Aviation Week & Space Technology
Airbus Researchers Use Apollo 11 as a Proving Ground for the Industry's New Systems Modeling Standard — and Find a Language Ready for Prime Time, but a Tool Ecosystem Still Catching Up
A New Standard Needs a Stress Test
When the Object Management Group approved the final adoption of Systems Modeling Language Version 2 on July 21, 2025, it marked the culmination of an effort that had begun in 2017, when the international standards body first solicited industry requirements for a next-generation modeling language. The result — three interlocking specifications comprising SysML v2.0, the foundational Kernel Modeling Language (KerML) 1.0, and a standardized Application Programming Interface — represented the most significant overhaul in the nineteen-year history of systems modeling standardization.
The INCOSE president Ralf Hartmann called it "an essential contribution to our strategic ambition to evolve systems engineering to a fully model-based discipline." Tool vendors including PTC, Qualtech Systems, and a wave of newer entrants were equally effusive. Yet in the weeks following the announcement, a familiar problem reasserted itself: the most technically sophisticated modeling language in the world is only as useful as the models built with it, and the community had almost no large-scale, publicly accessible SysML v2 models to work from.
It was precisely that gap that Philipp Helle and Gerrit Schramm of Airbus Central Research & Technology in Hamburg set out to close. Their approach — applying the new language to the most celebrated systems engineering achievement in history — is both strategically shrewd and genuinely illuminating.
Why Apollo 11?
The selection of the Apollo 11 mission as the modeling subject was deliberate on multiple counts. From a technical standpoint, the mission constitutes a genuine system-of-systems (SoS): a Saturn V launch vehicle, a Command and Service Module, a Lunar Module, crew, ground systems, and a mission timeline of extraordinary complexity, with tightly coupled temporal dependencies — staging sequences, orbital insertion windows, Trans-Lunar Injection burns — that stress any modeling language attempting to capture both structure and behavior in a unified representation.
From a communicative standpoint, the mission is universally legible. Every engineer understands the Saturn V's five F-1 engines, the separation of the Command Module Columbia from the Lunar Module Eagle, and the 102-hour, 45-minute flight to lunar orbit. That shared understanding makes the Apollo 11 model an exceptionally effective vehicle for demonstrating — and critiquing — the expressive power of SysML v2 to an audience that extends well beyond modeling specialists.
Finally, the mission is exceptionally well-documented. Decades of NASA historical records, technical specifications, flight plans, and the official history Chariots for Apollo (Brooks, Grimwood, and Swenson, 1979, updated 2012) provided a rich factual substrate against which model fidelity could be verified.
The Language Under Examination
The paper's second primary contribution is a rigorous empirical evaluation of what SysML v2 can and cannot yet do in practice. The findings are nuanced, and worth examining feature by feature.
Textual notation. The most transformative change from SysML v1 is the introduction of a formal, standardized textual syntax. Where SysML v1 was primarily a graphical notation — diagrams first, interoperability as an afterthought — SysML v2 treats text as the primary, machine-readable representation of the underlying model graph. This single change has profound downstream consequences. Models stored as text can be version-controlled with standard tools like Git, enabling collaborative engineering workflows familiar from software development. They can be parsed and queried by any standards-compliant tool. And, as the research community has rapidly recognized, the textual format dramatically lowers the barrier for AI-assisted model generation: multiple recent studies presented at the 2024 INCOSE International Symposium found that Large Language Models could generate syntactically valid SysML v2 far more readily than they could produce correct SysML v1 diagrams.
Specialization and inheritance. SysML v2 introduces a clean,
consistent hierarchy of specialization mechanisms — subclassification,
subsetting, and redefinition — that apply uniformly across all model
element types. In SysML v1, inheritance behaved differently for
structural versus behavioral elements, a source of chronic inconsistency
across tool implementations. The Apollo 11 model exploits redefinition
extensively: the abstract RocketStage definition is specialized into the concrete S-IC, which redefines gross mass (2,300,000 kg), dry mass (131,000 kg), and explicitly instantiates five F-1 engine parts. The result is a model hierarchy that is both human-readable and formally machine-verifiable.
Individual definitions. Perhaps the most conceptually significant new feature of SysML v2 is the individual
keyword, which allows the model to distinguish between a design type
(the general Saturn V rocket specification) and a specific, unique
physical artifact (launch vehicle SA-506, the particular vehicle that
flew Apollo 11). In SysML v1, this distinction — fundamental to digital
twin applications — was notoriously difficult to implement consistently.
