A useful way to think about an aeroengine digital twin is not as a single 3D model but as a working connection between a physical engine and its digital counterpart. In one common formulation, the digital twin is a three-part model: a physical entity, a digital entity, and the link between them. For engine operation and maintenance, that idea expands into four elements: the real operating environment in physical space; the engine itself; the digital environment and digital entity in digital space; and the two-way exchange of data and information between the two spaces.
Physical space has two parts: the actual operating environment and the engine entity. The environment is defined by the flight envelope and aircraft operating state. The engine entity is the physical engine, including its control system. Digital space likewise has two parts: a digital environment and a digital entity, which are the digital counterparts of the operating environment and the engine. Data moving from physical to digital space includes real-time data and offline data. Real-time data means sensor data; offline data includes historical flight data, delivery data, and maintenance or fault records. Information moving back from digital to physical space is data with behavioral guidance, representing the behavior characteristics of the digital twin.
Constructing an operational digital twin follows a model-driven starting point and then a data-driven correction. First, an initial model is built from engine principles using a model-driven approach. Then different data sources are fused to correct that initial model. The resulting twin supports four functions.
Precise monitoring combines real-time sensor data with a performance model. As the operating environment changes and the physical engine degrades, the model adapts, allowing component-level and whole-engine performance to be monitored accurately. This addresses the problem that, after engine degradation, a model may no longer estimate whole-engine performance parameters accurately in real time.
Fault diagnosis takes fault modes from historical maintenance data and injects them into the three-dimensional physical model and the performance model to create a fault model. During operation, measured data are compared continuously with the model, and similar fault modes are extracted to predict faults. Maintenance and fault records from engines of the same batch are analyzed to form fault modes, which are then injected into the initial model.
Performance prediction combines historical flight data with the performance model and with data-driven methods to build a prediction model for whole-engine performance and remaining life. Operating histories from the same batch of engines can be recorded and fused with a baseline model for prediction.
Control optimization fuses a locally linearized model with an aircraft operating-state environment model to build a control optimization model, enabling engine control performance to be optimized. This addresses engine control optimization within integrated aircraft-engine control.
The construction and use of the twin can be summarized as follows:
graph LR
A[控制器]-->B(航空发动机)
B(航空发动机)-->C[自适应模型]
C[自适应模型]-->D1[推进模型]
C1[参数估计]-->C2[机载模型]
C2[机载模型]-->C[自适应模型]
D1[推进模型]-->D2[优化模型]
D2[优化模型]-->A[控制器]
Aeroengine simulation, which supplies much of the technical basis for these models, has developed along two useful axes. One is simulation by discipline: fluid simulation, or computational fluid dynamics; combustion simulation, or computational combustion; heat-transfer simulation, or computational heat transfer; structural strength simulation, or computational structural mechanics, which calculates structural performance from parts to components, assemblies, subsystems, and the whole engine; and material simulation, which models material composition, structure, properties, and service performance to study phenomena and characteristics across scales.
The other axis is simulation by product-structure level. This covers control-system simulation; hydraulic mechanical actuator simulation; combustor simulation; and simulation of mechanisms such as the compressor and turbine.
Reliability work adds a further way to organize the twin. A reliability digital twin can be built in four layers. The first layer is the complete engine geometry, including dimensions and tolerances. The second layer adds material properties for each component on top of that geometry. The third layer evaluates engine performance from component and system level up to the whole engine. The fourth layer covers fault diagnosis and reliability prediction. At that fourth layer, the twin is no longer only a monitoring tool; it becomes a way to reason about how the engine might fail and how reliable it will remain.