Software Failure Modes Effects Analysis Overview

© Softrel, LLC 2020
SoftwareFailureModesEffectsAnalysis
Overview

© Softrel, LLC 2020 2
Copyright
•This presentation is copy protected and
licensed to one person who has registered for
the class
•You may not
•Use any part of this presentation in a
derivative work including but not limited to
presentations, conferences, published
articles or books, theses, etc.
•Take snapshots of, scan this presentation or
convert to digital format
•Violators will be prosecuted to the fullest
extent of the law

Additional information
•“Effective Application of
Software Failure Modes Effects
Analysis” book.
• The SFMEA toolkit contains a
complete list of software failure
modes and root causes

About Ann Marie Neufelder
• Has been developing software and managing software engineers since
1984
• Has been applying software failure modes effects analysis since 1986 on
complex software intensive engineering systems
• Has conducted numerous software FMEAs in medical, defense, space,
energy and electronics industries
• Has reviewed numerous software FMEAs for large organizations
acquiring software intensive systems
• Has read more than 200,000 software problem reports and assessed the
failure mode and root cause of each
• Has identified more than 400 failure mode/root cause pairs for software
• Wrote the SFMEA webinar for NASA
• Has seen the good, bad and ugly with regards to effective software
FMEAs
• Chair-person for the IEEE 1633 Recommended Practices for Software
Reliability, 2016
• Published current industry guidance for software failure modes effects
analysis as referenced by IEEE 1633 and SAETAHB0009
• Invented the only industry accepted method to predict the number of
software failures before the code is even written

• The increase in size of F16A to F35 is just one example[1]
• With increased size comes increased complexity and
increased failures due to software
1.0 Introduction
0
5000000
10000000
15000000
20000000
25000000
30000000
1970 1980 1990 2000 2010 2020
SIZE IN SLOC OF FIGHTER AIRCRAFT SINCE 1974
[1] Delivering Military Software Affordably, Christian Hagen and Jeff Sorenson, Defense AT&L, March-April 2012.

• Addressing one failure mode could mean eliminating several failures
• i.e. if the software engineers failed to consider what the software shall do when there is a
hardware fault, this effects all hardware interfaces which can have wide reaching effect
• The SFMEA output can be used for any of these purposes
• Develop better software requirements and/or make requirements reviews
more effective
• Abnormal behavior/alternative flows that might be missing from the requirements or
design specifications
• Unwritten assumptions
• Develop a better design and/or make design reviews more effective
• Features that need fault handling design
• Defects that easier to see when looking at the design or code but difficult to see during
testing
• Design a health management system (HMS) or Built InTest (BIT)
• Identify defects that cannot be addressed by redundancy or other hardware controls
• Develop a test plan that tests for more than the happy day scenarios
• Identify tests that find failures as opposed to confirm requirements
• Identify failures in systems that are difficult or expensive to test (i.e. spacecraft, missiles,
medical equipment, etc.)
• Develop software/supporting documentation that minimizes user error
1.0 Introduction

• Treating the software as it’s a black box –Early guidance on software FMEA recommended
the black box approach which doesn’t provide value. The functional viewpoint has proven
effectiveness.
• Assigning the analysis to someone who’s not a software engineer. Reliability engineers
may facilitate but software engineers understand the failure modes.
• Assuming the software will work. Instead, one must assume that:
• Specifications are missing something crucially important
• Specifications are incorrect
• Specifications are sufficient but design/code doesn’t address them
• Not having sufficient written specifications
• SFMEAs are much more effective when conducted on use cases and design than high
level requirements
• Trying to analyze everything
• Most productive when they focus on riskiest software, riskiest functions, with most
likely failure modes. More effective when preceeded by a root cause analysis.
• Conducting the analysis too early or too late
• Spending too much time spent assessing likelihood.
• SFMEAs are NOT used to calculate failure rates. Once root cause of failure mode is
removed correctly, failure event doesn’t recur. That’s different than for hardware.
1.0 Introduction

• The sources of all software faults lie in the below three basic sources
• Software FMEA analyst must understand and consider all three sources
• It’s common for SFMEAs to assume that all software specifications are both complete
and correct and that requirements based testing will find all defects
• The only thing that can be assumed is that the customer specification is correct
1.0 Introduction
Top level software fault Why its not found prior to operation
The software specifications are
missing crucially important
details
These aren’t found in requirements based testing
because only the written part is tested.
The software specifications are
themselves faulty and hence the
code is faulty.
These are found in requirements based testing
because the code does what the stated requirement
says
The software engineer may not
always write the code as per the
written specifications
• Some specifications aren’t
coded at all.
• Some specifications are
coded incorrectly.
• These faults can happen with very large systems
and/or insufficient traceability.
• These can get through requirements based testing
undetected if the code works with some inputs but
not others

• Faulty functionality
• Software does the wrong thing
• Software fails to do the right thing
• Software feature is overengineered
• Faulty error handling - Inability to detect and
correctly recover from
• Computational faults
• Faults in the processor
• Memory faults
• Environment faults (Operating system)
• Hardware faults
• Power faults
• Power not checked on startup or power down
• Faulty I/O
• Faulty sequences/state
• Missing safe state and/or return to safe state
• Prohibited state transitions allowed
• Faulty timing
• Race conditions
• Timeouts are too big or too small
• Software takes too long to execute and misses
timing window
• Faulty data handling
• Data and interface conflicts
• Insufficient handling of invalid data
• Faulty algorithms
• Crucially important algorithm isn’t specified
• Algorithm is specified incorrectly
• Faulty logic
• Control flow defects
• Faulty comparison operators
• Faulty usability
• Insufficient positive feedback of safety and
mission critical commands
• Critical alerts aren’t obvious
• Faulty processing
• Software behaves erratically or can’t start
after a loss of power or user abort
• Endurance/peak load
• Safety and mission are degraded when
system remains on for extended period of
time
• Operational status isn’t monitored
Copyright Softrel, LLC 2019
1.0 Introduction

• Severity depends on the feature that has the failure mode
• All software failure modes can result in catastrophic failure and all can result in a non-critical failure.
• If a mission critical feature has one of these failure modes the effect will generally be severe,
however, non-mission critical features encountering these failure modes may still have a severe
consequence.
• Likelihood depends on the application type and the development practices
1.0 Introduction
Failure mode When it’s more likely
Faulty functionality Common if the software requirements are too high level
Faulty error
handling
Most common for nearly every organization because software engineers almost never
consider failure space
Faulty
sequences/state
More likely if the software engineers don’t do detailed state design at the lower levels and
highest level
Faulty timing More likely if there aren’t timing diagrams
Faulty data handling More likely if there aren’t well defined software interface design documents and data
definitions
Faulty algorithms Applies to mathematically intensive systems
Faulty logic Common for all software systems and is due to insignificant logic design at the lower levels
Faulty usability Common if the user has to make quick decisions that are important
Faulty processing Most common for systems that may encounter interrupted power supply
Endurance issues Most visible for systems that do have an uninterrupted power supply
Peak loading issues Applies to systems that can be used by more than one end user at the same time

