VWAY

Newsroom
Technical Resources


STPAAn STPA control structure, common cause and loss scenarios of the FAA ground stop, rebuilt in TRACEIT

An air route traffic control center at night: a dark console radar screen with faint green tracks, and a controller seen from behind holding a headset
TRACEIT Tech Brief
The FAA Nationwide Ground Stop: What Air Traffic Control Loses When One Network Wavers
An STPA control structure, common cause and loss scenarios of the FAA ground stop, rebuilt in TRACEIT
Category
VWAY Technical Article
Published
Nov 17, 2026
Read time
8 min
STPAControl StructureUCALoss ScenarioCommon CauseTRACEIT

01Summary

On November 13 (Eastern Time), the U.S. Federal Aviation Administration (FAA) held departures nationwide for about 4 hours and 20 minutes because of communication issues at multiple Air Route Traffic Control Centers. The FAA says its legacy equipment became unstable when connection attempts flooded the network, which disrupted aircraft communications and radar feeds. The final root cause is still under investigation. We built an educational STPA model from the published facts in TRACEIT. In that model, the two inputs a controller needs to issue a clearance, radar data and voice communication, pass through the same network equipment. When one network attribute is turned off in the abnormal analysis, the router, the radar site and the radio site all become abnormal together, and the clearance can no longer be generated. Two feedback paths that looked separate were tied to one path.

02Four hours and twenty minutes without departures

According to the FAA, the problem began at about 3:30 p.m. Eastern Time on Friday, November 13. At 3:54 p.m. the FAA issued a nationwide ground stop, citing "communication issues impacting multiple Air Route Traffic Control Centers," and lifted it at 8:15 p.m. The FAA said instability began at 17 of its 22 Air Route Traffic Control Centers (ARTCCs) and spread to some approach control facilities.

The FAA's questions-and-answers page of November 15 explains it this way. The FAA's legacy equipment became unstable when data began flooding the network at a rate of about 1,000 connection attempts every four to five seconds, and this disrupted aircraft communications and radar feeds. The FAA named two potential locations: a radar site in Amarillo, Texas, and a radio communications site in El Dorado, Arkansas. Blocking traffic from the two sites stabilized the network, and the FAA disabled the broadcasting function on the new routers as a precaution. It took the same step on roughly 60 similar routers.

This network, the FAA Telecommunications Infrastructure (FTI), links control centers, radar facilities and radio sites. The Secretary of Transportation said the contractor, L3Harris, owns the network. The work that day was described as part of routine maintenance that replaces copper lines at remote sites with fiber and moves them to IP-based communication. The FAA counted about 12,000 cancellations that day, noting that not all of them were directly caused by the ground stop.

Note

The FAA states that it "has not established the final root cause." Reports differ on the router maker and model, the role of the Amarillo site, and whether the cause was an installation or configuration error or a hardware fault; none of these has been officially confirmed. The model below is a simplified educational example based on information published through November 16. It is not an investigation finding.

03Drawing the event as a control structure

STPA (System-Theoretic Process Analysis) treats an accident as a failure of control rather than a chain of component failures. You draw a control structure that shows who sends which control action to whom and which feedback each controller uses to decide, and then you look for ways each control can become unsafe. In this case we used three control actions: the network operator's router replacement and feature activation, the Command Center's (ATCSCC) ground stop issue and release, and the en route controller's clearance.

TRACEIT control structure screen. The Command Center and airports and airlines at the top, the ARTCC controller and the FTI network router (orange) in the middle, the network operator on the left, and the radar site, radio site and aircraft at the bottom
Figure 1. The educational control structure drawn in TRACEIT. Radar data, voice communication and clearances all pass through the orange FTI router.
ElementRoleSignals
[E1] Command Center (ATCSCC)Manages national traffic flow; issues and releases ground stopsReceives center availability reports and departure queues; sends the ground stop (CA2)
[E2] En route controller (ARTCC)Gives altitude, heading and speed clearances in its sectorReceives radar track data and pilot voice; sends clearances (CA3)
[E3] Network operatorOperates the FTI network and replaces equipmentReceives network status; sends router replacement and feature activation (CA1)
[E4] FTI routerCarries signals between centers and remote sitesCarries radar data, voice and clearances
[E5] Radar siteTurns aircraft position into track dataRestates aircraft position and altitude as radar track data
[E6] Radio siteRelays voice between controllers and pilotsRelays pilot voice and clearances
[E7] Aircraft and pilotFly as clearedSend position, altitude and voice; receive clearances
[E8] Airports and airlinesReport departure queues and delaysSend departure queues and delays to the Command Center

TRACEIT records for every signal where it is generated, where it is restated and where it is executed. The radar site receives aircraft position and altitude (FB1) and sends it on as radar track data (FB9). The controller's traffic picture (PM1) is a process model with the expression PM1 = FB9 × FB2. The × means AND: the picture holds only when both radar data and pilot voice are present. The clearance (FB3) is generated from this picture. The Command Center's national decision (PM2) is built the same way from center reports and departure queues.

