Runtime Governance of Competing Actuation Requests in Multi-Service Digital Twins for Manufacturing Cells

Manufacturing-cell digital twins increasingly combine independently developed services that can influence shared physical resources. When several state-changing services are individually admissible but request incompatible values for the same control endpoint, runtime governance is required before execution. This work proposes a governance model that represents service–resource bindings, service and resource constraints, runtime context, arbitration-policy bindings, and enforcement relations for composed manufacturing-cell DTs. State-changing requests are evaluated individually before compatible requests are coalesced and incompatible admissible requests are resolved through a resource-specific governance policy. The model is implemented using ODRL for portable policy representation and OPA/Rego for executable runtime decisions. The evaluation examines functional validity, deterministic decision behavior, and request-to-enforcement latency using recorded CNC machining data and controlled replay scenarios in which two DT services share the feedOverride resource. All eight validation scenarios produced the expected decisions, including denial, delay, coalescing, and arbitration, while the simulated enforcement path produced at most one governed write per cycle. The OPA policy round trip showed a median latency of 4.962 ms (p95: 6.709 ms) and the complete governed request-to-enforcement path 15.003 ms (p95: 28.046 ms). The results demonstrate the feasibility of explicit runtime governance for competing DT-service actuation requests in the evaluated manufacturing-cell setting.