Satellite Systems Engineers: AI Requirements Verification in 2026

Satellite Systems Engineers: AI Requirements Verification in 2026
Contents
  1. Why a satellite requirement is never verified just once
  2. The verification methods and where each goes stale
  3. Mass, power, thermal and structural margins as live attributes
  4. System Architecture interfaces: where a requirement meets the part that satisfies it
  5. When a CAD change moves mass: change impact across requirements, margins and evidence
  6. How Tandem fits in 2026
  7. A verification checklist for your next design review
  8. Make your next review start from live attributes

Satellite requirements verification is only as current as the last input that fed it. The four verification methods (test, analysis, review of design and inspection under ECSS, or test, analysis, inspection, and demonstration in NASA's handbook) all produce evidence at a moment in time, and a satellite's mass, power, thermal and structural numbers keep moving after that moment. The fix is to verify requirements continuously against live attributes pulled from CAD, spreadsheets and telemetry. Stop treating the verification matrix as a document you close out at review.

This piece is for the systems engineer who owns the matrix and the mechanical, thermal and structural engineers who feed it. It covers where each method goes stale, how to hold margins as live attributes, and how Tandem keeps requirements, architecture, CAD changes and evidence linked so a change shows up in the matrix the day it happens.

Why a satellite requirement is never verified just once

A satellite requirement is verified against a design, and the design does not sit still. A bracket gets thicker after a structural analysis, a battery cell swaps vendors, a radiator grows by 40 mm. Each of those edits can invalidate evidence you already signed off.

The Verification Control Document records, for every requirement, the selected method, level and stage (ECSS-E-ST-10-02C, 2018 Rev.1). That record is a snapshot. It says how you intended to prove compliance and what you concluded on a given date. It has no idea the mass budget moved last Tuesday.

Treat verification as a state, not an event. A requirement is verified while the attributes it depends on still match the evidence behind it. The moment one drifts, the status should flip, and someone should be told which requirement and which piece of evidence is now in question.

The verification methods and where each goes stale

There are four basic methods of requirement verification. ECSS names them test, analysis, review of design and inspection, ordered roughly from greatest to least confidence (ECSS-E-ST-10-02C, 2018 Rev.1). NASA's Systems Engineering Handbook lists test, analysis, inspection, and demonstration (NASA, 2020). The sets overlap heavily, and a requirement can use a combination. For a longer walk through each, see the guide on verification methods: test, analysis, inspection and demonstration.

Here is where each one decays on a satellite program:

Test goes stale when the article under test is no longer the flight configuration. A vibration test on a structural model says little once a mass-model change shifts the center of gravity.

Analysis goes stale fastest. A thermal or structural model is built from inputs: dimensions, material properties, dissipated power. Change any input and the result is a number about a design that no longer exists.

Review of design goes stale when the drawing or report it cites is superseded. The approval stays in the matrix while the document underneath it is revised.

Inspection goes stale when the inspected feature changes after the inspection record is filed.

Analysis and review of design carry the most risk, because their evidence is derived from other data. Link those requirements to the source attributes, not just to a PDF.

Mass, power, thermal and structural margins as live attributes

Margins are the cleanest case for continuous verification, because each one is a number compared against a limit. NASA guidance treats mass and power margin as key technical performance measures tracked throughout development, with margin defined as allowance for uncertainty and risk (NASA, 2020).

The practical move is to store each margin as an attribute with a live source instead of a typed value in a slide. Mass comes from the CAD assembly. Power comes from a load-and-generation sheet. Thermal and structural results come from analysis outputs. After launch and during ground test, telemetry supplies measured values that replace estimates.

In Tandem, requirements are verified continuously against live attributes from CAD, spreadsheets and telemetry. For the mechanics of building these budgets, see mass and power budget tracking for hardware programs.

Test and telemetry data often lives in dedicated platforms. Tandem works alongside them. Their measured data can feed Tandem attributes that verify requirements, so you keep your test tooling and still close the loop to the requirement.

System Architecture interfaces: where a requirement meets the part that satisfies it

A requirement is only checkable if you know which part satisfies it. In Tandem's System Architecture module, you model system, subsystem and component nodes in tree and canvas views, with interfaces between subsystems, tags and custom attributes such as mass, cost and status.

That structure is where verification becomes concrete. A top-level requirement for total on-orbit mass flows down to subsystem allocations. The power subsystem node carries its own attributes, and the interface between the power subsystem and the avionics box carries the load it must support. When a requirement links to a node and its attributes, a change to that node reaches the requirement by structure, not by someone remembering.

Architecture also supports configurations and branching in a git-style way, which suits the reality that you carry a baseline and a candidate design side by side. You can see what a requirement looks like against the candidate before committing to it.

The agents in Tandem use this same graph for change-impact analysis. They search, create and update architecture nodes, requirements and attributes, and run impact analysis across them. Tandem is building toward more autonomous, program-wide automation on top of this graph, but the linked structure is what makes any of it trustworthy.

When a CAD change moves mass: change impact across requirements, margins and evidence

Take a plain scenario, offered as a hypothetical. A mechanical engineer thickens a structural panel to recover a stiffness margin. In CAD, the panel's mass goes up. That single edit now touches a mass requirement, the margin against it, the structural analysis that justified the change, and possibly a center-of-gravity requirement.

Without links, someone finds out at the next budget review, or at the next design review, or not at all. With links, the CAD change surfaces its consequences immediately. Tandem's Design Changes view gives a before and after 3D CAD diff with a plain-language summary and the affected requirements and annotations. The mass attribute updates from the CAD assembly, the mass-margin requirement recalculates, and the evidence tied to the old value is flagged as needing a look.

