Query-Defined Submodels: Publishing Topological Graph Analyses to the Asset Administration Shell

Abstract

The Asset Administration Shell (AAS) standardises how industrial assets are described, but its query language does not support traversal of references between assets, so a client must evaluate patterns spanning several assets on data retrieved in bulk. Graph databases evaluate such patterns natively, yet their results reach only consumers with graph tooling. We publish graph analysis results as AAS submodels, composing an analysis query with an export function. A template records the query, the source revision and the validity, references the affected assets and marks derived values as computed. The submodel’s extent follows from the query, not from a source document. We prove conformance, determinism and traceability, and that for a broad class of local queries the export cannot be inverted. An end-to-end implementation with Neo4j and Eclipse BaSyx measures the export side of a cost model. For 200 drives, the pipeline cuts latency 1.95-fold and volume 5.92-fold against a client filtering by value. A client that also trims, multiplexes and compresses reaches parity. Counting processing and transfer, the graph replica pays off above 0.32 reads per plant change if every tenth change triggers an export, and above 0.60 if every change does. The approach suits selective analyses read by several consumers.

The authors' abstract, as published at the source. Applied Sciences, 2026 · DOI ↗