HomeTechnical ResourcesSafety assessment guide

Technical insight / System safety

From FHA to
PSSA to SSA.

Build a safety argument that follows the design.

A failure condition, a fault tree and a test report only become useful evidence when the connections between them are clear. This guide follows one engineering question from the aircraft effect to the implemented system.

Aerospace Consulting16 November 2026Based on ARP4761A and ARP4754B

Start with the right level

The aircraft argument
and the system argument.

An FHA identifies the effects of losing a function or providing it incorrectly. A PSSA examines whether a proposed system architecture can meet the safety objectives and establishes the requirements needed to make that argument work. An SSA brings together evidence for the system that was actually implemented.

ARP4761A also makes the aircraft-level work explicit. The AFHA, PASA and ASA address aircraft functions, allocation and interactions that may cross system boundaries. The system SFHA, PSSA and SSA sit within that wider assessment. These are connected, iterative processes.

Aircraft level
AFHAAircraft effects and objectives
PASAAircraft architecture and allocations
ASAIntegrated aircraft evidence
System level
SFHASystem effects in aircraft context
PSSAArchitecture and derived requirements
SSAEvidence for the implemented system

AFHA: Aircraft Functional Hazard Assessment · PASA: Preliminary Aircraft Safety Assessment · ASA: Aircraft Safety Assessment · SFHA: System Functional Hazard Assessment. The diagram is an orientation to the relationships, not a complete process flow.

Keep the distinction clear.

A failure condition describes an aircraft or system effect. A failed sensor, seized mechanism or erroneous software output may be a cause. Listing equipment failures does not, by itself, establish the failure conditions or their severity.

An original illustrative example

Uncommanded flap asymmetry
during approach.

Consider an aircraft with powered left and right flap actuation. One side moves without a valid command while the other side remains in position. The engineering question is whether the resulting asymmetry can compromise the aircraft, and what the design needs to do about it.

This example is hypothetical. It assigns no severity, probability target or development assurance level. Those conclusions require the aircraft characteristics, operating assumptions and applicable certification basis.

Define the effect

What happens to the aircraft?

Describe the effect of the asymmetry over the relevant speed, flap position, loading and flight phase. Consider controllability, flight path, structural consequences, crew workload and the margin available for recovery.

The FHA should identify the assumptions behind the classification. If crew recognition or corrective action is credited, establish the information available, the time allowed and the actions assumed. The reasonableness of that credit needs supporting evaluation.

The SFHA connects the allocated flap functions and system failure conditions to this aircraft context. A new interaction discovered at system level is fed back to the aircraft assessment.

Make the connections usable

A requirement needs an owner,
a baseline and evidence.

A useful safety requirement says what the design must achieve and under what conditions. “Provide monitoring” is incomplete. For this example, the team needs to establish the allowable asymmetry, the detection and stopping response, the relevant operating conditions and the failures that the protective path must tolerate.

Keep the chain from the failure condition to the safety requirement, allocated implementation requirement, verification activity and result. A document reference alone is often insufficient: identify the section, test case or analysis conclusion that supports the claim.

Claim in the safety argumentWhat needs to be establishedExample evidence
Movement is arrested in timeDetection, processing and physical stopping delays fit within the aircraft's validated limit.Timing analysis plus representative end-to-end response tests.
The protection remains availableCredited independence covers shared resources, interfaces, installation and relevant development errors.Architecture review, common-mode analysis and installation-related assessments.
Latent loss is controlledThe diagnostic or maintenance action detects the relevant failure with the assumed coverage and exposure.Detection-coverage evidence and an agreed inspection or test interval.
Crew action limits the consequenceDetection, interpretation, workload and response time support the particular credit.Applicable human-factors and operational evaluation linked to the assumption.
The evidence applies to the aircraftHardware, software, installation, requirements and procedures match the assessed configuration.Configuration index, change-impact review and relevant integration results.
Safety assessment and development assurance support each other.

ARP4754B connects requirements development, validation, implementation verification and configuration control. The safety assessment supplies requirements and depends on evidence from those processes. An SSA does not replace software or electronic-hardware development assurance.

What a technical review should challenge

Where the argument
can break.

These are practical review questions. Their relevance and depth depend on the system and its safety objectives.

The failure condition and the analysis top event differ

Check that losses, erroneous behaviour, combinations and operating conditions are defined consistently. A tree for loss of actuation does not automatically cover uncommanded movement, a jam or a mechanical disconnect.

Redundancy is treated as proof of independence

Identify what can affect both paths. Shared requirements, sensing, power, environment, installation or implementation can defeat the claimed protection. Quantitative combinations depend on justified dependencies and exposure assumptions.

A numerical result hides an unsupported assumption

Check the failure data, units, operating time, latent exposure, coverage and model boundaries. Development assurance levels are not failure rates. Development errors should not be assigned invented numerical probabilities to make a fault tree pass.

A monitor is credited without a complete response path

Detection is only part of the argument. Establish the consequence of the indication, the authority to intervene, the response delay and whether the physical action remains effective for that failure.

Verification proves the component but not the integration

Review interfaces, combined delays, test configuration and the difference between real and simulated functions. Identify the evidence that is still required at system or aircraft level.

The design changes but the safety argument does not

Revisit affected failure conditions, allocations, independence, maintenance and verification when the design or operation changes. A previous conclusion remains useful only where its assumptions and configuration still apply.

Take it into your next review

Ten questions
to work through.

Mark the questions you have reviewed. This is a personal working aid; the count is not a compliance result. Selections stay in this page session.

0 of 10 questions marked as reviewed.

References and application. This is original explanatory material informed by ARP4761A, particularly Sections 3–5 and Appendices A–F, and ARP4754B requirements, validation, verification and configuration-management processes. It is not a reproduction of either standard or a complete compliance procedure. Use the editions and means of compliance applicable to the programme.

Need a second view of your safety argument?

Bring the failure condition, architecture, assumptions and the evidence available today.