Knowledge Engineering for Embedded Configuration

This thesis presents a way to simplify setup of complex product systems with the help of embedded configuration. To achieve this, one has to focus on what subsystems need to communicate between themselves. The required internal knowledge is then structured at three abstraction levels. Simplifications of the internal workings are both due to hardware- and application-induced configuration taking place both within the overall system and in each subsystem. By relating parameters in such a way, the number of user inputs or decision variables should decrease drastically, thus increasing the overall usability of the installation. In our case, we have rationalized that this should be done with embedded configuration, and the expected result is enhanced usability. The suggested method is deeply rooted in system theory. It draws on the emergent properties expected from the system, and tries to embed into the system the knowledge needed to achieve them. In order to understand the system, one draws simplified functional streams and identifies archetypes from the product assortment, and then one maps the two together into a system breakdown model. The system model indicates how many encapsulation models (EMs) should be made and the first decomposition in their tree centrepiece. the encapsulation model describes the archetype on the three abstraction levels: application, function, and the physical artefact. All levels are connected through relational matrixes both for internal and mapping relations, and predefined relation types are suggested. The models are stringent and thought out so they can be implemented in software. They should allow both import and export of product knowledge from the knowledge-based system. The purpose of this work is to simplify the installation process of product systems that have been treated with extreme postponement, meaning that variance is given with variables while installing. Variables here can be both software- and hardware-like in nature. These variables are defined as decision variables, and it is the reduction of these variables that is the overall goal. The next step can be said to be two-fold: first, to construct a system based on this philosophy and to show that it actually leads to the expected results. And second, to further develop the modelling tools and methods for supporting the making of embedded configuration systems, or in essence, a distributed artificial intelligence system.