Following one clearance to the end

Pick the clearance (FB3) in the signal list of the control structure, and TRACEIT highlights in yellow the path of every signal used to generate it. In Figure 2 the clearance starts at the aircraft, passes through the radar site and the radio site into the router, goes through the controller's decision, and returns to the aircraft through the same router and radio site. The whole control loop passes through one router twice.

TRACEIT trace highlight screen. The path of the clearance signal is highlighted in yellow, linking the aircraft, radar site, radio site, FTI router and ARTCC controller in one loop
Figure 2. Trace of the clearance (FB3). Radar data, voice and clearances all pass through the FTI router.

04Turning off one shared network

In the control structure diagram the radar site and the radio site stand apart, so the two feedback paths look independent. Whether they really are depends on what they share. TRACEIT's Abnormal Analysis addresses this question. You attach attributes to elements, and when you turn off one attribute value, TRACEIT assumes every element with that value loses its function at the same time and computes the effect. It is a way to simulate common cause failure (CCF) on the control structure.

In the example we created an attribute named Network and gave the FTI router, the radar site and the radio site the same value, "FTI legacy network (shared)," meaning all three are connected to the legacy FTI network. Then we selected all three control actions and ran the simulation.

TRACEIT Abnormal Analysis screen. The top shows the signal flow graph of clearance CA3; the result table below marks the ARTCC controller, FTI router, radar site, aircraft and radio site as Abnormal with their cause paths
Figure 3. Abnormal Analysis result for the clearance (CA3). For each element it shows which broken signal stops the clearance from being generated.

The result table shows what breaks at each element. If the router cannot transfer radar track data (FB9), the controller's traffic picture (PM1) cannot be calculated and the clearance (CA3) cannot be generated. Losing pilot voice (FB2) at the radio site leads to the same conclusion. Because PM1 is an AND expression, losing either input is enough to stop the clearance.

TRACEIT Abnormal Analysis result by attribute. One row for the Network value FTI legacy network shows Abnormal, the affected elements E4, E5 and E6, and the cause path
Figure 4. Result by attribute. One network value makes E4, E5 and E6 abnormal together and blocks the clearance.

The result by attribute says it more briefly. When the value "FTI legacy network" is turned off, E4, E5 and E6 become abnormal together. Radar and voice use different equipment at different sites, but because they are tied to the same network they behave like one feedback path in this case. This matches the FAA's own description right after the event, which named aircraft communications and radar feeds together.

When two feedback paths share one route, they are really one. Abnormal Analysis shows that route on the control structure.

05Summarizing with UCAs and loss scenarios

We set two losses. L1 is loss of life or aircraft due to a collision or accident, and L2 is loss to passengers, airlines and airports due to large-scale flight disruption. This event led to L2, and the ground stop was a control meant to prevent L1. If that control comes too late, the hazards on the L1 side (H1, H2) grow; if it lasts too long, the hazard on the L2 side (H3) grows.

UCAStatementHazard
UCA1-P-1The network operator replaces a router or activates a feature on the operating shared ATC network without reviewing the impact and isolating it.H1
UCA1-S-2The network operator leaves the faulty feature on longer than necessary after abnormal connection attempts surge.H1
UCA2-T-2The Command Center issues the ground stop late, after control at many centers has already become unstable.H1
UCA2-S-3The Command Center keeps the ground stop longer than necessary after control has recovered.H3
UCA3-N-1The controller must issue a clearance when two aircraft approach minimum separation, but cannot because radar and voice are lost.H2

Loss scenarios explain why a UCA occurs. Three scenarios in this model deserve attention.

