Abstract Postdisaster inspection of concrete structures demands rapid, reliable assessment tools, yet conventional manual surveys are slow, subjective, and impractical under emergency conditions. Although deep learning has advanced crack detection, most prior work is limited to binary segmentation and neglects other critical surface damages such as spalling. In addition, postdisaster environments present two major challenges: (1) severe class imbalance at the pixel level; and (2) constrained computing resources that hinder deployment of computationally intensive models. The objective of this study is to systematically evaluate U-Net variants for multiclass concrete damage segmentation under class imbalance and computing-resource constraints. This study systematically evaluates four U-Net variants, the original U-Net, dilated U-Net, attention U-Net, and spatial and channel squeeze-and-excitation (SCSE) U-Net, alongside a well-established external benchmark, DeepLabv3 with a ResNet-101 backbone. All models were trained, validated, and tested on an annotated dataset of cracks and spalling and assessed using both quantitative metrics and qualitative inspections of segmentation outputs. Computational efficiency, including training time, inference time, and memory usage, was also monitored to gauge practical feasibility for on-site deployment. Results show that all U-Net variants achieved strong segmentation performance with only marginal differences between them. Attention U-Net delivered the most balanced class-specific performance, while the original U-Net combined balanced accuracy with the lightest computational footprint. Compared to the benchmark, U-Net variants achieved 5.9% faster training, 49.8% faster inference, and overall reduced memory usage, underscoring their suitability for practical deployment in postdisaster environments. This study provides a systematic evaluation of U-Net variants for multiclass concrete damage segmentation. It offers empirical insights into the persistent challenge of class imbalance and emphasizes computational efficiency as a core evaluation dimension. These findings lay the groundwork for future frameworks that link automated damage maps with structural models to accelerate disaster reconnaissance.
Despite rising labor costs and growing labor shortages, adoption of machine harvesting in fresh market blueberry production has been limited by issues such as high rates of bruising and the significant loss of fruits on the ground. Machine settings strongly affect fruit quality and harvesting performance, yet they are manually configured based on operator experience, with no real-time performance feedback. Towards addressing these limitations, this paper presents a sensing proof-of-concept for analyzing berry flow through the harvester by estimating the pixel-wise proportion of blue vs green berries and volumetric flow rate from overhead RGB-D images of the harvester's conveyor. An automated labeling pipeline combining a domain-specific YOLOv8 patch detector with SAM2-based mask generation was developed to generate a realistic large-scale berry dataset from unlabeled images captured during commercial over-the-row blueberry harvesting operations. Compared to a baseline SegFormer model trained on a smaller dataset of manually annotated images, our fine-tuned SegFormer-trained and validated on the large-scale automatically generated dataset-achieves superior performance, with green berry IoU improving from 0.377 to 0.817 and recall increasing from 0.414 to 0.911. A scene-level YOLOv8-l detection model trained and evaluated on the same dataset achieves mAP(50-95) of 0.966 and 0.898 for blue and green berries, respectively. Both models' performance is further confirmed by an independent evaluation on a manually annotated subset of the validation images. We also compared two approaches for predicting harvest yields across five machine-setting trials: a depth-based method using the RGB-D sensor and a detection-based method using only RGB images, achieving R2 = 0.989 and R2 = 0.894, respectively, against ground-truth load cell measurements from these trials. To support future research, we publicly release our datasets and models.
