Incident analysis
 

Does anybody have any guidance as to suitable tools to use for incident analysis by a flight safety officer?

If you haven't heard yet about them, I suggest you take a look at the 5M-model and Edward's SHEL model.

HFACS (Human Factors Analysis and Classification System) is supposed to be quite good. There's a book by Douglas Wiegman that sees a lot of use in my company.

Also, Sidney Dekker's Field Guide to Understanding Human Error might be a good starting point.

I hope it helps :-)

Thank you for your reply and your suggestions.

The top level document is the ICAO Manual of Accident and Incident Investigation Doc 9756-AN/965
Via aerolearner. ICAO DOC 9756 Part 2 and 3 [Archive] - PPRuNe Forums

As suggested in #2, a good grounding in HF is required, ideally over and above that given in the usual CRM / HF courses.
I don’t favor HFACS or similar systems as there is a risk that once the incident is placed in the appropriate box then the matter is closed; deeper analysis should always be undertaken.

The best tool is your mind – your ability to think about the incident in context and based on experience, yours and others; but of course remaining impartial and avoiding hindsight. I judge avoiding hindsight the most important and the most difficult aspect of any investigation.
Information gathering will probably focus on the people involved (the crew). They will remember much more than they think they have done, although their time-line can be incorrect and distorted; thus the task is to extract the relevant memories.

If you have FDR or similar records, first study them and construct a generalized view. Then interview each crew member individually without use or sight of this data. Update your model of events as the process progresses. If possible interviews should be in private and away from work connotations. I had some successes on coffee shops.
Many operators / people wish to include unions etc. IMHO these are more of a hindrance than a help; any external involvement must be strictly as an observer. If a crew does not have confidence in an operator’s safety system and impartiality, then you have your first finding of the investigation.
Any other inputs such as specialist technical questions, should be put to the interviewer away from the interview.

Repeat the process with all crew present, first without, and then with data.
Look at the crew’s body language. Arms and legs can indicate control actions. Arms, hands, finger positions and looks can indicate instrument indications and switching activities. Listen to the content and intonation of what is said; note the confidence of statements given.
Avoid leading questions, let the crew ‘construct’ their view of events from memory, don’t update or comment on this, but use it as a basis for future enhancing questions. The crew will have different memories and perceptions of events; don’t side with one or the other as both are valid. The investigators task is to ‘keep the peace’, explaining that everyone will see and remember things differently, and that their view, their opinion is ‘correct’; the objective is to learn and correct ‘the system’, not to allocate blame. Avoid the use of the term ‘error’ – try performance variability (search for Hollnagel).
The opinions – individual’s perception of events, can unlock aspects of understanding, training, experience, and organizational inputs – SOPs, safety culture.

For info:
Becky Milne explains the cognitive interview - OpenLearn - Open University

http://www.cs.virginia.edu/~cmh7p/iet2006-reading.pdf

http://www.skybrary.aero/bookshelf/books/228.pdf

http://brainimaging.waisman.wisc.edu...nts%202002.pdf

A Layman's Introduction to Human Factors in Aircraft Accident and Incident Investigation

http://www.it.uu.se/research/project...evention%20%22

HFACS | Wiegmann and Shappell Publications

This is pretty much the industry standard technique at the moment, and marries really well with Swiss Cheese / Accident Chain explanations. Used properly it should help identify all of the causal factors, not just put it in a convenient box (which as alf rightly says, does sometimes happen, and is inappropriate).

I have a colleague slightly at a loose end before he starts a new job as a university lecturer, who used the Wiefmann and Shappell model extensively on his PhD thesis into air accident causality. I imagine he'd be delighted to sell you a couple of days of his time and run a training course for you in-house. Drop me a PM if you wish and I'll give you his contact details.

G

My view is that the various different tools are better suited to different types of events. The trick is to use the right one.

One I have found very useful is Tap Root. I know a few airlines are using it successfully.

Many thanks to all of you. It would appear that I have a lot to learn.

I agree very much with ALF, taxonomies like MEDA/ HFACS etc are great ways to classify the findings of an investigation but the danger is the it becomes excuse maker pro. I would suggest a good error investigation course, I did one with Baines Simmons in the UK that was very good. Couple this with a well thought out taxonomy and a system like Safetynet/ AQD or QPulse to manage the information in a sensible way and to track and manage the resulting actions. You're almost a culture change away from a proper SMS.

I would like to reinforce my concerns over ‘box’ type investigation systems. Few if any provide insight as how to gather information prior to allocation, how to asses relevancy, or which box to place it in. Then with the ‘evidence’ in the box, it is easy to allow the conclusion to follow on from the contents.

As an example; the NTSB report into the LEX ‘wrong runway’ accident concluded that there was a crew ‘failure’ which implied ‘blame and train’ action. In this instance a ‘box’ type of investigation was used; this has been critiqued by Hollnagel in a presentation of an alternative process used for evaluating complex systematic issues. I am not suggesting that an operator requires the skills of a FRAM analysis for incidents, but the example (pages 19-24) shows how a different type of investigation (thinking) can identify contributing causes which arise from the different viewpoint. Thus at LEX - the crew’s performance varied as it would in normal operations, but it was the confluence of at least 6 contributing weaknesses which culminated in the accident. The Changing nature of Risk.

For more on FRAM see Course materials for FRAM - erikhollnagel2 - ‘FRAM Background’. (The LEX accident is expanded in ‘FRAM AA’)

I suspect that a flight safety officer will not require this type of tool, but the knowledge associated with it could (should) be applied during routine investigations to see what lurks in the complexity of interacting safety weaknesses and human performance found in most incidents.
A benchmark for most of this would be what happens in normal operations, and what differences are there between what is expected and what actually happens in operation – check your SOPs and the general understanding / interpretation of them – in context.

The US DOE publishes a comprehensive reference on human performance, CRM, and SMS aspects which would apply to aviation. There is a guide to investigation in Vol 2 starting on page 85, but beware of asking leading questions.

`Human performance improvement handbook'; `volume 1: Concepts and Principles'.
http://www.hss.doe.gov/nuclearsafety...09_volume1.pdf

`Volume 2: Human performance tools for individuals, work teams, and management'.
http://www.hss.doe.gov/nuclearsafety...09_volume2.pdf

Tripod Beta a reasonable incident analysis tool. You have to put some work in to use it but gives the user a clear model of what happened and a framework for a report.

PM