Schedule a Software Walkthrough
SAE ARP5580 — FMEA Practices for Non-Automobile Applications Issued 2001, Reaffirmed 2020 Topic — Software FMECA, Software FMEA contrasted Dated Approach
5580
SAE ARP5580 · Software FMEA · Software FMECA

An Aging Approach, and What Replaced It

ARP5580 was reaffirmed in 2020 without updating its 2001 software content. It still frames software around hardware-derived units, its detailed FMEA is expensive and still misses most of what breaks, and the 6D CDE Software FMEA was built specifically to fix what it leaves on the table.

1Scope

What SAE ARP5580 Software FMECA covers

Issued in July 2001 and reaffirmed — unchanged — in August 2020, ARP5580 lays out functional, interface, and detailed FMEA, plus pre-analysis planning and post-analysis documentation, applied across hardware, software, and process design. It was a genuine step forward at the time: earlier guidance like MIL-HDBK-338B barely mentioned software at all. ARP5580 at least introduced the idea of distinct software viewpoints.

But "introduced" is the operative word. It names a handful of software failure modes and gives limited examples — and twenty-plus years later, that's still all it offers.

2CSCI

The CSCI approach — and its faulty assumption

SAE ARP5580 also lays out the analysis in terms of the Computer Software Configuration Item (CSCI) — a hardware-derived unit borrowed straight from systems engineering's parts-and-assemblies mindset. The implicit assumption is that software fails the way hardware does: at the level of a discrete, physically-bounded "part," where you can point to the component that broke.

That assumption doesn't hold. A CSCI is a packaging and configuration-management boundary, not a failure boundary — software failure modes routinely cut across CSCI lines through shared data, timing, and interface behavior that a component-by-component breakdown was never built to see. Treating software like an assembly of parts is the same category error the rest of this series keeps running into: it's the hardware-centric habit that SAE1025 Section 7 was written specifically to break.

3Gaps

Three real problems with the approach

Where ASAE RP5580 software FMEA falls short Dated methodology
01
Detailed FMEA is expensive, and incomplete anyway
Running a "detailed" FMEA line-by-line at the code level is a huge effort for any real system — and even at that cost, it only ever captures roughly a third of what can actually go wrong.
02
A short list of overused failure modes
The same handful of generic modes — "incorrect output," "no output," "timing" — get reused across every analysis, regardless of what the software actually does, because the standard never gave analysts a broader, evidence-based list to draw from.
03
No real guidance on occurrence
ARP5580 doesn't tell analysts how to actually rate how likely a software failure mode is — leaving occurrence to guesswork, the same gap SAE1025 and IEEE 1633 were later written to close.
4Compare

ARP5580 vs. the 6D CDE Software FMEA

SAE ARP5580 — detailed software FMEA

Code-level, line-by-line analysis. Expensive to run, and even when finished, catches roughly a third of real failure modes — drawn from a short, generic, frequently-reused list.

6D CDE Software FMEA

Seeded from a Common Defect Enumeration built from nearly a million real software failure events, examined across six dimensions instead of code alone — identifying on the order of 100 times more critical failure modes for comparable effort. Automated in Requs AI Software FMEA.

56D

The 6D CDE model

The core insight: software doesn't fail in a silo. A failure mode is rarely just a code defect — it's the product of how software, hardware, and the world around it interact. The 6D CDE evaluates each candidate failure mode across six dimensions:

Software
Electronics / hardware
System
User
Mission
Environment

Where ARP5580 stops at the code, the 6D model deliberately asks how a candidate failure mode plays out against the hardware it runs on, the system it's part of, the person using it, the mission it supports, and the environment it operates in. The 6D CDE software FMEA is fully automated in Requs AI Software FMEA.

6Payoff

What it actually catches that SAE ARP5580 Software FMEA doesn't

Where defects are assumed to be

Code-level analysis implicitly assumes most defects originate in the code itself — which is exactly what a detailed, line-by-line FMEA is built to find.

Where defects actually originate

Root-cause analysis of real failure events shows roughly 70% of defects don't originate in the code at all — they trace back to requirements, design, or the interactions the 6D model is built to examine.

7Source

Where this comes from

Ann Marie Neufelder — Mission Ready Software 6D CDE Software FMEA

The 6D CDE Software FMEA is the same practitioner lineage running through the rest of this series — the Common Defect Enumeration behind SAE1025's Section 7, and the occurrence model IEEE 1633 adopted. It was built directly in response to gaps like SAE ARP5580 software FMEA: too expensive at the code level, too generic a failure-mode list, and no real answer for occurrence. Requs AI Software FMEA automates the entire software FMEA process providing even more value in an agile environment.