American Systems (United States)

Enhancing decision-making accuracy and inter-pretability is a critical issue in the current financial information systems environment. This study proposes FiT-XGSHAP, a novel hybrid AI structure with a Financial Transformer for high-dimensional sequence learning, XGBoost for structured learning, and SHAP for post-hoc interpretability. The ‘Give Me Some Credit’ dataset, which comprises over 150,000 real credit records with a range of financial characteristics, including DebtRatio and Monthly Income, is used to train and test the algorithm. Predicting two-year severe delinquency-a binary classification challenge at the core of financial risk prediction-is the goal. FiT-XGSHAP is 96.74% accurate. precision = 96.00%, recall = 97.00%, F1-score = 96.00%, AUC = 0.98, particularly outstanding to the conventional single-model approaches. The experiments demonstrate the power and usefulness of hybrid architectures for financial predictive tasks, presenting high performance and interpretability. The study supports advancing intelligent support systems for financial decision-making by integrating the power of ensemble learning with the explainability of explainable AI.

Accurate fish disease detection is essential for sustainable aquaculture, but existing machine learning models often require too much computational power for practical use. This study evaluates lightweight ResNeXt models (ResNeXt50 and ResNeXt101) for efficient fish disease classification using a comprehensive freshwater fish disease dataset. ResNeXt’s residual connections enable stable training and fast convergence, while its split-transform-merge design effectively learns diverse lesion patterns. ResNeXt50 achieved 95% accuracy, outperforming established models with significantly lower computational requirements. The results show that ResNeXt50 provides an effective solution for real-time fish disease detection, balancing efficiency and accuracy.

Recent advances in battery risk assessment methodology can be difficult to understand and apply. This article presents a series of example risk assessments on real battery systems of different sizes and chemistries. We walk through work planning and the control process for energized work on batteries, from the initial work order to project completion. We elaborate on how different engineering controls, such as a ground fault detector and indicator, impact battery risk assessment and what to do when you do not know whether they are functioning correctly. We present case studies in several types of battery systems, including lead–acid, lithium-ion, and vanadium redox. The article concludes with an assessment of training, policy, and code shortfalls that may have contributed to past accidents.

Digitization introduces new complexities in cyberattacks for healthcare and critical infrastructure sectors due to risks associated with sensitive data such as Protected Health Information. This article examines AI-enabled cybersecurity in healthcare for anomaly detection, prevention of intrusions, and threat intelligence employing machine learning and deep learning for autonomous real-time defense. AI-based intrusion prevention systems targeting endpoint protection and mitigation of behavior-based threats are reviewed alongside overcoming model accuracy, tolerance, and privacy issues. It is posited an inclusive framework that employs AI and blockchain for secure verifiability of data emphasizes the credibility of trusted data, however, still requires substantiation. We focus on comparative analysis of AI models using real-world datasets MHealth, CICIDS 2017, and NSL-KDD, achieving higher detection rates coupled with fewer false positives and reduced processing times. Increased dataset and real-world simulated cyberattack validations are suggested as a next step to validate anticipatory effectiveness claims.

Bridge load capacity, defined as the maximum load a bridge can sustain without structural failure, is a critical determinant of infrastructure safety and operational reliability. Traditional assessment methods, including analytical modeling, static load testing, and empirical estimation, often fail to capture the dynamic, nonlinear, and time-dependent behavior of bridges subjected to evolving environmental and operational conditions. This study presents a predictive modeling framework that integrates machine learning (ML) techniques with real-time sensor data to enhance the accuracy and adaptability of load capacity prediction. The research utilized continuous data streams collected from strain gauges, accelerometers, displacement sensors, and fiber Bragg gratings, complemented by environmental and traffic datasets capturing temperature, humidity, axle loads, and vehicle volumes. Data preprocessing involved noise reduction, normalization, dimensionality reduction, and feature selection using principal component analysis (PCA) and LASSO regression. Multiple machine learning algorithms, Random Forest, Gradient Boosting, Support Vector Regression, and Deep Neural Networks, were developed and evaluated against traditional regression models. Descriptive and correlational analyses revealed that strain, displacement, and axle load exhibited the strongest positive relationships with load capacity, while temperature and humidity had significant negative effects. The regression model achieved high explanatory power (R² = 0.86), with ensemble learning methods further improving predictive accuracy (R² = 0.91–0.92) and reducing error metrics (RMSE ≈ 27 kN, MAE ≈ 20 kN). Reliability analyses demonstrated high internal consistency (Cronbach’s α ≥ .85) and measurement stability, while validity tests confirmed strong convergence with established load rating standards. Collinearity diagnostics and regularization techniques effectively mitigated redundancy among predictors, enhancing model interpretability. Findings indicate that integrating real-time sensor data with ML enables continuous, adaptive modeling of structural behavior under varying loads and environmental influences. This approach captures complex interactions that static models cannot, providing early indicators of structural deterioration and supporting predictive maintenance strategies. The study concludes that data-driven predictive modeling represents a paradigm shift in bridge engineering, offering superior precision, scalability, and operational insight for infrastructure management. It recommends expanded sensor deployment, hybrid physics-informed modeling, and interdisciplinary collaboration among civil engineers, data scientists, and transportation authorities to advance resilience, safety, and sustainability in bridge infrastructure systems.

