Background <p>In wind tunnel tests, the requirement for structure pressure field monitoring poses significant challenges to simulations and measurements. Machine-learning data fusion could address this, yet demands an efficiency-precision balanced innovative framework.</p> Objective <p>This paper aims to propose a digital twin framework for real-time high-precision pressure field monitoring in wind tunnel tests and utilize historical datasets for smart post-experiment pressure field prediction.</p> Methods <p>Four stages constitute this framework: (i) Offline stage I: establishing a surrogate mapping from airflow conditions to reduced-order pressure fields by the proper orthogonal decomposition and Gaussian process regression, (ii) Online stage I: predicting simulation data at target airflow conditions in real time and building pre-trained model, (iii) Online stage II: fusing simulation and experiment data efficiently by transfer learning and generating digital twin for pressure field monitoring, (iv) Offline stage II: constructing digital twin database and predicting pressure fields at new airflow conditions by another surrogate.</p> Results <p>In the example of the ONERA M6 wing, the sub-minute established digital twin provides a three-dimensional high-precision pressure field compared with experiment data. Precision of pressure fields for the same wing at new airflow conditions predicted with the digital twin database also surpass simulations. The performance of core methods is investigated and the generalization assessment is conducted on another wing.</p> Conclusions <p>The proposed framework efficiently generates the digital twin in high precision and enables smart prediction for new airflow conditions.</p>