ABSTRACT. The PRODUTECH R3 Agenda — short for Recovery, Resilience, and Reindustrialization—is a strategic initiative aimed at transforming Portugal’s Production Technologies Sector (FTP). It focuses on enhancing the sector's capacity to lead the green and digital transitions, reducing external technological dependence, and increasing the value added within the country. This initiative is funded by the European Union through the Recovery and Resilience Plan (PRR) under the NextGenerationEU program and it is organized in programs and projects.

In this session, the key objectives of the Agenda will be presented:

Industry Transformation: Develop and transfer R&D into new products and services, particularly production technologies, to empower companies in innovation processes.

Qualification for Resilience and Growth: Support companies in adopting and internalizing new technologies, contributing to competitiveness and sustainability.

Capacity Building and Training: Enhance professional skills in innovative production technologies.

Internationalization: Promote the global presence of Portuguese companies.

Dissemination: Widely share the initiative's results and best practices.

ABSTRACT. Esta apresentação dá a conhecer o Observatório da Fileira das Tecnologias de Produção enquanto instrumento permanente de inteligência estratégica ao serviço das empresas e do ecossistema da FTP. Serão apresentadas a sua estrutura, principais áreas de informação e funcionalidades, bem como a forma como integra e transforma dados sobre a Fileira, mercados, tecnologias, competências e tendências em conhecimento relevante para a tomada de decisão. Mais do que uma plataforma de informação, o Observatório pretende constituir uma capacidade permanente de acompanhamento, antecipação e apoio à definição de prioridades estratégicas, contribuindo para uma Fileira mais informada, competitiva e preparada para responder aos desafios futuros.

ABSTRACT. Esta apresentação sintetiza os principais resultados dos estudos de mercado desenvolvidos para a Fileira das Tecnologias de Produção (FTP), analisando o posicionamento e as oportunidades da Fileira em diferentes geografias internacionais. A partir da caracterização dos mercados, setores utilizadores, tendências de procura, concorrência, barreiras e oportunidades de negócio, serão identificados os mercados e segmentos com maior potencial para as empresas portuguesas. Mais do que caracterizar geografias, os estudos procuram transformar informação de mercado em inteligência estratégica, apoiando a seleção de prioridades, a diversificação das exportações e a definição de estratégias de entrada e posicionamento internacional.

ABSTRACT. Esta apresentação sintetiza o Roadmap de Inovação da Fileira das Tecnologias de Produção, identificando as principais tendências, desafios e oportunidades que deverão moldar a evolução da FTP na próxima década. A partir da análise das transformações tecnológicas, industriais e de mercado, serão apresentadas as áreas prioritárias de inovação e as trajetórias consideradas mais relevantes para reforçar a capacidade de antecipação e adaptação da Fileira. O Roadmap pretende constituir um instrumento de orientação estratégica para empresas e restantes atores do ecossistema, apoiando decisões de inovação, investimento e desenvolvimento de capacidades e contribuindo para uma FTP mais tecnológica, sustentável, resiliente e competitiva

ABSTRACT. Esta apresentação analisa o papel atual e futuro da Inteligência Artificial na Fileira das Tecnologias de Produção, identificando tendências tecnológicas, níveis e possibilidades de adoção, principais áreas de aplicação e desafios associados à sua integração nas empresas. Serão abordadas as oportunidades proporcionadas pela IA para a inovação, produtividade e transformação dos processos industriais, bem como as necessidades de competências, adaptação organizacional, segurança e governação que condicionam a sua adoção. O estudo procura, assim, apoiar as empresas da FTP na compreensão do potencial da IA e na definição de trajetórias de adoção que permitam transformar esta tecnologia numa efetiva fonte de competitividade e criação de valor.

ABSTRACT. Esta apresentação caracteriza a realidade das empresas familiares na Fileira das Tecnologias de Produção, analisando o seu desempenho, governação, sucessão, profissionalização e principais desafios de transformação. Para além do diagnóstico e da comparação entre empresas familiares e não familiares, serão apresentados o Modelo de Maturidade Estratégica, a ferramenta de autodiagnóstico e a Agenda de Transformação desenvolvidos no âmbito do estudo. A apresentação procurará evidenciar como estes instrumentos podem apoiar as empresas familiares na preparação da sucessão, reforço da governação e desenvolvimento da sua competitividade e sustentabilidade a longo prazo.

