This paper proposes an integrative and flexible categorisation framework for early-stage material exploration in design-oriented contexts. The framework responds to the growing complexity of contemporary material landscapes, where conventional taxonomies and material databases often fail to adequately represent emerging, non-conventional, and sustainability-oriented materials. To develop the proposal, the authors conducted a critical review of 35 international material libraries, both physical and digital, and identified significant heterogeneity in categorisation criteria, as well as major gaps in the treatment of life-cycle-related information, aesthetic and sensorial characteristics, and less conventional material classes. The framework was then iteratively drafted, tested, refined, and consolidated through collaborative discussion among the authors and a pilot application involving 110 material case studies, with particular attention to recent and innovative materials. The resulting structure is designed to guide the organised and consistent comparison of material case studies in the early stages of the design process and, eventually, support the organisation of repositories of knowledge and material references. To do so, the framework integrates general information, material classification, technical data, aesthetic-sensorial characteristics, composition, life-cycle-related aspects, certifications, and links to the Sustainable Development Goals. Unlike late-stage selection tools focused mainly on engineering performance, the proposed framework is intended to support holistic exploration when uncertainty is high and information is often incomplete. Its modular and adaptable structure makes it suitable for use in design education, academic research, professional practice, and corporate R D. By promoting clearer terminology, greater comparability, and more comprehensive material descriptions, the framework aims to become a practical tool for navigating the rapidly evolving field of materials and for supporting more organised case studies exploration and categorisation.
Generative Artificial Intelligence (AI) and Large Language Models are increasingly considered for integration into engineering design workflows, yet their actual capabilities on the analytical and optimization tasks that arise in engineering design remain poorly understood. In this study, we examine Large Language Model-generated answers to 86 text-based questions covering three task categories chosen for their direct correspondence to typical engineering design activities: 50 problem-solving tasks from the Kangaroo Math Competition, 20 structural engineering design tasks involving stress analysis and cross-section selection, and 16 engineering optimization tasks involving conflicting constraints and trade-offs. Across the evaluated models, we observe that performance is consistently higher on the general Kangaroo problem-solving questions than on the engineering geometry and optimization tasks, even though these draw on a comparable set of underlying skills. The study further exposes systematic failure modes across the examined task types, including geometric reasoning, multi-step numerical computation, and optimization under conflicting constraints. These findings characterize the specific capabilities that AI-enabled design systems must develop to be reliable in practice, and establish an empirical baseline for evaluating future augmentation strategies such as tool integration and structured design workflows.
This study introduces a dual functional connectivity (FC) framework utilizing the Weighted Phase Lag Index (wPLI) and Mutual Information (MI) to analyze electroencephalography (EEG) data from 28 graduate engineering students. Participants completed a modified Torrance Test of Creative Thinking (TTCT-F) designed to capture four cognitive states: idea generation (IDG), idea evolution (IDE), idea rating (IDR), and rest (RST). The primary objective of this study is to develop a computational framework that leverages brain network dynamics derived from functional connectivity measures, combined with effective feature-selection and classification methods, to enable reliable and efficient recognition of design-creativity cognitive states. From the resulting connectivity matrices, we extracted graph-based features—Strength (S) and Betweenness Centrality (B)—and applied a robust feature-selection pipeline to identify discriminative connectivity markers. Statistical analyses validated the selected features, confirming their statistical significance and discriminative relevance while supporting the effectiveness of the proposed feature-selection approach. The results indicate that the selected features are concentrated mainly over frontal and parietal scalp regions, with wPLI and MI providing complementary but descriptive connectivity information. Specifically, wPLI identified 14 features with balanced hemispheric contributions, whereas MI yielded five statistically significant features. To validate the framework, we evaluated six well-established classification models using hold-out, 10-fold, and one-subject-out cross-validation. A voting ensemble achieved high discriminative performance, reaching 97.35
With the knowledge economy and artificial intelligence advancing, engineering design in Industry 5.0 is shifting toward knowledge-driven intelligent product design (KDIPD). Although knowledge-based engineering and AI-assisted design have been reviewed, KDIPD still lacks a systematic account of its development, hotspots, and thematic evolution. To bridge this gap, this study retrieved 732 publications (2005–2025) from Scopus and Web of Science. Combining Bibliometrix, CiteSpace, and BERTopic, this study systematically examines publication dynamics and document types, geographic and institutional landscape, sources and authorship structure, research hotspots and emerging trends, along with topic modeling and thematic evolution of KDIPD. Publication trends show expansion in KDIPD since 2005, surging around 2025. China emerges as a dominant contributor, particularly Zhejiang University. Publications span interdisciplinary sources, mainly mechanical engineering and computer science. Collaboration involves a few highly productive scholars and broader small-scale participation. Hotspots have moved from collaborative design and product development to artificial intelligence, especially knowledge graphs. Further analysis identified 14 keyword clusters and their emerging trends. BERTopic extracted 16 topics synthesizing into four research areas, including human-centered intelligent decision-making, knowledge modeling computational methods, domain-oriented knowledge applications, and product process systems engineering. Topic evolution further developed three continuous streams. Knowledge graph-related research became increasingly prominent, concurrent with developments of product–process systems and human-centered intelligent decision-making. The interpretive synthesis of these findings suggests that KDIPD is extending its foundation in explicit knowledge formalization toward the externalization and operational use of tacit, contextual, and experience-based knowledge, with implications for designers, enterprises, and industry.
