Integrating multi-layer digital twins into engineering education: a design-based research framework for industry 4.0 competency development

1 Introduction

The rapid digital transformation of manufacturing (Subhedar and Dasnurkar, 2026) has fundamentally changed the knowledge and competencies expected of engineering graduates. The emergence of Industry 4.0 (Yaqub and Alsabban, 2023) has introduced highly interconnected manufacturing environments in which cyber-physical systems (Tao et al., 2019), Industrial Internet of Things (IIoT) (Mon et al., 2025), Digital Twins (Speicher et al., 2026; Fuller et al., 2020), cloud computing (Adel, 2022), artificial intelligence (AI) (Ahmed et al., 2022), and advanced analytics are integrated to optimize production processes and support data-driven decision making. More recently, the industry 5.0 (Yaqub and Alsabban, 2023) paradigm has extended this vision by emphasizing human-centered, resilient, and sustainable manufacturing systems (Subhedar and Dasnurkar, 2026), reinforcing the need for engineers who can combine technical expertise with collaborative problem-solving and critical thinking.

These technological advances present significant challenges for engineering education (García-Ávila et al., 2026). Traditional lecture-based instruction often provides students with theoretical knowledge of automation and manufacturing systems but offers limited opportunities to experience the multidisciplinary integration required in modern industrial environments. As a result, many graduates encounter difficulties when applying concepts from manufacturing, automation, simulation, control systems, artificial intelligence (Nantoi et al., 2025), and systems engineering to solve authentic industrial problems. Consequently, universities are increasingly adopting active learning strategies that promote experiential learning, project-based learning (PBL) (Barak and Yuan, 2021), and authentic industrial collaborations to bridge the gap between academic instruction and professional engineering practice.

Among the technologies supporting this educational transformation, Digital Twins have emerged as one of the most promising educational tools (Speicher et al., 2026; Lu et al., 2020). A Digital Twin is a dynamic virtual representation of a physical system that integrates simulation (Thelen et al., 2022), real-time operational data, three-dimensional modeling, and intelligent analytics to monitor, predict, and optimize system behavior throughout its lifecycle. Within engineering education, Digital Twins provide students with opportunities to develop systems thinking, analytical reasoning, simulation skills, interdisciplinary collaboration, and data-driven decision-making while working on realistic engineering problems that closely resemble industrial practice.

Recent studies have demonstrated the educational value of simulation-based learning, virtual laboratories, and Digital Twin environments for teaching manufacturing (Subhedar and Dasnurkar, 2026), automation, robotics, and industrial engineering. However, many reported educational implementations focus on isolated software tools or simplified laboratory exercises. Few studies describe comprehensive instructional frameworks in which students progressively design, simulate, validate, monitor, and optimize an integrated Digital Twin using professional industrial software and modern Industry 4.0 technologies within the context of a real industrial challenge (Yaqub and Alsabban, 2023). Moreover, relatively little research has examined how Design-Based Research (DBR) can support the iterative development of Digital Twin-based learning experiences that combine engineering design, simulation, supervisory control, artificial intelligence (Nantoi et al., 2025; Wang et al., 2018), and authentic industrial decision making.

Design-Based Research provides an appropriate methodological framework for addressing these challenges because it combines the iterative design of educational interventions with systematic evaluation in authentic learning environments. Rather than separating educational theory from engineering practice, DBR enables instructional strategies to evolve through successive refinement cycles while simultaneously generating practical solutions and educational knowledge. When combined with Project-Based Learning, this approach encourages students to engage in open-ended engineering problems requiring investigation, modeling, experimentation, validation, collaboration, and continuous reflection.

To address these educational needs, this study presents a Design-Based Research framework implemented in an undergraduate engineering course through the development of a Digital Twin for a real food manufacturing process. The industrial challenge involved redesigning the manual preparation process for chocolate bars used in cake production by proposing a semi-automated production cell capable of improving productivity while maintaining food safety and operational flexibility. Throughout the project, students progressively developed five complementary Digital Twin layers: (i) process analysis and process mapping; (ii) automation concept development; (iii) discrete-event simulation using Siemens Tecnomatix Plant Simulation; (iv) three-dimensional validation using Siemens Process Simulate and CAD; and (v) SCADA 4.0 implementation with Node-RED (OpenJS Foundation, 2025), Google Firebase (Google LLC, 2025a), and artificial intelligence-assisted decision support using the Gemini API (Google LLC, 2025b).

Unlike conventional engineering design projects that primarily emphasize the final technical solution, the proposed instructional framework uses the Digital Twin development process itself as the primary learning mechanism. Students are required to integrate knowledge from manufacturing systems, industrial automation, simulation, mechanical design, industrial communication networks, supervisory control, and artificial intelligence while continuously evaluating technical alternatives using engineering evidence. This multidisciplinary approach promotes the development of competencies aligned with current Industry 4.0 and Industry 5.0 workforce requirements (Yaqub and Alsabban, 2023), including systems thinking, digital manufacturing, simulation-based decision making, interdisciplinary collaboration, and data-informed engineering practice.

The contributions of this study are threefold. First, it proposes a structured educational framework that integrates Design-Based Research, Project-Based Learning, and Digital Twin technologies within undergraduate engineering education. Second, it demonstrates how professional industrial software including Siemens Tecnomatix Plant Simulation, Process Simulate, Node-RED, Google Firebase, and generative artificial intelligence can be incorporated into an authentic learning environment centered on a real industrial case study (Paul et al., 2026). Third, it illustrates how Digital Twin development can simultaneously support engineering problem solving and competency development, providing a potentially transferable model for preparing future engineers to address the technological and organizational challenges of Industry 4.0 and Industry 5.0.

The remainder of this paper is organized as follows: Section 2 reviews recent literature on Digital Twins, Design-Based Research, Project-Based Learning, Industry 4.0, and engineering education. Section 3 presents research methodology and instructional design. Section 4 describes the Digital Twin development process, educational implementation, and results. Section 5 discusses the implications of the proposed framework for engineering education, and Section 6 concludes the paper by summarizing the principal contributions, limitations, and opportunities for future research.

2 Literature review

2.1 Engineering education for industry 4.0 and industry 5.0

The digital transformation of manufacturing has significantly influenced engineering education by redefining the competencies expected of future engineers. Industry 4.0 integrates cyber-physical systems (Monostori, 2014), Industrial Internet of Things (IIoT) (Mon et al., 2025), cloud computing (Adel, 2022), artificial intelligence (AI) (Ahmed et al., 2022), advanced robotics, and Digital Twins into highly connected production environments. More recently, Industry 5.0 has expanded this technological perspective by emphasizing human-centered manufacturing, sustainability, resilience, and collaboration between humans and intelligent systems (Breque et al., 2021; Xu et al., 2026).

These developments require engineering graduates to master not only disciplinary knowledge but also systems thinking, interdisciplinary collaboration, digital literacy, simulation-based decision making, and data analytics. Consequently, universities are increasingly redesigning engineering curricula to incorporate authentic industrial projects that expose students to professional software environments and multidisciplinary engineering problems. Recent studies emphasize that active learning methodologies better prepare students for these challenges than traditional lecture-centered instruction because they allow learners to integrate theoretical concepts with practical engineering decision making.

Educational institutions are therefore moving toward competency-based curricula in which students progressively develop technical, analytical, and professional skills through realistic engineering experiences. Digital manufacturing environments have become particularly valuable because they allow students to explore complex industrial systems without the financial, operational, or safety constraints associated with physical production facilities.

2.2 Design-based research and project-based learning in engineering education

Design-Based Research (DBR) has become an increasingly important methodology for developing and evaluating educational innovations in authentic learning environments. Unlike conventional experimental research, DBR combines iterative instructional design with systematic evaluation, enabling researchers to refine educational interventions while simultaneously generating theoretical knowledge about learning processes.

Within engineering education, DBR (Crompton et al., 2026) aligns naturally with Project-Based Learning (PBL) (Marini et al., 2026; Hagedorn et al., 2023; Barak et al., 2021), where students solve open-ended engineering problems that require investigation, design, experimentation, reflection, and continuous refinement. Rather than emphasizing memorization of isolated concepts, PBL (Funa, 2026) encourages learners to integrate knowledge from multiple engineering disciplines while developing communication, teamwork, and critical thinking skills.

Recent studies have demonstrated that DBR and PBL (Jelodari et al., 2026) improve student engagement, promote deeper conceptual understanding, and strengthen problem-solving abilities when authentic industrial challenges are incorporated into course activities. Furthermore, iterative design cycles encourage students to evaluate engineering alternatives using evidence generated through simulation, experimentation, and collaborative reflection, closely resembling professional engineering practice.

Despite these advantages, relatively few studies have reported comprehensive instructional frameworks that combine Design-Based Research, Digital Twin technologies, simulation software, supervisory control systems, and artificial intelligence within a single engineering education project. Most reported implementations focus on isolated laboratory exercises or individual software tools rather than multidisciplinary learning experiences integrating multiple Industry 4.0 technologies.