ABSTRACT. Esta apresentação analisa os custos de contexto associados à formação profissional com impacto nas empresas da FTP, identificando os principais constrangimentos legais, modelos de governação pública, administrativos, organizacionais e operacionais que condicionam o investimento na qualificação dos trabalhadores. Para além da caracterização destes custos e do seu impacto nas empresas, serão apresentadas as principais oportunidades de simplificação e melhoria dos mecanismos de formação. O estudo procura contribuir para um modelo mais eficiente e ajustado à realidade empresarial, capaz de reduzir encargos de contexto e, simultaneamente, reforçar a qualificação, a produtividade e a competitividade das empresas.

ABSTRACT. Esta apresentação, baseada num estudo detalhado realizado para a FTP, propõe uma arquitetura de instrumentos financeiros orientada para responder às principais necessidades de financiamento da Fileira das Tecnologias de Produção. A partir do diagnóstico das lacunas existentes, são apresentados quatro fundos complementares dirigidos à inovação e I&D colaborativo, modernização tecnológica, aquisição de tecnologias desenvolvidas pela FTP e internacionalização. Serão sintetizadas as propostas dos respetivos modelos de financiamento e governação, a complementaridade com os instrumentos públicos existentes e o roadmap proposto para a sua implementação, destacando o potencial destes mecanismos para apoiar o investimento, a inovação, a competitividade e o crescimento da Fileira.

ABSTRACT. Objective of the article Multi-material Additive Manufacturing (AM) enables the integration of materials with distinctly different mechanical properties within a single component [1]. In medical applications, such as orthoses, prosthetics, wearables, and implants, there is a critical trade-off between structural integrity and patient comfort. For instance, orthopedic braces play a crucial role in the rehabilitation and correction of musculoskeletal issues by providing essential external support, alignment, and stabilization to the affected areas [2]. However, despite their clinical importance, traditional orthopedic braces manufactured via manual thermoforming still face significant limitations regarding comfort, breathability, personalization, and manufacturing. Issues such as mechanical stiffness mismatch, where hard plastics press against soft, compliant tissue, and the accumulation of metabolic heat under impermeable shells frequently cause skin irritation and discomfort [3]. Ultimately, these ergonomic drawbacks can severely reduce treatment effectiveness and patient compliance over long-term use [3]. To ensure its effective operation, a brace must utilize rigid materials to ensure load-bearing capacity and durability, while simultaneously incorporating soft materials to enhance ergonomics, adaptability and pressure distribution [4]. Therefore, the design and development of orthopedic braces needs to consider user’s anatomical needs aligned with materials interfaces properties [5]. Advances in multi-material AM offer new opportunities to overcome these limitations by allowing for the design and production of customized orthopedic devices that better match patient anatomy and functional requirements [6]. To achieve this, stiff materials are applied to ensure load-bearing capacity and durability, acting as a structural core. Conversely, soft materials are integrated as flexible, skin-contact outer layers to enhance ergonomics, adaptability, and pressure distribution [7]. Unfortunately, current design approaches for these devices often rely heavily on empirical material selection rather than the systematic optimization of material distribution and interface behavior. Specifically, this research focuses on the development of an orthopedic vest brace, developed in collaboration with the orthopedic company MultiOrthos. By utilizing a structured Design of Experiments (DOE) approach and mechanical tensile testing, this work demonstrates how integrating a stiff core with a flexible outer layer can significantly improve both the structural reliability and the personalized comfort of the patient. Objective: The study proposes to investigate design strategies, interfacial behaviour and process parameters in multi-material AM to optimize the compromise between mechanical performance and comfort in medical applications, supported by quantitative data and the fabrication of a functional prototype.