• Faulty error handling – Apollo 11 lunar landing,ARIANe5, Quantas flight 72, Solar
Heliospheric Observatory spacecraft, Denver Airport, NASA Spirit Rover (too
many files on drive not detected)
• Faulty data definition - Ariane5 explosion 16/64 bit mismatch, Mars Climate
Orbiter Metric/English mismatch, Mars Global Surveyor, 1985 SDIO mismatch,
TITANIV wrong constant defined
• Faulty logic– AT&T MidAtlantic outage in 1991
• Faulty timing - SCUD missile attack Patriot missile system, 2003 Northeast
blackout
• Race condition - Therac 25
• Peak load conditions - Affordable Health Care site launch
• Faulty usability
• Too easy for humans to make mistakes – AFATDS friendly fire, PANAMA city
over-radiation
• Insufficient positive feedback of safety and mission critical commands – 2007
GE over-radiation
The above illustrates that history keeps repeating itself because people assume
root causes from other industries/applications are somehow not applicable.
Lesson to be learned – the root causes are applicable to all software. It’s the
hazards/effects that result from the root causes that are unique
1.0 Introduction

Loss of $370 million payload
On June 4th, 1996 launch vehicle veered off course
and self-destructed 37 seconds into the flight, at an
altitude of 4km and a distance of 1km from the
launch pad.
Guidance system shut down and passed control to
an identical redundant unit which had the identical
software defect.
Unnecessary course correction.
Primary defect–Faulty data -The guidance system’s
computer attempted to convert the sideways velocity of
the rocket from a 64 bit format to a 16 bit format. The
result was an overflow.
Related defect- Faulty error handling -The guidance
software had not been designed to trap for overflows
because it was thought that the overflow could never
happen.
Related defect- One size fits all error handling. When the
overflow occurred the computer reboot by design. It
should handled the fault differently.
1.0 Introduction
Example of mismatched
size formats from past
history [ARIANE5]

Whether the life cycle model is waterfall, incremental, spiral, etc.
The best time to do the particular SFMEA is shown above. With
incremental models, there may be more than one iteration.
1.0 Introduction
SW increment 1 SW increment 2 SW increment n…
Within an increment
some or all of these
Activities take place
SW Requirements
SW Design
SW Implementation
and unit testing
SW IntegrationTesting
SW SystemTesting
SystemTesting
Final
SystemTesting
…
Serviceability
SFMEA for
update and
install scripts
Maintenance
generally once
software is
deployed
Functional, Usability
Interface
Detailed ,
Vulnerability

Prepare Software FMEA
Define scope Tailor the SFMEA
Generate
CILMitigate
Analyze failure modes and
root causes
Identify
resources
Identify
equivalent
failure
modes
Identify
consequences
Identify local/
subsystem/
system
failure effects
Identify
severity
and likelihood
Identify
corrective
actionsIdentify
controls
Identify
compensating
provisions
Analyze applicable
failure modes
Identify root causes(s)
for each failure mode
Generate a
Critical
Items List
(CIL)
Identify
boundary
Set
ground
rules
Select
View
points
Identify
what can
go wrong
Gather
artifacts
Define
likelihood
and
severity
Select
template
and
tools
Revise RPN
Identify
riskiest
functions
For each
use case, use case steps,
SW requirement, interfaces,
detailed design, user manuals,
Installation scripts …
(as applicable based
on selected view point)

1. The software engineer will not write code that’s not specified.
2. The software test engineers may test the written
requirements but won’t test what’s not in writing.
3. The software engineer may not always write the code as per
the stated requirements or specifications.
• It cannot be assumed that even when requirements are complete and
unambiguous that the code will be written correctly.
4. The software engineer may “guess” if presented with more
than one option for implementing a requirement.
• The wrong guess can result in code that causes a system failure.
5. It cannot be assumed that “thorough testing” will uncover all
software failures prior to operation.
6. The failure modes from a past similar system are likely to be
relevant for the current system
2.0 Prepare for the SFMEA

This is a specification pertaining to the initialization of a system:
The software shall check for voltages within <x> and <y>
• There are no other statements regarding the voltages.
• The software development plan indicates that there will be “requirements”
based testing. There is no indication of stress testing or fault injection testing.
This is what’s wrong…
• The software specification doesn’t say what the software shall do if the voltages
are out of range.
• The software specification doesn’t explicitly say that the software does not
continue if voltages are out of range.
• Technically the specification is passed if the software checks the voltages regardless
of what it does when the voltages are out of range.
• Hence there’s no reason to believe that there will be code to handle what
happens if the voltages aren’t within <x> and <y>.
• Since there is only requirements based testing it is very likely that voltages out
of range won’t be tested.
This is what can happen if the voltages are out of range…
• The initialization completely stops (this is called a dead state)
• The initialization proceeds to the next state when it should not (this is called an
inadvertent state transition) Copyright Softrel, LLC 2019

This is the specification for the logging feature for a mission and
safety critical system:
1) The software shall log all warnings, failures and successful missions.
2) At least 8 hours of operation shall be captured
3) Logging to an SD card shall be supported in addition to logging to the
computer drive
This is what you know about the software organization and software
itself
1) Logging function will be called from nearly every use case since
nearly every use case checks for warnings, failures and successes
2) Testing will cover the requirements. But no plans to cover stress
testing, endurance testing, path testing, fault insertion testing.
3) Software engineers have discretion to test their code as they see fit.
4) There is a coding standard but there is no enforcement of it through
automated tools and code reviews only cover a fraction of the code

• These are the faults that can/will fall through the cracks
• No checking of read/write errors, file open, file exist errors which are
common
• No rollover of log files once drive is full (may be beyond 8 hours)
• No checking of SD card (not present, not working)
• Logging when heavy usage versus light or normal usage (might take
less than 8 hours to fill drive if heavy usage)
• Is both SD card and drive to be written to or does user select?
• This is why these faults aren’t found prior to operation
• No one is required to explicitly test these faults/scenarios
• No one is required to review the code for this fault checking
• No one is required to test beyond 8 hours of operation
• This is the effect if any of these faults happens
• Entire system is down because it crashes on nearly every function once
drive is full, SD card removed, file is open or read/write errors
• When conducting a SFMEA one cannot assume that best practices
will be followed unless there is a means to guarantee that

• Even if previously discussed issues are specified - these things
can happen with the logging function due to coding mistakes
• logs the wrong status
• fails to log any status
• takes too long to log the status (log file has stale timestamps)
• logs corrupt data
• deletes the entire log file
• fails to acquire 8 hours of operation under any scenario

FMEA When this viewpoint is relevant
Functional Any new system or any time there is a new or
updated feature/use case/requirements/design.
Interface Anytime there is complex hardware and software
interfaces or software to software interfaces.
Detailed Almost any type of system is applicable. Most
useful for mathematically intensive functions.
Maintenance An older legacy system which is prone to errors
whenever changes are made.
Usability Anytime user misuse can impact the overall
system reliability.
Serviceability The software is mass distributed and remotely
installed/updated as opposed to installed in a
factory, authorized service personnel or controlled
environment.

analyze what
the software does
• Hardware FMECA focuses on the configuration items.
Software FMECA is less effective with that approach.
• Focus should be on what the software is required to do and not
the CSCI unit itself.
• A CSCI typically performs dozens, hundreds or even
thousands of functions so the below is too open ended
Common Viewpoint Mistake
LRU Failure mode Recommendation
Executive
CSCI
CSCI fails to
execute
Doesn’t address states, timing, missing
functionality, wrong data, faulty error
handling, etc.
Executive
CSCI
CSCI fails to
perform required
function
CSCI performs far too many features and
functions. List each feature and what can
go wrong instead.