2.3 Digital twins as learning environments

Digital Twins have emerged as one of the most significant enabling technologies for Industry 4.0 and are increasingly being adopted within engineering education (Speicher et al., 2026). A Digital Twin (Fuller et al., 2020) is commonly defined as a dynamic digital representation of a physical system that continuously exchanges information with its real-world counterpart to support monitoring, simulation, prediction, optimization, and decision making (Tao et al., 2019).

Recent reviews describe Digital Twins as educational environments capable of integrating multiple engineering domains, including manufacturing systems, automation, mechanical design, industrial communication, control engineering, data analytics, and artificial intelligence (Fuller et al., 2020; Lu et al., 2020). Unlike conventional simulation exercises, Digital Twins require students to consider the complete lifecycle of an engineering system, from conceptual design and process analysis to operational monitoring and continuous improvement.

Educational implementations have shown that Digital Twin projects improve systems thinking by requiring students to understand interactions among physical equipment, virtual models, process data, and operational decision making. Students also develop greater confidence in interpreting manufacturing data, validating engineering assumptions, and proposing evidence-based improvements before physical implementation (Correia et al., 2026; Impelluso, 2026).

However, most educational Digital Twin studies remain limited to virtual laboratories or isolated simulations. There is comparatively little research describing instructional approaches in which students design an entire Digital Twin ecosystem that integrates process modeling, discrete-event simulation, three-dimensional validation, supervisory control, cloud-based data management (Adel, 2022), and AI-supported decision making within a real industrial context.

2.4 Simulation, SCADA, and artificial intelligence for experiential learning

Simulation has long been recognized as one of the most effective instructional tools in engineering education because it allows students to evaluate complex systems without the cost or risks associated with physical experimentation. Discrete-event simulation environments such as Siemens Tecnomatix (Siemens Digital Industries Software, 2024) Plant Simulation enable learners to investigate production flow, resource utilization, bottlenecks, waiting times, and throughput while testing multiple process alternatives before implementation.

Three-dimensional simulation environments, including Siemens Process Simulate, complement discrete-event models by allowing students to validate equipment layouts, ergonomic conditions, spatial constraints, and mechanical feasibility. Together, these tools encourage students to connect operational performance with physical system design, reinforcing multidisciplinary engineering reasoning.

The educational value of simulation is further enhanced when combined with SCADA technologies (Jia et al., 2026) and Industrial Internet of Things architectures (Mon et al., 2025). Platforms such as Node-RED (OpenJS Foundation, 2025) and Google Firebase (Google LLC, 2025a) provide accessible environments for implementing supervisory monitoring systems, enabling students to visualize process variables, develop human-machine interfaces (HMIs) (Xu et al., 2026), analyze alarms, and interpret production data in real time.

Recent advances in generative artificial intelligence have introduced new opportunities for engineering education (Paul et al., 2026). Rather than replacing engineering judgment, AI systems can support learning by assisting students in interpreting operational data, identifying anomalies, generating maintenance recommendations, and evaluating engineering alternatives. When integrated into Digital Twin environments, AI encourages learners to justify decisions using process evidence while maintaining critical evaluation of automated recommendations. Such environments promote higher-order cognitive skills associated with engineering problem solving and data-informed decision making.

A growing body of work has begun to examine more directly how AI itself, as opposed to simulation or SCADA technologies more broadly, shapes the competencies engineering students develop. Ribeiro et al., 2026, in a literature review focused specifically on this question, report that AI-supported learning experiences are associated with gains in problem-solving, data literacy, critical evaluation of automated outputs, and adaptability, while also cautioning that over-reliance on AI recommendations can weaken independent engineering judgment if not counterbalanced by structured reflection. This distinction is relevant to the present framework: in our implementation, the Gemini API was deliberately positioned as a decision-support layer nested within a broader, human-supervised Digital Twin architecture (simulation, 3D validation, and SCADA), so that students were required to interpret and validate AI-generated recommendations against simulation evidence and physical constraints rather than accept them at face value. This design choice reflects the concern, echoed by Ribeiro et al., 2026, that competency development from AI exposure depends heavily on how AI is embedded within the broader learning task, not merely on whether AI tools are present.

2.5 Research gap and contribution

Although previous research has demonstrated the educational value of Project-Based Learning, Design-Based Research, simulation technologies, and Digital Twins independently, relatively few studies integrate these approaches into a unified educational framework based on an authentic industrial problem. Existing studies frequently focus either on educational methodology without extensive industrial technology integration or on technical Digital Twin development with limited discussion of student learning.

Furthermore, there is limited evidence describing how undergraduate engineering students can progressively develop Industry 4.0 competencies through the complete lifecycle of Digital Twin development, including process analysis, automation concept generation, discrete-event simulation, three-dimensional validation, SCADA implementation, cloud-based data integration, and AI-assisted operational decision support.

This gap is particularly pronounced with respect to artificial intelligence. Ribeiro et al., 2026 note in their review that, despite rapidly increasing interest in AI within engineering curricula, empirical accounts rarely trace how AI-assisted tasks interact with the other technical layers of a realistic engineering workflow (e.g., simulation-derived evidence, physical/3D validation, and supervisory control) to shape specific, observable competencies such as data-informed decision-making or critical evaluation of automated recommendations. Most reported interventions instead treat AI as a stand-alone tool or a short add-on exercise, evaluated in isolation from the broader engineering design process. The present study addresses this more specific gap by embedding AI-assisted decision support as the final, integrative layer of a five-stage Digital Twin architecture, allowing us to examine how students' interpretation of AI outputs was conditioned by evidence generated in the preceding simulation, 3D validation, and SCADA stages, and thereby to provide a more contextualized account of AI's contribution to competency development than is typically available in the literature identified by Ribeiro et al., 2026.

To address this gap, the present study proposes a Design-Based Research framework in which undergraduate engineering students collaboratively develop a multi-layer Digital Twin for a real food manufacturing process. The instructional approach combines Project-Based Learning with professional industrial software, including Siemens Tecnomatix Plant Simulation, Siemens Process Simulate, Node-RED, Google Firebase, and generative AI tools. By integrating engineering design, simulation, supervisory control, and intelligent decision support within a single authentic industrial project, the proposed framework contributes both to engineering education research and to the development of educational strategies that prepare students for the technological, organizational, and human-centered challenges of Industry 4.0 and Industry 5.0 (Xu et al., 2026).

To guide this investigation, the following research questions were formulated:

• RQ1: How does a Design-Based Research framework integrated with Project-Based Learning support the development of Industry 4.0 and Industry 5.0 competencies in undergraduate engineering students?
• RQ2: What engineering learning outcomes emerge from the iterative development of a multi-layer Digital Twin within an authentic industrial context?
• RQ3: How do professional engineering tools including discrete-event simulation, three-dimensional validation, supervisory control, and artificial intelligence contribute to students' evidence-based engineering decision-making?

3 Methodology

3.1 Research design

This study adopted a Design-Based Research (DBR) methodology to design, implement, and evaluate an educational intervention aimed at developing Industry 4.0 and Industry 5.0 competencies among undergraduate engineering students. The study was guided by the three research questions introduced in Section 2.5: RQ1 (How does a Design-Based Research framework integrated with Project-Based Learning support the development of Industry 4.0 and Industry 5.0 competencies in undergraduate engineering students?), RQ2 (What engineering learning outcomes emerge from the iterative development of a multi-layer Digital Twin within an authentic industrial context?), and RQ3 (How do professional engineering tools including discrete-event simulation, three-dimensional validation, supervisory control, and artificial intelligence contribute to students' evidence-based engineering decision-making?).

DBR was selected because it supports the iterative design, implementation, analysis, and refinement of instructional innovations within authentic educational environments while simultaneously contributing to educational theory and practice.

The instructional intervention was implemented through a Project-Based Learning (PBL) approach (Pomboza Floril et al., 2026; Barak and Yuan, 2021; Wei et al., 2023) in which students addressed a real industrial automation challenge provided by an industrial partner in the food manufacturing sector. Rather than solving an artificial classroom exercise, evidence from project artifacts and instructor observations suggests that students developed a complete Digital Twin for an existing manufacturing process (Ponnusamy et al., 2024), integrating engineering analysis with educational reflection throughout multiple iterative design cycles.

The research followed the three major phases commonly associated with Design-Based Research: The overall research design sought to investigate how authentic industrial projects supported by Digital Twin technologies can contribute to competency development in undergraduate engineering education as presented in the framework in Figure 1.

• Analysis and exploration, during which the educational needs, industrial problems, expected competencies, and learning objectives were identified.
• Design and implementation, where students progressively developed the Digital Twin using professional engineering software and collaborative engineering practices.
• Evaluation and reflection, during which engineering performance and student competency development were analyzed to refine the instructional framework.

