Digital Workflows for Circular Manufacturing: Mapping the Landscape, Choosing an Architecture

Circular manufacturing is often discussed as a materials problem: recover this alloy, redesign that joint for disassembly, close the loop on polymer streams. In practice, the harder problem is a data and workflow problem. A component that leaves a factory floor must carry enough structured information for a remanufacturing facility in another country to decide whether to refurbish, disassemble, or recycle it. A design engineer must be able to query what happens to a fastener in year twelve of product life before committing to a geometry. A supply-chain coordinator must trace whether a batch of recovered aluminium actually meets the specification assumed in the next production cycle. None of these decisions are solved by a sustainability report or a policy statement. They require digital workflows — connected sequences of data capture, exchange, transformation, and action that span the full product lifecycle and the organisational boundaries between stakeholders.

The European policy environment has made this urgency concrete. The Circular Economy Action Plan, the Ecodesign for Sustainable Products Regulation, and the concept of a Digital Product Passport all point toward a future where circularity claims must be backed by traceable, machine-readable lifecycle data. For a mid-sized manufacturer in Austria, Germany, or the Netherlands, the question is no longer whether to prepare for circular workflows but which digital architecture to adopt, how to sequence implementation, and what organisational capabilities are needed to make the workflow actually function rather than exist as a compliance artefact.

What Circular Manufacturing Actually Requires from a Workflow

Before comparing technologies, it is worth being precise about the operational demands. A linear manufacturing workflow follows a single direction: design, procure, produce, deliver, support, discard. A circular workflow introduces multiple feedback loops and reversals. Design-for-disassembly decisions made in CAD must propagate to bill-of-materials structures that a disassembly planner can later query. Material composition data captured at production must remain accessible at end-of-life, potentially decades later, and in a format the recovery facility can parse. Remanufacturing instructions must be versioned, linked to specific serial numbers, and updated when a field failure reveals a design weakness. Product-as-a-service models require that ownership of data shifts from the customer to the manufacturer while the customer retains operational telemetry access.

These are not single-system problems. They involve at minimum: a product lifecycle management layer that holds design intent and configuration history; a manufacturing execution layer that records actual production parameters and material inputs; a service and telemetry layer that captures use-phase data; an end-of-life and recovery layer that documents disassembly, material sorting, and re-entry into production; and a governance layer that defines who owns, updates, and accesses each data object across organisational boundaries. The digital workflow is the connective tissue between these layers. Its architecture determines whether circularity is achievable at production scale or remains a pilot-project curiosity.

Alternative Workflow Architectures

Several distinct architectural patterns have emerged, each with different assumptions about data ownership, integration depth, and organisational reach. They are not mutually exclusive, and many mature implementations combine two or three, but understanding their individual logic is necessary before making a selection.

PLM-Centric with Extended Lifecycle Modules

The most common starting point for manufacturers already running a product lifecycle management system such as Siemens Teamcenter, PTC Windchill, or Dassault Systèmes ENOVIA is to extend the PLM’s data model to cover post-production phases. Material passports are stored as structured attributes on the product or component object. Disassembly instructions, remanufacturing procedures, and recycling pathways become linked documents or workflow tasks within the same system that already manages engineering change orders. The advantage is deep integration with the existing engineering and configuration-management backbone: a design change automatically propagates to the material passport, and the bill of materials that a recovery facility reads is the same authoritative structure the design team maintains.

The limitation is reach. PLM systems are designed for the engineering organisation, not for a third-party remanufacturing partner or a municipal waste-sorting facility. Extending access to external stakeholders requires either costly integration middleware, a federated data model, or a separate lightweight interface. Additionally, most PLM platforms were not architected for high-frequency telemetry ingestion or for the kind of event-driven workflow triggers that a service-phase circularity loop demands. The system becomes a repository rather than a live coordination engine.

MES-Integrated Production Traceability

Manufacturing execution systems sit closer to the shop floor and capture what actually happened during production: which heat lot of steel went into which serial number, what welding parameters were applied, which quality inspection was passed or failed. For circular manufacturing, this production-level provenance is critical. A remanufacturer deciding whether to reuse a machined housing needs to know not just the nominal alloy grade but the actual heat treatment cycle it underwent, because that determines residual stress and fatigue life.

An MES-centric circular workflow extends traceability records beyond the factory gate. Serial-number-linked production data is published to a downstream access point that a recovery or remanufacturing partner can query. Some implementations use the MES as the authoritative source for a Digital Product Passport, pushing selected fields to a public or consortium-facing registry. The strength here is factual accuracy at the unit level. The weakness is that MES data is inherently production-scoped; it knows nothing about design intent, use-phase behaviour, or end-of-life decision logic unless explicitly integrated with other systems.

