Digital twin-based enhancement of vibration measurements for gearbox monitoring under data scarcity

Ensuring the reliability of vibration-based measurements in wind turbine gearboxes is essential for minimizing measurement uncertainty and reducing maintenance downtime. One of the main challenges lies in the scarcity and variability of experimental data, which hinder accurate signal interpretation and uncertainty quantification. To address this issue, this study proposes a Digital Twin (DT)-based measurement enhancement framework that explicitly models the sensor–structure interaction and quantifies the contribution of uncertainty sources, including modeling errors, environmental variability, and sensor noise. The DT acts as a digital measurement model, generating synthetic vibration signals that replicate real measurement conditions with improved repeatability and fidelity. By combining an energetic model of gearbox dynamics with Feature Mode Decomposition (FMD), the framework isolates mechanical components from acquisition noise and reconstructs realistic vibration responses. The extracted noise signatures are then recombined with simulated signals to produce high-fidelity synthetic measurements that statistically reduce variance and enhance repeatability. A 1D Convolutional Neural Network (1D-CNN) is employed as a measurement estimator, improving signal denoising and uncertainty assessment rather than fault classification. Statistical indicators such as standard deviation, expanded uncertainty, and Pearson correlation are used to evaluate uncertainty reduction and measurement consistency. A wind turbine gearbox case study demonstrates that the proposed DT-based framework significantly enhances vibration measurement reliability under data scarcity, providing a scalable approach for uncertainty-aware monitoring of complex rotating machinery. • Digital twin framework enhancing gearbox vibration signals for reliable monitoring. • Feature mode decomposition adapting simulated healthy signals to real data. • Synthetic signals boosting CNN diagnosis accuracy under data-scarcity conditions.