Skip to main content

Digital thread, data and the supply chain

📍 Where we are: Part V · Industrial Reality — Chapter 13 of 19. The previous chapter showed why qualification is the central bottleneck; this one follows the data that qualification depends on, and the supply chains that data is meant to unlock.

The simple version

Think of a restaurant recipe that has to be cooked identically in a hundred kitchens. The recipe is not just a list of ingredients: it also says which oven, how hot, how long, and who checked the result. Additive manufacturing (AM) is like that, but the "recipe" is scattered across the designer's CAD file, the slicer, the machine, the sensors, the inspection lab, and a customer's database, each in its own language. The digital thread is the idea of one unbroken string connecting all of it. Today the string is mostly knotted, and the parts that cross company borders are the most tangled.

What this chapter covers​

We define the digital thread for AM, review the standards and data models that try to implement it, and then explain the structural reasons it stays incomplete: data volume, proprietary formats, trust between companies, and security. We then turn to the supply-chain promise built on the thread, spare parts on demand and distributed manufacturing, and ask where evidence supports it. Evidence levels are marked as lab-scale, production, or contested.

Problem​

A conventional machined part is described by a drawing, a material certificate and an inspection report. An AM part is also defined by the process history of its own build: powder lot and reuse count, machine and its calibration state, scan strategy, layer-by-layer sensor signals, post-processing, and inspection. Because properties depend on the process, that history is part of the evidence that the part is good (see qualification and certification).

Two quantities make this a data problem rather than a paperwork problem. First, volume: a NIST report cites an estimate of about 3000 TB of data for the qualification of one additively built aircraft component, and in-situ monitoring of powder bed fusion can generate data at gigabytes per second [1]. Second, fragmentation: the data comes from design, materials, machine, process and inspection domains, each with its own informatics systems, which the same report identifies as a technical barrier to finding, integrating and comparing data [1].

The digital thread is the connected, traceable record that links these domains for a given part [2] (the AM-informatics overview that introduced the term to AM; we found no more highly cited AM-specific definition). The same idea underpins the supply-chain vision: if a validated part definition and its process recipe can be transmitted faithfully, the part can in principle be made where it is needed.

State of the art​

Formats and data models​

Demonstrated at lab and pilot scale; partially in production. The workhorse exchange format, STL, carries only a triangle mesh. Successors aim to carry materials, colours, and metadata. The 3MF Consortium announced in July 2025 that its XML-based format has been recognised as an ISO standard (ISO/IEC 25422:2025) [3]; the claims about regulated-sector adoption and traceability benefits come from the consortium itself and should be read as vendor-side statements [4]. A 2025 review of STEP-NC (the ISO 14649 / STEP AP238 family) for AM argues that the core problem is still the lack of standardised data exchange between lifecycle stages, with reliance on legacy formats such as STL and G-code causing information loss, and proposes an architecture and data model [5].

Data models go further than file formats. NIST and partners describe a common data dictionary (ASTM F3490), a common data model, and exchange formats, with classes for in-situ monitoring devices, reference frames, build simulation and analysis results, framed around making AM data findable, accessible, interoperable and reusable (FAIR) [1]. A NIST-led XML "data pedigree" with five models (project management, feedstock, build and post-processing, microstructure and property measurement, simulation) was tested with laser powder bed fusion data from America Makes; this is a proof of concept, not an adopted industry schema [6].

Standards also set expectations on the production side. ISO/ASTM 52920:2023 defines qualification requirements for industrial AM processes and production sites, layered on top of a quality management system [7]. It is the reason sites need disciplined records in the first place.

Supply chain concepts​

Mostly modelling and case studies; limited production evidence. The foundational model of AM spare-part supply (Khajavi and colleagues, using an aircraft environmental control system as the case) concluded that distributed production was held back by slow build rates relative to shipping from a central warehouse, high machine investment, and labour-intensive steps [8]. Earlier conceptual and modelling work was similarly cautious: Holmström and colleagues judged centralised deployment by the original manufacturer, or deployment by service providers near the point of use, the most feasible, and Liu and colleagues' model of aircraft spares found reduced safety inventory to be the main opportunity [9] [10]. A system-dynamics comparison of conventional, centralised-AM and distributed-AM spare-part supply chains found the AM-based chains better on total variable cost and carbon emission in its modelled scenarios [11]; that is a simulation, not an operating record. A 2023 case study of a manufacturer and a customer identifies both advantages and barriers of AM for spare parts [12].

Digital thread across the AM lifecycle with data silos, trust boundaries and the print-anywhere supply chain. The digital thread links design, build, inspect and certify records; the gaps (proprietary formats, trust boundaries between companies, file security) sit exactly where data must cross between systems or organisations. Original diagram by the authors, created with AI assistance.

Why it is hard​

Interoperability is a many-to-many problem. Each machine vendor, software tool and inspection system produces its own data. A 2024 NIST workshop report on small and medium-sized enterprises (SMEs) lists, among pain points, process qualification and part certification, access to machine vendor software, and control of intellectual property; it also quotes aerospace and defense survey barriers to digital-thread adoption: IP protection, system interoperability, lack of real-time supplier and production data, and the high cost of curating and maintaining data quality [13].