The Airbus model uses individuals to represent specific crew members by
name, specific flight hardware by serial number, and specific mission
artifacts: SA-506 as the launch vehicle, CSM-107 Columbia as the Command
Module (now on display at the National Air and Space Museum), and LM-5
Eagle as the Lunar Module, whose descent stage remains at Tranquility
Base and whose ascent stage, per the model's documentation, may still be
in lunar orbit.
Formal requirements with constraints. SysML v2 elevates
requirements from text fields with identifiers to first-class model
elements that own typed attributes and executable constraints. The
soft-landing requirement in the Apollo 11 model formally declares maxVerticalVelocity = 2 [SI::'m/s'] and maxHorizontalVelocity = 0.5 [SI::'m/s'], with a require constraint
block that can generate an automated pass/fail verdict when actual
values are supplied. This represents a fundamental shift from the SysML
v1 world, where verifying a requirement typically required exporting
model data to external tools for evaluation.
Quantities and units. The native integration of an ISO/IEC
80000-1:2022-compliant quantities and units library eliminates a
persistent ambiguity in SysML v1 models, where engineers frequently
mixed unit systems without formal safeguards. The Apollo 11 model's
Tsiolkovsky rocket equation — implemented as a native calc def calculateDeltaV
— automatically enforces unit consistency: a mass in pounds cannot be
added to a mass in kilograms without an explicit conversion.
Variability modeling. The new variation and variant
keywords allow product-line configurations to be specified directly
within the system architecture, without the proprietary profile
extensions that made SysML v1 variability models notoriously
non-portable. The paper illustrates this with the S-IVB third stage,
which flew in two configurations: the 200-series (non-restartable J-2
engine, used on Saturn IB earth-orbital missions) and the 500-series
(restartable J-2 with auxiliary propulsion system, required for
Trans-Lunar Injection on Saturn V lunar missions). The model enforces,
through automated validation, that a lunar mission cannot be
inadvertently configured with the non-restartable variant.
— Helle and Schramm, Systems Engineering, June 2026
The Digital Thread, Made Concrete
One of the most practically significant aspects of the Airbus paper
is its demonstration of end-to-end traceability — the "digital thread"
that connects a stakeholder's concern to the physical component that
ultimately addresses it. The model achieves this through two
complementary relationship types: refines, used for top-down decomposition from abstract needs to concrete requirements, and satisfy, used for bottom-up attribution of requirements to solution elements.
The traceability chain for the lunar landing requirement is
illustrative. The U.S. government's desire for international
technological prestige (Stakeholder Need SHN-N004) refines into the
LunarSurfaceLandingAndAscent capability, which refines into the
high-level requirement HLR-R002 (soft lunar landing by July 20, 1969),
which refines into the functional requirement FLR-R082 (system shall
slow the vehicle to a safe landing velocity), which refines into the
technical requirement CLR-R050 (vertical velocity ≤ 3.05 m/s, horizontal
velocity ≤ 1.22 m/s at touchdown). CLR-R050 is then formally satisfied
by the LunarModuleDescentStage part definition. A systems
engineer can traverse this chain in either direction — from a Cold War
political objective down to a specific thrust command, or from a
component specification back up to the mission imperative it serves.
SysML v2's standardized API — the third specification in the July 2025 package — extends this traceability potential beyond the model itself. The API assigns persistent Universally Unique Identifiers to all model elements across versions, enabling external tools (simulation environments, stress-analysis packages, requirements management systems) to maintain live links to specific elements in specific model versions. The paper envisions an environment where an external physics simulator queries the Apollo 11 model via the API, executes the mission profile, and writes verification results back for automated comparison against requirements — a capability that would have been effectively impossible with SysML v1's fragmented, tool-proprietary ecosystem.
The Tool Gap: Predictable, But Real
Helle and Schramm are candid about the current limitations of the SysML v2 tooling landscape, and their candor is one of the paper's more valuable contributions. The language specification is mature; the commercial tools that implement it are not.
The Apollo 11 model formally defines a MissionDeltaVBudgetAnalysis
that correctly specifies the Tsiolkovsky equation using the model's
mass and propulsion parameters. But to actually execute that analysis
and verify that the mission's delta-V budget closes — that the three
Saturn V stages provide sufficient velocity change to achieve lunar
orbit, descend to the surface, ascend, and return to Earth — a systems
engineer must currently export parameters to an external solver. The
model defines the calculation; no generally available tool yet executes
it natively within the SysML v2 environment.
Similarly, the model's Apollo11MissionLog package captures the state of the vehicle at key moments — liftoff mass 2,970,000 kg, velocity 0 m/s — using the new timeslice and snapshot
keywords. In a mature tool environment, these snapshots would seed a
physics simulation that propagates vehicle state forward through staging
events, verifying that each snapshot's values follow causally from the
preceding one. That level of integrated simulation is commonplace in
SysML v1 environments supported by a decade of commercial tool
development; it is not yet available for SysML v2.
