Agent-based systems have been widely adopted in industries such as communication, business, and manufacturing, to support distributed decision-making, scalability, and responsiveness in complex environments. In manufacturing, conventional agent-based systems automate tasks ranging from product design to supply chain management, but they remain constrained by fixed rules and narrow-domain models, limiting their ability to adapt to changing conditions or reason over unstructured and multimodal data. Recent advances in artificial intelligence, particularly Large Language Models (LLMs) and multimodal Foundation Models (FMs), introduce new capabilities such as contextual analysis, flexible task execution, and natural language interaction. However, their systematic role within manufacturing agent-based systems remains insufficiently explored. In this paper, we investigate Agentic AI, defined as LLM-driven, goal-oriented agents operating within structured workflows, and analyze how it extends the roles and behaviors of agents in the manufacturing ecosystem. We present a system-level framework that formalizes agent roles and enables role-constrained orchestration with human-in-the-loop oversight in manufacturing systems. Our approach integrates Agentic AI through explicit role definitions, structured task decomposition, and closed-loop human supervision. The approach is illustrated through a melt-pool image analysis case study in powder bed fusion additive manufacturing, and contrasted with a conventional agent-based pipeline. The results highlight functional differences in adaptability, interpretability, and decision support, indicating that Agentic AI systems can complement conventional agents by providing contextual analysis and actionable recommendations, where human oversight is important. Our framework outlines role-constrained AI capabilities in manufacturing, illustrating how Agentic AI can transform, extend, and integrate with existing manufacturing agent-based systems.

This paper reviews the design of customized 3D printed (also referred to as additively manufactured) implants. A focus is placed on the information flow of design as it is processed, starting from a patient’s scan and culminating with the 3D printing compatible customized implant’s design. We discuss the challenges related to the introduction of 3D printing technologies into the design of the implant, the variabilities encountered, and opportunities for standardization. The paper identifies research and standardization gaps in four stages of a 3D printed customized implant’s design process, namely, medical imaging, constructing CAD (3D) model of VOI, design, and 3D printing compatible file formatting. We hope the paper will help drive research to overcome future challenges encountered in the design process of 3D printed customized medical implants.

Refractory alloys (RAs) are promising materials due to their exceptional physicochemical properties, but most research remains at the laboratory scale. For broader adoption, advancements in manufacturing are essential. Because their high stability makes conventional methods like machining and casting difficult, additive manufacturing (AM) is emerging as an effective approach for fabricating refractory alloy components. However, AM's repeated non-equilibrium thermal cycles introduce undesired features (e.g. defects, anisotropic microstructures, and residual stresses), which are magnified due to RAs’ unique properties. This paper comprehensively reviews the state-of-the-art methods of AM for refractory alloys. It explores data analytics techniques to establish design rules based on multi-fidelity experimental and computational methods. Furthermore, it investigates integrated, collaborative efforts to harmonise standalone databases, information, knowledge, and predictive models at multi-physics, multi-stage, and multi-scale. Unlike the existing literature that focuses primarily on material systems or process fundamentals, this work provides an integrated perspective on AM of refractory alloys from a data analytics standpoint, highlighting the roles of integrated computational materials engineering (ICME), verification, validation, and uncertainty quantification (VV&UQ), and digital twin-driven qualification in overcoming data scarcity and accelerating rapid qualification.

Variability in the additive manufacturing process and powder material properties affect the microstructure which influences the macro-scale material properties. Systematic quantification and propagation of this uncertainty require numerous process-structure-property (P-S-P) simulations. However, the high computational cost of the P-S simulation (thermal model), which relates the microstructure to the process parameters, necessitates the need for inexpensive surrogate models. Moreover, the P-S simulation generates a high-dimensional microstructure image; this presents a challenge in constructing a surrogate model whose inputs are process parameters and output is the microstructure image. This work addresses this challenge and develops a novel approach to surrogate modeling. First, a dimension reduction method based on combining the concepts of image moment invariants and principal components is used to map the high-dimensional microstructure image into latent space. A surrogate model is then constructed in the low-dimensional latent space to predict the principal features, which are then mapped to the original dimension to obtain the microstructure image. The surrogate model-predicted microstructure image is verified against the original physics model prediction (thermal model + phase-field) of the microstructure image, using Hu moments. Developing this surrogate modeling approach paves the way for solving computationally expensive tasks such as uncertainty quantification and process parameter optimization.