A methodological clarification is warranted regarding the DBR framing used throughout this manuscript. The iterative refinement most directly evidenced in this study occurred at the level of students' engineering designs within a single semester-long implementation (e.g., teams revising conveyor speeds, buffer capacities, and workstation configurations after interpreting simulation results, Section 4.2). This iterative student engineering work should be distinguished from iteration of the educational intervention itself, that is, macro-level design-implement-evaluate-redesign cycles in which the instructional design (staging, rubric, tool sequencing, assessment criteria) is itself revised between successive offerings of the course based on evidence from a prior cycle. The present study implemented and evaluated a single offering of the instructional design; the three DBR phases described above (analysis and exploration; design and implementation; evaluation and reflection) were carried out once, in sequence, rather than across multiple iterated cycles of the intervention with documented before/after instructional changes. Section 5.6 reports lessons learned from this implementation that are intended to inform a subsequent cycle, but that subsequent cycle has not yet been implemented or evaluated. Accordingly, this study is more accurately characterized as a project-based educational intervention informed by Design-Based Research principles (Design-Based Research Collective, 2003; Anderson and Shattuck, 2012) than as a fully executed, multi-cycle DBR program, and claims throughout the manuscript should be read with this scope in mind.

3.2 Educational context and participants

The study was conducted within an undergraduate engineering course focused on Industrial Automation, Digital Manufacturing, and Industry 4.0 technologies at a higher education institution.

The educational intervention was integrated into the regular course activities over one academic semester. Students worked collaboratively in multidisciplinary teams to analyze, redesign, simulate, validate, and monitor a real manufacturing process involving the preparation of individually packaged food products. The project was developed in collaboration with an industrial partner from the food manufacturing sector, providing students with an authentic engineering context while preserving the confidentiality of the participating company.

The participants in this study were undergraduate engineering students enrolled in the course. A total of 60 students participated in the educational intervention and were organized into 18 project teams. All participants engaged in the same instructional activities and worked toward developing a Digital Twin following the proposed Design-Based Research framework, although completion of the later SCADA and AI-assisted decision-support stages varied across teams (see Section 4.5). Students were formed into teams of 3–4 members, with each team including at least one student from each engineering discipline (BME and IMT) to promote interdisciplinary collaboration.

Participation complied with institutional ethical guidelines for educational research. Students were informed about the objectives of the study, and all educational data were anonymized prior to analysis.

All 18 teams addressed the same authentic industrial challenge provided by the partner: the manual preparation of individually wrapped chocolate bars for cake decoration. While each team worked toward an independent Digital Twin solution for this shared manufacturing process, the consistency of the industrial problem allowed for comparative analysis of solution approaches and competency development across teams, without introducing variability from different manufacturing contexts. As detailed in Section 4.5, all 18 teams completed the process-analysis, conceptual-design, simulation, and 3D-validation stages, whereas completion of the SCADA and AI-assisted decision-support stages varied across teams (15/18 and 16/18, respectively) within the available course timeframe.

3.3 Educational objectives

The educational intervention was designed to promote the development of competencies required in contemporary digital manufacturing environments. At the conclusion of the project, students were expected to demonstrate the ability to: These objectives were aligned with competency-based engineering education and current Industry 4.0 and Industry 5.0 workforce expectations. Table 1 lists the institutional outcome codes, descriptions, and required mastery levels against which these objectives were formally assessed.

• Analyze manufacturing processes using systems thinking principles;
• Identify operational bottlenecks and justify automation opportunities;
• Construct discrete-event simulation models for production analysis;
• Validate manufacturing layouts using three-dimensional digital environments;
• Integrate Digital Twin technologies with industrial communication and supervisory systems;
• Interpret production indicators to support engineering decision-making;
• Collaborate effectively within multidisciplinary engineering teams; and
• Communicate engineering solutions through professional technical documentation and presentations.

Table 1. Institutional learning outcomes targeted by the automation of manufacturing systems course, including outcome codes, descriptions, and required mastery levels (A–C scale).

To clarify the assessment scale referenced above, the institutional A-C mastery scale denotes increasing depth of demonstrated proficiency within a given outcome, where A corresponds to introductory exposure (the student recognizes and describes the relevant concept or technique), B corresponds to intermediate application (the student applies the technique correctly in a guided or partially structured context), and C corresponds to advanced, autonomous mastery (the student independently selects, justifies, and validates the technique against evidence in an open-ended, authentic engineering problem, as required in this course). All outcomes in Table 1 required at least level B, and the three outcomes requiring level C (SMR0101, SMR0104, SMR0302) represent the competencies most central to the project's open-ended design character.

Regarding the relationship between Table 1 and the industry 4.0/industry 5.0 distinction, the outcomes are not individually labeled as “4.0” or “5.0” in the institutional framework, because most of them are dual-purpose, foundational engineering competencies that underpin both paradigms. That said, they can be characterized by primary orientation. SMR0103, SMR0104, SMR0201, SMR0202, SMR0302, and SMR0304 are primarily Industry-4.0-oriented, in that they concern the technical core of cyber-physical integration: simulation-based component selection, mechatronic system interaction, automation/control system design, and implementation using current industrial technology. SEG0702 (evaluation and selection of cutting-edge information technologies to optimize professional-field processes) and SMR0101 (establishing requirements, including economic and safety constraints, and generating alternative solutions) are dual-purpose but lean toward Industry 5.0 in this study, because their assessment in the illustrative-case team's work explicitly incorporated human-centered and socio-economic considerations, food-safety and operator-clearance requirements, and return-on-investment reasoning that weighs automation against human labor roles, rather than pursuing technical optimization alone. Beyond the eight formally coded outcomes, the interdisciplinary collaboration and engineering communication competencies assessed through our rubric (Section 3.6) are the two dimensions most directly aligned with Industry 5.0's emphasis on human-centered, cross-disciplinary teamwork, since they required BME and IMT students to negotiate shared understanding across disciplinary boundaries rather than optimize a single technical subsystem.

3.4 Educational intervention

The educational intervention was structured as a progressive Digital Twin development project composed of five instructional stages. The first stage focused on process analysis, during which students observed the existing manufacturing operation, documented material flow, identified bottlenecks, and proposed automation opportunities.

The second stage involved conceptual engineering design, where students evaluated alternative automation strategies, selected industrial components, developed preliminary layouts, and justified engineering decisions. The third stage introduced discrete-event simulation using Siemens Tecnomatix Plant Simulation. Students modeled the production process, analyzed operational indicators, compared alternative production scenarios, and used simulation results to support engineering decisions.

During the fourth stage, students employed CAD tools and Siemens Process Simulate to verify equipment positioning, operator accessibility, spatial constraints, and production flow before physical implementation.

Finally, students completed the Digital Twin by integrating SCADA concepts through Node-RED dashboards, cloud-based data exchange using Google Firebase, and AI-assisted operational analysis using the Gemini API (Google LLC, 2025b). This final stage demonstrated how real-time monitoring and intelligent decision support can complement simulation-based engineering design.

Regarding the AI-assisted decision-support layer, teams accessed the Gemini API (Google LLC, 2025b) directly rather than through a shared, centrally fixed configuration; consequently, the specific model version, temperature/sampling parameters, and prompt wording could vary by team and were not centrally logged or held constant across teams. Each team constructed prompts that combined a short natural-language instruction with simulation or SCADA-derived numeric process variables (e.g., cycle-time anomalies, tool-wear estimates, conveyor throughput) as input, and the API's free-text response was displayed to students as decision-support output within their Node-RED/Firebase dashboard rather than being parsed into a constrained output schema. The authors did not retain a consolidated log of exact model versions, prompts, and parameter settings across all 18 teams; future implementations of this framework should fix and record these elements centrally to support reproducibility.

Each stage built upon the previous one, progressively increasing the technical complexity while reinforcing multidisciplinary engineering competencies.

3.5 Data collection

The study employed multiple sources of evidence to evaluate both the educational intervention and the engineering project outcomes.

Educational evidence included: project reports; instructor assessment rubrics; student design reviews; oral presentations; reflective discussions; engineering documentation produced throughout the semester.

Engineering evidence included: process maps; simulation models; CAD assemblies; Digital Twin architectures; Process Simulate validation; SCADA dashboards; performance indicators generated during simulation.

The combination of multiple evidence sources enabled methodological triangulation, strengthening the credibility of the study.

3.6 Data analysis

Engineering performance was evaluated using production indicators generated by the simulation environment, including throughput, utilization, waiting times, bottleneck identification, and production capacity.

For clarity, the simulation KPIs reported throughout this manuscript are defined as follows. “Output” (or “conveyor output”) is the total count of units that completed processing and exited the modeled line within the simulated time horizon. “Utilization” is the percentage of the simulated horizon during which a given station was actively processing, expressed per station. “Blocking rate” is the percentage of the simulated horizon during which a station was prevented from releasing a completed unit because the immediately downstream buffer or station was full (reported separately for the source/upstream stations and for downstream stations). “Waiting time” is the mean time a unit spent queued at a station before processing began. Unless otherwise noted, all KPIs were computed over the full simulated horizon configured for each team's model (a fixed simulated period, not wall-clock experiment duration), as generated directly by the Plant Simulation software's built-in statistics.

Educational outcomes focused on competency development rather than solely on the engineering artifact. Student performance was analyzed according to demonstrated proficiency in systems thinking, engineering design, simulation modeling, industrial automation, interdisciplinary collaboration, and data-informed decision-making. Evidence from project deliverables, instructor evaluations, and the illustrative-case team's reflective discussions (Section 3.7) was used to identify learning achievements across the iterative Design-Based Research cycles; reflective discussions from the broader cohort were not systematically collected as individually attributable research data, consistent with the ethics and consent basis described in Section 3.7.

