complexity
Keynote Lectures: Theme-4
Keywords: CAPE tools, software engineering, integration
The development of CAPE tools requires experts from the field of chemical engineering as well as software engineering. A major reason is that the software developed is expressed in program code that mixes aspects of the chemical engineering and software engineering domains. This code is hard to understand for chemical engineers in order to support validating or extending this code. Therefore, program code is one of the worst representations of domain knowledge the author can think of. This drawback is making tasks like construction, validation, maintenance, extension, or integration of CAPE systems difficult and costly as the presentation will explain.
A better approach is to consider the development of CAPE tools from two orthogonal aspects: On the one hand, we need a domain-specific problem description which is posed in concepts familiar to the domain expert. On the other hand, we need applications which can be fed with problem description from our domain and are able to compute the desired results. Mathematical modelling environments are a good example where this approach is already applied: Engineers are nowadays developing models in tailored modelling languages rather than coding subroutines for numerical solvers in Fortran as they did not too long ago.
An interesting area to learn from is knowledge engineering which has evolved from an active field of research into a standard technology applied in software development. Techniques such as model-driven architectures, domain specific languages, or ontologies aim at an explicit representation of knowledge about a domain (such as chemical engineering in this case). Due to the formal and precise nature of these representations, the computer is able to exploit this knowledge and make use of it in problem solving. The underlying techniques are maturing and their use in commercial software is increasing because programming efforts can be reduced or avoided at all.
As a consequence, applications of these techniques have the potential to tremendously change the way CAPE systems are being developed. Not only will required skills of individuals change, but also the lifecycle of CAPE systems from research towards commercialization may undergo a radical change. This contribution presents some of these approaches and discusses chances to achieve a new generation of CAPE systems that would be impossible to construct using traditional programming.
Presented Thursday 20, 14:40 to 15:20, in session Keynote Lectures: Theme-4 (T4-K3, K4, K5).