• Common but ineffective approach - Analyze every software
CSCI against total failure of that CSCI to execute
• Hardware engineers assume that since hardware fails when it
wears out or breaks that software failures when it doesn’t
execute at all
• Sounds good-but history shows that the most serious hazards happen
when the software is running
• Your book has many real world examples
• If the software doesn’t execute at all that’s likely to be identified in
testing hence this is akin to a “CaptainObvious” failure mode
• The concept of a partial failure causing a total system failure is
often difficult for hardware/systems engineers to grasp

System
requirements
Software
requirements
Software interface
design
Software design – state
diagrams, timing diagrams,
sequence diagrams, DB design,
GUI design
Module and class design
Line of code
Functions, procedures (code)
Not enough
coverage across
the software and
not enough
coverage of
design or
software only
requirements
Analyzing one
line of code at a
time has
potential to miss
the design and
requirements
related faults.
Won’t work for
code that’s
“missing”.

Functional root causes Process related root causes
Data conversion from 64 to
16 bit
• Faulty assumption that since code didn’t change from
ARIANE4 then it is OK for ARIANE5. In fact preparation
sequence and workload were different from ARIANE5 than
ARIANE4.ARIANE 5 could handle a heavier payload.
• Insufficient review of existing design against new ARIANE5
environment.
Conversion error
unprotected (not handled)
• Faulty assumption that since overflow didn’t happen on
ARIANE4 it is “impossible” for ARIANE5.
• Since there were no controls for overflow “impossible” was
wrong assessment.
One size fits all defect
exception handling -
Software shut down when
unhandled events detected
• No requirement for ARIANE5 for the protection of the
unhandled conversion.
• Since only requirements are tested, no tests new
environment or for overflow were run in simulation
• Functional root cause – defect in the software requirements, specifications, design,
code, interface, usability, etc.
• Process root cause – the reason why the software product root causes weren’t
detected or mitigated prior to system failure
• Example – Ariane 5 disaster 1996

Contrary to popular myth it’s not feasible to conduct a software FMEA on
all of the software. Even small applications have thousands of lines of code and
countless combinations of inputs.
Some options:
1. Broad and shallow
• Choose one or two known common failure modes and apply them
throughout the mission and/or safety critical software specifications/use
cases.
• Example:There have been many problems with alternative flows and
handling of faulty hardware, computations, input/output, etc. Apply “missing
fault handling” to every mission and safety critical use case.
2. Deep and narrow
• Choose the most critical function or use case and analyze it against every
applicable software failure mode
• Example: A missile must be ejected so as to not hit the platform that
launched it. The specifications related to the initial ejection are analyzed
against faulty state, sequences, timing, data, error handling.

Software FMEA is not a “one person” activity. Inputs from software and
systems engineering are required.
Reliability engineers who haven’t developed software cannot perform the
analysis effectively by themselves.
Software
engineering
Systems
engineering
SFMEA
facilitator
Brainstorm what can go
wrong/past failure modes
x x x
Gather all use cases/
identify riskiest
x x x
Tailor the SFMEA x
Analyze failure modes x x
Analyze consequences x x x
Identify mitigations x x x
Execute mitigations x
Create CIL x

Generate CILMitigate
Analyze failure modes and root
causes
Identify
resources
Identify
equivalent
failure
modes
Identify
consequences
Identify local/
subsystem/
system
failure effects
Identify
severity
and likelihood
Identify
corrective
actionsIdentify
controls
Identify
compensating
provisions
Analyze applicable
failure modes
Generate a
Critical
Items List
(CIL)
Identify
boundary
Set
ground
rules
Select
View
points
Identify
what can
go wrong
Gather
artifacts
Define
likelihood
and
severity
Select
template
and
tools
Revise RPN
Identify
riskiest
functions
For each
SW specifications, interfaces,

© Softrel, LLC 2020 293.0 Analyze failure modes and root causesFailureModeNo.
SoftwareItemUnderConsideration
SoftwareItemFunctionality
DesignRequirement(RequirementIDtag)
PotentialFailureMode
PotentialRootCause
PotentialEffect(s)ofFailure
EffectLevel(E)
DetectionMethod(s)
OccurrenceLevelofFailureMode(O)
DetectionLevelofFailureMode(D)
RiskPriorityNumber(RPN)
SoftwareCTQ(DesignDetails)
CTQRationale
RecommendedAction(s)
ResponsibleIndividual(s)/Function(s)
TargetCompletionDate
Action(s)Taken
ResidualEffectLevel(E)
ResidualOccurrenceLevel(O)
ResidualDetectionLevel(D)
ResidualRiskPriorityNumber(RPN)
SystemHazardID(s)

© Softrel, LLC 2020 303.0 Analyze failure modes
Viewpoint Sub-view Failure modes/root causes
Functional What can go wrong with
most or all of the
specifications?
Missing crucially important details,
overengineered, under-engineered,
Missing error handling, one size fits all
error handling, wrong error recovery, no
recovery, missing essential timing
requirements, missing fault state, missing
transitions to fault state, prohibited states
allowed, implied dead states
What can go wrong with a
feature?
What can go wrong with a
specific specification?
Interface What can go wrong
between two components?
Conflicting units of measure, type, size.
Unable to handle corrupt, missing or null
data.
Detailed,
maintenance
What can go wrong with
the detailed design for a
module. What can go
wrong when the module is
changed after a baseline.
Algorithms themselves are faulty or are
implemented faulty. Faulty logic, faulty
sequences, faulty data definition, faulty
error handling, faulty I/O
Usability What can go wrong with
the user?
Unnecessary fill in the blanks, faulty
assumption user is always there or looking
at software, overly cumbersome

• Each of the viewpoints has templates which facilitate analysis
at that viewpoint with most relevant and likely failure modes
and root causes
• HIGHLY recommended –Convert text based specifications to
flow, sequence, state transition, timing diagrams when
analyzed
• MUCH easier to see the failure modes and root causes when
there is a diagram or table
• In this presentation the functional viewpoint is presented
3.0 Analyze failure modes

• This is a typical text based software design
• The software shall Initialize
• The software shall execute <ProcessA>
• The software shall execute <Process B>
• The software shall execute <ProcessC>
• In writing there doesn’t appear to be anything
wrong
• However, when it’s illustrated as a flow it’s clear
that some things are missing from specification
• Initialization may not be successful
• ProcessA may not be successful
• Process B may not be successful
• ProcessC may not be successful
• Do previous processes have to be successful for
next one to execute?
• What should software do if any of these tasks
are not successful?
More pictures – more efficiency

• Text based software design specification
• The software shall Initialize.
• If initialization is unsuccessful the software will
make X attempts to initialize.
• If initialization not successful after x attempts the
software shall log an error.
• The software shall execute <Process A>. If not
successful it shall log an error.
• The software shall execute <Process B>. If not
• The software shall execute <Process C>. If not
• These are the problems which are easily to see in
diagram form
• Errors are logged but there is no action defined
beyond that
• As written the software will simply stop at the first
error event
• This is a very common problem within and outside
of defense industry
• It’s due to very little thinking about the failure
scenarios
• This is also an example of “one size fits all” error
handling.
• If error – log event.
• Not applicable for all errors.