Digital twins (DT)s can be understood as fit-for-purpose technologies that link the physical to virtual counterparts. In principle, the fit-for-purpose concept enables DTs to support and advance a variety of existing technologies, including additive manufacturing (AM). Many DTs have been proposed to improve several purposes within the AM process chain, assuming the DT developer has free communication with the AM user. However, in the scenario of restricted communication, it is unclear how the AM user can identify and evaluate existing DTs without its developer. It is for this scenario that the governing fit-for-purpose and DT purpose are defined to provide clarification. In this paper, the DT purpose defined by both the developer and the user align to achieve fit-for-purpose requirements. A five-step conceptual methodology to realize the alignment is proposed that is both traceable and can be partially automated. Illustration of the proposed methodology is performed to identify a product, equipment, and facility DT that align with three user defined metrics of functional suitability, performance efficiency, and interaction capability. From this case, four suitable DTs from 32 candidates are identified. This paper aims to provide a direction to achieve fit-for-purpose between AM users and DT developers; whereby helping to advance both technologies.

In this paper, we briefly discuss the evolving concepts of the digital economy, digital threads, and product realization (manufacturing) in the digital economy-enabled futuristic world. We posit that if we need to become self-sufficient, it is imperative that small and medium manufacturing enterprises need to be equitably integrated into the digital economy. We lead this discussion to the most important research questions related to representation, search, and composition to achieve such a realization.

Additive Manufacturing (AM) is gaining popularity in the industry for its cost-effectiveness and time-saving benefits. However, AM encounters challenges that need to be addressed to enhance its efficiency. While Machine Learning (ML) can tackle various AM challenges, it is often limited to specific issues, necessitating multiple models. In contrast, Generative Artificial Intelligence (GenAI) has the potential to mitigate instance-specific bias due to its broader training. This paper presents a comprehensive methodology for evaluating the capabilities of various existing GenAI tools in addressing diverse AM-related tasks. We propose three categories of metrics, totaling 35 metrics, namely agnostic, domain task, and problem task metrics. Additionally, we introduce a scoring matrix, a practical tool that can be used to assess the responses of different GenAI tools. The study involves data collection from diverse published papers, which are used to create inquiries for GenAI tools. The results demonstrate that transformer-based models, such as multi-modal GPT-4 and Gemini (prev. BARD), can handle both AM image and text data. In contrast, uni-modals such as GPT-3 and Llama 2 are proficient in processing AM text data. Furthermore, image-based models such as DALL center dot E 3 and Stable Diffusion can accept AM text data and generate images. It is also observed that the performance of these models varies across different AM-related tasks. The variation in their performance may be due to their underlying architecture and the training dataset.

Recent advances in Additive Manufacturing (AM), particularly in production scenarios, have been largely driven by insights achieved through data analytics. AM has greatly benefited from the increasingly large amounts of data generated during the design to product transformation. Despite the large amounts of data that can be generated from each build, the variations that can occur between builds create challenges in data reuse. Advances in data analytics are coming in the form of advanced machine learning algorithms, and are often associated with deep learning, where large amounts of data are analyzed, and interpretations are made. While such algorithms are increasingly adept at solving complex problems, solutions are highly dependent on the data used to train the algorithms and thus often subject to unwanted bias. Contextualization of data can help limit unintended bias and improve the probability of attaining viable results. Data contextualization often comes through domain context. This work describes the development of an AM ontology to support context-aware data analytics. The Additive Manufacturing Data Analytics Ontology, or AMDA Ontology, is developed to facilitate the contextualization of AM data through the representation of explicit AM concepts throughout the design to product transformation and the encoding of inhering relationships within the concepts. The AM concepts are accompanied by a suite of concepts representative of the necessary modeling, simulation, and analytics terms to create links between AM data, AM data analytics opportunities, and appropriate machine learning algorithms. The early results indicate that AMDA ontology has the ability to facilitate key correlations between AM data and the analytic opportunities to enhance the design to product transformation of AM parts.

The rapid pace of maturation of metal additive manufacturing (AM) technologies, makes them an excellent candidate for the fabrication of nuclear power plant (NPP) components. However, the current levels of process variations create numerous challenges yet to be overcome for industrial acceptance of AM for NPP applications. One of those challenges is associated with the qualification and certification of fabricated components for mission-critical applications such as NPPs. To reduce the cost of qualification and certification, in-process monitoring and in-process non-destructive evaluation (NDE) are considered essential tools. This report aims to support the efforts of the U.S. Nuclear Regulatory Commission (NRC) to meet the challenges of the certification of the NPP components fabricated by AM technologies. It provides a review of the current state-of-the-art in the areas of in-process monitoring and in-process NDE methods, instruments, and relevant standards, identifying gaps in knowledge, technologies, and standards to achieve the goal of using them as robust tools for the AM process and part qualification.

