General-purpose large language models (LLMs) often struggle to generate reliable responses in specialized engineering domains due to limited domain grounding and insufficient exposure to structured technical knowledge. This study investigates practical strategies for adapting a foundation LLM to the additive manufacturing (AM) domain in order to improve answer accuracy, relevance, and usability for expert-level question answering. AM knowledge is distributed across heterogeneous sources such as academic literature, manufacturer documentation, technical standards, and procedural guides. Although general LLMs demonstrate strong linguistic capabilities, they frequently fail to retrieve and contextualize such domain-specific information. Two common approaches to address this limitation are domain-specific fine-tuning and retrieval-augmented generation (RAG). We construct a curated AM corpus and evaluate three configurations based on LLaMA-3-8B: (1) the pretrained baseline model, (2) a RAG system that retrieves relevant document chunks from a vector database, and (3) a model fine-tuned on raw domain text. Performance is evaluated using 200 expert-designed AM questions assessed by mechanical engineering experts for accuracy, relevance, and overall preference. Results show that the RAG model consistently outperforms the baseline. Among the 200 questions, 75.5
In production engineering, the identification of optimal process parameters is essential to advance product quality and overall equipment effectiveness. Optimizing and adapting process parameters through experimental design is relevant for different phases of the life cycle of a production process: (i) design and development of new processes, (ii) failure analysis and optimization, and (iii) adaptation and calibration in series production. Existing experimental design approaches tend to be inefficient because they comprise static, non-adaptive methodologies that separate experiment design from execution and analysis. Instead, Bayesian Optimization (BO) offers an adaptive and data-efficient methodology for experimental design termed Bayesian experimental design (BED). In BED, the selection of an experiment is re-evaluated in each iteration based on previous experiment results according to an acquisition function that aims to maximize the informational content of each experiment. However, the configuration of BO algorithms for specific optimization problems requires extensive knowledge of both BO and process characteristics. The mean and covariance functions of the surrogate model, the acquisition function, and initial data sampling must be individually configured and significantly influence overall optimization performance, preventing widespread adoption in production engineering practice. To guide the configuration of BO algorithms for optimizing production processes, in this paper, we perform an extensive benchmark study with a total of 15,360 experiments. We evaluate the performance of a variety of BO algorithm configurations (including kernels, acquisition functions, and initial sampling sizes) on a total of eight optimization problems with a noiseless and a noisy variant each. The performance and robustness analysis reveals significant performance differences between individual BO algorithm configurations. The results of our benchmarking serve as empirical references based on which we derive actionable guidelines for the application of BED in production engineering.
Integrating AI-driven applications in manufacturing processes holds immense potential for enhancing production performance. However, adopting AI technology in manufacturing presents significant challenges, including the frequent need for human and AI work harmonisation and the knowledge gap among diverse stakeholders. This paper presents a development framework for human-AI systems that effectively integrates human work in manufacturing settings. The framework is specially designed for development project leaders from the manufacturing domain who play a critical role in integrating the diverse expertise required to realise such systems. The framework was derived based on a literature review, and its practical utility was examined through empirical applications. The resulting framework contains 147 task recommendations associated with six conceptual development progression stages. The empirical studies, including external reviews and preliminary empirical validation of the framework through pilot applications to two real-world projects, indicate the framework’s practical utility, such as facilitation of deep multi-domain dialogues for early phase project planning and mid-term reflective planning, as well as remarks, such as its use in iterative development environments. The present study extends the existing manufacturing research by offering deeper insights into and advancing a structured approach to developing human-integrated AI systems in industrial environments, which the study found to be immature.
This study investigates the influence of infrared (IR)-assisted automated fiber placement (AFP) process parameters on the mechanical performance of carbon fiber reinforced thermoplastic composite to fabricate defect-free structures. The effect of process parameters on the processability of carbon fiber-polycarbonate (CF/PC) was analyzed and presented. A structured design of experiments approach was adopted to study the effects of the compaction force and tool temperature on the flexural and lap-shear strength of the manufactured CF/PC parts. The fractography and the interlaminar adhesion were analyzed using digital microscopy. Further, the CF/PC laminated were successfully laid onto a rotating 3D printed polyetherimide cylindrical mold. This work shows, for the first time, the ability of IR-assisted AFP to manufacture CF/PC laminates with improved mechanical properties, demonstrating its potential as a promising material for semi-structural applications.