Design thinking is widely acknowledged as an innovative approach to addressing complex issues, yet the trait-related correlates associated with its disposition remain insufficiently analyzed from a quantitative perspective. This study aims to identify core trait factors influencing young adults’ design thinking disposition and characterize their nonlinear predictive attribution patterns. Integrating exploratory/confirmatory factor analysis with the Random Forest-Shapley Additive exPlanations framework, data were collected via multidimensional scales to construct and validate a trait correlation model. The results indicate that model-predicted design thinking disposition was associated with cognitive, empathy-related, and personality traits, each exhibiting distinct operating patterns and non-linear interactions. Cognitive flexibility showed the largest model-based attribution to design thinking disposition. Leveraging explainable artificial intelligence helps address some limitations of conventional linear modeling approaches by characterizing complex nonlinear model-based attribution patterns among traits. This study develops a multidimensional trait-correlation framework, providing a novel quantitative interpretation of the predictive trait architecture underlying design thinking disposition.
Artificial Intelligence Generated Content (AIGC) is transforming early-stage industrial product ideation, but the joint effects of human and technological factors on creative outcomes remain underexplored. Using a motorcycle concept-design task, this study examined how design experience, AIGC tool proficiency, image-generation alignment strategy, and prompt type shape creative quality, novelty, diversity, and perceived usefulness, with a 3 × 2 × 2 × 2 mixed experimental design involving 60 participants. Midjourney V6 and ChatGPT-4o served as operational representatives of aesthetic-prioritised and instruction-following alignment strategies respectively. Expert ratings and self-reports showed that alignment strategy and prompt type significantly predicted creative quality and perceived usefulness; AIGC tool proficiency was associated with novelty; and diversity was primarily shaped by alignment strategy. These findings provide empirical evidence for design education, tool development, and human-AI collaborative workflow optimisation.
Greater product reliability is required by the U.S. Food and Drug Administration (FDA) for medical devices designed especially for emergency-use drug delivery systems such as autoinjectors (AIs; EpiPens). To achieve high levels of reliability, manufacturers are implementing design frameworks and their associated methods, which leverage design outcome prediction approaches such as Design for Six Sigma (DFSS) in lieu of the ‘Build, Test, Fix’ (BTF) frameworks that are widely used in the medical device industry. In this context, these two design frameworks, which are currently used by design teams and fit this purpose, are analyzed in the present study. The objective of this study is to compare the frameworks relative to their ability to yield suitably reliable designs and to understand whether changes can be made to these methodologies to improve the time and cost of meeting these high standards. For this purpose, a fundamental shift in how fault tree analysis (FTA) is used to model failures is leveraged to drive the system reliability level to the design targets meeting the mandates of the FDA. The results show the study design outcome prediction framework decreased the time needed to converge on device reliability targets on average by 61
Modern interactive products require tighter integration between UX research and engineering design. However, the translation of scenario-based UX representations into engineering specifications is often informal, weakening traceability and making it harder to preserve the contextual rationale for technical decisions. We conceptualize this problem as the experience–specification gap: a representational discontinuity between experiential user scenarios and formal engineering representations. To address it, we develop Scenario-Oriented Specification (SOS), a five-step design method that structures translation from user scenarios to engineering specifications through linked artifacts anchored by persistent identifiers. SOS is grounded in five meta-requirements and three design principles. The method was evaluated through longitudinal enactment in an interdisciplinary industrial development project across 15 user scenarios and four predefined dimensions: practical enactability, output quality, process quality, and process efficiency. The results show that SOS supported traceable specification development while making under-specification and feasibility constraints explicit, and that a sixth step, spatial integration, emerged during enactment as an extension needed in this embodied product context. The study contributes to engineering design research by conceptualizing the experience–specification gap as a representational problem, introducing SOS as a reusable representational translation logic, and demonstrating a combined empirical and structural validation approach for evaluating design methods under realistic industrial conditions.
