The UK manufacturing industry is undergoing digital transformation to improve productivity
and support the production of high-value components required for sectors such as energy,
transport, and healthcare. In ultra-precision manufacturing, productivity is limited by dynamic
errors caused by machine motion, structural compliance, and inertial effects. Many of these
errors occur outside the machine servo control loop and cannot be measured or corrected using
conventional encoder-based feedback systems, limiting achievable machining accuracy,
automation, and production speed.
This research investigates a digital twin-driven approach for measuring, predicting, and
compensating for dynamic errors in ultra-precision manufacturing systems. The proposed
framework integrates a collaborative robot (COBOT) and a CNC hybrid milling machine,
enabling automated workpiece handling and machining within a coordinated digital twin
architecture.
A novel sensing methodology using triaxial accelerometers is developed to measure machine
dynamic behaviour outside the control loop. Machine displacement is estimated through signal
processing techniques, including fast Fourier transform (FFT) filtering and double integration
of acceleration data. Experimental validation using ball bar testing demonstrates that this
approach improves COBOT tracking accuracy, achieving an average improvement of 41.3%
compared with encoder-based measurements.
To enable real-time implementation, explainable artificial intelligence (XAI) models are
introduced to predict dynamic error directly from acceleration signal features. Using the
QLattice modelling approach, interpretable mathematical expressions are generated to estimate
dynamic errors with reduced computational requirements. Experimental evaluation shows that
the COBOT digital twin reduces placement error in automated pick-and-place tasks by 31.5% for MoveJ motions and 4% for MoveL motions. The Move Joint (MoveJ) commands, where
the robot follows the most efficient joint path between two points, and Move Linear (MoveL)
commands, where the Tool Centre Point (TCP) moves in a straight line between two positions.
Application of the approach to the hybrid mill demonstrated the ability to detect machine
dynamic behaviour; however, reliable dynamic error correction at micron-level precision was
limited by sensor resolution, accelerometer frequency response, and environmental variation.
A proof-of-concept smart manufacturing demonstrator integrating the COBOT and hybrid mill
was implemented using a CubeSat end cap component. While full closed-loop corrective
machining was not achieved, the system successfully demonstrated coordinated robotic
handling and machining within a digital twin framework.
Overall, this research demonstrates that accelerometer-based digital twins combined with
explainable AI provide a viable approach for measuring and compensating dynamic errors
outside the control loop, particularly for robotic workpiece handling. The work establishes a
foundation for future development of predictive digital twin systems capable of improving
precision, automation, and productivity in ultra-precision manufacturing.
| Date of Award | 24 Apr 2026 |
|---|
| Original language | English |
|---|
| Awarding Institution | - University Of Strathclyde
|
|---|
| Sponsors | University of Strathclyde |
|---|
| Supervisor | Xichun Luo (Supervisor) & Erfu Yang (Supervisor) |
|---|