Automated Planning for Production Routines in Semiconductor Manufacturing

Abstract

Semiconductor device manufacturing involves complex production routines comprising hundreds of sophisticated steps, e.g., Lithography, Epitaxy, and Ion Implantation. Formulating these routines is challenging, as it requires a deep understanding of the technological interdependencies between operations and the ability to anticipate the consequences of long operation sequences. Currently, this process remains largely manual, with minimal tool support. This paper presents a knowledge-driven architecture for automated manufacturing routine composition, integrating expert knowledge acquisition through ontologies and Knowledge Graphs. The system enables the automatic generation of classical and hierarchical planning domain specifications, which are then used by automated planners for routine construction. Domain experts utilised the developed prototype to effectively model real-world semiconductor manufacturing use cases and compute corresponding routines, thereby demonstrating the practical applicability of the proposed approach.