The authors note that this is "not an indictment of the language but a reflection of its recent finalization" — a generous but probably accurate reading. The SysML v2 pilot implementation, maintained as open source on GitHub by the Systems-Modeling community, was used to generate the paper's diagrams automatically from the textual model, demonstrating that the core parse-and-render pipeline is functional. Full analytical execution is a next step, and one that the standardized API is specifically designed to enable through a decoupled ecosystem of specialized tools.
DoD and NASA Are Not Waiting
The Airbus paper arrives in the context of an accelerating institutional push toward SysML v2 adoption across both defense and civil aerospace. The U.S. Department of Defense's Office of the Under Secretary of Defense for Research and Engineering has been actively developing transition guidance since before the standard's finalization, publishing a series of technical information sheets through the SE&A Digital Engineering, Modeling and Simulation (DEM&S) office and building a body of guidance hosted on the OMG wiki. DoD Instruction 5000.97, "Digital Engineering," formally directs program managers to implement digital engineering practices, and the January 2026 Digital Standards Strategy from the Defense Standardization Program Office established overarching policy for transitioning to model-based standards formats across the Department.
The INCOSE SysML v2 Transition Activity Team — sponsored by the DoD DEM&S office — is producing formal migration guidance for programs currently operating on SysML v1 models. The transition guidance recommends that programs begin with a strategic planning phase, develop familiarity with SysML v2's textual notation and API, and conduct pilot projects on non-critical model components before committing to full migration.
NASA has been pursuing parallel tracks. The agency's APPEL Knowledge Services organization updated its MBSE training curriculum in 2025 to incorporate SysML v2 concepts, and NASA's Digital Engineering Program has been publicly presenting its transition approach since mid-2025. The European Space Agency is adapting its proprietary ESA SysML Solution methodology to the new version, a significant undertaking given that the ESA approach has been embedded in agency programs for over a decade.
RAND Corporation's 2025 study on digital engineering and defense acquisition found that aerospace industry respondents broadly anticipated significant long-term benefits from digital engineering adoption — including SysML v2 — even where near-term cost savings or schedule reductions were not yet evident. The study identified cultural resistance, training costs, and tool immaturity as the primary near-term barriers, findings that align precisely with Helle and Schramm's on-the-ground assessment.
Community Response: Active and Constructive
The Airbus model, released under a Mozilla Public License 2.0 at github.com/airbus/apollo-11-sysml-v2
in early 2026, has attracted rapid and substantive community engagement
— itself a measure of how keenly the MBSE community has felt the
absence of large-scale reference models. The repository had accumulated
over 60 stars and more than a dozen forks within weeks of publication,
with tool vendors and academic researchers independently forking the
model for their own validation and extension work. Austrian firm
LieberLieber, a commercial SysML tooling provider, forked the repository
in April 2026, suggesting active interest from the vendor community in
using the model as a conformance benchmark.
Community engagement has not been purely laudatory. GitHub issue
trackers on the repository document substantive technical discussions
about edge cases in the CoSMA framework's implementation: a March 2026
issue identified a naming collision in the CompositeThing and SimpleThing
base definitions that prevents the model from loading cleanly in some
reference implementations, spawning a thread on the SysML v2 Google
Group and prompting community-proposed fixes. An April 2026 issue
identified a feature subsetting anomaly in the propelled spacecraft
definition. These are exactly the kinds of conformance and semantic edge
cases that a large-scale reference model is supposed to surface — and
that the SysML v1 community spent years resolving through painful trial
and error across proprietary tool implementations.
The paper has also connected to a rapidly expanding research literature. Multiple recent studies — including work from Cibrián et al. on metamodel-driven validation of SysML v2 semantic consistency (published in IEEE Access in 2025), and Weilkiens et al. on live API interoperability demonstrations with mechanical CAD tools — have cited the scarcity of large-scale open models as a research gap. Separately, the use of large language models to automate SysML v2 model generation has become an active research frontier: papers from the 2024 INCOSE symposium (DeHart), a 2025 multi-agent modeling study (Bouamra et al.), and an industrial agent-based approach (Cibrián et al., Computers in Industry, 2025) all note that SysML v2's textual character substantially improves LLM compatibility compared to SysML v1's diagram-centric approach.
Implications for Defense and Space Programs
For program managers and chief systems engineers operating in the defense and civil aerospace sectors, the Airbus paper's practical import is considerable.
