Digital twin validation and maintenance across the lifecycle
A digital twin is not a static product delivered at a given point in time, but a computational representation that must be kept aligned with the physical system it models. In critical infrastructure environments, this alignment is an operational condition, not an optional enhancement. Continuous validation of the twin consists in periodically comparing its outputs, predictions and internal states against observed data from the real system, whether field measurements, onboard telemetry or data from external sources such as Earth observation imagery or high-precision GNSS readings. When the divergence between the model and reality exceeds thresholds defined by the operator, this is referred to as model drift, a phenomenon that may originate in physical changes to the environment, sensor degradation, updates to operational parameters, or simply the accumulation of numerical errors over time. The process of updating the twin involves distinguishing between two types of intervention. The first is data assimilation, whereby the model incorporates new observations to correct its internal state without altering its underlying structure. The second is structural revision, which requires modifying the twin's own algorithms, causal relationships or geospatial databases when the physical system has undergone changes that the model can no longer represent adequately. Both interventions must be documented and traced, which leads directly to the need for a robust versioning system that makes it possible to audit what the twin represented at each point in time and under what conditions it was validated. The versioning of the digital twin has direct implications for the chain of responsibilities. In regulated contexts, such as the management of transport infrastructure, energy networks or defence systems, operational decisions may be supported by the twin's outputs. This requires that each version of the model be associated with a set of validation data, a record of the sources used and an explicit statement of its conditions of applicability. The absence of this traceability framework reduces the twin to an internal analysis tool, but disqualifies it as a decision-support instrument in environments subject to audit or certification requirements. From the perspective of operators and institutions of the European Union, the governance of the digital twin raises questions about the ownership of validation data, the minimum frequency of comparison against reality and the acceptance criteria for a new version of the model. These questions have no universal technical answer, but depend on the application domain, the criticality level of the system being modelled and the applicable sectoral regulatory frameworks. What can be established as a general principle is that a twin without a documented validation process and without a defined update policy progressively loses its operational utility, regardless of the sophistication of its initial architecture.
NASSAT - Network Satellite Systems