Physics-informed neural networks (PINNs) encounter ill-posed optimization, loss competition, and parameter compensation in partial differential equation (PDE) inverse problems. Transfer learning can reuse representations from source tasks, but direct fine-tuning may introduce negative transfer when dominant physical mechanisms, governing parameters, or observation noise differ between source and target domains: the model achieves low field error yet recovers incorrect target physical parameters. To mitigate, we propose Target-Guided Selective Reweighting PINN (TGSR-PINN), a target-evidence-driven representation correction method for PINN inverse transfer learning. TGSR-PINN transfers only the weights and biases from the source PINN, while target physical parameters are independently initialized; after a short target-adaptation phase, the method computes neuron target scores using first-order Taylor sensitivity and pre-activation variance on fixed scoring batches, and converts evidence associated with low-scoring neurons into continuous weak-adaptation signals via a Gaussian mixture model (GMM) with rank fallback. TGSR-PINN then applies selective soft decay to input weight rows and biases of low-scoring neurons instead of hard pruning or random resetting. In experiments, TGSR-PINN improves target parameter recovery while maintaining comparable field accuracy in the high-Péclet 2D advection-diffusion task and in the Allen–Cahn to Burgers cross-PDE-family transfer task; a 5
To meet the high reliability and fast isolation requirements of main protection in MMC-HVDC grids, this paper proposes a single-ended transient protection principle based on Spectrum-Guided Adaptive Variational Mode Decomposition (SGA-VMD) and high-frequency energy spectrum relative entropy. First, analyzing the boundary filtering effect of smoothing reactors and the inter-pole electromagnetic coupling mechanism of bipolar lines rev eals high-frequency energy attenuation laws and morphological differences in inter-pole energy spectra during internal and external faults. Second, utilizing the frequency-domain sparsity of transient signals, an SGA strategy is designed for one-step adaptive locking of VMD parameters. Furthermore, integrating the Teager Energy Operator (TEO) enables high-frequency instantaneous energy extraction within a short data window. On this basis, an energy ratio criterion is constructed using the reactors' blocking effect on high-frequency components for internal and external fault identification. Simultaneously, relative entropy is introduced to quantify the spatial distribution asymmetry of positive and negative high-frequency energy spectra, establishing a reliable fault pole selection criterion. Finally, a 500 kV two-terminal half-bridge MMC-HVDC simulation model in PSCAD verifies the method under various transition resistances, signal-to-noise ratios, and internal/external faults. Results demonstrate that the proposed method achieves reliable line protection with low computational overhead and exhibits good robustness against high transition resistances and strong noise.
Orthogonal frequency division multiplexing (OFDM) has attracted extensive attention in Internet of Underwater Things communications owing to its superiority in achieving high data rate transmission as well as its multi-user and multiple access provisions. However, the time-varying Doppler effects induced by platform mobility significantly limit the demodulation performance of OFDM in underwater acoustic (UWA) channels. Conventional Doppler estimation strategies that search for the best match between the received signals and local copies generate mismatches in adverse dynamic UWA channels. Therefore, this study proposes a time-varying Doppler estimation algorithm based on an optimal bit-error information criterion for suppressing time-varying Doppler-induced distortion via symbol-by-symbol Doppler estimation and compensation. To this end, numerical simulations and sea trial experiments are performed to demonstrate the effectiveness and superiority of the proposed estimation algorithm relative to those of block estimation, cross-ambiguity function estimation, cross-correlation quasi-gradient estimation, bit error rate search estimation, and cyclic prefix-based differential phase estimation strategies.
Natural gas hydrate (NGH) is a promising clean unconventional energy resource, with global research evolving from early field investigations to offshore field trials. However, no breakthrough in commercial NGH exploitation has been achieved to date, constrained by core bottlenecks including insufficient single-well production capacity, challenges in long-term stable production, high engineering costs, and difficulties in marine geological risk prevention and control. Existing reviews mainly focus on isolated technical branches, lacking a systematic cross-paradigm comparison and a unified evaluation framework for stimulation technologies, and fail to clearly address the core theoretical barriers to scaling laboratory results to field applications. To fill this research gap, this review systematically summarizes production technologies for exploitation of marine NGH, covering six core directions: (i) improvement of in situ dissociation efficiency, (ii) expansion of the hydrate dissociation area, (iii) reservoir modification for exploitation environment optimization, (iv) multi-gas co-production, (v) solid-state fluidization mechanical exploitation, and (vi) innovative non-drilling penetrating exploitation technologies. In addition, we elaborate on the stimulation mechanisms, applicable reservoir conditions, advantages, and limitations of various technical routes, and establish an analytical framework of technologies for NGH production at commercial scale. Finally, we summarize the key challenges restricting NGH commercialization and propose several directions for future research on sustainable marine NGH development.
This study proposes a multi-scale framework integrating experimental characterization, microstructural analysis, and molecular dynamics (MD) simulations to evaluate the fatigue performance of asphalt mastic in which recycled concrete powder (RCP) replaces limestone powder (LP). Linear amplitude sweep tests reveal a systematic shift from stiffness to toughness dominated behavior with increasing RCP content: the initial shear modulus decreases from 91.89 kPa to 28.92 kPa, while the critical strain increases from 7.7
