A multi-agent system to facilitate component-based process modeling and design

摘要

Component-based software technology and software interfaces standardization initiatives, such as CAPE-OPEN, have made it possible to model chemical processes and to perform model-based engineering tasks by combining components of process modeling software from different sources, hence providing the potential of exploiting the "best of breed" offered by the CAPE community. In this context, software component libraries, possibly located on a local computer, on the intranet of an organization, or on the Internet, have to be searched to find the most suitable components for a particular engineering task at hand to be integrated into the engineers' computing environment. This paper proposes to address this issue through a multi-agent software system which facilitates the engineers to find and to integrate software components and aims at reducing the engineers' effort to the minimum. Within this system, a directory facilitator serves as the "yellow pages" such that an undetermined set of software component libraries located anywhere may be registered with the system. A matchmaker is used to match the specification of a desired software component with the potential candidates in the relevant libraries. The integration of a matching component into the computing environment is handled by an integration manager. A prototype of such a system, called COGents, has been developed employing an existing multi-agent platform. The ontology OntoCAPE defines the chemical engineering and modeling concepts required for specifying desired software components and for characterizing existing ones. OntoCAPE also provides a shared semantic basis for communication between the software agents. Details of the implementation of COGents are presented and the re-usability of the parts of the COGents system is discussed. Three successful demonstrative applications of COGents are reported, each dealing with different types of tasks, specifically flowsheeting, detailed modeling and process design.