A digital twin is only as good as the measurements behind it

Almost every aerospace manufacturer now has a digital twin on its roadmap. The promise is attractive: a virtual copy of the product and the assembly line that predicts problems before they reach the shop floor. What gets less attention is where the "twin" part comes from. A CAD model on its own is not a twin. It becomes one when it is updated with data from the real part, and in assembly that data comes from metrology.

What a digital twin actually is

The AIAA and the Aerospace Industries Association define a digital twin as "a set of virtual information constructs that mimics the structure, context and behavior of an individual / unique physical asset, or a group of physical assets, is dynamically updated with data from its physical twin throughout its life cycle and informs decisions that realize value" [1].

Three words in that definition matter most for manufacturing:

• Individual. The twin represents this wing, this fuselage section or this jig, not the nominal design.
• Updated. It changes as real data arrives. A model that is never measured against reality is a simulation, not a twin.
• Decisions. It exists to change what someone does next, such as the shim to cut, the hole to drill or the fixture to adjust.

Where twins are already paying off in assembly

The clearest results so far are in predicting gaps between large parts before they are joined. Traditionally, panels are trial-fitted, gaps are measured, shims are made and the assembly is refitted. Each loop takes time and floor space.

• Researchers working with Boeing data from 54 commercial aircraft used historical gap measurements to decide where to measure. Their method predicted 99% of shim gaps within the required tolerance using 3% of the laser scan points normally collected [2].
• A physics-driven digital twin for aircraft skin panels combined in-line measurement with variation simulation. It reported about 98% correlation between predicted and measured shim thickness, and a 75% time saving from fewer measure-fit-adjust loops, which the authors note can each take 10 to 14 days [3].
• Robot cells benefit too. A 2024 study used a laser tracker to build the digital twin of a robotic station, because an offline program is only as accurate as the model of where the robot, fixture and part really are [4].

Airbus's Wing of Tomorrow programme points the same way at industrial scale. Its digital smart factory solution, developed with Capgemini, combines IoT sensors, geolocation, automated guided vehicles, robotics, machine learning and digital twins to improve first-time assembly and reduce cycle time [5].

The metrology questions most twin projects skip

A twin fed with poor data will give confident, wrong answers. Before trusting one with production decisions, these questions need clear answers:

1. What is the measurement uncertainty? Every value entering the twin should carry an uncertainty estimated in line with the GUM [6]. A predicted 0.3 mm shim means little if the measurement behind it is uncertain by 0.25 mm.
2. Is the reference frame the same everywhere? Laser trackers, scanners, robots and CAD often use different coordinate systems. Errors in how they are tied together go straight into the twin.
3. Is the instrument still in specification? Interim checks, such as the laser tracker test in ASME B89.4.19 and ISO 10360-10, give evidence that the data is trustworthy on the day it was taken.
4. Is the part in the state the model assumes? Large thin-walled structures deflect under gravity, clamping and temperature. Measuring a part in a different condition from the one the model represents builds in error.
5. Is the data traceable and audit-ready? If a twin is used to accept a part or select a shim, the data trail needs to survive a customer or regulatory audit.

Where to start

• Pick one decision, not a platform. Shim prediction, robot cell calibration or jig health monitoring each give a measurable return. Start with the one costing you most rework.
• Audit the measurement process first. Map every instrument, frame transformation and uncertainty contribution feeding that decision.
• Measure less, but better. As the shim study shows, well-chosen measurements combined with historical data can beat dense scanning.
• Close the loop. Track the twin's predictions against what actually happened on the line, and use the difference to improve both the model and the measurement plan.

Digital twins hold real value for aerospace manufacturers, especially as production rates rise. The ones that deliver are built on measurement processes that were designed for them.

HK Robotics and Metrology helps manufacturers design the measurement processes that digital twins depend on, from uncertainty budgets and coordinate frame strategy to laser tracker and robot cell integration. Book a discovery call to talk through your assembly programme.

References

1. Digital Twin: Definition & Value. An AIAA and AIA Position Paper. AIAA Digital Engineering Integration Committee, December 2020.
2. Predicting shim gaps in aircraft assembly with machine learning and sparse sensing. Journal of Manufacturing Systems, 2018.
3. Virtual shimming simulation for smart assembly of aircraft skin panels based on a physics-driven digital twin. International Journal on Interactive Design and Manufacturing, 16, 753–763, 2022.
4. Creating Digital Twins of Robotic Stations Using a Laser Tracker. Electronics, 13(21), 4271, 2024.
5. Shaping the wing of tomorrow with Airbus. Capgemini, 2025.
6. JCGM 100:2008 Evaluation of measurement data: Guide to the expression of uncertainty in measurement. BIPM, 2008.