Literature Review Traditional orthopedic brace production relies on manual vacuum thermoforming over milled polyurethane molds, a labor-intensive process that generates substantial material waste and introduces human variability into the anatomical fit. In contrast, Fused Deposition Modeling (FDM) offers a fully digital workflow utilizing 3D patient scanning and computer-aided design (CAD) to fabricate patient-specific devices. This digital transition eliminates the need for physical molds, reduces production time, and enables precise topological customization. Effective orthoses require stiff materials for structural load transmission and flexible materials for skin-contact compliance. Multi-material additive manufacturing (MMAM) addresses this by allowing the simultaneous deposition of distinct polymers within a single continuous build, replacing failure-prone manual assembly boundaries with integrated structural joints. Within medical applications, synthetic biocompatible polymers are heavily utilized for this purpose. Stiff thermoplastics provide the flexural modulus and impact strength required for spinal stabilization, while flexible elastomers offer high elasticity and appropriate shore hardness to manage pressure distribution and patient comfort. Applying Design for Additive Manufacturing (DfAM) strategies shifts brace design from uniform, single-material shells to functionally graded architectures. This approach mitigates the mechanical impedance mismatch by utilizing a stiff core to manage corrective forces and a compliant interface to reduce localized peak pressures and shear stresses. Furthermore, DfAM enables the integration of specific internal infill patterns, such as zigzag or gyroid geometries, and porous topologies. These structures significantly improve convective heat transfer and microclimate management compared to traditional impermeable plastic shells. Despite these geometric advantages, FDM components exhibit inherent mechanical anisotropy, meaning their ultimate strength and stiffness vary based on raster angle, layer adhesion, and build orientation. The structural integrity between dissimilar polymers, such as rigid structural plastics and soft elastomers, relies heavily on macro-mechanical interlocking and thermal diffusion at the boundary layer. Consequently, mechanical characterization using standardized tensile testing, such as the ISO 527-2 standard, is essential to validate interfacial bond strength and guarantee that the hybrid structure can withstand cyclic physiological loads without delamination.

Methodology The experimental methodology is grounded in the careful selection of complementary material pairs using the Ashby method. This involved combining a stiff thermoplastic, specifically polyethylene terephthalate glycol (PETG) with a flexible elastomer, thermoplastic polyurethane (TPU 85A). Because of this stark contrast in stiffness, achieving integration can be a complex task since it is necessary to consider different properties of the materials (e.g., microstructure, polarity, melting temperature) to guarantee interlayer adhesion. Once the materials are selected, the study utilized Fused Filament Fabrication (FFF) with a multi-toolhead 3D printer (Original Prusa XL) to prevent material cross-contamination. This was followed by the configuration of the specific printing parameters through a Fractional Factorial Design to optimize variables like interlocking depth, interlocking width, extrusion temperatures, and layer thickness, assuring optimal interfacial bonding. To ensure structural integrity and quantify the bonding strength between the rigid and soft materials, the resulting multi-material structures underwent mechanical testing. Specifically, standardized tensile tests (ISO 527-2 Type 1A) were conducted on both single-material and multi-material specimens to determine the Ultimate Tensile Strength (UTS) and elongation at break across the interface. Combined with this testing program, targeted comfort-related assessments, were conducted on the flexible layer. This involved evaluating various elastomeric infill patterns (e.g., cubic, zigzag, honeycomb, gyroid) and densities through manual compression and tactile feedback. These dual assessments are crucial to measure the device's practical efficacy, demonstrating exactly how the structure safely supports operational loads while simultaneously adapting to the patient's anatomy to prevent localized pressure points.

Results The mechanical characterization initially established the baseline properties of the isolated polymers, with PETG exhibiting a mean ultimate tensile strength of 37.70 MPa and a Young's Modulus of 1825.46 MPa, while TPU demonstrated extreme ductility by reaching the testing machine's physical limits at nearly 500% elongation. When evaluating the integrated multi-material assemblies, all tested specimens exhibited pure interfacial failure. This confirmed that because PETG and TPU lack natural chemical affinity, the structural integrity of the joint relies entirely on physical interlocking and localized thermal adhesion. Statistical analysis of the Fractional Factorial Design revealed that the physical geometry of the joint, specifically the interaction between interlocking depth and width, was the primary factor dictating interfacial bond strength. Utilizing a predictive Response Optimizer model, the ideal processing parameters were identified as an interlocking depth of 1 mm, an interlocking width of 1.2 mm, a TPU extrusion temperature of 220°C, a heated bed temperature of 40°C, and a layer thickness of 0.20 mm. This optimized configuration maximized the structural reliability of the boundary layer, achieving an ultimate tensile strength of 6.02 MPa and an elongation at break of 29.92%. Beyond interfacial strength, the study successfully optimized the ergonomic comfort and weight of the medical device by defining a hybrid structural architecture comprising a 4 mm PETG structural core and a 6 mm TPU flexible layer. Experimental tactile and compression assessments of various elastomeric geometries demonstrated that a 15% zigzag infill provided the optimal balance of cushioning compliance and stable manufacturability. When applied to the physical prototype, this specific infill strategy yielded an absolute weight reduction of approximately 35% compared to a solid elastomeric layer, significantly improving wearability without compromising the required mechanical boundaries. Finally, dimensional assessments of the fabricated prototype confirmed a high degree of process repeatability, with wall thicknesses deviating by a mean of only 0.15 mm from the digital model. Theoretical implications This research offers substantial theoretical contributions by providing a strong foundation on how multi-material printing can effectively balance mechanical resistance with patient comfort. By demonstrating that the structural integrity between polymers lacking natural chemical affinity relies fundamentally on physical interlocking and localized thermal adhesion, the study contributes to the definition of process parameters that ensure optimal interface performance. Specifically, the application of a Fractional Factorial Design expands current understanding of the thermo-mechanical coupling and synergistic interfacial bonding phenomena that occur during complex multi-material extrusion processes. This work also drives the advancement of functionally graded material (FGM) concepts, moving away from failure-prone discrete boundaries toward interwoven hybrid structural architectures, specifically as they are applied to the design of patient-specific medical devices. Additionally, the orthopedic brace design will follow a design centered on the patient, highlighting the advantages of additive manufacturing in the development of orthoses.