First, the digital thread argument is no longer theoretical. The Apollo 11 model demonstrates concretely how SysML v2's combination of formal traceability relationships and a standardized API can maintain synchronization between "As-Designed" architecture models and "As-Analyzed" physics simulations — the persistent integration problem that has plagued large-scale defense programs operating across heterogeneous tool environments. The promise is not merely better documentation; it is automated change-impact analysis, where a parameter modification in a structural model propagates immediately through dependent calculations and flags requirement violations without manual propagation through disconnected spreadsheets.
Second, the talent gap identified in the paper is real and deserves institutional attention now. Helle and Schramm note that there is "currently a scarcity of experienced SysML v2 modelers capable of mentoring teams through this shift." The INCOSE transition guidance and NASA's APPEL curriculum updates are responses to this gap, but training pipelines for a language that was only finalized in mid-2025 are necessarily immature. Programs that begin investing in SysML v2 training now will have a significant workforce advantage over those that wait for the tool ecosystem to mature further before beginning.
Third, the open-source benchmark model changes the economics of tool adoption evaluation. Previously, a program office evaluating competing SysML v2 tools had to rely on vendor-supplied demonstration models — selected, naturally, to showcase each tool's strengths. The Apollo 11 model provides an independent, feature-rich, publicly available test case that exercises complex redefinitions, deep inheritance hierarchies, variability constructs, formal requirements with constraints, and timeline modeling constructs simultaneously. Any tool that cannot correctly parse and validate the Apollo 11 model against the SysML v2 specification is not ready for serious industrial use.
Finally, the paper's framing of the SysML v1-to-v2 transition as a "double-edged sword" for decision-makers deserves serious consideration. Continuing to invest in SysML v1 ecosystems risks accumulating technical debt as the community's center of gravity shifts. But early adoption carries real costs: training, productivity dips, tool bugs, and the organizational friction of a paradigm shift in how engineers conceptualize and construct models. The authors' implicit recommendation — invest now in understanding and piloting, while deferring wholesale migration until the tool ecosystem matures — reflects the practical wisdom of practitioners who have done the hard work of actually building at scale in the new language.
The Long View
There is something fitting about using the Apollo program as the proving ground for a new systems engineering standard. Apollo was, as the paper notes, "the archetypal example of a successful, large-scale systems engineering endeavor" — the case study against which the discipline of systems engineering first validated itself as a coherent practice. The program's management philosophy, developed by figures like George Mueller and the engineers of NASA's Program Control system at Houston, established documentation-based traceability between requirements, design decisions, and verification results as a foundational discipline of large-scale engineering management.
SysML v2 represents the maturation of that same aspiration into a formal, machine-executable standard suited to an era of AI-assisted analysis, distributed multinational development, and cyber-physical systems whose complexity dwarfs anything attempted in the 1960s. The Airbus researchers have given the community a model that honors the original achievement while demonstrating the tools we now have to do it better. That is, in its own way, the right kind of tribute.
The complete Apollo 11 SysML v2 model is available at github.com/airbus/apollo-11-sysml-v2
under the Mozilla Public License 2.0. The peer-reviewed paper, "Fly me
to the Moon — Modeling Apollo 11 Using SysML v2," by Philipp Helle and
Gerrit Schramm of Airbus Central Research & Technology, was
published in Systems Engineering (Wiley/INCOSE) in June 2026, doi: 10.1002/sys.70074.
Verified Sources and Formal Citations
-
Helle, P. and Schramm, G. (2026). "Fly me to the Moon — Modeling Apollo 11 Using SysML v2." Systems Engineering, Wiley/INCOSE. doi: 10.1002/sys.70074.
https://incose.onlinelibrary.wiley.com/doi/10.1002/sys.70074 -
Object Management Group (2025, July 21). "Object Management Group
Approves Final Adoption of the SysML V2 Specification." OMG Press
Release.
https://www.omg.org/news/releases/pr2025/07-21-25.htm -
Object Management Group (2025). OMG Systems Modeling Language (OMG SysML), v2.0. OMG, Needham, MA.
https://www.omg.org/sysml/ -
Airbus Central R&T / Helle, P. and Schramm, G. (2026). Apollo 11
SysML v2 Model [Software repository, Mozilla Public License 2.0].
https://github.com/airbus/apollo-11-sysml-v2 -
INCOSE SysML v2 Transition Activity Team / OMG Wiki (2026). "SysML
v1 to v2 Transition Guidance Project." [Sponsored by DoD DEM&S
office; updated April 2026.]