IoT and Edge Platforms for Use-Phase Feedback

Circularity in product-as-a-service models, or in any model where the manufacturer retains a relationship with the product during its use phase, depends on continuous or periodic telemetry. Vibration signatures, thermal cycles, load profiles, and failure events collected by embedded sensors or gateway-connected equipment feed back into the design and remanufacturing loops. An IoT platform such as Siemens MindSphere, PTC ThingWorx, or an open-source stack built on MQTT and time-series databases provides the ingestion, storage, and rule-engine layer for this data.

The workflow contribution is specific: telemetry triggers condition-based maintenance decisions, informs whether a returned unit is a candidate for remanufacturing versus recycling, and generates design feedback that closes the loop between field experience and next-generation CAD models. The challenge is data governance and interoperability. Telemetry formats are often proprietary to the sensor or gateway vendor. Normalising this data into a lifecycle context that a PLM or MES can consume requires transformation logic that is frequently underestimated in project scoping.

Digital Twin Architectures as Coordination Layers

A digital twin, in its most operationally meaningful sense, is not a 3D visualisation but a synchronised simulation model that reflects the current state of a physical asset and can be queried for what-if scenarios. For circular manufacturing, a product-level digital twin can integrate design geometry, material composition, production history, and use-phase telemetry into a single queryable entity. A recovery engineer can ask: given this serial number’s actual thermal history and the alloy composition recorded at production, what is the predicted remaining fatigue life, and does it meet the threshold for remanufacturing?

Platforms such as Dassault Systèmes’ 3DEXPERIENCE, Ansys Twin Builder, or open frameworks combining CAD kernels with physics solvers and data connectors enable this. The workflow advantage is decision support: the twin transforms scattered lifecycle data into a coherent analytical context. The disadvantage is maturity and cost. True operational digital twins that maintain synchronisation across the full lifecycle are still rare outside aerospace, automotive, and energy-sector applications. For a typical European SME, the investment in twin infrastructure may not be justified by the current volume of circularity decisions, and a well-structured data model with rule-based queries may deliver 80 percent of the value at 20 percent of the cost.

Distributed Ledger and Provenance Registries

Blockchain and other distributed ledger technologies have been proposed for circular manufacturing primarily as a provenance and trust mechanism: a tamper-evident record of material composition, production events, and chain-of-custody transfers that multiple stakeholders can verify without a central authority. Projects in the EU research landscape have explored DLT for material passports, supply-chain transparency, and cross-organisational certification of recycled content.

It is important to distinguish the stated scope of these research efforts from their operational readiness. DLT solves a trust and immutability problem, not a data-quality or workflow-orchestration problem. If the data written to the ledger is incomplete or inaccurate at the point of capture, the ledger faithfully preserves that inaccuracy. For circular manufacturing workflows, DLT is most relevant where multiple independent parties need shared assurance of a record — for example, a recycled-content certification that a downstream OEM must accept without re-testing. In single-enterprise or tightly coupled consortium contexts, a conventional database with role-based access control is usually simpler, cheaper, and sufficient. DLT should be considered when the trust topology genuinely requires it, not as a default.

ERP-Embedded Circularity Modules

Some enterprise resource planning vendors have begun offering circularity or sustainability modules that extend the traditional procurement-production-sales data model to include recovery, re-entry, and secondary-material tracking. SAP’s circular economy capabilities, for instance, allow a material flow to be tracked from primary procurement through production, sale, return, and re-introduction as a secondary input. The workflow advantage is financial and logistical coherence: the same system that calculates product cost can account for recovered material value, and the same inventory logic that manages primary stock can manage return-stock and remanufacturing queues.

The limitation is depth of technical lifecycle data. ERP systems are optimised for transactional and financial accuracy, not for storing CAD geometry, material microstructure data, or engineering change history. They work best as the coordination layer for material flows and commercial decisions, with PLM and MES providing the technical detail behind each transaction.

Comparison Criteria and Trade-Offs

Selecting among these architectures is not a matter of ranking them. It is a matter of matching the workflow pattern to the specific circularity model, organisational structure, data maturity, and integration budget at hand. The following criteria provide a structured way to evaluate fit.

Data scope and lifecycle coverage. Does the architecture cover only production, or does it span design, use, end-of-life, and re-entry? A PLM extension covers design-to-manufacture well but requires bolt-ons for use-phase and recovery. An IoT platform covers use-phase well but knows nothing about design intent. A digital twin aspires to full coverage but is operationally complex. The circularity model determines which phases are critical: a remanufacturing business needs production and end-of-life depth; a product-as-a-service model needs use-phase telemetry and design feedback.

