The dispersion characteristics of fuel particles over the bed have always been a key consideration in the design of fluidized beds. However, the lateral fuel dispersion coefficients (D_sr) reported in the previous literature widely ranged from 10⁻⁴ to 10⁻¹ m²/s, which were hard to be directly applied in large-scale CFB (circulating fluidized bed) boilers with wider bed width (close to 30 m) and variable coal feeding modes. To solve this problem, field tests were conducted on the first 600 MW CFB boiler in this paper, and the distributions of flue gas composition and flue gas temperature were obtained near the coal feeding port under four working conditions. The dispersion characteristics of fuel particles were analyzed in detail combined with the dispersion model. In addition, a new method based on bed heat balance was also proposed to estimate D_sr and applied in a 300 MW CFB boiler with a different coal feeding mode. The results revealed that D_sr through the local heat balance method ranged from 0.10–0.35 m²/s, which were 9–18% higher than the values through the dispersion model. Compared with the coal feeding port immersed in the bed, D_sr of the fuel particles fed into the splash zone increased by about 34.7%. D_sr under the coal feeding mode of a screw feeder with pressurized air ranged from 0.12–0.16 m²/s, which were lower than D_sr under the coal feeding method of importing fuel into circulating ash. The method and results in this paper could expand the applicability of fuel dispersion, and provide the most direct guidance for the design and optimization of coal feeding ports in large-scale CFB boilers.

Forest fire monitoring in remote sensing imagery has long relied on traditional perception models that primarily focus on detection or segmentation. However, such approaches fall short in understanding complex fire dynamics, including contextual reasoning, fire evolution description, and cross-modal interpretation. With the rise of multi-modal large language models (MLLMs), it becomes possible to move beyond low-level perception toward holistic scene understanding that jointly reasons about semantics, spatial distribution, and descriptive language. To address this gap, we introduce FireMM-IR, a multi-modal large language model tailored for pixel-level scene understanding in remote-sensing forest-fire imagery. FireMM-IR incorporates an infrared-enhanced classification module that fuses infrared and visual modalities, enabling the model to capture fire intensity and hidden ignition areas under dense smoke. Furthermore, we design a mask-generation module guided by language-conditioned segmentation tokens to produce accurate instance masks from natural-language queries. To effectively learn multi-scale fire features, a class-aware memory mechanism is introduced to maintain contextual consistency across diverse fire scenes. We also construct FireMM-Instruct, a unified corpus of 83,000 geometrically aligned RGB–IR pairs with instruction-aligned descriptions, bounding boxes, and pixel-level annotations. Extensive experiments show that FireMM-IR achieves superior performance on pixel-level segmentation and strong results on instruction-driven captioning and reasoning, while maintaining competitive performance on image-level benchmarks. These results indicate that infrared–optical fusion and instruction-aligned learning are key to physically grounded understanding of wildfire scenes.

This study developed a two-dimensional multiphysics-coupled model for co-electrolysis of CO₂ and H₂O in solid oxide electrolysis cells (SOECs) using COMSOL Multiphysics, systematically investigating the influence mechanisms of key operating parameters including temperature, voltage, feed ratio, and flow configuration on co-electrolysis performance. The results demonstrate that increasing temperature significantly enhances CO₂ electrolysis, with the current density increasing over 12-fold when temperature rises from 923 K to 1423 K. However, the H₂O electrolysis reaction slows beyond 1173 K due to kinetic limitations, leading to reduced H₂ selectivity. Higher voltages simultaneously accelerate all electrochemical reactions, with CO and H₂ production at 1.5 V increasing by 15-fold and 13-fold, respectively, compared to 0.8 V, while the water–gas shift reaction rate rises to 6.59 mol/m³·s. Feed ratio experiments show that increasing CO₂ concentration boosts CO yield by 5.7 times but suppresses H₂ generation. Notably, counter-current operation optimizes reactant concentration distribution, increasing H₂ and CO production by 2.49% and 2.3%, respectively, compared to co-current mode, providing critical guidance for reactor design. This multiscale simulation reveals the complex coupling mechanisms in SOEC co-electrolysis, offering theoretical foundations for developing efficient carbon-neutral technologies.

The closed-loop testing of autonomous driving systems critically depends on large-scale libraries of diverse and realistic 3D vehicle assets, yet current pipelines still rely on labor-intensive modeling or multi-view capture, making efficient construction a key bottleneck. To overcome this bottleneck and enable convenient, cost-effective 3D asset generation, we propose a semantic prior-guided framework for accurate and robust vehicle point cloud reconstruction from casually captured single-view photographs. Our framework is built on a diffusion backbone but is fundamentally driven by two forms of prior knowledge: First, geometric and appearance priors from camera-aware image features, masks, and distance-transform maps are projected onto the evolving point cloud, compensating for the severe information loss in single-view inputs. Second, we introduce distillation-style regulators—pretrained neural networks that encode vehicle type and model semantics; they act as teacher networks that impose high-level constraints on the generated point clouds, transferring rich semantic knowledge and effectively regularizing the learning process. With these priors, our model infers vehicle-specific semantics from limited observations and reconstructs high-quality 3D point cloud assets. On the 3DRealCar++ dataset, our method clearly surpasses state-of-the-art point cloud baselines in both F-score and Chamfer Distance.