Laser powder bed fusion (LPBF) is a popular additive manufacturing process with many advantages compared with traditional (subtractive) manufacturing. However, ensuring the quality of LPBF parts remains a challenge in the manufacturing industry. This work proposes the use of unsupervised learning, specifically, the k-means clustering method, to identify unique melt pool shapes produced during LPBF manufacturing. Melt pools are a key process signature in LPBF and can assist in the evaluation of process quality. k-means is employed multiple times sequentially to produce clusters of melt pools, and the silhouette value is used to identify the optimal number of clusters. The clusters produced by k-means are used as labels to train a deep neural network to classify the melt pool shapes. By inputting the melt pool image and the corresponding LPBF machine process parameters into the neural network, the neural network identifies the melt pool shape to aid human analysis and provide insight into part quality. The trained neural network is interpreted using explainable artificial intelligence (XAI) methods to investigate the relationships between process parameters and the melt pool shape. Using layer-wise relevance propagation, the process parameters that most significantly influence the melt pool shapes are identified. The relationship between process parameters and melt pool shapes can be useful for selecting the process parameters to produce the desired melt pool shapes. In summary, this study describes an approach that combines unsupervised machine learning and XAI methods to effectively enable the analysis and interpretation of melt pools.

This paper investigates a novel approach to efficiently construct and improve surrogate models in problems with high-dimensional input and output. In this approach, the principal components and corresponding features of the high-dimensional output are first identified. For each feature, the active subspace technique is used to identify a corresponding low-dimensional subspace of the input domain; then a surrogate model is built for each feature in its corresponding active subspace. A low-dimensional adaptive learning strategy is proposed to identify training samples to improve the surrogate model. In contrast to existing adaptive learning methods that focus on a scalar output or a small number of outputs, this paper addresses adaptive learning with high-dimensional input and output, with a novel learning function that balances exploration and exploitation, i.e., considering unexplored regions and high-error regions, respectively. The adaptive learning is in terms of the active variables in the low-dimensional space, and the newly added training samples can be easily mapped back to the original space for running the expensive physics model. The proposed method is demonstrated for the numerical simulation of an additive manufacturing part, with a high-dimensional field output quantity of interest (residual stress) in the component that has spatial variability due to the stochastic nature of multiple input variables (including process variables and material properties). Various factors in the adaptive learning process are investigated, including the number of training samples, range and distribution of the adaptive training samples, contributions of various errors, and the importance of exploration versus exploitation in the learning function.

Metal additive manufacturing machines are complex and inherently digital and often cyber-physical systems. As the adoption of this manufacturing technology increases and it becomes increasingly industrialized, concerns about security are evermore prevalent. Both cyber and non-cyber related attacks on critical infrastructure such as additive manufacturing production systems are causes for concern for industry and government. Both the public and private sectors need to focus on securing their information systems to reduce the risk of security attacks and their adverse effects. This research aims to apply the National Institute of Standards and Technology's Risk Management Framework to the metal additive manufacturing production scenario. The Risk Management Framework defines a rigorous, yet flexible and repeatable, process for managing security risk. A model-based assessment approach is proposed to leverage the digital nature of this manufacturing technology. A case study is performed to demonstrate this approach for a commercial laser powder bed fusion machine in its operating environment. This case study focuses on the technological security risks. This study demonstrates how a model-based approach maximizes the benefits of the Risk Management Framework by improving information and decision traceability for addressing metal additive manufacturing security risks.

In metal laser powder bed fusion, a laser beam rapidly scans through a thin layer of powder. The laser heats up and liquefies the powder forming a melt pool. The melt pool characteristics affect the material property and physical performance of the part. Physical simulations of melt pool dynamics are computationally expensive and are often limited to a short laser scan. Alternatively, Finite Element (FE) simulations for part-scale thermal history are less computationally intensive but do not capture the detailed melt pool temperature profile and geometry in general. Accurate modeling of the high spatial and temporal thermal gradients inside and near the melt pool is essential to many applications of the thermal history including design and optimization of the laser scan path, material microstructure characterization, and deformation prediction.In this work, we developed MeltpoolGAN, the first conditional Generative Adversarial Network (cGAN) to predict the melt pool image based on the simulated thermal history along the laser scan path. The integration of generative deep learning and thermal history simulation achieves melt pool-level accuracy while significantly reducing the physical and computational complexity. The data pair of thermal history and melt pool images for neural network training and validation are obtained through Contact-Aware Path Level (CAPL) thermal simulation and a co-axial melt pool monitoring system developed by NIST, respectively. The predicted melt pool morphology is validated against experimentally acquired melt pool images through geometric characteristics, including the length and width, of the melt pool.