• Text based software design specification
• The software shall Initialize.
• If initialization is unsuccessful the
software will make X attempts to
initialize.
• If initialization not successful after x
attempts the software shall log an
error.
• The software shall execute <Process
A>. If not successful it shall log an error.
B>. If not successful it shall log an error.
C>. If not successful it shall log an error.
• These are the problems which are easier
to see in a diagram
• In this example, order doesn’t matter
for process A, B or C.
• If not stated as such, the code might be
written to require sequential
processing which might result in timing
requirements not being met.
• Similarly the reverse can happen if the
specification is not clear.

The “Not operational” state is the missing safe state. The transitions to it are the
safe state returns. If the system cannot revert back when the faulted state is
entered that’s an example of no fallback or recovery. (In this example it should
transition back to initialization)
Initializing
Remote
operations
Online
operations
Pre-
launch
Launch
Not
operational

A system has a set of
3 lights modeled after
a traffic light.
This status light is attached to mission critical equipment in a human-less
factory. It is imperative that when the system is in a faulted state that
the status is set to allow persons watching from above to send out a
service technician before the fault causes a loss of product or equipment.
These are the specifications for the software:
• SRS #1 -The software shall display a red light if at least one system failure
has been detected.
• SRS #2 -The software shall display a yellow light if there are no system
failures and at least one system warning has been detected.
• SRS #3 -The software shall display a green light if there are no system
failures and no system warnings detected.
• SRS #25 -The software shall detect all system failures as per appendix B
within 2 seconds of the onset of the system failure
• SRS #26 -The software shall detect all system warnings as per appendix B
within 2 seconds of the onset of the system warning
Informative – Physically the light tower also has a blue light in addition to
red, yellow, green. The blue light isn’t used for this feature but it used by
other software features. There are 4 total light bulbs.
Functional FMEA – top level color display
Detailed FMEA – all alarms in appendix B
Interface FMEA – interfaces from each device to light tower
Status light failure analysis example
System wide
failure state
Warning state
Operational
state
D
e
v
i
c
e
1
D
e
v
i
c
e
2
D
e
v
i
c
e
3
D
e
v
i
c
e
n
…

• First identify the severity of all known hazards
Status Light failure analysis example
Severity Events Immediate effect Company effect
Catastrophic There is a system
failure but green light
is on or no light at all
No service person is sent to
equipment to fix system
failure
Loss of product and/or loss of
equipment. Potential loss of
productivity for entire factory.
Critical There is a system
failure but yellow light
is on
A service person is sent to
the equipment but not as
quickly as if the light
displays red
Loss of product for several
minutes.
warning but green light
is on or no light at all
A service person is not sent
to the equipment
There will eventually be a
failure that requires immediate
attention
Moderate There is no failure or
warning but red light is
on
this equipment when not
needed
Major inconvenience if it
happens at all
Moderate There is a system
on
this equipment sooner than
need be
happens regularly.
Moderate All of the lights are on,
or more than one light
is on
the equipment and doesn’t
know what’s wrong
It can take longer to service.
happens regularly.

• Brainstorm what can go wrong with implementation with light example regardless of
whether the specifications are correct
• More than one light is on at a time
• The lights blink
• No light is displayed
• The software displays a red light when no fault has occurred
• The software displays a yellow light when no warning has occurred
• The software displays a green light when there is a warning or failure
• The software displays the wrong light color when there are multiple failures
• The software displays the wrong light color when there are multiple warnings
• The software displays green after one failure is corrected even though there are
multiple failures
• The software displays green after one warning is corrected even though there are
multiple warnings
• The software takes a long time to change the light color when the status changes
• Place the above failure modes and root causes on the SFMEA template as implementation
faults
• Next, let’s analyze the software design specifications for faults in the specifications
themselves

• This is a stateful system, hence we know that faulty state management is
relevant
• Brainstorm all of the ways that this system can be problematic with regards to
sequences and statement management.
• Since this is a state-ful system draw a state transition table or state diagram
(see next slide)
• What’s not in the design specifications?
• Refer to the below list of possible root causes
• Add any additional root causes that you identify in brainstorming
Failure mode Generic root cause
Faulty
sequences
Required operations are specified in the wrong order
Required state transitions are incorrect or missing
Specification implies a dead state
Specification implies an orphan state
Prohibited transitions aren’t explicitly specified

• This is the required system state diagram for the status light
Initializing Operational Maintenance
Shut downPower
applied
Initialization
success
Operator Command
or faulted state detected
Fault
addressed Operator
command
Many of the faults are related to
jammed or misaligned material.
A common maintenance action is to
unjam the material. This can be done
without rebooting the equipment.
Other maintenance actions may
require a reboot of hardware. If the
reboot time exceeds 15 minutes that
will affect the factory production.
Red
light on
Yellow
light on
Green
light on
All faults
Addressed but
at least one
warning
All faults
addressed
and no
warnings
At least one
system fault
At least on
system
warning
All warnings
addressed

• The state diagram on the previous page is the required state for the status light
• The written specifications are converted to a state diagram and shown on bottom right
• The specifications specify how the lights are set, but don’t explicitly cover what
happens when the warning or failure is resolved
• Since software engineers are obligated to trace their code directly to a written
specification, it can’t be assumed that their code will clear the lights when the warning
or failures are cleared.
• Notice that there is a timing requirement for the setting of lights when an error is
detected but not for the resetting of the lights once error is cleared.
• This is an example of when a picture is most efficient for finding gaps in specifications
Red
light on
Yellow
light on
Green
light on
All faults
Addressed but
at least one
warning
At least on
system
warning
All warnings
addressed
Red
light on
Yellow
light on
Green
light on
At least one
system
warning
detected
At least one
system fault At least one
system fault
detected
This is what’s specifiedThis is what’s required

• Brainstorm this generic template against the illustration
• Identify specific root causes for the status light
Failure mode Generic root cause Specific root causes
Faulty
sequences/
state
transition
Required operations are
specified in the wrong
order
Doesn’t appear to be relevant from state
diagram
Required state transitions
are incorrect or missing
The state transition from yellow to green
is missing.
The specifications don’t state what lights
are displayed – if any – when in the
initializing mode
Specification implies a
dead state
Once the red light is on, it’s always on
until reboot of software
Orphan state See below
Prohibited transition isn’t
explicitly specified.
In this case the blue light is required to be
an “orphan”.A transition for the blue light
isn’t prohibited.

