Generative design only produces manufacturable designs if design and manufacturing knowledge enters the generation process. However, this knowledge is mostly unstructured, for example in design methods, design guidelines and manufacturing guidelines. Formalizing it manually into executable rules is laborious and does not scale.

The institute has developed a rule-based generative design algorithm for bent sheet metal parts that systematically integrates design and manufacturing knowledge into the generation process. Its rule base has so far been derived and implemented manually from technical literature and expert knowledge.

Large Language Models (LLMs) can process large volumes of technical documents, interpret domain knowledge and generate program code. This opens up the possibility of automatically translating documented design knowledge into executable generation strategies and continuously extending the knowledge base of generative algorithms.

Tasks

The goal of this thesis is to develop an AI-based approach that automatically processes knowledge sources and derives executable strategies for the existing generative design algorithm.

Possible tasks include:

• Selection and preparation of suitable knowledge sources (e.g. design methods, design and manufacturing guidelines) for automated processing
• Design and implementation of an LLM-based pipeline for extracting and interpreting design knowledge (e.g. using Retrieval-Augmented Generation)
• Derivation of generation strategies and their translation into executable code for the existing algorithm
• Integration and testing of the generated strategies in the algorithm's Python environment
• Evaluation of the strategies regarding correctness, executability and impact on generation quality
• Analysis of the potential and limitations of LLM-based knowledge integration in generative design

The specific focus of the thesis can be adapted to your interests and field of study.

Profile

• Studies in mechanical engineering, mechatronics, computer science or a related field
• Interest in Large Language Models, knowledge processing or generative design
• Programming skills in Python

If you are interested, feel free to contact me: christoph.wittig∂kit.edu