Traceable Compilation of Multi-Output Total-Process Inventories into Product-Resolved Boundary Models for Life Cycle Assessment

Industrial inventory records commonly describe batches, facilities, unit operations, or reporting periods that combine total inputs, multiple products, wastes, and direct emissions. Product life cycle assessment (LCA), however, requires inventories referenced to a selected product and functional demand. This paper presents a traceable compilation framework implemented in the open-source Nebula LCA software. It preserves unallocated total-process records while deriving product-resolved computational representations through explicit exchange classification, allocation evidence, and unit-aware transformations. The resulting process columns are assembled into a signed technology system and solved to obtain activity levels, inventories, and environmental impact results. Internal process networks can further be compiled as partially terminated systems (PTS), exposing declared products and selected boundary exchanges without requiring every internal process to appear in the surrounding model. The method is evaluated through five progressively composed, refinery-inspired benchmark projects covering a multi-output unit process, a balanced foreground chain, a normalized electricity supplier, an internal recycle loop, and PTS compilation. All five API-executed audits passed. Independent numerical calculations reproduced the classified source inventory, internal PTS activities, declared boundary exchanges, and elementary-flow components; the expanded and PTS-integrated models also produced identical reported EF 3.1 results for the tested demand. The benchmark quantities are hypothetical rather than representative refinery data. Complete ILCD/TIDAS round-trip retention, equivalence across every possible PTS demand, and quantitative privacy guarantees remain outside the demonstrated scope.