  • LS1 (UCA1-P-1) The network operator believes the replacement affects only that site (PM3), but the new router floods the shared network with connection attempts and radar and voice at many ARTCCs become unstable at once. The causes are not checking in advance how the new equipment's feature affects legacy equipment (an analysis assumption) and managing legacy and new network work schedules separately.
  • LS3 (UCA3-N-1) Radar track data (FB9) and pilot voice (FB2) pass through the same router, so one equipment fault removes both inputs of the controller's traffic picture (PM1) at the same time. The cause category is common cause.
  • LS4 (UCA2-T-2) The Command Center forms its national decision (PM2) from center reports that arrive one by one, so the decision to issue can lag until the common cause becomes visible.

The schedule cause in LS1 is based on the U.S. Government Accountability Office (GAO) report of September. GAO found that the air traffic control modernization effort has no integrated master schedule and that 11,389 separate project schedules lead to installations at the same sites being scheduled without regard to air traffic operations. The report predates the event, and no direct link to this event has been confirmed.

TRACEIT Trace View screen. Losses, hazards, UCAs, causes, loss scenarios and countermeasures are linked from left to right in color-coded boxes
Figure 5. Trace View. Follow the chain from losses to hazards, UCAs, causes, loss scenarios and countermeasures on one screen.

06Six countermeasures

Countermeasures are attached to each loss scenario. The first three have importance 1.

1
Signal path impact analysis before a change

Attach to the work approval the radar and voice signals that pass through the equipment being replaced and the list of affected centers. The trace highlight and abnormal analysis results become that list directly. (LS1)

2
Staged activation and automatic rollback

Turn new features on one site at a time, and switch them off automatically when connection attempts exceed a threshold. (LS1, LS2)

3
Diverse paths for radar data and voice

Keep the two feedback paths on different equipment and routes even during the network transition. You can check this by whether the two paths get different attribute values in the abnormal analysis. (LS3)

4
Link the network operations view to ATC anomaly reports

Share each center's availability report with the network operator in real time, so work and ATC anomalies are connected quickly. (LS2, LS5)

5
One schedule for legacy and new network work

Coordinate work at the same site on a single schedule. (LS1)

6
Predefined criteria for multi-center instability

Decide in advance under which conditions to issue and release a ground stop when several centers waver at once. (LS4, LS5)

The same question in other industries

Redundant feedback built on shared infrastructure is not unique to air traffic control. Safety sensors and control signals in a factory that pass through the same industrial switch, two vehicle sensors on the same power supply or gateway, and several reactor units that rely on the same off-site power all have the same structure. Once you draw the control structure and attach attributes such as network, power and location to each element, turning off a single value can reveal hidden common causes early.

07Frequently asked questions

Q1Can STPA be used on an accident whose cause is not yet established?

Yes. STPA looks for the conditions under which control can fail rather than for a single root cause. Build the control structure and UCAs from confirmed facts, and mark what is not yet confirmed as analysis assumptions. When the investigation reports, revise the assumptions and causes and trace again.

Q2If radar and voice are run separately, aren't they redundant?

Even with different equipment and sites, they can fail together from one cause if they share the same network, router or power supply. In this example, turning off one network attribute made the radar site, radio site and router abnormal together, and no clearance could be generated. Redundancy holds only after you check that nothing is shared.

Q3How does TRACEIT's Abnormal Analysis differ from FTA?

FTA starts from a defined top event and searches downward for combinations of causes. TRACEIT's Abnormal Analysis uses the signal flow and process model expressions of the control structure as they are, and computes forward which signals break and which control actions cannot be generated when one attribute value is turned off. It works well as a supporting step for checking common causes within an STPA analysis.

Note

The control structure, signals, UCAs, loss scenarios, causes and countermeasures in this example are simplified educational assumptions based on reports and FAA materials published through November 16. They are not investigation findings and do not judge the responsibility of any agency or company.

References

· FAA, "Nationwide Ground Stop: Questions and Answers" (2026-11-15) — FAA page

· FedScoop, report on the FAA nationwide ground stop and its investigation (2026-11) — Read

· ABC News, report on the Transportation Secretary's remarks on the routers' broadcast function (2026-11) — Read

· GAO-26-107992, "Air Traffic Control Systems: Ambitious New Modernization Effort Needs to Improve Cost and Schedule Planning" (2026-09) — GAO page

· Leveson & Thomas, "STPA Handbook" (MIT, 2018) — PDF

· MIT PSASS, "STAMP Tools" — Page

Want to check whether your redundancy is really independent?
TRACEIT is an STPA tool co-copyrighted by VWAY and the Korea Atomic Energy Research Institute (KAERI). We can walk you through drawing a control structure with attributes and finding common causes, using an example from your own system.
VWAY Co., Ltd. | sales@vwaycorp.com