Trine University

Microplastics (MPs), both conventional and biodegradable, are emerging contaminants of significant concern in marine ecosystems. While biodegradable plastics such as polylactic acid (PLA) and polybutylene adipate-co-terephthalate (PBAT) are marketed as eco-friendly alternatives, growing evidence suggests that their degradation products may exhibit comparable toxicity to conventional polymers like polyethylene (PE) and polystyrene (PS). This review critically examines the physicochemical properties, degradation behavior, and ecotoxicological impacts of both biodegradable and conventional MPs on the marine microalga Chlorella vulgaris. The analysis highlights that both MP types inhibit algal growth, disrupt photosynthesis, and induce oxidative stress through reactive oxygen species (ROS) generation, enzymatic imbalance, and cellular membrane damage. Mechanistic pathways involve particle aggregation, surface adsorption, and the release of toxic leachates and nano-sized derivatives, amplifying their ecological risks. Furthermore, biodegradable MPs often fragment faster, increasing MP particle counts and pollutant adsorption potential. The findings challenge the perceived environmental safety of “biodegradable” plastics and underscore the need for realistic field-based assessments. Understanding MP-algae interactions is crucial, as disruptions at this trophic level may cascade through marine food webs, affecting productivity and biogeochemical cycling. The review calls for comprehensive ecotoxicological frameworks to guide sustainable polymer design and effective marine pollution management.

Incorporation of AI and ML into healthcare monitoring systems has driven the enhancement of the most significant progress in patient care, such as early detection, personalized care, and cost efficiency. This paper presents a new AI solution of a predictive analytics framework specifically designed for healthcare systems with an emphasis on the respiratory imbalance severity. Our experimental setup includes a combination of different machine learning algorithms, including Support Vector Machines (SVM), AdaBoost, and state-of-the-art advanced ensembles like Stacked Ensembles, as well as deep learning algorithms, including TabNet. Applying strict feature selection methods, data preprocessing with feature scaling, SMOTE, and other balancing methods guarantees the efficiency of the developed framework in terms of predictive performance and model interpretability. Concerning the prioritization of features, key findings show the importance of the respiratory rate and the type of first diagnosis when developing clinical models. The Stacked Ensemble classifiers, based on SVM, KNN and AdaBoost, provided nearly perfect accuracy of 99.92%, which was far superior to individual classifiers. Exploratory data analysis or EDA, and reducing dimensions with the help of PCA, gave additional specificity in data representation for building performance and scalable, efficient predictive models. Performance metrics integral to the confusion matrices and AUC scores provide additional validation that the framework is valuable in minimizing sources of misclassification error and providing actionable insights for operational decision-making. A use case of this research is the innovative capabilities of AI-enabled systems to revolutionize healthcare observation by identifying severe states at an early stage for certain conditions using revolutionary clinical procedures. It also shows the limitations of the study, including ethics issues, data normalization, and interdisciplinarity, to advance recommendations for future research to incorporate more sophisticated approaches to deep learning. Thus, the work demonstrated in this study brings together technological advancement and clinical ideas to make healthcare more intelligent in the future.

The intersection of predictive analytics, Industrial Internet of Things (IIoT), and cloud-based systems is also changing sophisticated manufacturing in the United States. With increased global competition and the emergence of operational resilience as a strategic goal, manufacturers are turning toward data-driven systems, increasing productivity and driving innovation. This paper investigates the operational performance and innovation of advanced sectors of the U.S. manufacturing industry with the help of predictive analytics by using the Smart Manufacturing IoT-Cloud Monitoring Dataset which includes 100,000 real-time sensor observations of 50 industrial machines. The theoretical framework of Industry 4.0 and resource-based perceptions of technological capability make use of the notion of machine learning models that forecast the maintenance demand and approximate the remaining useful life dependent on sensor variables and indicators of anomalies such as temperature, vibration, humidity, pressure, energy consumption, and the risk of downtime. The algorithms of supervised learning, which are Random Forest and Gradient Boosting, are applied to assess the predictive accuracy based on ROC-AUC, F1-score, and RMSE values. The results of empirical studies indicate that predictive analytics can substantially increase the effectiveness of failure detection and minimize unexpected downtime, which improves the effectiveness of devices in general, in addition to enhancing operational reliability. The findings also reveal that predictive maintenance systems are not only efficiency-enhancing systems, but also innovative enablers, as predictive maintenance systems enable real-time decision-making, optimizing processes and resources in an adaptive manner. Using the ideas of turning reactive maintenance systems into proactive, information-driven approaches, the advanced manufacturers will be in a better position to enhance production continuity, cut expenses on the operation, and become more technological. The research is relevant to academic literature because it forms a quantifiable connection between AI-based predictive maintenance and productivity increase and innovation abilities in the advanced manufacturing ecosystems. In terms of policy and management, the results can be used practically to guide the U.S. manufacturers who want to use predictive analytics to improve industrial competitiveness and maintain their long-term growth in the dynamic digital economy.