The maximum power formula (proposed by Dan Doan in a 2007 IEEE paper) for calculating dc arc-flash incident energy has been in NFPA 70E since 2012; however, repeated testing by multiple companies over the years has shown that it overestimates arc-flash energy by a factor of at least 2-10, especially at lower voltages (such as nominal 48, 125, or 250 Vdc). This paper analyzes the actual measured arc-flash energy from over 100 different tests at various voltages, and proposes modifications to the arc sustaining time in the maximum power transfer theorem equation(s) found in Annex D.5 of NFPA 70E. In addition, more realistic estimates for dc arc-flash boundary multipliers are developed for cases where the arc is "in a box", than what has been found in previous editions of NFPA 70E. These are based on observations of dc arc-flash tests, calculations of reflected energy from the rear of typical cabinets using the maximum power method, and previous studies of "arc-in-a-box" energy multipliers for ac arc-flash.

The National Aeronautics and Space Administration (NASA) has completed a flight test to evaluate the performance of an onboard prototype automation system operating in future high density urban airspace. The test was part of a research investigation of the Urban Air Mobility (UAM) concept, with a focus on a future environment having hundreds of simultaneous operations over a metropolitan area. The complexity of this future UAM airspace may require automation capable of replanning an aircraft’s path in the presence of traffic and other changing constraints. A live-virtual-constructive (LVC) approach was used to conduct the test. Prototype automation technology was integrated into one of the two live aircraft, which were combined with virtual traffic to create a mixed reality environment at the target airspace density. In-flight evaluation enabled verification of the automation’s functions and discovery of any unexpected behaviors resulting from its operation in an actual flight environment. The in-flight evaluation also provided data for validation of air traffic simulations. This paper discusses the design, methodology, and challenges overcome to conduct a successful flight test. Remaining challenges, future work, and recommendations to improve the flight test capability are also discussed.

This paper presents the results of a study that investigated if controller force compensations accounting for the inertial force and moment due to the aircraft motion during flight have a significant effect on pilot control behavior and performance. Seven rotorcraft pilots performed a side-step and precision hovering task in light turbulence in the Vertical Motion Simulator. The effects of force compensation were examined for two different simulated rotorcraft: linear and UH-60 dynamics with two different force gradient of the lateral stick control. Four motion configurations were used: large motion, hexapod motion, fixed-base motion, and fixed-base motion with compensation. Control-input variables and task performance such as the time to translate to the designated hover position, station-keeping position errors, and handling qualities ratings were used as measures. Control force compensation enabled pilot control behavior and performance more similar to that under high- or medium-fidelity motion to some extent only. Control force compensation did not improve overall task performance considering both rotorcraft models at the same time. The control force compensation had effects on the linear model with lighter force gradient, but only a minimal effect on pilots? control behavior and task performance for the UH-60 model, which had a higher force gradient. This suggests that the control force compensation has limited benefits for controllers that have higher stiffness.

To determine if construction and trades workers employed at U.S. Department of Energy (DOE) nuclear sites facilities are at significant risk for diseases associated with occupational exposures, we compared the mortality experience of participants in the Building Trades National Medical Screening Program (BTMed) to that of the US population. The cohort includes 24,086 BTMed participants enrolled between 1998 and 2016 and 5203 deaths. Cause-specific standardized mortality ratios were calculated based on US death rates. Mortality was elevated for all causes, all cancers, cancers of the trachea, bronchus, and lung and lymphatic and hematopoietic system, mesothelioma, chronic obstructive pulmonary disease, asbestosis, transportation injuries, and other injuries, particularly those caused by accidental poisoning, suggesting a possible effect of the opioid epidemic. Apart from other injuries, mortality patterns were very similar to those reported in the past in this population. Construction workers employed at DOE sites have a significantly increased risk for occupational illnesses. Risks are associated with employment during all time periods covered including possibly after 1990. The cancer risks closely match the cancers identified for DOE compensation from radiation exposures. The high risk of lung cancer supports the value of early lung cancer detection. Continued medical surveillance is important.