This paper investigates repeatability of additive manufacturing on the energy absorption capacity of polymer Body-Centered Cubic (BCC) lattice structures. As a class of cellular materials, lattice structures enable the tailoring of macroscopic properties through controlled microgeometry, where relative density, unit cell topology, print quality and manufacturing defects play a significant role. BCC lattice structures with a relative density of approximately 0.2 were designed in Rhinoceros using Grasshopper plugin and fabricated using fused filament fabrication (FFF) technology with PLA material. Experimental compression tests were conducted to evaluate the force – displacement response and determine the energy absorption characteristics of the structures. The results showed that lattice structures with uniform spatial distribution of BCC unit cells achieved an average peak force of 2529.89 N at an average compression displacement of 4.04 mm. Although manufacturing defects did not significantly affect the maximum load-bearing capacity, it strongly influenced the deformation behavior of the lattice structures. After reaching the peak load, diagonal collapse of the structures was observed, indicating limited post-elastic energy absorption capability. The average specific energy absorption of the tested structures was 0.41 J/g, with a coefficient variation of 11.32%. The results demonstrate that manufacturing defects contribute to the variability in energy absorption of FFF-fabricated BCC lattice structures. Their mechanical influence is governed not only by defect number but also by defect type relative to load-bearing struts and nodes. These findings highlight the importance of defects caused by FFF production in usable products for achieving favorable energy absorption characteristics.
Most structural codes focus on prismatic columns, often overlooking the benefits of non-prismatic shapes. This study investigates how reinforced concrete columns tapered along a circular arc behave compared to standard prismatic ones under axial loads. Four specimens were tested (2 prismatic and 2 non-prismatic). Using 1.5m height specimens with equal reinforcement (6φ10) and a constant depth of 150mm, the experimental program tested columns that varied their width from 300mm at the ends to 150mm at mid-height as non-prismatic columns, whereas the prismatic columns had a section of 150mm x 150mm. Results revealed that non-prismatic columns have a higher load carrying capacity than prismatic columns by 17.90% and improved axial stiffness by at least 74.03%. Notably, while prismatic columns failed in the upper half, tapered specimens failed at the necking region, demonstrating a more efficient use of the entire member. Additionally, the strain in the curved rebars was 66% of that recorded in prismatic members at equivalent load level. Although non-prismatic columns dissipate less energy than prismatic specimens (1306.71J compared to 2137.33J) and exhibited a more brittle response, they maintained superior lateral stability. Within the limits of the tested specimens, this study supports that the circularly tapered shape increases strength due to improved confinement provided by bi-axial compressive stress within the column, thereby offering a more efficient method for structural design.
Myocardial ischemia is a major cause of illness and death worldwide, and its early detection remains difficult. The electrocardiogram (ECG) is the most widely used test for ischemia, but manual interpretation is time-consuming and subject to inter-observer variability. This study develops a deep learning framework based on a Fully Convolutional Network (FCN) that classifies myocardial ischemia from 12-lead ECGs and explains its decisions with Local Interpretable Model-Agnostic Explanations (LIME). The public PTB-XL database was used, comprising 11,018 ECGs from 10,135 patients (2,441 ischemia and 8,577 normal). Signals were cleaned, segmented into beats and summarized as lead-wise median beats. Training, validation and independent test sets were separated at the patient level, performance was estimated over ten patient-level bootstrap iterations, and the model was externally validated on 2,033 ECGs from the Georgia 12-lead ECG Challenge database. On the independent test set, the FCN achieved an accuracy of 97.82% ± 0.19%, precision of 95.98%, recall (sensitivity) of 94.23%, specificity of 98.86%, F1-score of 95.10% and an area under the receiver operating characteristic curve (AUC) of 0.995 (95% CI 0.995–0.996). In external validation, the AUC was 0.959 and the accuracy was 90.31%. The FCN was compared with an attention-based encoder, a simple convolutional neural network, a Transformer encoder, random forest, support vector machine and XGBoost under the same protocol. LIME explanations placed 64.7% of the positive importance of ischemic ECGs in the ST–T region (chance level 48%) and 57.3% in the cardiologist-annotated ischemia territory (chance level 47.9%), and removing the most important segments reduced the predicted ischemia probability significantly more than removing random segments. The framework therefore combines competitive performance with explanations that are quantitatively consistent with clinical knowledge. Prospective, multicenter validation is required before clinical use.
Soil-structure interaction (SSI) strongly influences the seismic performance of shallow-founded structures, particularly when liquefaction occurs in stratigraphically variable deposits. This study develops machine learning (ML) models to emulate three-dimensional nonlinear effective-stress finite element (FE) predictions of permanent foundation settlement at substantially lower computational cost. The models are trained using the DesignSafe PRJ-5802 database, comprising 1676 simulations of unmitigated sites and 1272 companion simulations with dense granular column (DGC) mitigation, representing 96 soil-foundation-structure configurations subjected to 59 recorded ground motions. Eight ML algorithms are evaluated using configuration-grouped 10-fold cross-validation, ensuring that no soil-foundation-structure configuration appears in both training and test sets. Gradient-boosted ensemble models provide the best performance; the Gradient Boosting model achieves an out-of-fold R² of 0.535 for natural-log settlement, a mean absolute error of 6.74 cm, and 68% of predictions within a factor of two of the FE simulations. The study also demonstrates that the exceptionally high R² = 0.9965 reported in the original analysis resulted from methodological errors, including prediction of foundation contact pressure rather than settlement and data leakage caused by noisy replicated records. Reproducing the original procedure yields R² = 0.9984, whereas eliminating leakage reduces it to 0.884. SHAP analysis identifies ground-surface slope, sand relative density beneath the footing, cumulative absolute velocity, and foundation contact pressure as the dominant predictors. DGC mitigation reduced permanent settlement in 87% of paired simulations, with a median settlement ratio of 0.74. An explicit applicability domain and lognormal prediction interval are provided to support reliability-based design applications.
The status of devices in substation main control rooms is jointly characterized by fine-grained components such as indicator lights, pressure plates, rotary knobs, display screens and alarm windows. To overcome the limitations of traditional single-frame detection in small object recognition, state continuity and component logical relations, this study proposes a device status identification model integrating improved YOLO target detection, bidirectional temporal modeling and graph relation reasoning. Based on fixed monitoring videos, the model performs state component annotation, local feature enhancement, multi-scale feature fusion and device status correlation graph construction to achieve device status judgment from component detection. Experimental results show that the proposed model achieves a device status accuracy of 96.28%±0.29%, an anomaly recall of 95.36%±0.41%, and a false alarm rate of 2.74%±0.18%. In real scenarios, the recall of state components reaches 96.34%±0.35% at a monitoring distance of 4 m, and the anomaly recall reaches 94.21%±0.46% under 20% occlusion. The proposed method provides an effective intelligent approach for substation video inspection and anomaly detection.