Rather than evaluating only the final Digital Twin, the analysis considered the complete engineering design process as evidence of student learning.

Because evidence in this study spans multiple levels, we distinguish them explicitly here to avoid conflating team-level, cohort-level, and individual-level claims. (a) Individually attributable, in-depth evidence (Sections 4.3, 4.4, Tables 2, 3) comes from the illustrative-case team only, whose members are co-authors of this manuscript. (b) Anonymized, team-level cohort evidence (Section 4.5, Table 4) covers all 18 teams but reports only aggregate team counts, not individual student outcomes. (c) No individual-student-level competency assessment, separate from team-level artifacts, was conducted in this study; where the manuscript describes “students” developing a competency, this refers to evidence generated by a team (or, for the illustrative case, by the co-authoring team specifically) rather than to individually verified learning by each of the 60 participating students.

Table 2. Illustrative engineering validation results from the plant simulation and process simulate/CAD analyses (illustrative team example).

Table 3. Mapping of competency dimensions to institutional learning outcomes, with artifact-based evidence and fulfillment status from the illustrative-case team report.

Table 4. Cohort-level instructor impressions of competency engagement across eighteen project teams (illustrative context, not validated competency assessments).

Given the scope of this study, artifact-based evidence is presented illustratively through an in-depth examination of one illustrative-case team deliverable, consistent with Design-Based Research practice of using rich documentation of authentic work as a legitimate evidence source alongside quantitative instrument.

Competency development was assessed using a structured rubric developed for this study (see Supplementary Material S1). The rubric evaluated six competency dimensions (systems thinking, simulation modeling, industrial automation, interdisciplinary collaboration, engineering communication, and data-informed decision-making) on a 1–5 scale (this dimension, “data-informed decision-making,” is used interchangeably with “evidence-based engineering decision-making” elsewhere in this manuscript to refer to the same competency), where 1 = not demonstrated, 3 = adequately demonstrated, and 5 = excellently demonstrated. Each dimension was assessed through three sources: (a) the final technical report (40% of weight), (b) milestone presentations (30%), and (c) instructor observations during laboratory sessions (30%). Two instructors independently evaluated a randomly selected subset of five of the eighteen team deliverables (27.8%, reported elsewhere in this manuscript as approximately 30%), achieving inter-rater reliability of κ = 0.78 (Cohen's kappa) on that subset. Disagreements were resolved through consensus discussion. For the illustrative-case analysis presented in Sections 4.3, 4.4, scores were verified against the rubric criteria to ensure consistent interpretation.

To clarify the scope of application: the full six-dimension, 1–5 numeric rubric described above was applied to the illustrative-case team and to the five-team inter-rater subset used to establish κ = 0.78. It was not applied, dimension-by-dimension with numeric scores, to all 18 teams. The cohort-wide counts reported in Table 4 (Section 4.5) instead reflect a coarser, binary (demonstrated/not clearly demonstrated) holistic judgment made by the course instructor for each of the 18 teams, informed by, but not equivalent to, the structured six-dimension rubric. These two forms of evidence numeric rubric scores for a subset of teams, and binary holistic impressions for the full cohort should not be conflated, and this manuscript reports them separately (Sections 3.6, 4.4, and 4.5) for that reason.

Because the rubric used in this study was developed locally for this course and has not undergone independent psychometric validation (Section 5.5), future implementations should adopt a validated or externally benchmarked competency rubric to enhance assessment consistency, transparency, and comparability across studies. Because the present study prioritized rich artifact-based qualitative analysis over quantitative measurement, competency evidence should be interpreted as descriptive rather than inferential.

It should also be noted that this manuscript does not report the aggregate numeric (1–5) rubric scores by competency dimension for the five-team inter-rater subset described above; only the resulting inter-rater reliability statistic (κ = 0.78) and the illustrative-case team's qualitative fulfillment status (Table 3) are presented. Reporting descriptive statistics (e.g., means and standard deviations) for each of the six competency dimensions across the scored subset would strengthen the quantitative grounding of the competency-development claims made in this study and is recommended as a priority addition for future revisions of this dataset.

Figure 2 summarizes the methodological workflow adopted in this study, the instructional intervention generated multiple sources of evidence, including student artifacts, instructor observations, design reviews, and reflective discussions, which were triangulated to identify competency development and address the research questions. Consistent with the Design-Based Research methodology, evidence obtained during each implementation cycle informed reflection and refinement of the instructional framework.

Table 5 illustrates the methodological alignment between the research questions, the sources of evidence collected during the educational intervention, and the analytical procedures employed in this study. Rather than relying on a single source of data, each research question was addressed through the triangulation of complementary evidence, including project artifacts, technical reports, instructor observations, milestone presentations, and reflective discussions. This approach enhanced the credibility of the findings by allowing competency development and educational outcomes to be interpreted from multiple perspectives. The matrix also provides methodological traceability by explicitly linking the collected evidence to the analytical procedures used to answer each research question, thereby strengthening the transparency and trustworthiness of the Design-Based Research framework.

Table 5. Alignment between the research questions, evidence sources, and analytical procedures adopted in the design-based research framework.

3.7 Ethical considerations

Because the corresponding author served both as course instructor and researcher, competency evaluations used for grading were conducted independently of, and prior to, the research analysis reported in this manuscript, following the institutional assessment rubrics described in Section 3.3. This separation addressed the fairness of grading; a related but distinct question is whether students' participation in the underlying project activities, and the subsequent decision to use their project work as research material, were free of any influence arising from the instructor's evaluative authority over them. Participation in the Digital Twin project was a required course activity, assigned and graded independently of any research use of the resulting work: no student's grade, academic standing, or course experience was contingent in any way on whether their project was subsequently discussed, analyzed, or published in this or any other research output. Consistent with Mexican regulation of the general health law regarding research, article 17, paragraph I, this study was reviewed under institutional policy governing pedagogical scholarship based on anonymized educational artifacts and routine course assessment and was determined to be exempt.

The study's ethical risk profile is low by design. No experimental intervention, deception, or manipulation of students' educational experience was introduced for research purposes: the instructional activities analyzed here (the five-stage Digital Twin project) constitute the normal course curriculum that all enrolled students would complete regardless of any research use of their work. The research use of this material was limited to secondary analysis of educational artifacts (technical reports, simulation outputs, design reviews) and aggregated, anonymized observational data; no additional risk, burden, or intervention beyond standard coursework was imposed on any student.

Two distinct consent bases apply to the two tiers of evidence reported in this study. For the individually attributable evidence in Sections 4.3, 4.4, drawn from the deliverables of Team PLCake (Hugo Emilio Padrón Gómez, Saúl Eduardo Tobías Rodríguez, Emilio Marroquín Treviño, and Juan Pablo Grajeda Sánchez), these four students are co-authors of this manuscript. Their consent rests on two separate and sequential bases: first, their participation in the course project itself, which occurred under the standard terms described above and independently of any research use; and second, their subsequent, voluntary decision as co-authors to have their specific project deliverables described, analyzed, and published in this article, evidenced by their review and approval of the manuscript's description, analysis, and interpretation of their work prior to submission. This second, publication-specific consent addresses the substantive elements ordinarily expected of informed consent to participate in research, a clear description of how their material would be used and presented, awareness of the analysis and interpretation applied to it, and a voluntary, documented decision to proceed, notwithstanding that no separate signed consent instrument was executed, since authorship review and approval served that function. For the broader cohort of eighteen teams (Section 4.5; Table 4), no individual student or team is named or otherwise identifiable. These observations were collected as part of routine pedagogical assessment that would have occurred regardless of this study, and are presented only as aggregate counts (e.g., “15/18 teams”). Because no identifiable data from these students is used, individual research consent was not separately sought for this cohort-level material, consistent with standard treatment of anonymized, non-identifiable secondary use of routine educational records.

Because students occupy a position of academic dependency relative to their instructor, this safeguard is addressed directly. Participation, non-participation, or withdrawal from any research use of a student's coursework had no bearing on that student's grade, academic standing, or relationship with the instructor: the only students whose material appears individually in this study are the co-authors, who made an affirmative, separate decision to do so after course completion and grading had been finalized. No student was identified, approached, or evaluated differently on the basis of participating or declining to participate in research use of their work, and the cohort-level data in Section 4.5 involved no individual decision point at all, since it was collected and reported only in aggregate as part of standard instructional practice.

To protect the confidentiality of the industrial collaborator, the participating company and its manufacturing process are described only in generic terms. No proprietary operational information or commercially sensitive data are disclosed in this manuscript.

Regarding data sent to the cloud-based Gemini API itself: prompts were constructed from simulation and SCADA derived numeric process variables (Section 3.4) rather than from student personal data; no student names, identification numbers, grades, or other personally identifiable information were included in API prompts. Because the manufacturing process and industrial partner are described only in generic terms throughout this study (see above), and because the numeric process variables reflect a simulated, semi-automated concept rather than the partner's actual proprietary production data, disclosure risk to the industrial partner was likewise limited. That said, the authors did not implement a formal, centrally enforced data-loss-prevention check on every team's prompt prior to submission to the third-party API, and this is acknowledged as a limitation in Section 5.5; future implementations should adopt an explicit, auditable protocol confirming that no personal or proprietary data reach third-party AI services.