This is where the CAD plugin matters. Watch records a CAD session and groups feature-tree actions into described design goals, so the reason for the thickening is captured alongside the edit. Chat gives the engineer an in-CAD agent with the program's full context, and Annotations surfaces design-review comments inside the CAD tool. Tandem works through its plugin alongside SolidWorks, Onshape, CATIA, NX and Fusion. It replaces none of them.

The thermal engineer sees the same change. A heavier panel with different conductivity affects the thermal model. Because the requirement, the attribute and the analysis are connected, that dependency is visible instead of buried. For a related walkthrough on one change cascading through requirements, see propulsion engineering requirements traceability.

How Tandem fits in 2026

Tandem is an AI-native platform that connects design intent, requirements, system architecture, CAD changes, design reviews and validation evidence in one system, so the reasoning behind a verification status survives reviews, ECOs and handoffs. It is not a CAD application and it is not PLM software.

For a satellite program, four pieces work together. The CAD plugin brings mass and geometry changes into the graph. Live sheet attributes pull budget values from the spreadsheets your team already owns. Telemetry and test data from tools like Nominal, Sift and Revel can feed attributes that verify requirements. And Requirements carry a parent and child hierarchy, approvals and a redlined change history, with FMEA and risk tables trace-linked to the requirements they affect.

If your program runs a PLM such as Teamcenter or 3DEXPERIENCE, Tandem runs alongside it. Custom integrations are built as part of deployment, pushing and pulling data so the PLM stays your system for what it does well while Tandem holds the requirement-to-evidence links. Tandem also has its own PDM with check-in and check-out, soft revisions and a release workflow, live for SolidWorks, which teams can adopt without a PLM change.

Adoption is modular. Start with requirements or design reviews alone, or use the suite. Deployments include an integration setup phase with Tandem engineers to connect CAD, PDM or PLM, spreadsheets and telemetry.

A verification checklist for your next design review

Run these before the review, not at it.

  1. Every requirement names its method and level. Test, analysis, review of design or inspection, plus the verification level and stage, as the VCD expects.

  2. Every analysis-based requirement links to its input attributes. If a thermal result cannot be traced to a mass, power or geometry value, it is a snapshot waiting to go stale.

  3. Margins are attributes with a named source. CAD assembly, sheet cell or telemetry channel. Hand-typed values in slides don't count.

  4. Estimates are marked and replaced as measurements arrive. Measured mass and measured current should overwrite predicted values.

  5. Each requirement links to the architecture node that satisfies it. Interfaces carry the loads and allocations that cross subsystem boundaries.

  6. A CAD change triggers a change-impact pass. List the requirements, margins and evidence touched since the last baseline.

  7. Review-of-design evidence points at current documents. Superseded drawings should not still back an approved status.

  8. Evidence stays with the requirement. Test results, analysis reports and review records sit in the same place as the requirement they support.

  9. The redlined history is readable. A reviewer should see what changed, when, and why.

If you cannot check three of these today, that is your agenda for the review.

Make your next review start from live attributes

A satellite requirement is verified only as long as the numbers beneath it hold. Link requirements to live mass, power, thermal and structural attributes, and a CAD change stops being a surprise at the next review. To see this on your own program, book a Tandem demo and walk one requirement from its architecture node through a CAD change to the evidence it depends on.

Visit Tandem

Tandem is the AI platform for hardware engineering. It connects requirements, system architecture, CAD design changes, design reviews, PDM and validation evidence in one system, so design intent survives every review, change and handoff. It works inside SolidWorks, Onshape, CATIA, NX and Fusion and alongside PLM, Jira, Slack and Teams, for hardware teams building complex products from prototype through production.

Get started

Sources

Frequently asked questions

What are the four fundamental methods of requirement verification?

ECSS defines test, analysis, review of design and inspection. NASA's Systems Engineering Handbook lists analysis, inspection, demonstration and test. A requirement can use one method or several. Tandem links each requirement to the attributes and evidence behind its method, so a status changes when those inputs move.

How do you keep satellite requirements verified as the design changes?

Link each requirement to live attributes instead of static documents. Mass can come from CAD, power from a spreadsheet, and measured values from telemetry. Tandem verifies requirements continuously against those attributes and surfaces affected requirements and evidence when a CAD change lands.

What is a verification control document?

The Verification Control Document holds the verification matrix. For each requirement it records the chosen method or methods, the verification level and the stage (ECSS-E-ST-10-02C). It is a snapshot, so pair it with live links to the attributes and evidence each entry depends on.

Does Tandem replace our PLM or CAD tools?

No. Tandem is not a CAD application or PLM software. It works through a plugin alongside SolidWorks, Onshape, CATIA, NX and Fusion, and runs alongside a PLM such as Teamcenter or 3DEXPERIENCE through custom integrations built during deployment.

Can test and telemetry data feed requirement verification?

Yes. Tandem works alongside test and telemetry platforms such as Nominal, Sift and Revel. Their measured data can feed Tandem attributes that verify requirements, so estimated margins get replaced by measured values as ground test and operations data arrive.

Related reading

Written by

Tandem

Tandem is the AI platform for hardware engineering. It connects requirements, system architecture, CAD design changes, design reviews, PDM and validation evidence in one system, so design intent survives every review, change and handoff. It works inside SolidWorks, Onshape, CATIA, NX and Fusion and alongside PLM, Jira, Slack and Teams, for hardware teams building complex products from prototype through production.