Failure mode and root cause Section
No.
Software
itemunder
consideration
Design requirement
(Requirement IDTag)
Related Design
requirement
(Requirement IDTag)
Software item
functionality
Failuremode
GenericRoot
Cause
Potential root cause
1
Statuslighttower
SRS 1 -The software
shall display a red
light if at least one
system failure has
been detected.
SRS 25The software
shall detect all system
failures as per appendix
B within 2 seconds of
the onset of the system
failure
The red light
is on when
the system
is in a failed
state
FaultySequences/StateTransitions
Specifications
implyadead
state
Dead state when
transitioning from red to
yellow- (there are no
specifications for
transitioning from failed
state to warning state)
2 SRS 2 -The software
shall display a
yellow light if there
are no system
failures and at least
one system warning
has been detected.
SRS 26The software
warnings as per
appendix B within 2
seconds of the onset of
the system warning
The yellow
light is on
when the
system is in
a degraded
state
Missingstate
transition
No transition from yellow
to green- (there are no
specifications for
transitioning from
warning state to
operational state)
3 SRS 1 -The software
shall display a red
light if at least one
system failure has
been detected.
SRS 25The software
failure
The red light
is on when
the system
is in a failed
state
Specifications
implyadeadstate
Dead state when
green- (there are no
specifications for
state to clear state)

Status light example
Failure mode and root cause Section
No.
Software
under
consideration
Design requirement
(Requirement ID
Tag)
Related Design
requirement
(Requirement IDTag)
Software item
functionality
Failuremode
GenericRoot
Cause
Potential root
cause
4 Status
light
tower
#1The software shall
display a red light if at
least one system
failure has been
detected. #2The
software shall display
a yellow light if there
are no system failures
and at least one
system warning has
been detected. #3 -
The software shall
display a green light if
there are no system
failures and no system
warnings detected.
SRS_25The software
failure. SRS_26The
software shall detect all
system warnings as per
appendix B within 2
seconds of the onset of
the system warning.
SRS_25The software
failure
The red light is
on when the
system is in a
failed state.The
yellow light is
on when the
system is in a
degraded state
The red light is
on when the
system is in a
failed state
FaultySequences/StateTransitions
Prohibitedtransitionsaren’t
specified
There is no
specification to
explicitly
prohibit
changing the
blue light.
5
Requiredstate
transitionsare
incorrectormissing
There is no
specification
for what light,
if any, is on
when
initializing at
startup

Generate CILMitigate
Analyze failure modes and root
causes
Identify
resources
Identify
equivalent
failure
modes
Identify
consequences
Identify local/
subsystem/
system
failure effects
Identify
severity
and likelihood
Identify
corrective
actions
Identify
preventive
measures
Identify
compensating
provisions
Analyze applicable
failure modes
Generate a
Critical
Items List
(CIL)
Identify
boundary
Set
ground
rules
Select
View
points
Identify
what can
go wrong
Gather
artifacts
Define
likelihood
and
severity
Select
template
and
tools
Revise RPN
Identify
riskiest
functions
For each
SW specifications, interfaces,

© Softrel, LLC 2020 464.0 Analyze consequencesFailureModeNo.
PotentialRootCause
EffectLevel(E)
DetectionMethod(s)
CTQRationale
Action(s)Taken
SystemHazardID(s)

© Softrel, LLC 2020 474.0 Analyze consequences
Severity Events Immediate effect Company effect
Catastrophic There is a system failure
but green light is on or no
light at all
No service person is sent to
equipment to fix system
failure
Loss of product and/or loss of
equipment. Potential loss of
productivity for entire factory.
Critical There is a system failure
but yellow light is on
A service person is sent to the
equipment but not as quickly
as if the light displays red
Loss of product for several
minutes.
warning but green light is
on or no light at all
A service person is not sent to
the equipment
There will eventually be a failure
that requires immediate
attention
Major There is no failure or
on
A service person is sent to this
equipment immediately when
not needed
Major inconvenience if it happens
at all
Minor There is no system
warning but yellow light
is on
equipment when not needed
regularly.
Minor There is a system
on
equipment sooner than need
be
regularly.
Minor All of the lights are on, or
more than one light is on
A service person is sent to the
equipment and doesn’t know
what’s wrong
It can take longer to service.
regularly.

No. Design requirement
(Requirement ID
tag)
Potential root cause Potential
effects of
failure
Potential effects
of failure
Effect
level (E)
1 The software shall
least one system
failure has been
detected.
Dead state when
yellow- (there are no
specifications for
state to warning state)
There is no
failure or
warning but red
light is on
A service person
is sent
immediately to
this equipment
when not needed
Major
display a yellow light
if there are no system
failures and at least
one system warning
has been detected.
No transition from
yellow to green- (there
are no specifications for
transitioning from
warning state to
operational state)
There is no
system warning
but yellow light
is on
A service person
is sent to this
equipment when
not needed
Minor
least one system
failure has been
detected.
Dead state when
green- (there are no
specifications for
state to clear state)
There is no
failure or
warning but red
light is on
A service person
is sent
immediately to
this equipment
when not needed
Major
4.0 Analyze consequences

Status light example
No. Design requirement
(Requirement ID tag)
Potential root
cause
Potential effects
of failure
Potential effects
of failure
Effect
level (E)
least one system failure
has been detected.
The software shall
display a yellow light if
there are no system
failures and at least one
system warning has been
detected.
The software shall
least one system failure
has been detected.
There is no
specification to
explicitly
prohibit
changing the
blue light.
Blue light is changed
when it shouldn’t be
Process status isn’t
known to factory.
Potential for mis-
processing.
Critical
5 There is no
specification for
what color, if
any, is displayed
at the initial
state.
If the equipment is
in failed stated,
factory can’t see the
status at all on
startup
Delay in sending
service person to
equipment
Critical
These two hazards weren’t covered in the original FDSC.
• If the blue light isn’t correct it will effect whether the factory knows the state of the
material being processed by the equipment. There could be an undetected
misprocess as worst case.
• If there is no light when equipment starts up and there will be a delay in sending
service person to equipment. That could lead to a backup in the factory.