3.8 Instructional materials availability

To support replication and adaptation by other educators, the instructional materials developed for this intervention are available for academic use. The materials include: project guide and milestone descriptions; team formation guidelines and role assignments; instructor assessment criteria and evaluation rubrics; software tutorials (Plant Simulation, Process Simulate, Node-RED, Firebase); template files and starter architectures; example project deliverables.

Educators interested in adapting the framework for their own courses are invited to contact the corresponding author for further guidance and collaboration. Future work will include publishing a comprehensive educator's guide to support wider adoption of the framework.

4 Results

4.1 Overview of the educational intervention

The Design-Based Research framework was implemented over one academic semester through a sequence of project-based learning activities centered on the development of a Digital Twin for a real manufacturing process. Students were guided through five instructional stages that integrated manufacturing systems analysis, automation design, discrete-event simulation, three-dimensional validation, supervisory control, and artificial intelligence-supported decision making. Completion of the final two stages (SCADA and AI-assisted decision support) varied across teams within the available course timeframe (see Section 4.5).

Rather than developing isolated technical exercises, students worked collaboratively on an authentic engineering challenge that required continuous integration of concepts from manufacturing engineering, industrial automation, digital manufacturing, computer-aided design, industrial communication, and data analytics. Each instructional stage introduced additional technical complexity while reinforcing competencies associated with Industry 4.0 and Industry 5.0 engineering practice.

The resulting Digital Twin served simultaneously as an engineering solution and as a learning environment through which students applied theoretical concepts to realistic industrial scenarios.

4.2 Competency development throughout the design-based research cycles

The iterative Design-Based Research approach enabled students to progressively construct engineering knowledge while refining their technical solutions. During the initial project stage, students focused on understanding the manufacturing process, identifying production bottlenecks, documenting material flow, and recognizing opportunities for automation. These activities primarily supported the development of systems thinking and engineering problem formulation.

As the project evolved, students demonstrated increasing proficiency in conceptual engineering design by evaluating alternative automation strategies, selecting industrial components, and preparing preliminary manufacturing layouts. The iterative nature of the project encouraged continuous reflection and revision, allowing engineering decisions to be supported by analytical evidence rather than intuition.

The introduction of discrete-event simulation represented a notable transition from conceptual reasoning to quantitative engineering analysis. Students developed simulation models capable of evaluating production throughput, equipment utilization, waiting times, blocking conditions, and production capacity. Simulation results enabled comparison of alternative process configurations while illustrating the importance of evidence-based engineering decision making.

The integration of CAD models and three-dimensional simulation further strengthened multidisciplinary reasoning by requiring students to validate spatial constraints, operator accessibility, equipment positioning, and production flow before considering physical implementation. Students recognized that successful automation depends not only on process optimization but also on mechanical feasibility and ergonomic considerations.

Finally, the implementation of SCADA concepts, cloud-based data communication (Adel, 2022), and AI-assisted operational analysis introduced students to contemporary digital manufacturing architectures. Through the development of supervisory dashboards and intelligent decision-support tools, students explored how operational data can support predictive maintenance, alarm management, and process optimization within Digital Twin environments.

Collectively, these iterative learning experiences promoted the development of competencies in systems thinking, simulation modeling, industrial automation, interdisciplinary collaboration, engineering communication, and data-informed decision making.

4.3 Engineering validation of the digital twin

The Digital Twin developed during the educational intervention successfully represented the complete manufacturing process through the integration of multiple digital technologies.

To ground this analysis in concrete process evidence, the illustrative-case team's baseline documentation quantified the manual operation being redesigned: individually wrapped units were grouped into six-unit packages (approximately 290 g), delivered in master cases of approximately 9.3 kg, with manual opening batches of approximately 350 units (17 kg) requiring approximately 30 min and two operators per batch, at a rate of two batches per twelve-hour shift. These baseline figures (package/case masses, batch size, and current cycle time) were provided directly by the industrial partner as documented operating parameters; the 50%-reduction performance target was likewise the partner's stated business objective, not a value measured or assumed by the students. The resulting semi-automated process metrics reported below (Table 2), by contrast, are simulation outputs produced by the student teams’ Plant Simulation models, not measurements of an implemented physical system.

The discrete-event simulation model demonstrated that the proposed semi-automated process achieved continuous product flow while enabling quantitative evaluation of production indicators such as throughput, resource utilization, waiting time, and bottleneck occurrence. Students used these performance indicators to compare alternative process configurations and justify engineering decisions using objective production data rather than subjective assumptions.

This section presents technical validation results from one illustrative-case team (Team PLCake), selected as an illustrative case because the reporting team are co-authors of this manuscript and therefore made their full underlying project documentation directly available for in-depth analysis; this selection was made for reasons of data access rather than through a pre-specified, objective sampling procedure, and the team is not claimed to be statistically representative of the cohort (see Section 3.6 for further discussion). Across all teams, simulation output ranged from 720 to 850 units (mean = 789, SD = 38), and utilization rates ranged from 75%–92% (mean = 85%). Team PLCake's results (Table 2) fall within this central range. Notably, their model revealed a source blocking rate of 89.78%, a finding mirrored in 14 of 18 teams (77.8%), suggesting a common capacity mismatch challenge across the cohort. This pattern illustrates how simulation evidence supported evidence-based engineering decision-making across multiple student teams, not merely in the illustrative case.

Three-dimensional validation confirmed the geometric feasibility of the proposed production cells. In the same illustrative case, Process Simulate and CAD analysis confirmed a minimum operator clearance of 600 mm at the loading position and verified that the primary kinematic motions conveyor transport, cutter descent, and roller extraction were geometrically compatible and non-conflicting within the assembled station footprint, while maintaining physical separation between the operator access zone and the guarded cutting/extraction hazard zone. Equipment selection for this illustrative-case concept was grounded in commercial technical documentation rather than generic assumptions, including a food-grade sanitary belt conveyor (Dorner AquaGard 7350 series), a vibratory bowl feeder for package orientation (RNA Automation), and a pneumatic compact cylinder actuating the cutting mechanism (Festo ADVU), illustrating how students justified component choices against functional suitability and food-contact hygienic requirements rather than arbitrary specification.

The completed Digital Twin also incorporated supervisory monitoring through SCADA technologies. Process variables generated by the simulation environment were transmitted to cloud-based databases and visualized through interactive dashboards, with operational KPIs refreshed at short, configurable polling intervals (e.g., a five-second interval in the illustrative-case team implementation). Artificial intelligence tools demonstrated how production information could be analyzed to support operational recommendations: in one team's implementation, the AI layer generated context-appropriate responses to four distinct operational conditions, including alarm grouping for repeated cycle-time anomalies, predictive maintenance alerts based on cumulative tool-wear estimates, numerical setpoint recommendations (e.g., adjusting conveyor speed to balance throughput against quality), and SOP-style decision support for multi-condition abnormal states, demonstrating functional decision-support behavior aligned with the SCADA 4.0 paradigm.

These four AI-generated examples were judged “context-appropriate” by the course instructor(s) through comparison against the corresponding simulation and process evidence available to the team at the time (e.g., whether a flagged anomaly matched an actual simulated blocking or utilization spike, and whether a numerical setpoint recommendation moved the process in the direction indicated by the simulation results), rather than through a separately documented scoring instrument or a second independent rater. This evaluation was therefore qualitative and instructor-mediated rather than based on a pre-specified accuracy protocol; the manuscript does not claim a systematic count of hallucinated or incorrect AI outputs across the four examples or across teams more broadly, and this is noted as a limitation in Section 5.5. Students were explicitly instructed to treat AI outputs as recommendations to be validated against simulation and process evidence rather than as authoritative conclusions, consistent with the human-supervised decision-support design described in Section 2.4.

Although the Digital Twin was developed within an educational environment, the resulting architecture closely reflects current industrial practices associated with smart manufacturing (Mittal et al., 2018) and provides students with experience using professional engineering software commonly employed in modern manufacturing organizations.

4.4 Student learning outcomes

The project demonstrated that integrating Digital Twin technologies into engineering education provides opportunities for students to connect theoretical knowledge with authentic engineering practice.

Throughout the intervention, students engaged as active engineering problem solvers rather than as passive recipients of technical content; because no pre-intervention baseline was collected (Section 5.5), this reflects the students' engagement within the course as observed, not a measured before/after transition. The project required them to integrate concepts from manufacturing systems, simulation, industrial automation, mechanical design, supervisory control, and artificial intelligence while continuously evaluating engineering decisions using quantitative evidence.

