Maintenance work orders are the largest written record of how industrial equipment fails, but in most plants they are close to unusable for analysis. Short texts are abbreviated and misspelt, and asset identity is buried in free-form fields, making it hard to extract structured failure knowledge.

The paper, titled "Plant Memory: Technical Language Processing and a Verified Maintenance Ontology for Agent-Accessible Failure Knowledge from SAP PM Work Orders," proposes a method to process this technical language and build a verified maintenance ontology. The ontology is intended to make the failure knowledge accessible to software agents, enabling automated reasoning about equipment failures.

The approach combines technical language processing with a maintenance-specific ontology, which is verified to ensure consistency. While the abstract does not provide evaluation details, the contribution is a step toward turning unstructured maintenance records into a structured resource for industrial AI systems.