Evolving interdependencies across institutions and markets drive systemic financial risk, yet most forecasting models either treat assets independently or rely on static correlation structures. This limitation becomes particularly salient as cryptocurrency markets increasingly interact with traditional banking systems amid financial stress. Ignoring time-varying cross-market network structure risks understating tail risk ‎precisely during periods when accurate systemic risk assessment is most critical. This study proposes a dynamic graph neural network ‎‎(GNN) framework for systemic risk forecasting that models time-varying financial networks spanning banking institutions and major cryptocurrency assets. Nodes represent financial entities, while edges are constructed using rolling-window dependency measures that adapt to ‎changing market conditions. Node dynamics are modeled through temporal neural architectures, and stress regimes are explicitly identified ‎to evaluate performance under market turmoil. The empirical design includes strong temporal baselines, static-graph ablations, and cross-market removal experiments to isolate the contribution of network dynamics and crypto-market integration. Results indicate that a strong ‎LSTM baseline achieves superior volatility forecasting accuracy in both overall and stress-period evaluations, demonstrating the competitiveness of purely temporal models. However, within the class of graph-based models, dynamic GNNs substantially outperform static-graph variants during stress periods, demonstrating the importance of time-varying network structure for capturing volatility amplification. ‎Bank-only and full-system dynamic GNNs exhibit comparable stress-period performance, suggesting that cryptocurrency assets contribute ‎limited incremental information to bank-specific forecasts, while remaining informative for system-level stress characterization. The findings ‎suggest that dynamic graph representations enhance stress sensitivity and structural interpretability relative to static network models, even ‎when they do not surpass strong temporal baselines in raw predictive accuracy. The results support a restrained view of crypto–banking ‎contagion, emphasizing its conditional relevance during periods of market stress rather than unconditional systemic dominance.

The growing number of global shocks, which include geopolitical instabilities, pandemics, and climate shock, among others, has revealed major weaknesses in the U.S. manufacturing supply chains. The need to increase supply chain resiliency and, at the same time, improve sustainability and resource efficiency has become a strategic necessity. The proposed study is an AI- and machine learning-controlled predictive analytics-based framework created to enhance the reliability and resilience of the U.S. manufacturing supply chain systems. The framework incorporates heterogeneous data sources such as operational, environmental, market, and logistics data to facilitate the proactive identification of risks, prediction of demand, and optimization of resources. Developed machine learning models are used to embrace non-linear relationships with complexities across the supply chain nodes to assist in real-time decision-making and adaptive response strategies. The proposed framework enables the identification of disruption early in the supply chain planning and execution, better management of inventory and capacity, reduced waste, and increased energy and material efficiency by integrating predictive analytics into the supply chain planning and execution. In theory, this study will be a step forward in the interplay of artificial intelligence, sustainable supply chain management, and resilience engineering. In practice, it provides decision-makers with a scalable and evidence-based instrument in order to enhance the continuity of operations, environmental performance, and competitiveness in the long run of the U.S. manufacturing supply chains.

Sustainable operations management has gained prominence in the world's industrial strategy, particularly in the manufacturing industry, due to the apprehension of the resources and the environment that the world is experiencing. Practices of circular economy (CE) with their focus on reuse, recycling, and remanufacturing, and low consumption provide a structural transition out of the conventional linear model of take-make-dispose. This study is a comparative research paper based on the analysis of the sustainable manufacturing practices using data about the 100 most sustainable corporations. The aim is to assess how the major international manufacturing corporations are applying the principles of the circular economy to their manufacturing process and governing systems. This paper lists some important data amongst the dataset, such as sustainability revenue percent, ESG (Environmental, Social, and Governance) ratings, CEO compensation ratios, and measures of leadership diversity like women in the executive management and board representation. These variables are used as proxies of circularity, social responsibility, and ethical governance. Filtered data is refined on the basis that only corporations that deal with manufacturing industries are covered, including industries dealing with electronics, waste management, and materials manufacturing. The analysis of descriptive analytics and visualizations (bar charts, scatter plots, and box plots) provides evidence of a lot of differences in circular economy performance across the manufacturing sub-sectors. The waste management and electronics companies increase the percentage of revenue on sustainability and have better ESG ratings, whereas many companies need to improve the gender diversity situation. There are also analytical findings indicating that there might be a relationship between inclusive leadership and a high level of performance in sustainability. The results show that although most manufacturing companies in the world are on the path to becoming circular, the amount of implementation is still variable and tends to vary with the industry, even geographical region, and policy maturity within the company. The challenges relate to technological constraints, an unbalanced regulatory framework, and investments in the sphere of circular infrastructure and disclosures. The contribution of this study to the scholarship is the presentation of empirical, comparative information on the practice of the circular economy in the manufacturing industries using a global perspective. It also provides strategic solutions on how sustainability performance can be improved by associating executive compensation with ESG targets, encouraging diversity at the leadership level, and incorporating circularity in designing the lifecycle of products. This study accentuates the importance of sustainable operating management in contributing to the realization of world climate objectives, enhancing firm responsibility, and long-threatened industrial stability.