The rapid advancement and expanding application of robotics in various sectors necessitate the development of versatile and efficient gripping systems. Robotic grippers, serving as the primary interface between robots and the physical world, are pivotal for the success of robotic applications ranging from industrial automation to delicate tasks in healthcare. This paper delves into the diverse landscape of robotic grippers, examining their design principles, material choices, and the intricate balance between flexibility and strength required to cater to a broad spectrum of operational demands. It further explores the spectrum of actuation mechanisms that empower these grippers, from traditional electric and pneumatic systems to cutting-edge soft robotics technologies that offer unprecedented adaptability and safety. The integration of sensors within grippers, enhancing their intelligence and autonomy, is also scrutinized, alongside a critical evaluation of control strategies that range from conventional PID controllers to sophisticated machine learning algorithms. In addition, this paper emphasizes the significance of real-time control and adaptation in ensuring robots’ ability to respond promptly to environmental changes, leveraging fast data processing and feedback mechanisms for improved performance. The aim of this comprehensive review is not only to provide a foundational understanding of the current state of robotic gripper technology but also to highlight key considerations and emerging trends that will shape the future of robotic manipulation. Through this exploration, the paper seeks to equip designers, engineers, and researchers with the insights needed to innovate and advance the field of robotic gripping systems.
IntroductionOver the past seven decades, a significant volume of research has been dedicated to manufacturing systems due to their importance in the worldwide economy. Much of this research has focused on using Markov stochastic modeling to formulate manufacturing systems problems. During the research effort, numerous numerical methods have been developed for solving such systems; however, relatively few formulas have been proposed. That is because even small systems are characterized by the well-known state explosion problem.MethodsIn short serial production lines, the underlying Markov chain is depicted as a graph of the transition diagram, which is constructed by implementing an algorithm. The steady-state probabilities are extracted in the symbolic form of two polynomial ratios. That is accomplished by employing a recently introduced method that assigns probabilities in symbolic form on the graph anti-arborescences. Finally, the performance metrics of the short production line can be obtained in exact closed-form expressions via its known definition from extant literature using straightforward algebraic operations.ResultsThe closed-form formulae for the performance metrics of short serial production lines (e.g., throughput, maximum utilization, work in process, blocking probability for the second station, probability of the third station being idle, etc.) with two, three, and four stages, absent buffers, are presented herein for the first time in the extant literature. The proposed algorithm results shed light on the well-known phenomenon of production lines known as the “bowl phenomenon”. Comprehending the formula structure enables the formulation of a straightforward model for throughput estimation for fully balanced short serial production lines using genetic programming for lines up to thirteen stages.DiscussionThe enormous size of the exact formulas highlights the need for more computational support for production lines larger than four stages without buffers. A comprehensive understanding of the underlying principles governing exact formulas will facilitate the implementation of innovative mathematical approaches to problem-solving. This understanding will also enable artificial intelligence to derive precise mathematical relationships with reduced complexity, thereby fostering intuition in production lines.
IntroductionMarkov chains are a powerful tool for modeling systems in various scientific domains, including queueing theory. These models are characterized by their ability to maintain complexity at a low level due to a property known as the Markov property, which enables the connection between states and transition probabilities. The transition matrices of Markov chains are represented by graphs, which show the properties and characteristics that help analyze the underlying processes.MethodThe graph representing the transition matrix of a Markov chain is formed from the transition state diagram, with weights representing the mean transition rates. A probability space is thus created, containing all the spanning trees of the graph that end up in the states of the Markov chain (anti-arborescences). A successive examination of the graph’s vertices is initiated to form monomials as products of the weights of the edges forming the symbolic solution.ResultsA general algorithm that commences with the Markov chain transition matrix as an input element and forms the state transition diagram. Subsequently, each vertex within the graph is examined, followed by a rearrangement of the vertices according to a depth-first search strategy. In the context of an inverted graph, implementing a suitable algorithm for forming spanning trees, such as the Gabow and Myers algorithm, is imperative. This algorithm is applied sequentially, resulting in the formation of monomials, polynomials for each vertex, and, ultimately, the set of polynomials of the graph. Utilizing these polynomials facilitates the calculation of the stationary probabilities of the Markov chain and the performance metrics.DiscussionThe proposed method provides a positive response to the inquiry regarding the feasibility of expressing the performance metrics of a system modeled by a Markov chain through closed-form equations. The study further posits that these specific equations are of considerable magnitude. The intricacy of their formulation enables their implementation in smaller systems, which can serve as building blocks for other methodologies. The correlation between Markov chains and graphs has the potential to catalyze novel research directions in both discrete mathematics and artificial intelligence.