Practical implications From a practical perspective, the application of these methodologies and processes leads to improved patient comfort, achieved through the implementation of controlled compliance zones within the manufactured orthosis. By utilizing a 6 mm flexible TPU layer with a 15% zigzag infill, this approach enhanced ergonomics and facilitated a significant 35% reduction in the overall weight of the prototype section while maintaining the necessary structural integrity of the 4 mm PETG core. Consequently, the successful execution of this work will play a critical role in highlighting the vast capability of additive manufacturing technologies for real-world medical applications. By replacing the highly variable and labor-intensive manual thermoforming process with a fully digital workflow, the industry can eliminate physical molds, reduce material waste, and bypass the limitations of chemical adhesives. Instead of relying on trial-and-error, this work provides a solid foundation to understand how design strategies, interfacial behavior, and automated process parameters actually work together, ensuring the reliable development of an orthopedic brace adapted to the patient's anatomical needs. This makes it much easier to optimize that critical balance between mechanical performance and patient comfort in future clinical applications. Transitioning from conventional manual thermoforming to multi-material additive manufacturing offers substantial sustainability benefits and environmental improvements. Traditional brace manufacturing is highly material-intensive, relying on the subtractive milling of disposable polyurethane foam molds and generating significant thermoplastic off-cut waste during the vacuum draping process. In contrast, the proposed digital FFF workflow entirely eliminates the need for physical molds and achieves high raw material efficiency by depositing polymers only where structurally required. Specifically, by utilizing a 15% zigzag infill for the elastomeric layer rather than a solid structure, the material consumption for the prototype section was reduced by approximately 35% (from 434.71 g to 282.22 g). This direct reduction in raw material usage correlates with a lower manufacturing environmental footprint and decreased energy consumption per unit. Furthermore, by replacing failure-prone chemically glued foam liners with an integrated, interlocking multi-material joint, the life-cycle durability of the orthosis is significantly improved, which reduces the frequency of device replacements and ultimately minimizes long-term medical waste.

Limitations Despite its promising scope, this study faces some limitations. The findings are currently restricted to the selected material combinations, primarily PETG and TPU, evaluated during the research. Additionally, while static tensile strength was successfully validated, the interface durability of these printed structures under long-term cyclic loading conditions requires further validation before widespread clinical use. The current scope is also constrained by being limited to mechanical tests without considering real-world assessments with patients, as the physical prototype was manufactured as a continuous closed structure without the final anatomical cuts required for patient entry and mobility. From a manufacturing perspective, the intentional omission of sacrificial supports led to localized geometric defects, specifically elastomeric filament sagging on steep internal overhangs. Furthermore, the slicing software's automated algorithm exhibited limitations when handling asymmetrical wall thicknesses (e.g., the 4 mm to 6 mm ratio), unexpectedly reducing the frequency of the mechanical interlocking beams. Finally, the scalability of these advanced printing techniques remains highly dependent on the AM processes utilized and the capabilities of available multi-toolhead printing equipment to manage inherent constraints like differential polymer shrinkage, moisture absorption, and extrusion buckling.

Innovation The core innovation of this research lies in introducing a multi-material additive manufacturing process that is still rarely applied in the development of custom orthotics. It establishes a comprehensive design process that successfully balances high mechanical strength with personalized ergonomic comfort by integrating standard mechanical testing protocols directly with specialized, comfort-oriented performance metrics. Furthermore, the study identifies and utilizes optimized process parameters, specifically the computationally generated mechanical interlocking mechanism and calibrated thermal profiles, that allow for reliable printing between different stiff and soft materials. Lastly, it significantly advances the practical application of hybrid structural architectures, utilizing advanced elastomeric infill patterns to offer a highly sophisticated, durable alternative to traditional, failure-prone discrete material transitions, such as glued foam liners.

