Addressing the urgent demand for large-capacity, long-distance passive acoustic field monitoring in complex industrial environments, this paper proposes and demonstrates a 20-km, eight-channel passive audio sensing system based on extrinsic Fabry-Perot interferometer (EFPI) arrays over a single-mode fiber. To break through the physical bottleneck of “bandwidth-cavity length-loss” imposed by high-density coarse wavelength division multiplexing (CWDM, with single-channel bandwidth compressed to ∼5 nm), we fundamentally reconfigured the probe at the physical level. The physical cavity length was extended to approximately 240.25 μm to compress the free spectral range (FSR), and an approximately 30-nm-thick gold film (∼90% reflectivity) was innovatively coated on the inner surface of the silicon nitride MEMS sensing diaphragm. This design, based on a near-optimal opto-mechanical balance, effectively compensates for the significant Gaussian beam divergence attenuation induced by the long cavity, and achieves favorable optical intensity matching with the fiber end-face. Furthermore, a concurrent adaptive three-wavelength demodulation algorithm based on an FPGA was introduced at the receiving end to eliminate signal fading caused by operating point drift. Experimental results demonstrate that the system achieves high-fidelity parallel transmission of eight-channel audio with an excellent frequency response between 20 Hz and 4 kHz. The maximum signal-to-noise ratio (SNR) reaches 67.55 dB, and the peak acoustic pressure sensitivity is 722.1 mV/Pa. The system also features dynamic channel routing (approximately 4.7 ms switching latency) and maintains stable performance over 20–60 °C. By doubling the multiplexing capacity while maintaining exceptional sound pickup performance, this study provides an ideal solution for deploying large-scale, high-density passive acoustic sensing networks.
Learning performance modeling has achieved strong predictive performance, but the decision processes of neural models often remain implicit and difficult to analyze. To address this issue, we propose SPARK, a neural-to-symbolic framework for recovering symbolic decision structures from neural predictors of learning performance. SPARK derives performance-related factors from educational interaction data, learns the decision function with a multilayer perceptron as the teacher model, transfers it to a Kolmogorov-Arnold Network (KAN), and then applies structural refinement and symbolic regression to recover compact symbolic expressions. The goal is to convert implicit neural decision behavior into an analyzable symbolic form while preserving its main predictive behavior. Experiments on benchmark educational datasets show that SPARK remains competitive in prediction, preserves most of the teacher model’s decision behavior, and reveals dataset-dependent interaction patterns among student ability, cognition-related factor, and item difficulty.
Carbon capture and storage (CCS) is critical for decarbonizing the fossil-energy power sector, the world's largest source of greenhouse gas emissions. However, how deployment responsibilities should be allocated across heterogeneous regions to deliver economy-wide economic and environmental benefits remains poorly understood. This study develops an integrated optimization–multi-regional input-output–computable general equilibrium (OP-MRIO-CGE) framework to evaluate the economic outcomes and emissions reduction effects of alternative regional CCS deployment policies, emphasizing cross-regional spillovers and implications for policy implementation. Using China as a case study, results show that outcomes are highly sensitive to regional responsibility allocation. Scenario 6 formulates deployment as a multi-objective problem maximizing decarbonized fossil-energy power generation and minimizing total CCS deployment cost. The objectives are evaluated separately through Pareto dominance, without normalization or objective weights, producing non-dominated allocations rather than a unique optimum. One allocation is randomly selected as the reference solution before its downstream MRIO-CGE outcomes are assessed, and sensitivity analysis covers the entire retained Pareto set. This reference solution reduces economy-wide CO2 emissions by approximately 0.379 billion tonnes in 2030, increases national GDP by approximately USD 0.14 trillion, and improves consumer welfare by approximately USD 610 million. Concentrated CCS deployment in the Bohai Rim and the Junggar Basin increases emissions in non-host regions, including Sichuan, Hubei, and Anhui, while economic output losses occur in Shaanxi, Guangdong, and Jiangsu. These findings provide a quantitative basis for coordinated multi-regional CCS deployment and targeted complementary policies, offering an analytical framework applicable to other large, spatially heterogeneous economies pursuing deep decarbonization.
Clinical ultrasound images often contain artificial markers, such as measurement calipers and text, to assist diagnostic interpretation and comparison. However, these markers can introduce shortcut bias in downstream automated analysis, encouraging deep learning models to rely on marker-related cues rather than clinically meaningful anatomy. Existing marker removal methods are either mask-dependent and vulnerable to error propagation, or mask-free deterministic restorers that may over-smooth ultrasound texture and perturb unaffected background regions. To address these challenges, we present Echo-DM, a framework for ultrasound marker removal via conditional latent diffusion and region-aware fusion. Echo-DM follows a common encoder-diffusion-decoder pipeline, where a DiT-based conditional latent diffusion network performs global restoration and a region-aware fusion module enforces preservation-aware image-space refinement under end-to-end mask-free inference. Building on this fixed core design, we further instantiate Echo-DM-V and Echo-DM-R with VAE-based and RAE-based latent modules, respectively, which demonstrates that the Echo-DM architecture is compatible with diverse latent-module instantiations. Extensive experiments on Echo-PAIR, a large-scale paired clinical ultrasound dataset, demonstrate superior marker removal and strong anatomical fidelity compared with representative two-stage baselines, while providing favorable quality–efficiency trade-offs across deployment settings. Data, code and models will be released at https://github.com/MiliLab/Echo-DM.
Microplastics (MPs) and heavy metals like lead (Pb) are pervasive pollutants in aquatic ecosystems, yet their combined effects on aquatic plants remain poorly understood. To address this gap, Hydrocotyle vulgaris was exposed to single lead nitrate (Pb (NO₃) ₂) (0-3.5 mg/L) and in combinations with polystyrene microplastics (PSMPs) powder (3 μm) (0-200 mg/L) for seven days. We evaluated the weight growth rate (WGR), photosynthetic pigments, chlorophyll fluorescence, soluble sugar and protein content, malondialdehyde (MDA), peroxidase (POD), superoxide dismutase (SOD), catalase (CAT) and lead bioaccumulation. Results revealed a hormetic response to Pb, with low concentrations (≤ 2.5 mg/L) enhancing growth, while higher doses (> 3 mg/L) caused toxicity, including leaf deformation, chlorosis and WGR reduction. Pb accumulation increased dose-dependently, peaking at 3.5 mg/L in both single and combined treatments. MPs exhibited a concentration-dependent modulation of Pb toxicity, exacerbating effects at 50 mg/L but slightly alleviating them at 200 mg/L. These findings demonstrate the complex interaction between MPs and Pb in influencing metal bioavailability and plant stress responses. It provides critical insights into the ecotoxicological risks of co-occurring pollutants in wetlands and highlights implications for invasive species management under anthropogenic pollution. Low Pb enhanced plant growth, while higher levels caused damage and toxicity. PSMPs modified Pb effects, increasing harm at low MP doses but reducing it slightly at high doses. H. vulgaris showed tolerance and adaptability to Pb, despite stress responses. All high Pb treatments caused toxicity, regardless of PSMP presence.