IndroductionPersonalized medical devices, especially scaffold-based implants, are increasingly important in medical care. One established manufacturing method for these products is extrusion-based 3D printing, also called 3D material extrusion (MEX) or extrusion additive manufacturing (EAM). According to the current state of the art, this technique lacks scalability, as many adjacent processes, such as material handling or quality control, are still carried out manually and no holistically automated solutions have been established.MethodsThis work examines the extrusion-based 3D printing process for manufacturing cell-free porous scaffolds. Based on a literature review, relevant process parameters for MEX and quality attributes of polymer-based scaffolds are analyzed to derive functional requirements for holistically automating the manufacturing process. A concept for an end-to-end automated production infrastructure is developed, to allow efficient and scalable manufacture of scaffolds. All process parameters are analyzed for their influence on the quality attributes, and requirements are specified. Based on this, the development of the production concept is systematically carried out.ResultsThe resulting technical system consists of a magnetic planar drive, which is used as an intralogistic transport system, but also forms the horizontal axis plane of the 3D printer. The resulting frictionless levitating print bed increases cleanroom suitability and enables the parallelization of print jobs and quality control steps for improved production flexibility and scalability. The central approaches of the concept are presented in a physical demonstrator.DiscussionAn initial proof of concept for planar drive-based MEX is provided and lays the foundation for further development and validation of the conceptualized production infrastructure.
This article comprehensively analyses AJP technology, with a greater focus on the areas that received scant attention in the previously published literature. Whereas various reviews so far discussed the basic principles of AJP and its comparison with other printing techniques, the present article goes further to discuss different types of functional inks being utilized in AJP, including conductive, dielectric, semiconducting, and biological inks. The minimum resolutions of micropatterns achieved with these inks are then reviewed, together with the specific printing recipes enabling their use, to give an overview of the performances of different materials within the AJP process. Furthermore, the article classifies the dimensionality of AJP-printed patterns into 2D-planar, 2D-nonplanar, and 3D parts, underlining the capability of the technology for the fabrication of both planar and non-planar geometries. This makes AJP a tool of major relevance in the newly emerging fields of electronics, sensors, and biotechnology, which strongly demand precise micro-patterning and substrate adaptability. The review, therefore, explains how AJP is bound to change manufacturing processes by exploring its new applications in those sectors. The article also covers the current limitations of AJP, including how to optimize printing processes and generalize them into more industrial uses. Synthesizing state-of-the-art research, this review not only describes the main achievements of AJP technology but also points out likely future tendencies and even disruptions that may occur within this field. This review aims to be an extensive source of information for both researchers and industry representatives interested in finding opportunities for further applications of AJP in various areas.
Digital transformation is critical for small and medium-sized manufacturing enterprises (SMMEs) in China to enhance competitiveness within the sharing economy, particularly in the production capacity domain, which accounts for 32.75% of the sharing economy market. However, SMMEs face significant technological, financial, organizational, and cultural barriers that hinder their digitalization efforts. This study conducts a comprehensive literature review to analyze barriers to digital transformation and the role of sharing economy platforms. It proposes strategies leveraging chain-centric, industry-brain, and shared service cloud platforms. A case study of Company F, an SMME in the elevator industry, is analyzed using quantitative financial data (2015–2022) and qualitative interviews, employing statistical methods (e.g., regression analysis) and thematic analysis with NVivo to evaluate the impact of a shared service platform provided by Company L. The case study reveals that Company F's digital transformation, facilitated by Company L's shared service platform, significantly improved profitability and operational efficiency. Post-transformation (2019–2022), the mean main business profit margin increased from 32.65% to 45.94%, and total asset turnover rose from 0.64 to 0.84. Regression analysis confirmed significant gains (p < 0.05 for profit margin, p < 0.01 for asset turnover). Qualitative findings highlight enhanced maintenance efficiency, streamlined spare parts management, and increased collaboration, despite a temporary decline in 2020 due to the COVID-19 pandemic. The sharing economy, through data-sharing and collaborative platforms, mitigates SMMEs' resource constraints and fosters digital transformation. These findings underscore the potential of shared platforms to overcome transformation barriers, offering actionable strategies for SMMEs and policymakers. While focused on China, the study suggests broader applicability to other middle-income countries, emphasizing the need for supportive digital ecosystems and regional considerations to ensure equitable transformation.