Integration depth versus organisational reach. A tightly integrated PLM-MES-ERP stack gives excellent internal coherence but poor external accessibility. A federated architecture with standardised APIs and a shared data model (for example, using ISO 10303 STEP as an exchange format or IEC 61360 for attribute dictionaries) reaches across organisational boundaries but requires governance agreements and interface maintenance. The trade-off is control versus collaboration. A vertically integrated manufacturer with in-house remanufacturing can afford tight integration. A distributed supply chain with independent recovery partners needs federated reach.

Data quality dependencies. Every circular workflow is only as good as the data it operates on. A material passport that lists a nominal alloy grade but not the actual heat treatment is insufficient for remanufacturing decisions. A telemetry stream that has gaps or uncalibrated sensors produces unreliable condition assessments. The architectural choice must account for the data-capture infrastructure already in place and the investment needed to fill gaps. An MES-integrated approach assumes production-level data capture is already robust. A DLT approach assumes the data written to the ledger is trustworthy at the source. Neither assumption is automatic.

Scalability and cost of coordination. A workflow that requires manual data reconciliation between three systems at every product return will not scale beyond pilot volumes. Automated triggers, standardised interfaces, and event-driven architecture reduce coordination cost as volumes grow. But they also increase the upfront engineering investment. For an SME processing a few hundred returns per year, a well-designed spreadsheet-plus-PLM workflow may be more cost-effective than a full IoT-to-digital-twin pipeline. For a manufacturer handling thousands of units with complex remanufacturing decision logic, the automation investment pays for itself in reduced errors and faster throughput.

Maturity and vendor lock-in risk. PLM and ERP ecosystems are mature with established integration patterns. IoT platforms are broadly available but often proprietary in their data formats and analytics layers. Digital twin tooling is advancing rapidly but operational full-lifecycle twins remain a frontier capability. DLT for industrial provenance is still largely in research and early-adoption territory. The risk profile differs: mature systems carry integration and licensing costs but lower uncertainty; emerging architectures carry capability uncertainty but potentially lower lock-in if built on open standards.

Selection Guidance: A Decision Framework

Rather than a single recommended stack, a practical selection logic proceeds through the following questions:

First, define the circularity model. Is the goal material recovery and recycling, component remanufacturing, product refurbishment, or a full product-as-a-service lifecycle? Each model places different demands on the workflow. Material recovery needs composition data and chain-of-custody. Remanufacturing needs production history, design drawings, and condition assessment logic. Product-as-a-service needs use-phase telemetry, maintenance scheduling, and end-of-life decision triggers.

Second, assess the current data landscape. What systems already hold relevant data? Is production traceability in an MES or only in paper quality records? Are design data in a PLM or in CAD files on shared drives? Is use-phase data available at all? The workflow architecture must build on what exists, not assume a greenfield environment. Gaps in data capture are often the binding constraint, not the choice of platform.

Third, determine the organisational boundary. Is the circular loop internal (the same company designs, manufactures, and remanufactures), or does it span independent partners? Internal loops can use tighter integration and shared databases. External loops require standardised interfaces, agreed data schemas, and governance arrangements. The Digital Product Passport concept, as discussed in EU policy contexts, implies a cross-organisational data exchange requirement that pushes toward federated or registry-based architectures.

Fourth, evaluate integration effort against expected return. A full digital-twin implementation across the lifecycle may take multiple years and significant specialist capability. A PLM extension with structured material passport attributes and a simple API for downstream partners may deliver 70 percent of the operational value in a fraction of the time. Sequence accordingly: establish the data model and capture infrastructure first, add analytical and automation layers second, and pursue advanced simulation or DLT trust mechanisms only where a specific collaboration need demands them.

Fifth, plan for standards and interoperability from the outset. Even if the initial implementation is a single-enterprise workflow, designing data structures around recognised standards (STEP for product definition, IEC 61360 for attribute classification, ISO 22400 for manufacturing operations data) reduces future integration friction when regulatory requirements or partner expectations evolve. The cost of retrofitting a proprietary data model to a standard format is typically higher than designing to the standard from day one.

Implementation Consequences

Choosing an architecture is the beginning, not the end. The implementation phase reveals organisational and technical realities that the selection exercise can only approximate. Several consequences deserve explicit planning.

