AutoPCM – An automated LLM-based approach to identify potentials for circular design in automotive electronics

The growth of electrical/electronic (E/E) systems in vehicles intensifies the need to address their environmental impacts in the automotive industry. Existing tools for E/E architecture (EEA) development focus mainly on technical implementation, while corresponding environmental frameworks remain insufficiently integrated at the product design level. This paper introduces AutoPCM, an automated Large Language Model-based approach that augments human expertise by transforming manufacturing documents into product architecture decompositions. AutoPCM generates the Physical Component Mapping, a visualisation method for representing automotive electronic product architectures, and evaluates applicable circular strategies following the Eco-Sensitivity Framework, a product-centered view of possible circular strategies for distributed and centralised EEAs. Large Language Models interpret manufacturing documents to extract component relationships and joining technologies, generate clear matrix-based representations of product structures, and convert them into JSON Linked Data (JSON-LD) for circularity assessment. AutoPCM, implemented in Palantir Foundry, is validated on two automotive case studies, a headlight electronic control unit and a camera sensor, achieving high performance with F1 scores of 0.93–1.0 for matrix heading and 0.82–0.95 for joint coding. The approach enables real-time sustainability feedback, supporting designers and decision-makers in optimising EEAs for the circular economy.