Questionnaires are often treated as instruments for collecting information, but they also function as cognitive artifacts that shape how respondents produce that information. Their wording, response format, ordering, and context influence how respondents comprehend questions, retrieve information, form judgments, and map those judgments onto answers. Questionnaire failure can therefore arise when the respondent's perceived answering environment differs from the meaning anticipated by the designer—a condition termed perception asymmetry. This paper proposes a two-sided, perception-centered approach addressing two complementary objectives: informational completeness, ensuring that the questionnaire elicits the information required by the study, and interaction reliability, ensuring that respondents can provide that information accurately and truthfully under cognitive, affective, and contextual constraints. Grounded in Designics and operationalized through its Environment-Based Design and TASKS developments, together with response-process theory, the approach comprises five steps: deriving information needs, identifying respondent capability conditions, generating an initial questionnaire, diagnosing interaction barriers and response-process vulnerabilities, and recursively refining the questionnaire. During diagnosis, observable bias manifestations are used to locate response-process vulnerabilities and trace them to underlying emotion, logic, knowledge, or resource barriers. The approach is demonstrated through three role-specific questionnaires for assessing lean transformation in a Canadian manufacturing company with 44 respondents. The case illustrates the approach’s feasibility, traceability, and diagnostic usefulness, while not constituting full psychometric validation. The paper contributes an upstream design perspective by showing how questionnaires can be redesigned to reduce perception asymmetry before data collection.
In complex engineering systems, the dependencies among components or development activities are often modeled and analyzed using the Design Structure Matrix (DSM). Reorganizing the sequence of elements within a DSM to minimize feedback loops, thereby improving process efficiency and reducing rework, constitutes a challenging Combinatorial Optimization (CO) problem in engineering design and operations. As problem sizes increase and dependency networks become more intricate, traditional optimization methods that rely solely on mathematical heuristics often fail to capture the contextual nuances and struggle to deliver effective solutions. In this study, we explore the potential of Large Language Models (LLMs) to address such problems by leveraging their capabilities for advanced reasoning and contextual understanding. We propose a novel LLM-based framework that integrates network topology with contextual domain knowledge for iterative optimization of DSM sequencing, a representative CO problem in this domain. Experiments on various DSM cases demonstrate that our proposed method consistently achieves faster convergence and superior solution quality compared to both stochastic and deterministic baselines. Notably, incorporating contextual domain knowledge significantly enhances optimization performance regardless of the chosen LLM backbone. This study demonstrates the potential of LLMs as a promising foundation for advancing knowledge-informed optimization in engineering design.
Design changes in complex products often arise from multiple sources, including shifts in customer demands, advancements in technology, and evolving design methodologies. These changes can trigger significant adjustments at various stages of product development and manufacturing. Such modifications not only affect the design and functionality of the product itself but also create ripple effects across the entire production process, including supply chain management, cost estimation, and delivery schedules. Managing these changes effectively is crucial for manufacturers to maintain competitiveness and meet dynamic market requirements. To address these challenges, this paper proposes a novel decision-making framework that integrates a product knowledge network, machine learning techniques, and multi-objective optimization to optimize design change propagation. The framework utilizes a GNN-Transformer model for accurate prediction of change propagation probabilities, coupled with a cascade simulation model to evaluate propagation risks. By balancing coordinated upgrade coverage against implementation cost, development time, and system-level propagation risk, the proposed approach supports both localized corrective changes and comprehensive product-upgrade decisions.
Although Digital Twins (DTs) as digital representations of products are adopted by industry, the seamless exchange of DT data across its product lifecycle, from design to engineering, manufacturing, and operation, remains challenging. In particular, the early stages of product development, where the DT and its data evolve, as well as the cross-company collaboration during those phases are affected by significant interoperability issues.The current situation forces engineers to carry out lengthy processes of manual data identification, conversion, and integration, often resulting in substandard data quality. Consequently, this review article examines how cross-company DT data interoperability can be achieved by analyzing existing challenges, in terms of relevant context factors, requirements to be met, and proposed solutions including frameworks as well as enabling processes. The results indicate the significance of the addressed research problem as a variety of conceptual solutions and frameworks have been developed. However, most of the published work lack empirically valid results which leads to a lack of practical adoption. Moreover, these studies predominantly address later lifecycle phases and mostly don´t consider interoperability dimensions. We suggest research encompassing the semantic, technical, and organizational dimensions, particularly in the early DT lifecycle stages of product development. It needs to include a deeper understanding of stakeholder roles and processes to achieve data interoperability through process alignments.
