Bayes-Netzwerke für die Kostenprognose in der frühen Phase der Produktentwicklung

  • Bayes networks for cost prediction in the early stage of product development

Dehen, Frank; Feldhusen, Jörg (Thesis advisor)

Aachen : Shaker (2012)
Dissertation / PhD Thesis

In: Schriftenreihe Produktentwicklung und Konstruktionsmethodik 13
Page(s)/Article-Nr.: XIII, 126 S. : Ill., graph. Darst.

Zugl.: Aachen, Techn. Hochsch., Diss., 2012

Abstract

Experience gleaned in the field of product development shows that the early development phase is the phase during which product developers are most apt to be plagued by an acute sense of insecurity. Many of the product’s characteristics have yet to be determined. The results of the work that has been carried out are at this stage not accurate enough, incomplete and sometimes even flawed. The product developer is nonetheless obliged to make conceptual decisions on the basis of these criteria and to define the product’s central characteristics. These decisions not only determine fundamental product characteristics but also have a direct affect on the product development time and the product development costs. If the product that is to be developed has a high degree of novelty, then there is only a limited number of reference points available which can be used to perform a cost calculation using established procedures. This paper presents a method which allows products with a high degree of novelty to be assessed at a very early stage of their development with regard to their manufacturing costs. This method precedes the actual cost calculation. It closes the gap to the calculation process which requires established functions and solution principles in order for the costs to be assessed. To achieve this end, this paper deals with the development of a methodical and systematic approach to cost prediction using Bayes networks. The purpose of a Bayes network is to predict and diagnose factors or events, which because of existing information are either not obviously apparent or too complex to determine. Bayes networks are a special form of theoretical probability models. Following a comprehensive introduction into the fundamental principles of Bayes networks, this paper goes on to present guidelines for the generation and implementation of own networks. Tools that are needed for each of the necessary steps that are connected with generating and working with a network are also presented. This method is not only excellently suited for classic elements of engineering, but also for other fields and sectors like mechatronics or electrical engineering. In addition, universal network structures are developed which can be used for almost all product domains thereby making it easier to set up an own network. A detailled example is used to illustrate the approach to the solution of this problem.

Identifier