There is growing research on the implementation of Digital Materials Passports (DMPs) in different industries, yet there is lack of guidance for preparing the human and organizational components within this ecosystem of change. To help fill this void, in this position paper, we develop propositions for dealing with the people and organizational challenges of implementing DMP’s within organizations and across supply chains. Applying a socio-technical systems approach, we highlight the interconnectedness between the human, organizational, and technical factors when designing and implementing DMPs. We also use the IGLOO framework which highlights that organizational support needs to occur at the individual, group, leader, organizational, and omnibus (interorganizational) levels. We draw on research from the literature on human behavior at work, covering areas such as social identity, trust, resilience in organizations, leadership, participatory job redesign, and training and learning as mechanisms to reduce socio-technical challenges and to reach important interorganizational goals. Understanding these mechanisms helps us to develop seven propositions that organizations and supply chains can put in place when implementing DMPs. These propositions can offer mutually reinforcing support for organizations when implemented, and can be adapted to consider both the long-term and the immediate implementation context. We also discuss the role of employee involvement in enhancing the benefit of the propositions for organizations and supply chains in moving towards Industry 5.0.
The integration of sustainable additive manufacturing (AM) within the framework of African industrialization presents a promising avenue for economic advancement while addressing environmental concerns. This review explores the convergence of sustainable AM practices with the industrial landscape of Africa, highlighting potential benefits and challenges. Through efficient resource utilization and localized production capabilities, AM holds promise for enhancing industrial resilience, stimulating employment opportunities, and fostering innovation. However, the realization of these benefits necessitates navigating infrastructural limitations, technological disparities, and regulatory complexities. By critically examining sustainable AM strategies and their relevance to African contexts, this review aims to delineate actionable pathways for leveraging the transformative potential of AM. The role of AM in industrialization as expressed in the African Union Agenda 2063 are highlighted. This has the potential to increase the staggering ∼11% contribution of manufacturing to gross domestic product of Africa. Collaboration through the triple helix approach focusing on government, industry and academia is highly pivotal for the success of such nascent and ubiquitous AM technology which is able to address the sustainable development goals. Africa can leapfrog and harness sustainable AM as a catalyst for inclusive industrial development and sustainable growth across the continent. The implications of AM for an industrialised Africa and areas for future research direction are briefly discussed.
In the context of Industry 4.0, autonomous and data-driven manufacturing processes are advancing rapidly, with wire arc additive manufacturing (WAAM) emerging as a promising technique for producing large-scale metal components. Ensuring quality control and part traceability in WAAM remains an area of active research, as existing process monitoring systems often require operator intervention and are tailored to specific machine setups and camera configurations, limiting adaptability across industrial environments. This study addresses these challenges by developing an angle-invariant melt pool analysis pipeline capable of recognising bead features in wire-based directed energy deposition from monitoring images captured using various camera qualities, positions, and angles. A new benchmark dataset, WAAM-ViD, is also introduced to support future research. The proposed pipeline integrates two deep learning models: DeepLabv3, fine-tuned through active learning for precise melt pool segmentation (Dice similarity coefficient of 95.90%), and WAAM-ViDNet, a regression-based multimodal model that predicts melt pool width using the segmented images and camera calibration data, achieving 88.71% accuracy. The results demonstrate the pipeline’s effectiveness in enabling real-time process monitoring and control in WAAM, representing a step toward fully autonomous and adaptable additive manufacturing systems.
Titanium alloy Ti6Al4V is extensively utilized in biomedical applications due to its excellent biocompatibility, corrosion resistance, and mechanical properties. The design of dental implant surface textures has changed throughout time to address issues with oral rehabilitation in both healthy and damaged bones. The longevity of an implant is significantly impacted by surface roughness. This study examines the use of laser shock peening (LSP) as a surface modification technique to improve the mechanical properties of implants. A numerical model is developed using the commercial finite elements software in ABAQUS/Explicit for simulating dynamic conditions. The aim of the study is to develop surface roughness parameters using computational methods such as studies have not yet been contemplated. The single shot angle, shot repeat, time, material orientation, and laser power are applied for the first time simultaneously to evaluate the impact of material orientation and loading angles on surface roughness parameters. The study showed that the developed computational model’s compressive residual stress was −578.45 MPa, while the experimental samples were −592.18 MPa. Consequently, the difference between the computational and experimental results was 2.32%. Without regard to material orientation or angle, the compressive residual stress of the samples under examination was found to be −578.450 MPa after three repetitions and to decreased to −1.620 MPa after four. These results demonstrate that by varying the material orientation and loading angle, the Ra value may be increased four times.
