Abstract: Seismic-while-tunneling (SWT) detection, serving as a pivotal technology for mitigating the imbalance between mining and roadway tunneling in coal mines, enhances the tunneling efficiency by simultaneous seismic detection and roadway tunneling. However, due to constraints such as highly random seismic sources generated by tunnel boring machines and severe interference from strong underground background noise, data denoising and extraction of effective waves have remained as technical challenges. Hence, this study proposes a dynamic imaging method for SWT detection based on multi-algorithm joint denoising and wavefield separation. First, variational mode decomposition is introduced for the multi-scale decomposition and mode selection of raw seismic signals. In combination with the independent component analysis optimized by an improved artificial bee colony algorithm, a joint denoising model is constructed to deeply suppress complex mechanical and electromagnetic noise. Second, based on the cross-correlation principle, continuous random vibration records are reconstructed into virtual pulse source records, and strong direct waves are removed using singular value decomposition, thereby enabling the precise extraction of weak effective reflected waves. Third, the fine-scale characterization of geological structures along the sidewalls and ahead of the roadway is achieved using a diffraction-scanning migration imaging algorithm. By constructing a SWT advance detection model with dimensions of 300 m × 100 m × 200 m using a three-dimensional finite-difference algorithm for elastic waves, this study verified the capability of the proposed method to reconstruct effective wavefields in environments with low signal-to-noise ratios (SNRs). In the field application at mining face 51210 of the Guojiawan coal mine, the proposed method successfully delineated concealed faults along the roadway sidewall, with the interpretation results highly consistent with borehole verification data. Therefore, the proposed method effectively overcomes the challenge of low SNRs in SWT detection, providing robust technical support for the real-time monitoring of hidden disaster-causing factors and intelligent rapid excavation of roadways in coal mines.