For instance, project teams were required to translate simulation and validation findings into implementation-level reasoning, including preliminary cost estimation and return-on-investment analysis; in one team's proposal, an estimated automation investment of approximately USD 19,000 was evaluated against projected annual labor savings of approximately USD 8,900 (based on reducing the operation from two operators to one system supervisor), yielding an estimated payback period of approximately 2.1 years and an estimated five-year return on investment of approximately 134% (accumulated net benefit of approximately USD 25,500 after recovering the initial investment), demonstrating that students connected technical simulation evidence to engineering economic decision-making rather than treating the Digital Twin as a purely technical exercise.

Consistent with the instructional objectives outlined in Section 3.3, the project deliverables examined in this study including the illustrative-case team report, instructor observations, and design review records indicated that students engaged with the following competency areas throughout the intervention: The iterative Design-Based Research framework encouraged continuous reflection throughout the project, allowing students to refine both their engineering solutions and their understanding of complex manufacturing systems.

• Analyze manufacturing processes using systems thinking principles (SMR0101/SMR0104);
• Identify operational bottlenecks and justify automation opportunities (SMR0103/SMR0202);
• Construct discrete-event simulation models for production analysis (SMR0302);
• Validate manufacturing layouts using three-dimensional digital environments (SMR0304);
• Integrate Digital Twin technologies with industrial communication and supervisory systems (SMR0304);
• Interpret production indicators to support engineering decision-making (SMR0201/SEG0702);
• Collaborate effectively within multidisciplinary engineering teams (interdisciplinary collaboration); and
• Communicate engineering solutions through professional technical documentation and presentations (engineering communication).

To complement the descriptive account above, Table 3 maps the competency dimensions assessed in this study to the institutional learning outcomes of the course and to the specific artifact-based evidence identified in the illustrative-case team report. This mapping illustrates how the assessment approach described in Section 3.6 was operationalized against concrete deliverable content.

This multidisciplinary composition was intentional to promote interdisciplinary collaboration, as BME (Biomedical Engineering) students contributed expertise in systems analysis and process understanding while IMT (Industrial and Mechatronics Engineering) students contributed expertise in automation and mechanical design (Salawu and Glen, 2026).

Of the eight institutional sub-competency codes evaluated against the illustrative-case team report, six were fully evidenced and two were partially evidenced for this team (Table 3). The partial cases correspond to dimensions intentionally excluded from the declared project scope (state-of-the-art literature benchmarking, formal controller-synthesis comparison, and physical prototype implementation) rather than to deficiencies in student performance and are discussed further as a limitation in Section 5.5. This analysis is consistent with the illustrative approach described in Section 3.6, which uses in-depth examination of illustrative-case work as a legitimate evidence source alongside quantitative instrument.

4.5 Instructor observations

Instructional observations were conducted throughout the semester by the course instructors as part of the Design-Based Research (DBR) implementation. The purpose of these observations was to document students' learning processes, identify recurring challenges during the Digital Twin development, and evaluate how successive project stages supported the development of engineering competencies. Observations were non-participant in nature and focused on students' interactions during laboratory sessions, project reviews, and milestone presentations. Except where they concern the co-authoring team (Team PLCake), these observations are reported only in aggregate and do not identify individual students or teams; throughout this section, they are treated as secondary, contextual evidence rather than as validated, individually attributable findings.

The observations were structured around the five instructional stages of the project: (1) process analysis, (2) automation concept development, (3) discrete-event simulation using Siemens Tecnomatix Plant Simulation, (4) three-dimensional validation with Siemens Process Simulate and CAD, and (5) SCADA 4.0 and Artificial Intelligence integration. During each stage, instructors recorded evidence related to students’ technical reasoning, collaborative problem solving, systems thinking, and decision-making processes.

At the initial stage, observations indicated that most teams concentrated primarily on describing the manufacturing process rather than identifying production constraints. As students progressed through process mapping and bottleneck analysis, they began using engineering terminology more consistently and justified automation decisions based on observed process limitations rather than intuition.

During the simulation phase, instructors observed a gradual transition from software-oriented experimentation to data-driven engineering analysis. Rather than treating Plant Simulation as a visualization tool, students increasingly interpreted key performance indicators including throughput, utilization, blocking percentage, waiting time, and work-in-process to compare alternative system configurations and justify design modifications. Teams frequently revised conveyor speeds, buffer capacities, and workstation configurations after interpreting simulation results.

The Process Simulate and CAD activities provided additional evidence of students' developing spatial reasoning and systems integration skills. Observations showed that students recognized that production performance depended not only on process flow but also on equipment layout, operator accessibility, safety considerations, and kinematic feasibility. Design discussions increasingly incorporated ergonomic constraints, equipment clearances, and material handling requirements that had not been considered during earlier project stages.

The final SCADA and Artificial Intelligence implementation demonstrated a notable shift in students' understanding of manufacturing systems. Rather than viewing automation solely as mechanical equipment, students discussed manufacturing as an integrated cyber-physical system combining sensors, control logic, real-time data acquisition, visualization dashboards, and AI-assisted decision support. During demonstrations, teams justified the selection of monitored variables, alarm strategies, predictive maintenance indicators, and operational recommendations generated through the AI component.

Instructor observations were conducted during 12 laboratory sessions (3 h each) and 4 milestone presentation sessions (2 h each) across the 16-week semester. Observations were structured around the five instructional stages, with instructors recording evidence related to technical reasoning, collaborative problem-solving, systems thinking, and decision-making processes using a standardized observation protocol (Supplementary Material S2). Thematic analysis of observation notes followed Braun and Clarke, 2006 six-phase approach: (1) familiarization with data, (2) generating initial codes, (3) searching for themes, (4) reviewing themes, (5) defining themes, and (6) writing up. Because the observation protocol (Supplementary Material S2) was itself structured around the five instructional stages and the six rubric-defined competency dimensions (Section 3.6), this analysis is more accurately characterized as a hybrid deductive-inductive thematic analysis: the broad coding categories were specified in advance from the instructional design, while the specific themes and sub-codes within each category were generated inductively from the observation notes rather than imposed beforehand. Two researchers (the corresponding author and a second course instructor) independently coded a subset of the observation notes representing approximately 20% of the total corpus, achieving 85% agreement, calculated as the proportion of coded units for which both researchers independently assigned the same theme label; a unit was counted as “agreement” when both researchers assigned an identical theme to the same excerpt of observation text. Remaining disagreements were resolved through discussion. The following recurring themes emerged across project stages.

These instructional observations complemented the analysis of project deliverables by providing contextual evidence of students' learning trajectories throughout the implementation. Rather than serving as direct measures of learning outcomes, the observations were used to triangulate evidence from project artifacts, technical reports, simulation models, and final presentations, strengthening the qualitative interpretation of competency development within the Design-Based Research framework.

A team was considered to have demonstrated a competency when the competency was evidenced in the final technical report, validated during milestone presentations, and confirmed by instructor observations.

Instructor observations and the illustrative-case team's reflective discussions were analyzed through the hybrid deductive-inductive thematic analysis described in Section 3.6. Observation notes were reviewed after each milestone, recurring patterns related to systems thinking, collaboration, engineering reasoning, and decision-making were identified, and themes were compared across instructional stages.

Table 4 provides supplementary, non-primary context alongside the detailed, individually attributable analysis of the co-authoring team discussed previously. These aggregate, anonymized instructor impressions suggest that similar competency-development patterns may have extended informally to other teams in the cohort, particularly around systems thinking and simulation modeling, which were central to the instructional design and were informally observed across most teams. Impressions regarding SCADA implementation and Artificial Intelligence integration were comparatively less consistent, which instructors attributed to differences in the degree of completion of the final project stage within the available course timeframe rather than to differences in student engagement. Because these technologies were introduced during the last phase of the intervention and required the integration of multiple software platforms, some teams completed only partial implementations within the available course timeframe. As these cohort-level impressions were not systematically documented, individually verified, or collected under a formal research consent process, they are offered only as illustrative context for interpreting the co-authoring team's experience and are not presented as a validated finding of this study.

For the cohort-level summary, competency was considered informally indicated when the instructor's aggregate impression aligned with observable evidence in a team's technical report, project deliverables, or milestone presentation. The frequencies reported in Table 4 are descriptive, anonymized, and non-primary; they are intended to provide illustrative context rather than statistical measures of competency attainment or individually verifiable findings.

4.6 Educational implications

Beyond the successful development of the Digital Twin, the findings suggest the value of combining Design-Based Research with Project-Based Learning in undergraduate engineering education.

Unlike traditional laboratory activities that frequently isolate individual engineering topics, the proposed instructional framework required students to integrate knowledge across multiple engineering disciplines while addressing an authentic industrial problem. This integration promoted meaningful learning by encouraging students to justify engineering decisions through simulation, experimentation, virtual validation, and data analysis rather than relying exclusively on theoretical calculations.

The project also exposed students to professional engineering software widely used in contemporary manufacturing environments and may help narrow the gap between university education and industrial practice, although this was not directly measured. The combination of simulation, Digital Twin technologies, SCADA architectures, cloud computing, and artificial intelligence created a multidisciplinary learning environment aligned with current Industry 4.0 and emerging Industry 5.0 competency requirements (Schleiss et al., 2026).

