Weak-Signal Reconstruction under Strong Noise in Borehole Transient Electromagnetics

The borehole transient electromagnetic (BTEM) method is a key technology for metallic mineral, groundwater, and related resource exploration, with measured responses exhibiting exponential decay over multiple orders of magnitude in dynamic range. However, late-time signals attenuate to extremely low amplitudes and become submerged in noise, creating a bottleneck for exploration depth. To address the limitations of conventional processing approaches under low signal-to-noise ratio conditions, we propose a reconstruction framework that exploits the non-stationary nature and time–frequency evolution of BTEM decay curves. The framework incorporates two complementary mechanisms. A time–frequency ultra-pyramid fusion module captures the evolving decay behavior in the time–frequency domain and enables robust separation of signal and noise. In parallel, a noise-aware gating mechanism learns point-wise reliability weights from feature statistics to regulate activations, suppressing noise-dominated late-time components while retaining informative signal content. Validation on synthetic and field datasets demonstrates that the proposed approach extends the effective observation window and reliably recovers weak signals from the noise floor, thereby enhancing the achievable exploration depth for BTEM.