Abstract This special section contains a selection of 14 manuscripts (2 Technical Briefs and 12 Research Papers) from the 43nd American Society of Mechanical Engineers (ASME) Computers and Information in Engineering (CIE) Conference that was held in Boston, Massachusetts, August 20-23, 2023, in conjunction with the International Design Engineering Technical Conferences (IDETC). Nominated by the four technical committees, namely Advanced Modeling and Simulation (AMS), Computer Aided Product and Process Development (CAPPD), Systems Engineering, Information and Knowledge Management (SEIKM), and Virtual Environments and Systems (VES) based on the conference paper review results, these papers reflect recent and relevant advancements in these technical areas.

Abstract This article summarizes the outcomes from a NIST-sponsored workshop that examined the additive manufacturing (AM) data management pain points experienced by small- and medium-sized enterprises in interacting with larger organizations and government procurement agencies. Three prominent themes emerged: cost of compliance, technology gaps associated with uncertainty and AM process qualification, and a desire for government leadership to address these issues. The article includes lists of specific action items proposed at the workshop to address these challenges.

In this paper we outline the development of a scalable PBF thermal history simulation built on CAPL and based on melt pool physics and dynamics. The new approach inherits linear scalability from CAPL and has three novel ingredients. Firstly, to simulate the laser scanning on a solid surface, we discretize the entire simulation domain instead of only the manufacturing toolpath by appending fictitious paths to the manufacturing toolpath. Secondly, to simulate the scanning on overlapping toolpaths, the path-scale simulations are initialized by a Voronoi diagram for line segments discretized from the manufacturing toolpath. Lastly, we propose a modified conduction model that considers the high thermal gradient around the melt pool. We validate the simulation against melt pool images captured with the co-axial melt pool monitoring (MPM) system on the NIST Additive Manufacturing Metrology Testbed (AMMT). Excellent agreements in the length and width of melt pools are found between simulations and experiments conducted on a custom-controlled laser powder bed fusion (LPBF) testbed on a nickel-alloy (IN625) solid surface. To the authors' best knowledge, this paper is the first to validate a full path-scale thermal history with experimentally acquired melt pool images. Comparing the simulation results and the experimental data, we discuss the influence of laser power on the melt pool length on the path-scale level. We also identify the possible ways to further improve the accuracy of the CAPL simulation without sacrificing efficiency.

Laser Powder Bed Fusion (LPBF) additive manufacturing has revolutionized industries with its capability to create intricate and customized components. The LPBF process uses moving heat sources to melt and solidify metal powders. The fast melting and cooling leads to residual stress, which critically affects the part quality. Currently, the computational intensity of accurately simulating the residual stress on the path scale remains a significant challenge, limiting our understanding of the LPBF processes. This paper presents a framework for simulating the LPBF process residual stress based on the path-level thermal history. Compared with the existing approaches, the path-level simulation requires discretization only to capture the scanning path rather than the details of the melt pools, thus requiring less dense mesh and is more computationally efficient. We develop this framework by introducing a new concept termed effective thermal strain to capture the anisotropic thermal strain near and around the melt pool. We validate our approach with the high-fidelity results from the literature. We use the proposed approach to simulate various single-island scanning patterns and layers with multiple full and trimmed islands. We further investigate the influence of the path-level thermal history and the layer shape on the residual stress by analyzing their simulation results.

Data analytics with Machine Learning (ML) using physics knowledge and big data offers high potential to continuously transform raw data to newfound knowledge of Process-Structure-Property (PSP) causal relationships. In Additive Manufacturing (AM), however, realizing the potential is still limited largely due to the lack of a systematic way to learn the PSP relationships for various AM processes. To address the limitation, this paper proposes a novel framework driven by physics-guided ML, which consists of three tiers: (1) knowledge of predictive PSP models and physics, (2) PSP features of interest, and (3) raw AM data. The framework defines a PSP-learning process with two sub-processes. The first uses a knowledge-graph-guided top-down approach to generate the requirements for predictive analytics and data acquisition. The second uses a data-driven bottom-up approach to construct and model new PSP knowledge. Together, these processes connect the proposed framework to decision-making and control activities and physical and virtual AM systems, respectively. The paper includes a case study based on Laser Powder Bed Fusion processes including AM Metrology Testbed at the National Institute of Standards and Technology (NIST). The case study introduces predictive ML models and PSP knowledge extracted from the models. We also demonstrate the framework using an ML-Integrated Knowledge Extraction module called MIKE in NIST's collaborative AM Material Database. The framework newly enables a systematic physics-guided data-driven approach for PSP in AM that can couple physics knowledge with the versatility of data-driven ML models. Using the approach, the framework continuously updates the models (1) to improve the understanding of dynamically generated AM data and (2) to link sub-models into coupled PSP models. Based on the improved understanding, the framework also facilitates decision-making and control activities for AM at multiple scales.