|M.Sc Student||Guttman Gil|
|Subject||Creating Teams in a Network by Means of Knowledge Maps|
|Department||Department of Mechanical Engineering||Supervisor||Professor Emeritus Moshe Shpitalni|
|Full Thesis text - in Hebrew|
The product life cycle is a comprehensive engineering research field encompassing a diverse set of topics and subtopics that can be organized into a tree structure map. The ontology method offers an appropriate solution to the problem of forming an optimal map. In this method, diverse participants can contribute to organizing the map in order to achieve wide-ranging agreement regarding how to represent the product life cycle. Questionnaires were distributed among the 25 laboratories comprising the VRL-KCiP Network of Excellence (participants), in order to determine their knowledge regarding the product life cycle. Based upon participant responses, a tree structure map was built describing the expertise of the participants. Ultimately, 800 topics and subtopics were identified.
The goal of the current research was to find an optimal group of participants from among a large group of potential participants in order to carry out a specific project or projects. The objective was to find such a group of participants within a reasonable amount of time, subjected to a complex set of constraints related to covering the required knowledge by number of participants. In short, the problem is a set covering problem, which is known to be difficult and non-polynomial (NP). Therefore, a number of heuristics were developed.
Another objective was to examine the use of genetic algorithms, determined to be a suitable heuristic approach with promising potential for effectively coping with coverage problems. The main parameters of the genetic algorithm were optimized over hundreds of trials in order to achieve the best solutions within a reasonable amount of time.
Within this research we developed and implemented a new adaptive genetic algorithm based upon changing the mutation probability. The innovation of this algorithm is in the dependence of the mutation probability on the fitness values of the two parents.
In the course of the research, software with a graphical interface was developed. The software is suitable for handling more complex constraints with larger number of potential participants. Thus, if the NoE grows, the software can still continue to find optimal groups of participants.