• Likelihood is assessed AFTER severity is assessed to ensure that
catastrophic failure modes aren’t prematurely pruned from the
analysis
• Likelihood is a function of four things
• Existence likelihood – likelihood that the root cause exists in
the software
• Manifestation likelihood - How likely are the conditions that
cause the root cause to manifest into a failure
• Whether or not the failure mode/root cause is controlled
• How detectable the root cause is during the development
process
• Final likelihood for risk matrix =
Existence Likelihood * Manifestation likelihood * Control
• Detectability will be assessed separately on the risk matrix

1. First, determine if it’s known for sure that the failure mode/root
cause does in fact exist in the specifications or code.
2. If it the specification or code is itself deficient then likelihood of
existence is set to “high”
3. Otherwise, the below table is used to assess likelihood of
existence
Likelihood
of existence
Affected software
design
Past history Domain expertise
High
Very complex or
problematic
Has happened in every past
system or is known to be
present.
No experience with
this feature or product
Moderate Average complexity
Has happened at least once
in the past
Some experience with
this feature or product
Low
Very simple, not
problematic
Hasn’t happened in the past
and there’s no reason to
believe it will happen on this
system.
Significant experience
with this feature or
product

1. Firstly, determine if the failure mode/root cause is contingent upon another
failure – say a failure in the hardware. If the particular root cause will only
manifest itself when there is another failure then it’s likelihood can be no worse
than the likelihood of the related failure.
2. Secondly, determine if the failure mode/root cause could effect multiple
installed sites. If the root cause affects a feature used by most customers then
it’s frequency in operation is greater than a root cause in a feature not
frequently used.
The manifestation likelihood is set to no worse than the likelihood of the root
cause being related to another failure
Likelihood of
manifestation
Related to other failure Install base
High Not related to any other failure Could effect many installed sites, users
Moderate
Will happen if there is one HW
failure
Could effect multiple installed sites, users
Low
Will happen with multiple HW
failures
Could effect a few installed sites, users

•Examples
• BIT/Health monitoring software that monitors other software
• Interlocks can ensure that if the software provides conflicting
commands that the safest command is always executed.
•Review each row in the SFMEA, if there are any known
controls for this failure mode and root cause, identify
them.
Controls
High Multiple controls for the root cause
Moderate One control
Low No controls

The detectability of the failure mode/root cause depends on the operating
conditions required for it to be manifested. If the root cause is visible with
any operating condition then it’s almost certain to be detected in testing. If
the root cause is visible under an operating condition that cannot be
reproduced in a test environment then it’s detectability is improbable.
Likelihood Detection level
5-IMPROBABLE The failure mode can’t be reproduced in a development or
test environment.
4-LOW PROBABILITY The failure mode is visible and detectable only with fault
injection (faulty hardware, unexpected inputs, etc.)
3-MODERATE PROBABILITY The failure mode is visible and detectable with off nominal
testing (pressing cancel buttons, entering invalid data)
2-HIGH PROBABILITY The failure mode is visible and detectable with requirements
based testing
1-ALMOST CERTAIN The failure mode is visible and detectable under any set of
conditions

• Risk Matrices are typically company/project specific
• Here is an example
Likelihood Severity
5 –
Catastrophic
4 – Critical 3 – Major 2 – Minor 1 -
Negligible
5 – Almost
certain
Mitigate Mitigate Mitigate Evaluate Mitigation
not required
4- High Mitigate Mitigate Evaluate Evaluate Mitigation
not required
3 –
Moderate
Mitigate Evaluate Evaluate Mitigation
not required
Mitigation
not required
2 - Low Evaluate Evaluate Mitigation
not required
Mitigation
not required
Mitigation
not required
1 –
Improbable
Mitigation not
required
Mitigation
not
required
Mitigation
not required
Mitigation
not required
Mitigation
not required

No. Design
requirement
IDTag
Potential root
cause
Potential
effects of
failure
Effect
level
(E)
Occurrence level of failure mode (O) Detect-
ion level
of
failure
(D)
ControlsExistence
likelihood
Manifestation
likelihood
1 #1The
software
shall display
a red light if
at least one
system
failure has
been
detected.
Dead state
when
transitioning
from red to
green- (there
are no
specifications
for
transitioning
from failed
state to clear
state)
A service
person is
sent
immediately
to this
equipment
when not
needed
3-
Major
5-
None
5- This root
cause is
guaranteed
because
specification
is incorrect
3 - Will
happen
when there
is one HW
failure
5 –
Won’t
be
found
in any
test
• Likelihood is determined by the specific root cause and the control for that root cause – not the effect.
• This specific root cause is guaranteed to effect the design because the specification is insufficient. So the
existence likelihood = 5.
• This root cause is directly related to a hardware failure so the manifestation likelihood is assessed at 3.
• There is no control for the root cause
• The average likelihood = Average(5,5,3) = 4.33
• Since only requirements based testing is planned, and this is a missing specification there is virtually no
chance it will be found during testing so detectability is also assessed at 5. RPN = 3*4*5 = 60.

No. Design
require-
ment ID
tag
Potential root
cause
Potential
effects of failure
Effect
level (E)
Occurrence level of failure mode (O) Detect-
ion level
of failure
(D)
ControlsExistence
likelihood
Manifestation
likelihood
4 SRS
#1,#2,
#3
There is no
specification
to explicitly
prohibit
changing the
blue light.
Process status
isn’t known to
factory –
Potential for
misprocessing
4-
Critical
5.
None
3- Moderate
-----------------
3. Average
complexity
3. Similar
problems have
happened at
least once in the
past
3. Average
domain
experience
5- Likely
5. Could
effect many
installed sites
5. Not
related to a
HW failure
5.This
won’t be
detected
in
testing
because
only
require-
ements
are
tested
• The specification doesn’t prohibit the setting of the blue light by the status light feature. So, the question
is whether the code has been designed to prevent this.
• If this code does allow the prohibited transition it won’t be found in testing because of exclusive
requirements based testing.
• In the past, there has been state related problems.
• The software team has average experience with the application.
• This root cause is not related to a hardware failure.
• There is no control for the root cause. However, there are controls that can be mitigated.
• Hence likelihood is = Average(5,3,5) = 4.33 which is rounded down to 4.
• Risk Product Number (RPN) = 4*4*5= 80

FailureModeNo.
PotentialRootCause
EffectLevel(E)
DetectionMethod(s)
CTQRationale
Action(s)Taken
SystemHazardID(s)
5.0 Identify Mitigation

•Corrective action is from the development standpoint
and will depend on type of FMEA being performed
• Functional FMEA -corrective action may include changing the
specifications
• Interface and detailed FMEAs -corrective action may include
changing the design, code to correct the failure mode
• Process FMEA -corrective action may be the execution of a
particular practice or avoidance of a particular obstacle
•Examples of corrective actions that don’t apply to
software
• Replacing the software unit with an identical failed unit
• Operator repair of the software unit on site

•Adjust the RPN based on the assumption that
the corrective action is employed
•Don’t override the original RPN assessment
•Likelihood will change if the problem is
eliminated
•Severity will change only if the problem is
downgraded
•Detectability will change if the failure mode is
reviewed in design, code or fault injection test
procedures are developed.

No. Potential
root cause
Potential
effects of
failure
Effect
level
Detect-
ion
level
Occur-
rence
level
RPN Recommended Action(s) Residual
RPN
4 There is no
specification
to explicitly
prohibit
changing the
blue light.
Process
status isn’t
known to
factory –
Potential
for mis-
processing
4.
Critical
5.
Won’t
be
found
in any
test
4.
Likely
to
highly
likely
4*
5*
4=
90
Add a specification
statement prohibited any
changes of blue light by
light status code. Review
code to ensure that status
light never changes the
blue light display.
Monitor during endurance
tests to ensure the blue
light is never changed by
the status light code –
only the code that is
allowed to change the
blue light
RPN =4=
4*1*1
since root
cause is
mitigated
and the
code
review is
in place to
detect it
The corrective action is to add a specification for the prohibited state transition,
review the code to ensure there isn’t a transition for the blue light in the status code
(only the feature that is supposed to change the blue light). Prohibited transitions are
difficult to test so it would need to be monitored over the testing period.
The original RPN was 4*5*4. With the corrective actions, the root cause is mitigated
so the likelihood = 1 and detectability = 1. The adjusted RPN is then 4*1*1=4.