References [1] D. Han and H. Lee, “Recent advances in multi-material additive manufacturing: methods and applications.” Accessed: Jan. 09, 2026. [Online]. Available: https://doi.org/10.1016/j.coche.2020.03.004 [2] “ISO 8549-3 Prosthetics and orthotics-Vocabulary-Part 3: Terms relating to orthoses Prothèses et orthèses-Vocabulaire-Partie 3: Termes relatifs aux orthèses ISO 8549-3:2020(E) ii COPYRIGHT PROTECTED DOCUMENT,” 2020. [3] D. F. Redaelli et al., “3D printing orthopedic scoliosis braces: a test comparing FDM with thermoforming,” The International Journal of Advanced Manufacturing Technology, vol. 111, pp. 1707–1720, 2020, doi: 10.1007/s00170-020-06181-1/Published. [4] J. B. . Webster and Douglas. Murphy, Atlas of orthoses and assistive devices 5th edition. Elsevier, 2019. [5] K. J. Walker et al., “Novel 3D-printed foot orthoses with variable hardness: A comfort comparison to traditional orthoses,” Med. Eng. Phys., vol. 115, May 2023, doi: 10.1016/j.medengphy.2023.103978. [6] A. Bełżek et al., “Advances in 3D printed orthotics for rehabilitation,” 2025. doi: 10.36740/WLek/202602. [7] Seymur Hasanov et al., “Review on Additive Manufacturing of Multi-Material Parts: Progress and Challenges.” Accessed: Jan. 09, 2026. [Online]. Available: https://doi.org/10.3390/jmmp6010004

ABSTRACT. Additive Manufacturing (AM) has undergone significant changes in recent years, evolving from simple rapid prototyping to a part of advanced product development and production. Modern engineering now employs techniques for processing flexible materials such as Thermoplastic Polyurethane (TPU), an elastomer used in sectors such as automotive, aerospace, and consumer goods. TPU three-dimensional (3D) printing has seen success on traditional small printers. Still, with the rise of pellet-fed AM for large-area printing, optimising elastomers for this application has become increasingly important. All operational variables must be considered and aligned with the polymer's rheological properties, including extrusion temperature, printing speeds, heated beds, and overlap, all of which significantly influence print quality. As large-format granular fabrication (L-FGF) is an emerging topic in the industry, there is a need to fully incorporate TPU and to understand the best methods for its effective printing. This project aims to design and validate a heated printing bed, conduct rheological tests to determine the material’s exact working limits, and evaluate how printing parameters impact the final product. The study concludes that, beyond printing settings, detailed monitoring and control of material and environmental conditions are essential. Additionally, rheological characterisation is crucial to establish the baseline of optimal printing parameters.

ABSTRACT. The growing deployment of renewable energy sources has accentuated the need for efficient energy storage solutions to address the inherent intermittency of these resources. Among the various storage technologies, Cryogenic Energy Storage (CES) systems have emerged as a promising option due to their high scalability, operational efficiency, and potential for integration with complementary systems. This paper focuses on cryogenic approaches, with particular emphasis on Liquid Air Energy Storage (LAES). Key aspects discussed include the characterization of CES systems, operating principles, with special attention to energy efficiency and thermodynamic representation in temperature-entropy diagrams, as well as design considerations, technical challenges, and system construction. LAES technologies exhibit energy efficiencies ranging from 45% to 70%, which can potentially reach up to 75% when integrated with auxiliary systems, with capital costs estimated between €900/kW and €1,750/kW. Carbon dioxide (CO₂)-based systems, although demonstrating higher energy efficiencies (40% to 60%), face significant challenges due to elevated infrastructure costs. Moreover, hybrid configurations that combine advanced thermal cycles and waste heat utilization can achieve efficiencies in the range of 55% to 80%, highlighting their flexibility in complex energy scenarios. Compared to alternative technologies such as batteries and Compressed Air Energy Storage (CAES), CES systems—despite certain economic and technological constraints—represent a promising avenue for the global energy transition, particularly as ongoing developments are expected to enhance both their performance and economic feasibility.