These findings suggest that Digital Twin-based educational projects can serve as an effective instructional strategy for preparing future engineers capable of designing, analyzing, and managing intelligent manufacturing systems while simultaneously developing critical thinking, collaboration, and evidence-based engineering decision-making skills.

5 Discussion

5.1 Research question 1: competency development

Before turning to the substantive findings, it is worth making explicit the evidentiary chain connecting the instructional design elements described in Section 3 to the observable learning processes reported in Section 4, since design description alone does not constitute evidence of impact. The rubric-based assessment architecture (Section 3.6) specified, in advance, which artifacts (final technical report, milestone presentations, instructor observations) would count as evidence for each of the six competency dimensions and assigned each dimension a decision rule (fulfilled when evidenced across all three sources). This decision rule was then applied stage-by-stage: at each of the five instructional stages, we identified a specific, observable shift in student behavior, for example, the transition from descriptive process narration to constraint and evidence-based justification at the process-analysis stage, or the transition from treating Plant Simulation as a visualization tool to using its KPIs to justify buffer and conveyor redesigns at the simulation stage (Section 4.2 and 4.5), and we cross-checked that shift against the corresponding artifact evidence in the illustrative-case team's report (Sections 4.3, 4.4; Table 3) and against the independently coded instructor-observation themes (Section 4.5). It is this triangulation across design intent, artifact evidence, and independently coded observation, rather than the design description in isolation, that supports the claim that the DBR-PBL structure, and not simply the passage of a semester, produced the reported competency evidence. We acknowledge, however, that this chain establishes association and plausibility rather than causal proof; see Section 5.5 for the corresponding limitation regarding the absence of a pre-intervention baseline.

This study demonstrates that integrating Digital Twin technologies into undergraduate engineering education through a Design-Based Research (DBR) framework can provide students with meaningful opportunities to develop competencies associated with Industry 4.0 and the emerging Industry 5.0 paradigm. Rather than learning simulation, automation, or supervisory control as isolated topics, students experienced these technologies as interconnected components of a complete engineering system developed to address an authentic industrial challenge.

The project-based structure encouraged students to move beyond procedural software operation toward evidence-based engineering reasoning. Throughout the iterative DBR cycles, students progressively integrated process analysis, automation design, discrete-event simulation, three-dimensional validation, supervisory control, cloud-based communication, and artificial intelligence into a coherent Digital Twin architecture. This progression reflects authentic engineering practice, where complex problems are addressed through iterative refinement, multidisciplinary collaboration, and continuous evaluation.

An important educational outcome of the intervention was the development of systems thinking. Students were required to consider interactions among mechanical components, manufacturing processes, digital models, operational data, and human decision-making rather than optimizing individual subsystems independently. This holistic perspective aligns closely with current competency frameworks for digital manufacturing and smart production systems.

The project also promoted analytical reasoning by requiring engineering decisions to be justified using quantitative evidence generated through simulation and virtual validation. Instead of relying on assumptions or intuition, students interpreted throughput, utilization, bottleneck analysis, and production-flow indicators before proposing engineering improvements. This evidence-based decision-making process represents a fundamental competency for contemporary engineering practice.

5.2 Research question 2: contribution of the digital twin framework

The findings of this study are consistent with recent literature emphasizing the educational value of Project-Based Learning, Design-Based Research, and experiential learning within engineering education. Previous studies have reported that authentic engineering projects increase student engagement, improve conceptual understanding, and strengthen interdisciplinary collaboration by connecting theoretical instruction with professional practice.

Similarly, recent investigations into Digital Twin-based education have demonstrated that virtual manufacturing environments provide students with opportunities to develop systems thinking and simulation competencies while reducing the logistical constraints associated with physical laboratories. The present study extends these findings by integrating multiple Industry 4.0 technologies including discrete-event simulation, three-dimensional validation, supervisory control, cloud computing, and artificial intelligence within a single educational framework rather than treating these technologies as independent instructional activities.

Unlike many previously reported educational implementations that emphasize only simulation or virtual laboratories, the proposed framework required students to complete the entire engineering design process, from process analysis and conceptual design to virtual commissioning and intelligent supervisory monitoring. This comprehensive approach more closely reflects contemporary industrial engineering workflows and therefore provides stronger preparation for professional engineering practice.

Furthermore, the Design-Based Research methodology enabled continuous refinement of both the instructional activities and the engineering solutions throughout the project. This iterative process is consistent with the educational philosophy underlying DBR, where instructional interventions evolve through repeated cycles of implementation, reflection, and redesign.

5.3 Research question 3: implications for engineering education

The results suggest that Digital Twin development can serve not only as an engineering design activity but also as a comprehensive pedagogical strategy for engineering education. By integrating professional industrial software into authentic project-based learning environments, students experience the multidisciplinary nature of modern manufacturing while simultaneously developing technical, analytical, and professional competencies.

The proposed instructional framework illustrates how simulation, CAD, supervisory control, cloud computing, and artificial intelligence can be introduced progressively throughout a semester without isolating individual technologies into separate laboratory exercises. Instead, each technology contributes to the development of a common engineering solution, reinforcing knowledge integration and promoting meaningful learning.

The increasing adoption of Industry 4.0 and Industry 5.0 technologies within manufacturing organizations requires universities to reconsider traditional engineering curricula. Future engineers must understand not only individual software platforms but also the interactions among cyber-physical systems, digital manufacturing, data analytics, and human-centered decision-making. Educational experiences that integrate these technologies within authentic industrial contexts may therefore contribute to narrowing the gap between academic preparation and industrial practice.

Additionally, the instructional framework is transferable to other engineering disciplines. Although the present study focused on a food manufacturing process, the same educational methodology could be adapted to applications involving logistics, assembly systems, robotics, automotive manufacturing, pharmaceutical production, or other industrial environments where Digital Twin technologies support process improvement and operational decision-making.

5.4 Contributions to design-based research

This study contributes to Design-Based Research by demonstrating how DBR can be applied to the development of multidisciplinary Industry 4.0 learning environments. Rather than evaluating a single instructional tool or software platform, the proposed framework integrates multiple engineering technologies into a coherent educational intervention that evolves through iterative design cycles.

The study also illustrates how authentic industrial collaboration can strengthen engineering education while preserving educational objectives. The industrial challenge served as a realistic context for student learning without shifting the primary focus away from competency development. In this sense, the engineering artifact, the Digital Twin should be interpreted as evidence of student learning rather than the principal outcome of the research.

Further contribution lies in the integration of generative artificial intelligence within Digital Twin-based learning. Instead of replacing engineering analysis, AI functioned as a decision-support technology that encouraged students to critically interpret recommendations using engineering evidence. This approach reflects current perspectives on responsible generative-AI adoption within higher education (Gəruslu et al., 2026), where artificial intelligence enhances learning while maintaining human judgment and professional responsibility.

5.5 Limitations and future research

In summary and stated explicitly at the outset of this section: this study did not include pre/post competency measures, a comparison group, individual-student-level competency assessment (Section 3.6), systematic artifact analysis of the full 18-team cohort (Sections 3.6, 4.3–4.4), independent psychometric validation of the competency rubric, or physical implementation of the proposed automation concept (Table 3). These absences jointly limit the strength of the causal, generalizability, and individual competency claims that can be drawn from this study; the findings should be read as descriptive and illustrative evidence of a promising instructional approach rather than as confirmatory evidence of its effectiveness. Each of these points is elaborated below.

Several limitations should be acknowledged. First, the study was conducted within a single undergraduate engineering course at one higher education institution, which may limit the generalizability of the findings to other educational contexts. Replication across different institutions, engineering disciplines, and cultural settings would strengthen the external validity of the proposed instructional framework.

Second, the educational evaluation primarily relied on project deliverables, instructor observations, and competency-based assessments. Future studies should incorporate additional evidence, including validated self-efficacy instruments, learning analytics, peer assessment, reflective journals, and longitudinal follow-up to examine the persistence of competency development beyond the duration of the course.

Third, and importantly, this study did not include a pre-intervention baseline assessment or a comparison group. Because participating students were advanced undergraduates with substantial prior engineering training, it is not possible from the present design to determine what proportion of the competencies demonstrated at the end of the course reflect skills students already possessed before the intervention, as opposed to skills developed through the DBR-PBL structure specifically. We therefore reframe our claims accordingly: the evidence reported here demonstrates that the framework provided students with an authentic, integrated setting in which to apply, combine, and demonstrate these competencies across a complete engineering lifecycle, rather than constituting proof that the competencies were newly acquired from a zero baseline. Future implementations should incorporate pre/post competency measures and, where feasible, a comparison group exposed to a more traditional, non-integrated laboratory sequence, to isolate the marginal contribution of the DBR-PBL Digital Twin framework from students' pre-existing ability.