IntroductionWith the much-needed transformation from linear to more circular resource flows, it is imperative for enterprises to understand their financial prospects. Transforming towards Product-as-a-Service (PaaS) with circular elements introduces new considerations that must be addressed to ensure profitability and sustainability. However, there is a lack of comprehensive financial assessments based on empirical cases to guide companies in assessing the lucrativeness of their transformations. This paper addresses this gap by proposing a simple-to-use and flexible financial assessment model for PaaS, helping practitioners identify the conditions necessary for financial viability from both provider and user perspectives.MethodsGrounded in transdisciplinary research, this study focuses on a construction machine manufacturer's transformation from a traditional one-off sales business model to PaaS. The transformation is analysed through the development of a life cycle costing financial assessment model that addresses relevant cost drivers.ResultsUsing discounting methods, the model can help practitioners generate scenarios to identify feasible solutions for profitable PaaS setups. Additionally, the paper presents an analytical procedure to identify conditions for a financially viable PaaS. The procedure includes scenario-based analysis that accounts for systemic changes often necessary for successfully realising PaaS.DiscussionBy following the proposed procedure, along with the financial assessment model, manufacturers can streamline their financial assessments to identify necessary changes. For the analysed case company, the modelling results indicate that it must redesign its products for PaaS to outperform its previous business model.
New supercritical carbon dioxide (scCO2)-based cutting fluids combining the cooling from scCO2 and lubrication from various additives are investigated in this study. Selected components, including ionic liquids, vegetable and mineral oils, water, and PEG, were introduced into supercritical CO2, and tribological tests were performed on a CNC lathe to analyze their influence on Ti6Al4V/WC-Co contact. Forces and temperatures were recorded to compare the perfomances of the scCO2-lubricant mixtures for future use in machining. The analysis of the apparent friction coefficient and sticking zone showed a noticeable decrease by ionic liquids when combined with scCO2 at a speed of 100 m/min and 4 mL/min delivery flow rate. The other lubricants (water and PEG vegetable oils) performed similarly to standard mineral oil and are less expensive, which could help in developing future low-cost yet effective cooling and lubrication methods for the machining industry.
Bioreactors are essential for the production of biopharmaceuticals and bioproducts, requiring continuous monitoring to ensure quality assurance. Manual processes in manufacturing plants often lead to anomalies such as out-of-trend and out-of-spec incidents, necessitating extensive root cause analysis that typically takes 2–8 weeks. This paper introduces an innovative methodology that uses the golden batch profile as a benchmark to identify deviations and root causes in subsequent industrial batches. The methodology involves normalizing the data and calculating the variances of a specified batch from the golden batch profile. By examining the contribution of each critical process parameter to these variances, the study highlights their importance in root cause analysis. The application of this methodology to the IndPenSim dataset demonstrated its effectiveness by significantly reducing false positives and negatives compared to traditional PCA-based methods. Emphasis on the deviations of critical quality attributes and critical process parameters from the specified batch compared to the golden batch profile offers valuable insights into industrial process analysis. This approach not only enhances anomaly detection accuracy but also improves the efficiency and reliability of biopharmaceutical and bioproduct manufacturing processes.
This study investigates the spheroidization of titanium Ti-6Al-4V powder particles using numerical models developed in Abaqus and OpenFOAM. Spherical particles are crucial in powder-based additive manufacturing due to their superior flowability, packing density, and mechanical properties, enhancing printing precision and the quality of final products. While conventional techniques such as gas atomization and plasma spheroidization have been extensively researched, the potential of laser spheroidization remains underexplored. To address this gap, detailed numerical analyses of laser spheroidization were conducted, modeling heat transfer from the laser to powder particles using a transient uncoupled heat transfer method with latent heat considerations, while particle deformation was simulated with a phase-fraction-based interface-capturing approach integrated with Navier-Stokes equations. The results, validated against analytical models, indicate that particles within the 20–80 μm range experience optimal spheroidization within a 0.005-second residence time under laser heating, with particles smaller than 30 μm reaching evaporation temperatures of 5,000°C, while larger particles reshape without evaporating under a typical heat flux of 94 MW/m2 (1.8 kW laser power). This study demonstrates that laser spheroidization of Ti-6Al-4V powder can potentially increase powder yield by 10%, offering higher power density and shorter melting times compared to plasma spheroidization, thus presenting a more efficient alternative for achieving spherical particles of specific sizes.