ABSTRACT. Industries like plastic bag manufacturing involve complex processes that can be difficult to monitor and optimize. The growing adoption of Internet-of-Things (IoT) sensors under an Industry 4.0 approach generates large volumes of operational data, but transforming this data into actionable insights for process optimization remains a significant challenge. This study examined a plastic bag manufacturer whose machines are equipped with sensors, testing whether the prerequisites of the SARIMAX model hold for the Critical Process Parameter with the greatest impact on product quality. The application of SARIMAX was constrained by the irregular recording patterns of several process variables and by the absence of a meaningful seasonal component in the data, leading to the adoption of a simpler ARIMA model instead. A correlation and causality analysis was also conducted to identify exogenous variable candidates, but their inclusion did not improve forecasting performance beyond the ARIMA baseline. This provides the company with a validated tool and methodology to anticipate process deviations and support more proactive and informed data-driven decision-making.

ABSTRACT. Sealing integrity is a critical factor affecting food safety and shelf life in high-throughput packaging operations. Conventional quality control approaches rely primarily on downstream visual inspection, which detects defects only after they become visible, limiting the ability to prevent production losses. This study proposes a predictive quality control approach for sealing integrity based on process temperature data acquired directly from industrial packaging equipment.

An industrial case study was conducted in a high-speed ham packaging line, where controlled thermal degradation experiments were performed to isolate temperature-induced sealing failures. Temperature signals from sealing machine heating zones were synchronized with operator-labelled defect occurrence, enabling the formulation of a supervised binary classification problem. A label-shifting strategy was applied to anticipate defects within a five-minute prediction horizon.

Among several evaluated models, a Stochastic Gradient Descent classifier demonstrated strong real-time applicability, achieving 93% accuracy and 96% recall for defect prediction. Results confirm the existence of a measurable temporal gap between thermal degradation and visible sealing defects, enabling proactive intervention before quality loss occurs.

The study demonstrates that integrating upstream sensor-based prediction with downstream computer vision inspection enables a proactive, data-driven quality assurance framework for industrial food packaging systems.

ABSTRACT. The manufacturing industry is setting forth new standards and requirements for quality inspection in order to respond effectively to the market's demands, with surface defect detection to reduce operator fatigue and improving inspection consistency. However, the inspection of ceramic tableware remains technically challenging due to reflective, curved, and non-flat surfaces that compromise conventional imaging and classification methods. This work proposes an automatic inspection system that combines a customised real-time image acquisition setup with a YOLOv8-based deep learning classifier fine-tuned for ceramic surface. The system categorises plates into five industrial quality classes (C0–C3 and Scrap) in accordance with established inspection protocols. Experimental evaluation over five independent runs, on 3200 images, demonstrates strong performance, with AUROC = 0.992 ± 0.008, AUPRC = 0.998 ± 0.002, and macro-F1 = 0.946 ± 0.051. The class-wise accuracies were C0 = 0.933 ± 0.091, C1 = 0.967 ± 0.075, C2 = 0.867 ± 0.139, C3 = 0.967 ± 0.075 and Scrap = 1.000 ± 0.000, demonstrating a balance between minimising false positives and false negatives. Deployment profiling on an NVIDIA A100 GPU confirms real-time feasibility with median inference latency of 4.59 ms (~218 FPS, p95 = 4.73 ms), with a minimum/maximum latency range of 3.56–7.14 ms. These results indicate that the proposed system achieves high detection accuracy on reflective ceramic surfaces while satisfying stringent throughput requirements for in-line industrial inspection, thereby enhancing production efficiency and supporting dynamic, data-driven manufacturing environments.

ABSTRACT. Retrieval-Augmented Generation (RAG) has become a standard approach for grounding Large Language Models (LLMs) in external knowledge sources. However, most existing RAG pipelines rely on single-stage retrieval and independently processed document chunks, which limits their effectiveness for complex document-level question answering where evidence is distributed across multiple sources. This paper proposes a multi-stage RAG architecture designed to improve evidence coverage, relevance, and grounding in document-centric and multi-hop question answering. The architecture integrates ensemble retrieval, LLM-based relevance grading, and adaptive query decomposition to dynamically refine retrieval when initial evidence is insufficient. This design enables more reliable synthesis of information across documents without requiring retriever fine-tuning or ranking supervision. We evaluate the proposed approach on open-domain benchmarks (HotpotQA, 2WikiMultiHopQA, TriviaQA), document-focused datasets (Qasper), and a realworld battery sustainability document corpus. Across all settings, the proposed system consistently improves Context Precision, Context Recall, Faithfulness, and Answer Relevancy compared to a strong Basic RAG baseline, while maintaining competitive task-level accuracy. These results demonstrate that inference-time retrieval orchestration can substantially enhance document-level question answering, particularly in scenarios requiring multi-document reasoning and reliable evidence grounding. The proposed architecture provides a practical foundation for deploying RAG systems in research, enterprise, and scientific analysis workflows.