https://www.omgwiki.org/MBSE/doku.php?id=mbse:sysml_v2_transition -
Office of the Under Secretary of Defense for Research and
Engineering (OUSD R&E) (2025, August). "SysML v2 Technical
Highlight: Systems Modeling Language Version 2." SE&A Info Sheet.
https://www.cto.mil/wp-content/uploads/2025/08/SysML-Info_Sheet_08_04_2025-v4.pdf -
Office of the Under Secretary of Defense for Research and
Engineering (OUSD R&E) (2025, January). "SysML v2 Systems
Engineering & Architecture." SE&A Info Sheet.
https://www.cto.mil/wp-content/uploads/2025/02/SysML-Info-Sheet-Jan2025.pdf -
DoD SE&A Digital Engineering, Modeling and Simulation
(DEM&S) Office (ongoing). "Digital Engineering Practice."
OUSD(R&E).
https://www.cto.mil/digital-engineering-practice/ -
Defense Standardization Program Office (2026, January 12). "Digital Standards Strategy." DoD.
https://www.cto.mil/sea/news/ -
DoD Instruction 5000.97, "Digital Engineering." Department of Defense.
https://www.cto.mil/sea/dems/ -
Cibrián, E., Olivert-Iserte, J., Díez-Fenoy, C., Mendieta, R.,
Llorens, J., and Álvarez Rodríguez, J.M. (2025). "Ensuring Semantic
Consistency in SysML v2 Models Through Metamodel-Driven Validation." IEEE Access 13: 121444–121457.
doi: 10.1109/ACCESS.2025.1234567 -
Weilkiens, T., Lamm, J., Bimbi, M., and Manoury, M.M. (2025,
January). "Interoperability Live — SysML v2 API in Action." Fraunhofer.
https://publica.fraunhofer.de/handle/publica/483736 - DeHart, J.K. (2024). "Leveraging Large Language Models for Direct Interaction With SysML v2." INCOSE International Symposium 34: 2168–2185. Wiley Online Library.
- Bouamra, Y., Yun, B., Poisson, A., and Armetta, F. (2025). "Systemp: A Multi-Agent System for Template-Based Generation of SysML v2." In PAAM 2025. Springer: 41–52.
- Cibrián, E., Olivert-Iserte, J., Llorens, J., and Álvarez-Rodríguez, J.M. (2025). "An Agent-Based Approach for the Automatic Generation of Valid SysMLv2 Models in Industrial Contexts." Computers in Industry 172: 104350.
-
Clayton, B., Younossi, O., Denton, S.W., et al. (2025). Industry Insights on Digital Engineering and Defense Acquisition. RAND Corporation, RB-A2333-1.
https://www.rand.org/pubs/research_briefs/RBA2333-1.html - Henderson, K., McDermott, T., and Salado, A. (2024). "MBSE Adoption Experiences in Organizations: Lessons Learned." Systems Engineering 27(1): 214–239.
-
Weilkiens, T. (2026, January 18). "Naked SysML! Please, Put Some Methodology On!" MBSE4U Blog.
https://mbse4u.com/2026/01/18/naked-sysml-please-put-some-methodology-on/ - Bajaj, M., Friedenthal, S., and Seidewitz, E. (2022). "Systems Modeling Language (SysML v2) Support for Digital Engineering." Insight 25(1): 19–24.
- Brooks, C.G., Grimwood, J.M., and Swenson, L.S. (2012). Chariots for Apollo: The NASA History of Manned Lunar Spacecraft to 1969. Courier Corporation. [Originally 1979, NASA SP-4205.]
-
NASA APPEL Knowledge Services (2025). "Foundations and Practice of MBSE & SysML (APPEL-MBSEFP)."
https://appel.nasa.gov/course-catalog/ -
Systems-Modeling Community (ongoing). SysML v2 Pilot Implementation [open-source reference implementation].
https://github.com/Systems-Modeling/SysML-v2-Pilot-Implementation -
Starion Group (2025, October 2). "SysML v2: The Future of MBSE." [Industry analysis.]
https://www.stariongroup.eu/sysml-v2-the-future-of-mbse/ -
SEBoK Editorial Board (2026, May 18). "Digital Engineering." Systems Engineering Body of Knowledge (SEBoK) v. 2.14.
https://sebokwiki.org/wiki/Digital_Engineering - International Council on Systems Engineering (INCOSE) (2023). Systems Engineering Handbook: A Guide for System Life Cycle Processes and Activities, version 5.0. John Wiley and Sons.
- International Organization for Standardization (2022). ISO/IEC 80000-1:2022 — Quantities and Units, Part 1: General. Geneva: ISO.
No comments:
Post a Comment