Zur Seitenansicht

Titelaufnahme

Links
Download
Klassifikation
Abstract
  • This article presents research work on a neuro-symbolic decision support system for product planning. It describes the motivation for the work and presents the conceptual framework and the status of implementation. The system to be developed is intended to support engineers in selecting and evaluating assembly and disassembly alternatives during product development. The chosen approach combines symbolic models, represented as Petri nets, ontologies and knowledge graphs, with learningbased, non-algorithmic graph transformation using small language models. Symbolic representations derived from CAD data are formally verified with a theorem prover to ensure feasibility, compliance with constraints and consistency. These verified models provide a reliable basis for decision-making for neural components that generate and evaluate AND/OR Graphs. Instead of programming fixed decision rules, the system learns transformation strategies from verified planning data and human feedback. This enables adaptive decision support for different product concepts. The resulting Decision Support System provides explainable, validated decision alternatives, thus supporting informed decisions in product development, particularly with regard to assembly and disassembly, taking into account circular economy goals
Statistik
  • Das PDF-Dokument wurde 10 mal heruntergeladen.
Lizenz-/Rechtehinweis