•Two failure modes are equivalent if the effect, severity,
likelihood and corrective action are the same
•During this step, identify the failure modes that are
equivalent to each other so as to consolidate the
corrective actions
•Don’t change the worksheet, just do the consolidation
in a separate worksheet
6.0 Generate a Critical Items List (CIL)

1. Add transitions from red to yellow, red to green, yellow to
green to specification and test plan
2. Add prohibited transitions of blue light from status light code
to specifications, code review and test plan.
3. Add specification for what lights should do upon initialization
and test plan.
The corrective actions could all be made at one time without
regard for RPN - or they could be implemented selectively to
address only the most critical RPN root causes.
If the code hasn’t been written yet, it’s often most efficient to
simply fix all defective requirements.
When the code is already written changing the defective
specifications poses a bigger risk from both a schedule and
software standpoint. Hence, the corrective actions are
implemented based on risk.

No
.
Design requirement ID
tag
Potential root
cause
Potential
effects of
failure
Effect
level
(E)
Detect-
ion level
(D)
Occur-
rence
level
(O)
RPN Recommended
Action(s)
Residual
occur-
rence
level (O)
Residual
effect
level (E)
Residual
RPN
1 The software shall display a
red light if at least one
system failure has been
detected. The software shall
display a yellow light if there
are no system failures and at
least one system warning
has been detected. The
software shall display a red
light if at least one system
failure has been detected.
There is no
specification to
explicitly prohibit
changing the blue light.
Process status
isn’t known to
factory.
Potential for
mis-
processing.
4.
Critical
5. Won’t
be found
in any test
4-Likely
to highly
likely
90 Add specification for
prohibited state
transition, review
the code to ensure
there isn't one,
monitor during
testing
1 1 4
2 There is no
specification for what
color, if any, is
displayed at the initial
state.
Delay in
sending
service person
to equipment
4.
Critical
3-Visible
with any
off
nominal
testing
3- Likely 36 Add specification for
what light is on, if
any, during
initialization
1 1 4
detected.
Dead state when
transitioning from red
to green- (there are no
specifications for
transitioning from
failed state to clear
state)
A service
person is sent
immediately
to this
equipment
when not
needed
3. Major4 – Visible
only with
fault
injection
testing
3- Likely 36 Change
specifications to add
transition from red
to green, add
transition to test
plan
1 1 3
yellow light if there are no
system failures and at least
one system warning has
been detected.
No transition from
yellow to green- (there
are no specifications for
transitioning from
warning state to
operational state)
A service
person is sent
to this
equipment
when not
needed
2. Minor4 – Visible
only with
fault
injection
testing
3- Likely 24 Change
transition from
yellow to green, add
transition to test
plan
1 1 2
detected.
Dead state when
transitioning from red
to yellow- (there are no
specifications for
transitioning from
failed state to warning
state)
A service
person is sent
immediately
to this
equipment
when not
needed
3. Major4 – Visible
only with
fault
injection
testing
3- Likely
3
36 Change
transition from red
to yellow, add
transition to test
plan
1 1 3

Severity Likelihood
5 –Very likely 4 – Likely 3 – Moderately
likely
2-Unlikely 1-
Mitigated
5 – Catastrophic
4- Critical #1
3- High #2 #3,#5
2 – Moderate #4
1- Negligible
Severity Likelihood
5 –Very likely 4 – Likely 3 – Moderately
likely
2-Unlikely 1-
Mitigated
5 – Catastrophic
4- Critical #1
3- High #3,#5 #2
2 – Moderate #4
1- Negligible
In the below table, the high RPN items are mitigated

What you
learned
• How to prepare for the software FMEA to
minimize cost and maximize effectiveness
• How to get in the right mindset
• How to analyze the failure modes that
apply to the entire software system
apply to a feature
apply to a specific software specification
• How to assess the consequences of each
failure mode
• How to assess the mitigations of each
failure mode
• How to track each failure mode to closure

More information
Software failure modes
effects analysis class
Online self guided training
Online instructor led training
On site training
Open session training
http://missionreadysoftware.com/training
Software FMEA toolkit http://missionreadysoftware.com/products
EffectiveApplication of Software
Failure Modes Effects Analysis
https://www.quanterion.com/product/
publications/effective-application-of-
software-failure-modes-effects-
analysis/

References
• [1] Delivering Military Software Affordably, Christian Hagen and Jeff
Sorenson, Defense AT&L, March-April 2012.
• [2]The Cold HardTruth About Reliable Software, AM Neufelder,
Version 2i, 2019.
• Mil-Std 1629A Procedures for Performing a Failure Mode, Effects
and Criticality Analysis, November 24, 1980.
• MIL-HDBK-338B, Military Handbook: Electronic Reliability Design
Handbook, October 1, 1998.
• Society of Automotive Engineers, “SAE ARP 5580 Recommended
Failure Modes and Effects Analysis (FMEA) Practices for Non-
Automobile Applications”, July, 2001.
• NASA, “NASA-STD 8729.1, Planning, Developing and Managing an
Effective Reliability and Maintainability (R&M) Program.”, 1998.
• NASA, “NASA Guidebook 8719.13B Software Safety”, July, 2004.
• W.EVesely, F.F. Goldberg, N.H Roberts, D.F. Haasl; “FaultTree
Handbook NUREG 0492”, US Nuclear Regulatory Commission, 1981

References for famous software related failures from history
• [DART]”4.2 DART Mishap, NASA, Dec. 2006” http://www.nasa.gov/pdf/167813main_RP-06-
119_05-020-E_DART_Report_Final_Dec_27.pdf
• http://www.nasa.gov/mission_pages/dart/main/index.html
• Overview of the DART Mishap Investigation Results For Public Release
• http://www.nasa.gov/pdf/148072main_DART_mishap_overview.pdf
• [DENAIR] Impact of the Delayed Baggage System, US Government Accountability Office, Oct 14,
1994
• http://www.gao.gov/products/RCED-95-35BR
• [DRUM] Dan Stockman, A Single ZeroTurnsTraining toTragedy-City built software linked to
deadly ’02 Army friendly fire, May 18, 2008
• http://www.journalgazette.net/apps/pbcs.dll/article?AID=/20080518/LOCAL10/805180378
• [GEMINIV] Barton C. Hacker, James M. Grimwood; “On the Shoulders ofTitans: A History of
Project Gemini, Chapter 11 –The Covered Wagon, 1977”
• http://www.hq.nasa.gov/office/pao/History/SP-4203/ch11-4.htm
• [INTEL]Associated Press, ”IntelTakes $475 Million ChargeTo Replace Flawed Pentium; Net
Declines”, Santa Clara, CA. Jan. 17, 1995.
• Intel’s discussion of the Pentium bug can be found here:
http://www.intel.com/support/processors/pentium/sb/CS-013007.htm
• The web site for Dr. Nicely (the person who originally detected the defect):
• http://www.trnicely.net/pentbug/pentbug.html
• http://www.apnewsarchive.com/1995/Intel-Takes-$475-Million-Charge-To-Replace-Flawed-
Pentium-Net-Declines/id-853864f8f1a74457adf39ab272181e08