Parameter design involves assigning values to interconnected design parameters that must satisfy requirements and constraints while remaining consistent. This is a cognitively challenging activity in engineering design. Yet little empirical research has examined how designers approach this task. We conducted a think-aloud study with 14 designers solving a photovoltaic system design problem. To analyze their behavior, we used a descriptive model comprising five elements: focusing on a parameter, collecting information, assigning a value, handling violations, and learning. The key finding is that designers do not treat all parameters equally—contrary to assumptions in prior research. Instead, they adapt their approaches based on each parameter’s feedback complexity, using rapid trial-and-error when feedback on changes can be understood directly, but shifting to more deliberate analysis when understanding feedback requires multiple reasoning steps. Observed behavioral patterns were formalized using network-based metrics derived from problem structure and knowledge state. For example, effective parameter focusing correlates with readiness of relevant information and with the potential to reveal new information by solving the parameter. Information sourcing, strategy choice, and violation prioritization also correlate with specific structural properties of the parameter network. These findings may inform the development of computational design support tools that provide context-sensitive guidance based on parameter network structure and enable design educators to teach adaptive problem-solving strategies.
Building Information Modeling (BIM) and Product Lifecycle Management (PLM) are widely adopted in industry, yet their approaches to managing change are not fully aligned. Understanding both their common ground and their differences is essential for transferring knowledge across domains. This study examines how change is handled in BIM- and PLM-supported industries, focusing on design change management (DCM) within BIM-supported environments and engineering change management (ECM) within PLM-supported environments. Through a literature-based review of definitions, terminology, processes, tools, and methods, we identify patterns that shape each approach. From this comparison, representative ECM and DCM processes are outlined as potential best practices. The analysis highlights similarities as well as distinct characteristics, offering insights into functionalities that could be adapted or transferred between BIM- and PLM-supported industries.
The ability to reduce risk and uncertainty in the operation and performance of complex adaptive systems is highly desirable across a range of applications and sectors, especially when the correct operation of those systems is critical. Recently, digital twins have delivered this capability for cyber-physical systems by combining modelling with real/relevant-time information flows. Increasingly, however, critical national infrastructure which were previously isolated networks (such as transport and energy networks) are accelerating their levels of integration, forming cyber-physical ecosystems (CPES) and creating new design challenges for digital twins. One fundamental challenge lies in synthesizing information from multiple sources in order to understand the ecosystem as a whole. In this paper, we propose a novel mapping framework to meet this need: the ‘domain relationship diagram’, which balances flexibility and interpretability with sufficient rigour to support model-based design. After establishing the method, we illustrate its usage through two use cases, both motivated by priority research in critical CPES: (i) flooding resilience in northern England and (ii) energy infrastructure at the Port of Dover. We also tie this approach into other crucial topics in digital twins for CPES, including ownership, ontologies and communication, and trust and resilience.
Design Failure Mode and Effects Analysis (DFMEA) traditionally relies on static tables, limiting its ability to capture failure propagation across design hierarchies or support reasoning under out-of-distribution conditions. These constraints hinder knowledge reuse and evidence-based decision-making in early design phases. To address this, we formalize failure knowledge as a directed Knowledge Graph grounded in fault-tree logic and introduce AfGNN (Adaptive Failure Graph Neural Network). AfGNN achieves robust prediction on rare, long-tail failure patterns through three integrated innovations: (1) Causal-Enhanced Soft-Label Embedding (CESLE), which integrates semantic similarity with causal weights to distinguish genuine relationships from statistical correlations; (2) a Depth-Adaptive Causal Propagation Framework, synergizing dynamic subgraph sampling with depth-decay attention to balance efficiency and fidelity while suppressing noise in deep layers; and (3) a formalized computational workflow that transforms DFMEA into a reusable causal graph, enabling systematic reasoning over incomplete failure records. Evaluated on five public datasets and a self-constructed automotive failure KG, AfGNN surpasses all GNN-based baselines and competes with LLM-based methods on general benchmarks, while substantially outperforming all baselines on the domain-specific FMEA dataset (MRR, Hits@1, Hits@10). This framework enables engineers to reason about rare multi-failure cascading effects without relying on complete historical data, advancing failure knowledge management and reliability-oriented design decision-making.
