Abstract: Underwater target direction finding remains a critical research focus for towed linear arrays. While Time Difference of Arrival (TDOA) methods using three-element arrays offer simpler implementation and lower costs compared to conventional beamforming, their direction-finding accuracy significantly deteriorates under low Signal-to-Noise Ratio (SNR) conditions. To address this limitation, this paper proposes a joint optimization strategy combining Variational Mode Decomposition (VMD) and Fast Independent Component Analysis (FastICA), incorporating dynamic Intrinsic Mode Function (IMF) selection through power spectral entropy thresholding. This approach eliminates signal distortion caused by fixed thresholds in traditional methods while effectively enhancing denoising performance for received signals. Compared with traditional EMD-based methods, VMD-FastICA demonstrates stronger robustness in simulations. In 50 simulation experiments, when the target signal is a narrowband signal or a frequency-modulated signal and the signal-to-noise ratio is 0 dB, the mean direction-finding error using a three-element array after VMD-FastICA denoising is within 0.2°, and the maximum error in a single experiment is 1.58°.