Data ownership and governance across the lifecycle is rarely straightforward. Who maintains the material passport when the product is in the hands of a third-party service provider? Who has authority to update the remanufacturing procedure after a field failure reveals a design flaw? Who bears the cost of sensor infrastructure in a product-as-a-service model? These are not technical questions but organisational ones, and they must be resolved in the workflow design phase, not discovered during pilot operations.

Skills gaps are a practical constraint. A circular manufacturing workflow that integrates PLM, MES, IoT analytics, and data governance requires people who can work across those domains. In practice, this means interdisciplinary teams: mechanical engineers who understand data structures, software developers who understand manufacturing processes, and sustainability analysts who can translate regulatory expectations into workflow requirements. For many European SMEs, building this capability internally is slow; partnering with a technical education and research institute, or engaging in R&D collaboration networks, accelerates the learning curve without requiring every competency to be hired permanently.

Change management in the recovery and remanufacturing segment is often underestimated. A disassembly technician at a recovery facility is not the same person who trained the PLM administrator at the OEM. The workflow must present information in a form that is actionable at the point of physical work: a clear disassembly sequence, a material identification that can be verified on-site, a decision tree for component disposition. If the digital workflow produces a 200-page PDF that requires a separate software licence to open, it will be ignored in practice. The interface design at the recovery end is as critical as the data architecture at the design end.

Validation and iteration are necessary. No circular manufacturing workflow is fully specified at the first design. Real products fail in ways the model did not anticipate. Material streams are contaminated in ways the composition data did not predict. Partners change their processes. The workflow architecture must accommodate updates without requiring a full system reimplementation. Versioned data models, modular interfaces, and feedback loops from operations to design are structural features, not afterthoughts.

Where Interdisciplinary Support Helps

The complexity of circular manufacturing workflows is not purely technical. It sits at the intersection of mechanical design, data engineering, supply-chain management, sustainability policy, and organisational change. A company attempting to implement a Digital Product Passport workflow, for example, needs CAD engineers who can structure design-for-disassembly data, software architects who can build the exchange interfaces, data analysts who can validate material composition records, and someone who understands the regulatory context well enough to scope what data is actually required versus what is aspirational.

This is the kind of interdisciplinary coordination that research and training institutes are positioned to support. Cademix Institute of Technology, operating in Austria with a focus on digital strategy, software, data analysis, CAD and 3D modelling, simulation, and technical documentation, can contribute to circular manufacturing workflow projects in several concrete ways: designing structured material-passport data models within CAD and PLM environments, building prototype data-exchange interfaces between production and recovery systems, developing training programmes that equip disassembly technicians and remanufacturing planners to work within digital workflows, and supporting R&D collaboration where a manufacturer needs to explore a specific workflow architecture before committing to a full implementation. The institute does not certify compliance or guarantee regulatory outcomes; it supports the technical and organisational groundwork that makes circular workflows functionable.

For a European SME evaluating its options, engaging with such a partner can reduce the risk of a costly architectural misstep — choosing a heavyweight digital-twin platform when a structured PLM extension would suffice, or building a proprietary data model that will require expensive rework when the Digital Product Passport standard matures. The value is not in the technology itself but in the interdisciplinary translation between technical possibility, operational reality, and regulatory direction.

Conclusion: Workflow Architecture as a Strategic Choice

Digital workflows for circular manufacturing are not a single technology purchase. They are an architectural decision about how lifecycle data flows, who owns it, how it is transformed into operational decisions, and how the system evolves as products, partners, and regulations change. The landscape of available approaches — PLM extensions, MES-integrated traceability, IoT telemetry platforms, digital twin coordination layers, DLT provenance registries, and ERP-embedded circularity modules — offers genuine alternatives rather than a single correct answer.

The right choice depends on the circularity model, the current data infrastructure, the organisational boundaries of the loop, the integration budget, and the maturity of the relevant standards. A remanufacturing-focused OEM with in-house recovery operations and an existing PLM-MES stack has a different optimal path than a distributed supply chain of independent SMEs trying to share material passports across borders. Neither is wrong. Both require that the workflow be designed as a living system with feedback, versioning, and clear ownership rather than as a static compliance document.

The practical starting point for most organisations is not the most advanced architecture but the most coherent data model: a structured, standards-aligned representation of what a product is made of, how it was made, how it is used, and what happens at end of life. Everything else — the IoT feeds, the twin simulations, the ledger entries — builds on that foundation. Get the data model right, integrate it with the systems that already exist, and add analytical and automation layers as the circularity operations scale. The workflow will grow into the architecture the organisation needs, rather than the architecture the vendor brochure promised.