ABSTRACT. Life Cycle Assessment practitioners frequently struggle to map ambiguous process descriptions to the strict, non-standardized nomenclature of Life Cycle Inventory databases. Existing retrieval solutions often fail to balance the need for high recall (handling linguistic variability) with the strict requirement for traceability (avoiding hallucinations). This study proposes a multi-stage hybrid retrieval architecture that progressively resolves user queries through a cascade of deterministic SQL matching, historical memory, fuzzy string alignment, and offline semantic enrichment. For queries that resist database-driven resolution, the system deploys a constrained online LLM fallback mechanism to bridge the semantic gap. Experimental evaluation demonstrates that this architecture achieves a 100% retrieval coverage and a grounded precision of 75.2%, significantly outperforming monolithic neural retrieval baselines. A dedicated sensitivity analysis challenges the efficacy of parameter scaling for constrained retrieval. Lightweight language models (e.g., Mistral 7B) are shown to offer the optimal trade-off, achieving higher grounding success rates while keeping average system latency within interactive limits (~0.27s) compared to large-scale reasoning models, which exhibit a tendency towards ungrounded over-generation. These findings suggest that embedding generative intelligence within structured, deterministic frameworks offers a scalable, auditable path toward next-generation sustainability data tools.

ABSTRACT. Traditional manufacturing sectors face a workforce transition crisis as experienced operators retire and interest in manual labour among younger generations declines. This research aims to provide insights into preserving tacit knowledge through intelligent automation, using the cork stopper industry as a case study. Two systems are proposed: a vision-guided punching machine to optimise the punching of cork strips, and an automatic grading system to assess the quality of cork stoppers. Both utilise deep learning models trained on customised datasets from controlled conditions. The punching machine analyses cork strips with a sliding-window method to detect defects and identify optimal punch regions. The grading system employs high-resolution cameras to classify cork stoppers into six quality grades. Results validated in industrial settings revealed accuracy challenges due to cork's natural variability and class overlap, but indicated that models learned relevant features similarly to human reasoning. Future work will focus on generating larger datasets and implementing multiple training cycles in production to address the complexity of cork analysis.

ABSTRACT. Coffee capsules remain a difficult waste stream for circular economy strategies: they are small, multi-material and non-standardised. This study examines the adoption and use of a Reverse Vending Machine (RVM) prototype designed specifically for coffee capsules, deployed for twelve weeks in the workplace of the company that developed it (~300 employees). Using behavioural microdata on individual deposits rather than self-reported intentions we explore whether an incentive affects the decision of adoption and / or the intensity of participation (the extensive and intensive margins) differently; to what extent initial adoption translates into sustained use; and how individual deposited volumes vary across users. The data show that coffee capsules pattern of adoption and use to be compatible with an S-shaped trajectory. The incentive, one euro per kilogram donation to a social institution, acted almost exclusively on the extensive margin: the number of active users increased from 15 to 33 and 70% of the total volume (157.6 out of 224.3 kg) was collected following the introduction of the incentive, while the average quantity deposited per user did not increase. Retention was low, with over 40% of users participating for only one week, while a disproportionate share of the total deposited volume was accounted for by a small group of heavy users. These findings show that incentives can effectively broaden participation in circular collection technologies for used coffee capsules, but that increasing adoption does not necessarily translate into more intensive or sustained use.

ABSTRACT. Biomass is considered a renewable and sustainable resource for producing energy and bio-based materials, contributing to a more diversified energy mix, reducing reliance on finite resources, lowering greenhouse gas emissions, and addressing issues related to material biodegradability (European Commission, 2014). However, the growing demand for biomass—driven by technological advancements and renewable energy mandates—intensifies competition for land and may pose risks to food security. As a result, the use of marginal lands for cultivating industrial crops has been proposed as a strategy to reduce competition with food production and to avoid conflicts associated with land use change (Lewandowski, 2015). In this study, the environmental impacts of cultivating and utilizing Miscanthus for bioenergy production and particleboard manufacturing are assessed and compared with those of conventional reference systems. Production of Miscanthus in marginal land represents an approach to restore and attenuate the marginality of the soils and the higher above and belowground biomass provides lodging, favoring biological and landscape diversity. The impacts associated with the conversion phase, to bioenergy production and to particleboard production, in terms of biodiversity, landscape, and soil quality are lower than those from the conventional system.

