Search for: All records

Abstract Simulation modeling plays an important role in improving the understanding and control of manufacturing processes. Traditional approaches, such as finite element analysis, rely on dense meshes of elements, which can be computationally expensive for complex geometries. Network representations encode complex structures into sparse adjacency matrices, which show great potential as a meshless alternative for thermal modeling in additive manufacturing (AM) processes. By representing heat transfer as a diffusion process over sparsely connected nodes, network models provide an opportunity for efficient simulation of thermal behaviors such as heat transfer and temperature gradients, thereby accelerating process-structure-property mapping and control of AM. However, very little has been done on constructing network simulation models that effectively capture the underlying part geometry. This article presents a new self-organizing network approach for thermal simulation in AM processes. First, the part geometry is encoded into a sparse adjacency matrix, from which the original structural geometry can be recovered through a self-organizing process. Second, the governing diffusion equations are formulated and solved directly on the adjacency matrix to simulate heat transfer. Third, a new Markov Chain Monte Carlo formulation is proposed to stochastically calibrate the network model for thermal simulation. The proposed methodology is evaluated and validated through case studies of multi-layer AM processes. Numerical simulation results show that the optimized self-organizing network not only reconstructs a geometry resembling the original part but also enables accurate and efficient simulation of thermal behaviors. This study establishes a generalized framework for optimal design and calibration of network-based thermal models, which demonstrates strong potential for improving the prediction and control of advanced manufacturing processes. 

Early detection of cognitive impairment in assisted living is hindered by time-intensive tools like the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). We present a 60-second voice-based screening model that analyzes picture descriptions to estimate dementia risk. While recent deep learning approaches have shown promise on similar tasks, their lack of interpretability, large data requirements, and computational complexity limit clinical adoption. Using transcripts from the DementiaBank corpus, our model integrates traditional linguistic features (pause rate, pronoun use, syntactic complexity) with latent semantic dimensions extracted from language model embeddings. These semantic axes, interpretable constructs like “Drift & Hesitation” or “Over-detailed Narration”, consistently emerged as top predictors and may represent novel linguistic biomarkers of early decline. The final ElasticNet classifier is sparse, interpretable, and outperforms known non–deep learning baselines (AUC = 0.858), exceeding MMSE. Its simplicity enables deployment in mobile apps or in-room monitors, offering scalable, low-burden screening for early dementia. This work supports a shift toward linguistically grounded, tech-enabled cognitive care in aging populations. 

Manufacturing systems often face unexpected disruptions such as machine failures or material shortages, which can severely impact the production performance. Traditional methods for addressing these disruptions tend to be time-consuming and resource-intensive, which cannot effectively maintain the resilience of manufacturing systems. Despite recent advances in digital twins (DTs) and artificial intelligence (AI), very little has been done to mitigate high computational demands and generate resilient designs of manufacturing system under uncertainty. Therefore, this paper presents a new Generative Adversarial Network (GAN) approach for the on-the-fly design of manufacturing systems in response to production disruptions. First, we propose a novel Generative Adversarial Network (D-GAN) to generate diverse, adaptive system designs that align production performance with target key performance indicators (KPIs). Second, DT models are coupled with statistical metamodeling to sequentially generate large amounts of training data samples under different scenarios. Experimental results show the high potential of the proposed D-GAN approaches to generate cost-effective system designs and enhance manufacturing resilience. 

STEM-based computational modeling (CM) activities vary in their instructional demands depending on the computational thinking (CT) practices they emphasize. This qualitative study examines upper elementary teachers’ perceptions and instructional experiences with three CM activity types: constructing CMs, interpreting and evaluating CMs, and exploring and evaluating simulations. Drawing on interviews with six fifth and sixth-grade teachers following professional learning and classroom enactment, we analyzed how teacher background, classroom context, and curriculum type shaped perceptions of instructional difficulty, effectiveness, and student affordances. Teachers consistently perceived simulation-based activities as easiest and most effective to teach, followed by model evaluation and then model construction. Prior STEM experience shaped teachers’ sensemaking, particularly in anticipating programming-related challenges. Teachers also perceived differential engagement and accessibility across activity types. These findings highlight the value of offering multiple CM activity types to support diverse teacher and student needs. 