Functional Analysis supports product decomposition in the conceptual design phase by defining system functions and flows before committing to specific solutions and technical implementation. Although widely recognized, Functional Analysis practice remains fragmented, with varied methodologies, inconsistent terminology, and limited software support. This study presents a Systematic Literature Review of Functional Analysis applications in engineering design, screening six databases using the PICOC framework and PRISMA 2020 guidelines. Forty peer-reviewed papers were analyzed to identify dominant methods, inputs, outputs, tool usage, application domains, and selection criteria. Results show four main approaches widely adopted in engineering design: Black Box models, Function–Means Trees, Systems Architectures, and hybrid methods. Typical inputs include design requirements, interface characteristics, and customer needs; outputs are mainly flow diagrams, hierarchical decompositions, and systems architectures. Only 20
Design and acquisition of large-scale complex engineered systems can use technical measures to compare system alternatives through setting constraints on those measures and/or providing objectives using the measures. Selecting a technical measure set can be uncertain, with little selection guidance available, and difficult to validate, potentially leading to omitting technical measures. This research examines the impact of omitting technical measures on system alternative selection using a case study of real-world technical measures and system alternatives. A requirements-based constraint framework and an optimization-based objective function framework are developed using a set of real-world technical measures. The research models how omissions of technical measures lead to choosing a different system alternative. The impacts are demonstrated through an application of the NASA Human Landing System (HLS) using 13 system alternatives, including the systems proposed to NASA by Blue Origin, SpaceX, and Dynetics. The research finds that omissions in the constraint framework open the design space, potentially changing the system alternative chosen. Omissions in the objective function framework alter the indicated ordinal preference for the system alternatives, changing the system alternative chosen. An omitted technical measure on the side of the acquirer may change the system alternative selected, directing millions of dollars towards a specific organization and system. This research highlights the practical impacts of omissions of technical measures for awarding contracts during system acquisition. The likelihood of omissions, the difficulty to validate outcomes, and the impacts of omissions demonstrated in this research form evidence that the connection amongst problem formulation, problem solving, and validation must be emphasized if used for system alternative selection.
This study investigates how gender alignment between a designer and a user persona and the persona’s conformity to gender stereotypes shape early-stage design concept characteristics. While past research has examined gender and stereotype bias in final design outcomes, less attention has been paid to how these biases emerge during the initial design stages and shape early-stage design characteristics, such as aesthetics and functionality. To address this, we conducted a factorial design study with 50 expert and quasi-expert designers, paired with personas under different gender-alignment and gender-conformity conditions. Guided by Gender Schema Theory, we employed a summative content analysis and a linear regression model to investigate how gender alignment and gender stereotypes predict early-stage design concept characteristics. The results show that women designers produced concepts that emphasized more functionality and less affective considerations when designing for girl personas; designers produced concepts that emphasized durability and longevity more when designing for masculine personas, and more observed when men designers worked with masculine-boy personas. Interestingly, women designing for non-conforming boy personas designed concepts with more emotional and experiential considerations linked with self-expression and confidence. Masculine-boy and feminine-girl personas appeared to prompt more overt gendered embellishments compared to the gender-non-conforming personas. This study highlights how gender alignment and gender conformity influence the characteristics of early-stage design concepts. This work calls for a reflective and critical approach to persona development and use in early stages of design.
Robustness against geometric deviations is a crucial success factor in product development. While early-stage Robust Design (RD) methods aim to improve robustness to reduce later costly iterations, they often fail to quantify the impact of concept-specific geometric deviations on the functional fulfillment of product concepts. This paper presents an Embodiment Function Relation and Tolerance (EFRT)-based method for quantitatively evaluating the robustness of product concepts against geometric deviations. Using the EFRT-based method, the Design Robustness Index is derived through a six-step process as a quantitative measure to evaluate and compare the robustness of different concepts. A case study of a coining machine demonstrates the practical application of the method. The indices generated by the method are initially validated through a comparison with empirical testing consisting of rapid prototyping and simulation with surrogate models. The results show a strong alignment between theoretical evaluation and empirical testing for the 16 different concepts. The proposed method can support design engineers in quantitatively evaluating how product concepts respond to geometric deviations in the early development stages, giving them a stronger foundation to make informed design decisions. It has the potential to reduce the risk of costly iterations caused by geometric deviations and ensure more reliable products.