ABSTRACT. This exploratory case study examines how sustainability requirements can act as a driver of functional integration and innovation in food-service packaging design. Centred on a two-beverage takeaway packaging system, the study investigates a manufacturable secondary-packaging architecture in which mechanical stability is achieved through geometric interference rather than additional plastic components, and explores how the same design strategy can convert existing printable surfaces into a brand-communication interface. Using an exploratory single-case approach that combines iterative design thinking with Design for Manufacturing (DfM) principles, structured into problem framing, concept generation, structural development and communication-interface development, the study produced full-scale paperboard prototypes evaluated qualitatively against structural, manufacturability and foldability criteria. The results show that a single multifunctional paper accessory can replace a conventional plastic divider, consolidate two disposable wrappers into one printed component, and free a defined communication panel without additional parts or production steps. These findings suggest that resource efficiency, manufacturability and brand value need not be treated as competing objectives, but can emerge from the same sustainability-driven design decision, offering a transferable method for early-stage ecodesign of food-service packaging.

ABSTRACT. As organizations transition to Industry 5.0, the assessment of their current alignment with organizational, technological, and collaborative dimensions become relevant to identify gaps and opportunities for adopting Industry 5.0’s human-centric, sustainable, and resilient principles. Although numerous studies have explored the disruptive technologies reshaping the industry, such as Artificial Intelligence and Collaborative Robots, and examined their integration and impact, there remains a critical need to understand how an industry’s maturity level influences this transition. This paper presents a systematic literature review focused on the three core pillars of Industry 5.0, highlighting key technologies being adopted across various industries and examining Industry 5.0 frameworks in the context of industrial maturity. This study is complemented by an experimental case study with a leading automotive company, assessing how the company’s maturity aligns with the key success factors of Industry 5.0. The results emphasise the importance of the highlighted factors related to the Industry 5.0 pillars, laying a crucial groundwork for upcoming research and innovation in this field.

ABSTRACT. Reconstructing the human pose of the upper limb from the position of the hand is a core operation in motion capture, since instrumenting every joint of the arm is not practicable. This paper presents an analytical two-link inverse kinematics solver, in closed form, with redundancy resolution by pole vector and explicit handling of singular configurations, integrated into a Virtual Reality system. Correctness is established through geometric validation over one thousand random configurations, yielding errors that are negligible within numerical precision and correct handling of all 681 out-of-reach cases. The workspace is characterised analytically and verified by Monte Carlo sampling, with agreement below 0.1%. A sensitivity analysis quantifies the propagation of measurement error to the reconstructed pose: the mean amplification factor is close to unity, but configurations are identified, in the interior of the workspace, where redundancy resolution becomes ill-conditioned and the error amplifies by more than twentyfold. The method is applied to a case study of human interaction with a vehicle console, evaluated with ten participants and five hundred trials (98.4% accuracy), in which calibration drift of the flexion sensor is identified and quantified as a robustness limitation.

ABSTRACT. The transition towards Industry 5.0 requires manufacturing process development to consider not only productivity and efficiency, but also worker safety, well-being and acceptance of new technologies. This paper proposes the use of human-centered simulation as an intermediate development step to predict the ergonomic impact and acceptance potential of a new automated picking system before its physical implementation. The case study concerns a manufacturing process in which manual picking from shelves and floor-level boxes is replaced by an integrated system composed of a vertical warehouse, a robotic arm and a human-machine interface. A baseline assessment of the existing process was performed through operational measurements, ergonomic analysis and technology acceptance evaluation. The current process revealed critical postures and high ergonomic risk, while the acceptance assessment showed low satisfaction with the current task and identified physical ergonomics as a key factor influencing worker satisfaction. A simulation model of the proposed process was then developed in Visual Components®, representing the main loading and picking operations, operator movements and workstation interactions. The simulated layout incorporated ergonomic recommendations for the trolley, pallet carrier and picking table. Results show a significant reduction of the ergonomic risk score. Operationally, an example picking cycle for 12 kits was reduced from 6min 51s to 4min, while a combined loading and picking scenario reduced total time from 23min 9s to 14min 40s, corresponding to an estimated productivity increase of 58% in this scenario. These findings demonstrate that simulation can support human-centered process development by anticipating ergonomic risks, identifying improvement opportunities and increasing the likelihood of technology acceptance before implementation.