• [KAL]Flight IntoTerrain Korean Air Flight 801, Boeing 747-300, HL7468, Nimitiz Hill,
Guam, August 6, 1997. National Transportation Safety Board Washington, D.C. 20594, Aircraft
Accident Report Controlled
• The NTSB report can be found here: www.airdisaster.com/reports/ntsb/AAR00-01.pdf
• [MARINER]Parker, P. J., “Spacecraft to Mars”, Spaceflight, 12, No. 8, 320-321, Aug. 1970
• http://nssdc.gsfc.nasa.gov/nmc/spacecraftDisplay.do?id=MARIN1
• [MARS] Douglass Isbell, Mary Hardin, Joan Underwood, “Mars Climate Orbiter Team Finds Likely
Cause of Loss, Sept. 30, 1999”
• “Mars Climate Orbiter Mishap Investigation Board Phase I Report" (Press release). NASA.
November 10, 1999. Retrieved February 22, 2013.
• http://sunnyday.mit.edu/accidents/MCO_report.pdf
• http://mars.jpl.nasa.gov/msp98/news/mco990930.html
• [NORAD] William Burr, “False Warnings of Soviet Missile Attacks During 1979-1980”, March 1, 2012.
• http://www2.gwu.edu/~nsarchiv/nukevault/ebb371/
• [PANAMA] “Investigation of an Accidental Exposure of Radiotherapy Patients In Panama, Report of
aTeam of Experts, 26 May–1 June 2001, International Atomic Energy Agency.”
• http://www-pub.iaea.org/MTCD/publications/PDF/Pub1114_scr.pdf
• http://www.fda.gov/Radiation-
EmittingProducts/RadiationSafety/AlertsandNotices/ucm116533.htm
• [PHOBOSI] Anatoly Zak, “Phobos-I Mission”, Jan. 15, 2012
• http://www.russianspaceweb.com/phobos.html
• http://nssdc.gsfc.nasa.gov/planetary/phobos.html

• [QANTAS72] AustralianTransport Safety Bureau “Qantas AirbusA330 accident Media
Conference”, Oct. 14, 2008.
• https://www.atsb.gov.au/newsroom/2008/release/2008_43.aspx
• [RAILCAR] Lena H. Sun, ”Metro replacing software that allowed train fire”,
Washington Post, April 13, 2007.
• http://www.washingtonpost.com/wp-
dyn/content/article/2007/04/12/AR2007041202061.html
• [SCUD] R.W.Apple Jr., “WAR INTHE GULF: Scud Attack; Scud Missile Hits a U.S.
Barracks”, Feb. 26, 1991.
• http://www.nytimes.com/1991/02/26/world/war-in-the-gulf-scud-attack-scud-
missile-hits-a-us-barracks-killing-27.html
• David Evans, “Software Blamed For Patriot Error, Army Says Scud Attacking
Barracks Couldn`t BeTracked”, June 06, 1991, ChicagoTribune.
• http://articles.chicagotribune.com/1991-06-06/news/9102200435_1_iraqi-scud-
missile-defective-software-tracking
• [SDIO] “The development of software for ballistic-missile defense, H. Lin, Scientific
American”, vol. 253, no. 6 (Dec. 1985), p. 51.
• [SF911] Phillip Matier & Andrew Ross, ”San Francisco 911 SystemWoes”, San
Francisco Chronicle, October 18, 1995, pgA1.
• http://www.sfgate.com/news/article/PAGE-ONE-S-F-Stalls-Shake-up-Of-911-
System-3022282.php

• [SOHO] “SOHO Mission Interruption Joint NASA/ESA”, Investigation Board, Final
Report,August 31, 1998.
• http://umbra.nascom.nasa.gov/sohOSOHO_final_report.html
• [SPIRIT] Ron Wilson, “The trouble with Rover is revealed”, EETimes, 2/20/2004.
• http://www.eetimes.com/document.asp?doc_id=1148448.
• [STS-126] NASA, “Shuttle System Failure Case Studies: STS-126, NASA Safety Center
Special Study, NASA,April 2009.
• [SURVEYOR]NASA, Mars Global Surveyor (MGS) Spacecraft Loss of Contact, NASA,
April 13, 2007.
• http://www.nasa.gov/mission_pages/mgs/mgs-20070413.html
• [TITANIV] J.Gregory Pavlovish, Colonel, USAF,Accident Investigation Board
President. “Titan IV B-32/Centaur/Milstar Report”.
• http://webcache.googleusercontent.com/search?q=cache:b7DaYU3zSOgJ:sunnyday.mit.edu
/accidents/titan_1999_rpt.doc+&cd=12&hl=en&ct=clnk&gl=us
• [THERAC]Angela Griffith, “Therac-25 Radiation Overdoses”, August 8, 2010.
• http://root-cause-analysis.info/2010/08/08/therac-25-radiation-overdoses/
• [USSR] David Hoffman, “I Had a Funny Feeling”, Feb. 10, 1999.
• http://www.washingtonpost.com/wp-srv/inatl/longterm/coldwar/shatter021099b.htm
• [WWMCCS] RichardW. Gutmann, ”ProblemsAssociated with the world wide military
command and control system” pp. 17, April 23, 1979.
• http://archive.gao.gov/f0302/109172.pdf

Guidance Comments
Mil-Std 1629A Procedures for Performing
a Failure Mode, Effects and Criticality
Analysis, November 24, 1980.Cancelled
on 8/1998.
Defines how FMEAs are performed
but it doesn’t discuss software
components
MIL-HDBK-338B, Military Handbook:
Electronic Reliability Design Handbook,
October 1, 1998.
Adapted in 1988 to apply to software.
However, the guidance provides only
a few failure modes and a limited
example. There is no discussion of
the software related viewpoints.
“SAEARP 5580 Recommended Failure
Modes and Effects Analysis (FMEA)
Practices for Non-Automobile
Applications”, July, 2001, Society of
Automotive Engineers.
Introduced the concepts of the
various software viewpoints.
Introduced a few failure modes but
examples and guidance is limited.
“Effective Application of Software Failure
Modes Effects Analysis”, November,
2014, AM Neufelder, produced for
Quanterion, Inc.
Identifies hundreds of software
specific failure modes and root
causes, 8 possible viewpoints and
dozens of real world examples.

Software Failure Modes Effects Analysis Overview

Recommended

More Related Content

What's hot (20)

Similar to Software Failure Modes Effects Analysis Overview (20)

Recently uploaded (20)

Software Failure Modes Effects Analysis Overview