Data are not just big, they are unlabelled. Linking in-situ signals to a physical location in the finished part requires registration in time and space, and the metadata that makes the alignment possible. Without it a terabyte of sensor logs cannot support a certification argument.

Trust boundaries are contractual, not technical. Whoever owns the build parameters owns what is effectively the process know-how. Suppliers, OEMs and customers disagree about who may see what; this is the "control of IP" pain point above [13].

Files are physical weapons. In AM, a file is both intellectual property and an instruction to make a safety-relevant object. An overview of cyber-physical risks in AM evaluates two classes of malicious modification of the AM process [14]. A human-subjects study of an attack on the STL file of a test specimen tested whether operators could detect and diagnose it and proposed prevention and detection measures [15]. The dr0wned experiment showed, at lab scale, that a design file on a compromised PC could be replaced with a version that shortened the fatigue life of a 3D-printed quadcopter propeller; it failed in flight [16]. A 2024 analysis of one commercial FDM printer found missing encryption, weak integrity checking and exploitable proprietary protocols, permitting model sniffing and replacement [17]. A 2025 survey proposes an attack taxonomy for the whole cyber-physical AM supply chain, listing IP theft, design manipulation and counterfeit production as principal threats [18]. These are research demonstrations; we are not aware of verified public data on how often such attacks occur in production.

Spare-part economics are unforgiving. Printing a spare near the point of need competes with warehouse shipping, which is cheap and fast, and with a qualification burden per part, per machine and per site [8].

Open questions​

  1. A shared, adopted data schema. Are the proposed data dictionaries and models (ASTM F3490-based, STEP-NC, 3MF extensions) going to converge, or will each large integrator keep a private model with translators? Contested.
  2. What is the minimum thread for certification? Which in-situ and inspection data must be stored, for how long, to support a qualification argument, and which can be reduced to summary statistics? This links directly to in-situ monitoring.
  3. Protecting IP while sharing process data. Can a customer verify a part was built within a validated window without revealing the full parameters? Technical proposals exist, but we found no evidence of routine production use.
  4. Securing the file path end to end. Encryption and signed files are straightforward in principle; retrofitting installed machines with weak protocols [17] is not.
  5. Where does "print anywhere" actually pay? The modelling and case-study literature finds AM spare-part chains competitive in some scenarios but not others; the size of the subset where it pays is contested and is rarely measured in operation [11][12].

Industry impact​

Evidence: mixed, mostly organisational rather than technological. The data-integration problems cost most for smaller companies, which may lack resources for a full digital thread and report cost as a significant pain point when working with large system integrators [13]. That favours vertically integrated or single-vendor ecosystems, with the risk of lock-in.

In government sustainment, a 2019 DoD Inspector General audit reported that military depots and field locations were already using AM, partly to offset obsolete parts, but that DoD lacked a standard policy for cataloguing AM parts (the report contrasts how two commands assign part numbers) [19]. The bottleneck there was not the printer but administrative data: part numbers, technical data rights and approved sources. We did not independently verify how this has changed since 2019.

The honest summary: the digital thread is a necessary enabler of qualified, distributed AM, and standards are maturing, but a seamless, trusted, cross-company thread is not yet routine production practice. The closing chapter, the honest verdict, ranks this among the industry's unsolved problems.

Why it matters​

Qualification, supply-chain resilience and sustainability claims all rest on records. Without a connected thread, every part must be proven from scratch, and "print on demand" collapses into "re-qualify on demand".

In the real world​

Standards bodies and consortia (ISO/ASTM, the 3MF Consortium, NIST) are all publishing data-format and data-model work; aerospace and defense users cite IP protection and interoperability as the main adoption barriers; and military maintainers use AM for obsolete parts while struggling to catalogue them [13][4][19].

Key terms​

  • Digital thread — the connected, traceable record linking design, material, build, inspection and certification data for a part.
  • Interoperability — the ability of different vendors' software and machines to exchange data without losing meaning.
  • FAIR data — data that is findable, accessible, interoperable and reusable.
  • STL — a legacy mesh-only file format that carries geometry but not materials or process settings.
  • 3MF — an XML-based AM file format, recognised as ISO/IEC 25422:2025, able to carry materials and metadata.
  • STEP-NC — an ISO standard family for describing manufacturing operations and product data in one model.
  • Data pedigree — the documented lineage of a dataset: where it came from and how it was processed.
  • Digital spare part — a stored design and process definition from which a replacement part can be printed on demand.
  • Distributed manufacturing — producing parts at several locations close to demand from shared digital definitions.
  • AM sabotage — deliberate alteration of a design or process so that a printed part fails early.

Where this leads​

Data and logistics also determine environmental claims: whether AM is greener depends on measured energy, material and transport, which is the subject of sustainability and life-cycle assessment.