Steam reforming of oxygenated hydrocarbons such as methyl linoleate offers a renewable pathway for hydrogen and synthesis-gas production from biodiesel-derived feedstocks. Wall-coated catalytic microchannel reactors provide the heat and mass-transfer intensification required for this highly endothermic reaction, eliminating secondary cracking and carbon deposition typically observed under diffusion-limited conditions in packed beds. Palladium was uniformly deposited on Fe-Al intermetallic substrates via electroless plating from dilute PdCl2 solutions (<0.5 mM), forming submicron catalyst islets with clean metal-substrate interfaces after reduction in an N-2-H-2 atmosphere and thermal conditioning at 903-1273 K. Steam-reforming experiments at a steam-to-carbon ratio (S/C) of 7.75 exhibited nearly isothermal operation and rapid gas-phase equilibration. The overall reaction rate was surface-controlled and first-order with respect to methyl linoleate partial pressure. The microchannel configuration effectively eliminated diffusion limitations and suppressed carbon formation at S/C ratios above the thermodynamic decomposition threshold. Experiments performed at short mean residence times (55-62 ms) maintained conversions below equilibrium levels under industrially relevant temperatures. The overall reforming rate was controlled by surface kinetics and showed a first-order dependence on the partial pressure of methyl linoleate. This confirms that both internal and external diffusion limitations were effectively eliminated by the microchannel design. At 1173 K, hydrogen yields of about 40% were achieved, in close agreement with the numerical simulations of the coupled convection-diffusion-reaction model. This strong agreement between experiment and model validates the extracted kinetic parameters and confirms the suitability of wall-coated microchannel reactors for intensified biodiesel steam reforming under industrially relevant conditions.
The structure of coal determines its reactivity and products in direct liquefaction, and the production of aromatic chemicals from the abundant aromatic structures in coal is an important route for high-value conversion. In this study, an atomic-scale molecular model was implemented under the dual constraints of elemental composition and carbon skeleton based on elemental analysis, FTIR, and C-13 NMR. Because the depolymerization of the coal macromolecular network was initiated by thermal cleavage of weak bridge bonds, key chemical bonding descriptors (XCal-Cal, XCar-Cal, and XCal-O) were extracted from the model, and a quantitative structure-performance relationship model correlating these descriptors with aromatic compound yield was established. The results indicated that the proposed model followed a weighted average equation (Y-aromatic = 62.4X(Cal-O) + 44.5X(Cal-Cal) + 40.0X(Car-Cal)), in which the coefficients corresponded to the "theoretical reference yield" of the respective bonds. All three bonding descriptors positively contributed to aromatic compound yield (C-al-O > C-al-C-al > C-ar-C-al). In addition, the established model showed high accuracy (R-2 > 0.98), and an accurate prediction of aromatic compound yields from coal structure was achieved. This suggests that the formation of aromatic compounds mainly relies on the abundance of bridge bonds. In particular, C-al-O dominated by ether bonds acted as a critical "network depolymerization switch", and its preferential cleavage directly triggered the primary depolymerization of the coal macromolecular network. Hence, C-al-O dominated the formation of aromatic compounds. Consequently, weak covalent bonds, such as C-al-O, were redefined from traditional gas precursors to a core structural factor governing aromatic compound formation. This study establishes a quantitative predictive relationship between aromatic compound yield in direct coal liquefaction and microscopically computable bonding descriptors, providing theoretical references and tools for molecular structure-based feedstock evaluation and targeted conversion process regulation.
Abstract Many large igneous provinces (LIPs) comprise once contiguous magmatic products that are now fragmented across multiple tectonic plates. Emplacement of these voluminous magmatic products is commonly coeval with, or shortly followed by, fragmentation into constituent domains. However, the Kerguelen LIP, predominantly within the southern Indian Ocean, rifted after an unusually protracted c. 45 Myr interval post‐emplacement. To investigate this protracted post‐emplacement rifting, a recent voyage by the RV Investigator sampled the conjugate margins of the Kerguelen Plateau, William's Ridge (WR) and Broken Ridge (BR) for the first time. Dredged volcanic rocks were characterized using petrographic, geochemical (major and trace element, Sr‐Nd‐Pb‐Hf isotopes) and geochronological (40Ar/39Ar) techniques. Here, we focus on the sampled 82–67 Ma tholeiitic volcanic rocks. The observed geochemical data in conjunction with rare earth element inversion modeling of these Late Cretaceous volcanic rocks reflect plume–ridge interaction with variable assimilation (≤10%) of continental crust. We propose that the Kerguelen plume transported melt to the WR–BR margin during episodes of extension. Extension and magmatism were localized along the rheologically weak margins of the WR microcontinent. The combination of these mechanically weak boundaries and earlier extensional events facilitated a final ridge jump into the WR–BR margin at c. 44 Ma. We demonstrate a pre‐existing history of extension in the Late Cretaceous prior to complete fragmentation in the Eocene, resolving the apparent protracted interval between emplacement and separation of the Kerguelen LIP.