Domain shift across heterogeneous data sources remains a critical challenge in pattern recognition and classification tasks, often leading to significant performance degradation in real-world applications. To address this issue, this paper proposes a Multi-Source Domain Adaptive Network with Multilevel Feature Fusion (MSDAN) that effectively learns domain-invariant and discriminative representations by jointly leveraging multiple source domains. The proposed framework employs a shared deep feature extraction backbone to capture hierarchical representations and integrates a multilevel feature fusion mechanism to combine low-level spatial details, mid-level structural information, and highlevel semantic features. An adversarial domain adaptation module is incorporated to align feature distributions between multi-source and target domains, enabling robust knowledge transfer without requiring labeled target data. Extensive experiments conducted on benchmark pattern recognition datasets demonstrate that the proposed MSDAN consistently outperforms conventional deep learning models and state-of-the-art domain adaptation approaches, achieving a classification accuracy of 91.6% with precision, recall, and F1-score values exceeding 90%. Ablation studies further validate the effectiveness of each architectural component, confirming that the synergistic integration of multi-source learning and multilevel feature fusion significantly enhances generalization performance under complex domain shift scenarios.

Advances in wearable and implantable medical devices have increased the demand for ultrasensitive, reliable, and secure strain sensors capable of continuously monitoring subtle physiological signals. Conventional sensors often suffer from limited sensitivity, mechanical mismatch, and security vulnerabilities, restricting their applicability in next-generation healthcare systems. In this work, we present a Secure Electronics Medical Ultra-Sensitive Core–Sheath Fiber Strain Sensor integrated with a nanogenerator for self-powered medical monitoring applications. The proposed core–sheath fiber architecture enables ultra-sensitive detection of micro-strains, achieving a sensitivity of 0.45 mV/με, low RMS noise (5.2 μV), and high linearity (98%), allowing precise monitoring of physiological motions such as pulse, respiration, and joint movements. Dynamic analysis demonstrates a rapid response time of 18 ms, minimal hysteresis (2.1%), and excellent repeatability over 10,000 cycles, ensuring long-term reliability. The integrated nanogenerator provides sufficient energy to autonomously power both the sensing and secure electronics modules, with peak power reaching 109.8 μW at 2% strain, eliminating the need for external batteries. A hardware-level security framework, combining integrity checks, PUF-based authentication, and signal masking, ensures 99.2% integrity detection, 100% authentication accuracy, and 99.5% data confidentiality with minimal power overhead (5.6%). The proposed system demonstrates a synergistic integration of ultra-sensitive sensing, energy autonomy, and secure data management, making it highly suitable for wearable and implantable medical devices. These results provide a robust foundation for next-generation, self-powered, and secure healthcare monitoring solutions.

The rapid growth of Internet of Things (IoT) devices has introduced new security concerns, offering more possibilities to cyber attackers to exploit their vulnerabilities, limited resources, and less secure frameworks. Classical machine learning and deep learning approaches provide adequate threat detection accuracy, but also require more computational power than is available in IoT environments with limited resources. In this paper, we evaluate the performance of lightweight transformer models – MiniLM, TinyBERT, MobileBERT, and ALBERT –for malware detection in IoT traffic. Each model is trained and tested on a benchmark dataset, and their performance is compared with the models proposed in the previous literature in terms of accuracy and training time. Experimental results demonstrate that while all lightweight models achieve competitive accuracy, ALBERT outperforms the rest, achieving 99.48%accuracy performance, making it suitable for real-world IoT deployments.

Efficiently predicting the nature of tropical cyclones through machine learning techniques has always posed a challenge in the quest to save human lives. While existing research has proposed various methods to accurately predict cyclone behavior and reduce its impact on humanity, this paper introduces a unique customized Support Vector Machine (SVM) model. Unlike existing models, this machine learning-based custom model enhances evaluation metrics, offering significant improvements in binary classification forecasting. The paper also presents a schematic diagram outlining an architectural design for cyclone nature detection utilizing satellite images. The proposed customized SVM model achieves impressive classification metrics, with accuracy at 95%, precision at 94.78%, recall at 94.5%, and an F1-score of 94.9%. In contrast, other models such as Random Forest (RF), SVM, decision tree (DT), and Logistic Regression (LR) fall short, failing to reach an accuracy exceeding 92%. Furthermore, future work may involve the development of hybrid models.