The rapid evolution of modern manufacturing systems is driven by the integration of emerging metaverse technologies such as artificial intelligence (AI), digital twin (DT), and different forms of extended reality (XR) like virtual reality (VR), augmented reality (AR), and mixed reality (MR). These advances confront manufacturing workers with complex and evolving environments that demand digital literacy for problem solving in the future workplace. However, manufacturing industry faces a critical shortage of skilled workforce with digital literacy in the world. Further, global pandemic has significantly changed how people work and collaborate digitally and remotely. There is an urgent need to rethink digital platformization and leverage emerging technologies to propel industrial evolution toward human-centered manufacturing metaverse (MfgVerse). This paper presents a forward-looking perspective on the development of MfgVerse, highlighting current efforts in learning factory, cognitive digital twinning, and the new sharing economy of manufacturing-as-a-service (MaaS). MfgVerse is converging into multiplex networks, including a social network of human stakeholders, an interconnected network of manufacturing things or agents (e.g., machines, robots, facilities, material handling systems), a network of digital twins of physical things, as well as auxiliary networks of sales, supply chain, logistics, and remanufacturing systems. We also showcase the design and development of a virtual learning factory for workforce training. Finally, future directions, challenges, and opportunities are discussed for human-centered manufacturing metaverse. We hope this work helps stimulate more comprehensive studies and in-depth research efforts to advance MfgVerse technologies. 

Both college educators and students are increasingly using artificial intelligence (AI) and large language models (LLM) in the classroom, but without consistent levels of thoughtfulness. Much has been written about cheating, plagiarism, and other unethical uses; some have raised concerns about how such tools might reduce opportunities for deep learning. This calls for rethinking how we use GenAI to achieve our learning objectives and create course assessments. Interdisciplinary domains offer a useful context for developing AI learning supports. Here we take a relatively new interdisciplinary domain—data storytelling—and walk through the steps toward designing useful intelligent support. This paper argues that, given the prevalence of such tools, educators can adapt and achieve positive outcomes if they initially document how learning takes place without such tools, what challenges they face, and design ways that intelligent tools can support improved learning. For data storytelling, we used an evidence-centered design method to understand how relevant teaching and learning unfold without intelligent technologies. Then we documented the types of intelligent instructional supports and interactions that educators and their students desired. The paper includes recommendations on how postsecondary educators, educational researchers, instructional designers, and professional developers may apply these ideas and methods in their courses. Students might still take shortcuts with GenAI, but if more postsecondary educators can take a critical look at the gaps in their current instructional methods, and find ways to use AI to address them by supporting the hard-to-implement tasks— such as tracking nuanced learning progress (formative assessment) and providing targeted learning supports—postsecondary educators can pave the way to constructive, humanistic use of GenAI tools. 

The approval of disease-modifying treatments for Alzheimer's disease demands a rethinking of cognitive screening. Drawing on over 180 stakeholder interviews from the NSF National I-Corps program, this perspective highlights barriers in current workflows, from time constraints in primary care to learning effects in long-term care, and presents innovation pathways centered on AI and digital biomarkers. Speech analysis, in particular, offers a scalable and cost-effective screening tool aligned with existing CPT codes. We outline implementation strategies and emphasize the urgent opportunity to align technological innovation with frontline clinical needs to ensure advances translate into meaningful patient and provider benefit. 

Abstract Dementia care is reaching a technological turning point as rapid advances in artificial intelligence (AI), digital biomarkers, and sensor‐based monitoring continue to outpace current clinical practice. Yet most tools remain episodic and disconnected from the everyday environments where cognitive change unfolds. Early cognitive drift often goes unnoticed, the mild cognitive impairment (MCI) window remains a major diagnostic gap, and care teams lack continuous, actionable insights. In this perspective, we introduce Agentic Neurotechnology, an integrated framework that combines adaptive multi‐agent AI systems with longitudinal cognitive modeling to support proactive and individualized dementia care. We define a five‐layer agentic stack consisting of (1) sensing and digital biomarker agents, (2) interpretation and reasoning agents, (3) longitudinal cognitive trajectory models, (4) care navigation agents, and (5) population‐level forecasting agents. We further outline the Cognitive Continuum Loop, a closed‐cycle process in which detection, interpretation, reasoning, monitoring, intervention, and adaptation occur continuously across the care environment. This framework highlights three actionable windows for cognitive intervention: the Preclinical Risk Detection window, the MCI Detection window, and the Functional Instability Response window. Each stage presents unique opportunities for agentic AI systems that operate continuously rather than through isolated assessments. We conclude by presenting a translational pathway for deploying agentic neurotechnologies across assisted living, primary care, telehealth, and home‐based settings. The goal is to enable earlier detection, greater clinical efficiency, and more personalized cognitive support. Agentic Neurotechnology reframes AI not as a passive screening tool but as an active partner in continuous cognitive care.