Fourth, the multidisciplinary team composition (BME and IMT students within each team) was designed to promote interdisciplinary collaboration, but it also means that students likely entered the project with different prior knowledge and, in practice, gravitated toward different responsibilities: instructor observations and the illustrative-case team's documentation suggest that BME students contributed more heavily to process analysis and requirements definition, while IMT students contributed more heavily to automation design, simulation modeling, and controls implementation. Because our competency rubric (Section 3.6) was applied at the team level rather than disaggregated by individual student or discipline, the present study cannot determine whether the six competency dimensions were developed to the same degree by BME and IMT students, or whether role specialization produced uneven, discipline-dependent depth of learning within each competency (for example, deeper simulation-modeling competency among IMT students vs. deeper systems-analysis competency among BME students). We flag this as a limitation of the current assessment design and recommend that future implementations collect individual-level, role-disaggregated evidence (e.g., individual reflective logs, role-specific rubric items, or peer-assessment of contribution) to examine whether interdisciplinary team-based Digital Twin projects distribute competency development evenly across disciplines or instead reinforce pre-existing disciplinary specialization.

Moreover, while inter-rater reliability was assessed for a subset of team evaluations (κ = 0.78; Section 3.6), this figure reflects agreement between the two raters applying the rubric, not the psychometric validity of the rubric itself. The competency rubric was developed locally for this study and was not independently validated through procedures such as expert panel review or pilot testing prior to use. This limits the ability to benchmark competency gains against standardized educational metrics.

Additionally, the engineering artifact evidence presented in this study is drawn from an illustrative analysis of one illustrative-case team deliverable from the eighteen teams that participated in the intervention. Therefore, the findings related to competency development should be interpreted primarily as exploratory and illustrative rather than as statistically generalizable across the full cohort.

While this report was selected as an illustrative case of the type and quality of work produced across the cohort, a systematic analysis of all eighteen team deliverables was not conducted within the scope of this study. Future research should perform a full corpus analysis of team artifacts to enable cohort-level claims about competency development, KPI ranges across teams, and layer-completion rates. The present study therefore treats the artifact evidence as illustrative of the framework's outcomes rather than as a comprehensive assessment of all participants.

Third, although the Digital Twin incorporated simulation, virtual commissioning, supervisory control, cloud computing, and artificial intelligence, emerging technologies such as augmented reality, virtual reality, digital thread architecture, industrial metaverse environments, and collaborative robotics were beyond the scope of the present study. Integrating these technologies into future educational interventions may provide additional opportunities to investigate advanced engineering competencies associated with Industry 5.0.

Finally, future research should explore comparative studies between Digital Twin-based instructional frameworks and more traditional laboratory-based engineering education to quantify differences in student engagement, competency development, and long-term professional readiness.

Overall, the findings indicate that Digital Twin-based Project-Based Learning implemented through a Design-Based Research framework represents a promising approach for preparing engineering students to address the multidisciplinary challenges of contemporary digital manufacturing. By combining authentic industrial problems with professional engineering software and iterative educational design, the proposed framework contributes to both engineering education research and the development of future-ready engineering curricula.

5.6 Student challenges and lessons learned

Analysis of student team deliverables and instructor observations revealed several recurring challenges that provide insights for refining the instructional framework. First, students faced significant technical integration challenges when connecting five disparate technologies (Plant Simulation, Process Simulate, Node-RED, Firebase, Gemini AI) into a coherent Digital Twin architecture. Interoperability between these tools, all of which were novel to students, required substantial troubleshooting and iterative debugging. Second, simulation modeling proved demanding, as students needed to balance model fidelity against computational performance while resolving the capacity mismatches revealed by the model (e.g., the 89.78% source blocking rate requiring buffer tuning). Third, team collaboration across engineering disciplines (BME, IMT) required intentional coordination, particularly when integrating simulation results from one team member's expertise with SCADA implementation from another's. Fourth, the steep learning curve associated with professional engineering software consumed significant time that could otherwise have been devoted to deeper engineering analysis.

These challenges suggest several refinements for future implementations. Introducing selected software tools (e.g., Plant Simulation fundamentals, basic Node-RED programming) earlier in the curriculum could reduce the technical learning burden. Providing starter templates for common architecture patterns (e.g., Firebase data structures, Node-RED flows) could accelerate implementation. Additionally, more structured team coordination tools (e.g., project management templates, milestone checkpoints) could improve interdisciplinary collaboration. These refinements would allow students to focus more on engineering analysis and design reasoning, strengthening the learning outcomes of the Digital Twin development experience.

6 Conclusion

This study presented and evaluated a Design-Based Research (DBR) framework that integrates Project-Based Learning (PBL) and Digital Twin technologies to support competency development in undergraduate engineering education (Kager et al., 2026; Kolmos, 2026). Motivated by the growing demand for engineers capable of designing, analyzing, and managing Industry 4.0 and emerging Industry 5.0 systems, the proposed instructional framework engaged students in solving an authentic manufacturing challenge through the progressive development of a multi-layer Digital Twin.

The educational intervention demonstrated that combining process analysis, automation design, discrete-event simulation, three-dimensional validation, supervisory control, cloud-based communication, and artificial intelligence within a single project creates a rich and authentic learning environment. Rather than studying these technologies independently, students experienced their integration throughout the complete engineering design process, reflecting the multidisciplinary nature of contemporary industrial practice. This approach encouraged learners to connect theoretical concepts with practical decision-making while strengthening competencies in systems thinking, digital manufacturing, simulation modeling, industrial automation, engineering communication, interdisciplinary collaboration, and evidence-based problem solving.

From a pedagogical perspective, the Design-Based Research methodology provided an effective mechanism for designing, implementing, evaluating, and refining the instructional intervention. The iterative learning cycles promoted continuous reflection, collaborative knowledge construction, and progressive improvement of both the engineering solution and the students' understanding of complex manufacturing systems. The results suggest that authentic industrial projects supported by professional engineering software may help narrow the gap between university instruction and industrial practice while fostering higher-order cognitive skills required in modern engineering environments; this narrowing was not directly measured in the present study and should be interpreted as a plausible, not demonstrated, outcome.

Although the Digital Twin developed during the project successfully demonstrated the technical feasibility of the proposed manufacturing solution, the primary contribution of this study lies in the educational framework rather than the engineering artifact itself. The Digital Twin served as a pedagogical vehicle through which students integrated knowledge from multiple engineering disciplines, applied professional engineering tools, and developed competencies aligned with current industrial expectations. This distinction reinforces the role of Digital Twin technologies as instructional environments that promote meaningful learning, rather than merely as technical design tools.

The proposed framework also contributes to the growing body of research on engineering education by illustrating how Design-Based Research and Project-Based Learning can be combined with Digital Twin technologies to create adaptable, multidisciplinary learning experiences. Because the instructional model is centered on the engineering design process rather than a specific manufacturing application, it can be adapted to diverse engineering domains, including robotics, logistics, energy systems, aerospace, automotive manufacturing, healthcare engineering, and smart infrastructure.

Several limitations should be acknowledged. The study was conducted within a single undergraduate engineering course and evaluated one educational implementation. Future research should investigate the transferability of the proposed framework across different institutions, engineering disciplines, and educational contexts. Additional studies incorporating validated competency assessment instruments, longitudinal evaluations, learning analytics, and comparative instructional designs would further strengthen understanding of how Digital Twin-based learning influences student development over time.

Future work may also extend the educational framework by incorporating emerging technologies such as augmented and virtual reality, collaborative robotics, industrial metaverse environments, digital thread architectures, edge computing, and advanced artificial intelligence techniques. Exploring these technologies within Design-Based Research interventions may provide further insights into preparing future engineers for increasingly connected, intelligent, and human-centered manufacturing systems.

In conclusion, this study demonstrates that Digital Twin-based Project-Based Learning, implemented through a Design-Based Research framework, represents an effective educational strategy for preparing undergraduate engineering students to meet the multidisciplinary challenges of Industry 4.0 and Industry 5.0. By integrating authentic industrial problems, professional engineering software, iterative instructional design, and competency-oriented learning, the proposed framework contributes both to engineering education research and to the development of future-ready engineering curricula capable of supporting the next generation of digital engineers.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI tools (including grammar checkers and reference management software) were used during the preparation of this manuscript. Google Gemini AI was used for preliminary literature search assistance and reference organization. However, all intellectual content, analysis, interpretation, and conclusions are the sole responsibility of the authors. All AI outputs were critically reviewed, validated, and modified as needed by the authors before inclusion. The use of AI tools does not alter the authors' accountability for the accuracy and integrity of the work.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/feduc.2026.1925782/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Sonsoles López-Pernas, University of Eastern Finland, Finland

Reviewed by: Liudmyla Gryzun, Simon Kuznets Kharkiv National University of Economics, Ukraine

Jorge Muniz Jr, Universidade Estadual Paulista Julio de Mesquita Filho, Brazil

Abbreviations AI, artificial intelligence; API, application programming interface; BME, biomedical engineering; CAD, computer-aided design; CMS, condition monitoring system; DBR, design-based research; GenAI, generative artificial intelligence; HMI, human-machine interface; IEEE, Institute of Electrical and Electronics Engineers; IioT, industrial internet of things; IMT, industrial and mechatronics engineering; IoT, internet of things; ISA, International Society of Automation; KPI, key performance indicator; PBL, project-based learning; PLC, programmable logic controller; ROI, return on investment; RQ, research question; SCADA, supervisory control and data acquisition; SME, small and medium-sized enterprise; STEM, science, technology, engineering, and mathematics.