Abstract: To overcome the challenge of real-time acquisition of in-situ stress during drill-and-blast tunnel construction, the authors develop a high-precision synchronous acquisition equipment for drilling parameters using an impact rotary drilling system and a three-dimensional controllable stress loading platform. Twelve while-drilling tests are conducted under different surrounding rock stress states, systematically revealing the response laws between drilling parameters and ground stress, and between ground stress and wave velocity. Based on this, a five-dimensional while-drilling feature-wave velocity sample library of 6000 multi-condition features is constructed, and a Bagging decision tree integrated model is designed to achieve intelligent prediction of triaxial wave velocity. The magnitude of principal stress is then inverted based on acoustoelastic theory. Meanwhile, an analytical algorithm for principal stress direction based on wave velocity ellipsoid regression is proposed. Results show that the prediction accuracy of the triaxial wave velocity model ranges from 84.44% to 88.21%, and the principal stress magnitude inverted using the predicted wave velocity agrees well with the loading value. The maximum principal stress direction error analyzed by the wave velocity ellipsoid method is ≤0.1°, and the errors of other principal stress directions are ≤18.3°. The research findings provide a new approach for the rapid and intelligent analysis of the stress field of tunnel surrounding rock and provide a methodological reference for stress analysis in high-stress tunnels.