Skip to content

Analysis, debriefing, training

Processing takes time. PHASE lets you see it.

A piece of information reaches the flight deck. It is picked up, or it is not. A response follows, after a delay. PHASE lays all of this out on a single timeline: what was bearing on the crew, what they were doing, what the aircraft was doing, and what was still holding.

PHASE chart : Unstabilised approach continued, A320, flight XA 342
InterceptInterceptFinal approachFinal approachTouchdownTouchdownPF: perceived schedule pressure, reportedPF: end-of-duty fatigue, reportedWorkload management: speed reduction started latePM monitoring: not documentedPF: decision to continue, gate moved to 500 ftEnergy management: speed not brought back to targetFlaps 3 at 1,500 ftLate thrust reduction, nose loweredFlare and touchdown, pitch 2.2°Approach continued outside criteria at 1,000 ftHigh and fast at localiser interceptSpeed above targetRate of descent 1,150 ft/min at 900 ftFirm landing: +2.15 gWind 240/18 kt, gusting 28 ktModerate rain, wet runway, nightVectors shortened at a late stageFourth sector of the duty, night arrivalT−3:30T−3:00T−2:30T−2:00T−1:00T−30 sT01,500 ft1,500 ft1,000 ft, stabilisation gate1,000 ft, stabilisation gate500 ft500 ftTouchdownTouchdownT−1:20

A PHASE chart. Time runs left to right. From the bottom up: threats, aircraft state, then the crew. Move the cursor to see what was active at that moment.

What the chart reveals

Three moves are all it takes.

An investigation report tells an event theme by theme. The reader rebuilds the timeline in their head, one sentence at a time. On the chart, time is already laid out: all you have to do is walk through it.

  1. 01

    Stop the cursor

    At the chosen moment, the chart gathers everything that was active, stratum by stratum. “The checklist was omitted” becomes “at that moment, three tasks were running”.

  2. 02

    See a delay

    A signal is available at the bottom of the chart. Nothing lights up above it for twenty seconds. The delay is read off the axis; nobody has to argue for it.

  3. 03

    Open a record

    Behind every shape, a record states where the fact comes from and how solid it is: source, status, level of evidence. The timeline shows; the record establishes.

From event to training

From source material to case, from case to session.

  1. 1

    Source material

    A public investigation report, an FDM event with its ASR, a simulator session or a constructed scenario. The first step is to state what it can support, and what it leaves out.

  2. 2

    The case

    The narrative is broken down into time-stamped facts, each placed in one of ten strata, each with its own record. An AI-assisted analysis chain proposes; an analyst checks and decides.

  3. 3

    The chart

    The case becomes a chart you can read without a legend. You walk through it with the cursor, open the records, and discuss one object, one window, one uncertainty at a time.

  4. 4

    Training

    The EBT behaviours at stake, a scenario to replay in the simulator, the questions to ask in the debrief. What worked is kept, as much as what gave way.

The grammar in thirty seconds

Three shapes, three fills, a handful of marks.

The diamond marks what happens, the bar what lasts, the delta what was not expected. White protects or recovers, grey weakens, hatched informs. No colour, no severity scale: the chart reads just as well printed in black and white.

The signs, one by one

Shapes

  • Diamondwhat happens at an instant
  • Barwhat lasts; length is time
  • Deltadeviation from the expected (Error stratum)

Fills

  • Whiteprotects or recovers
  • Greyweakens
  • Hatchedneither

Marks

  • Dashednot known to the crew
  • AbcItalicstatus not established
  • Three chevronscontinues beyond the chart
  • One chevronend set by rule
  • Cutfrom here, something changes
  • Cursorinstant T

No colour on objects, no severity scale. The length of a bar expresses neither intensity nor severity.

Where PHASE comes from

Born in the field, formalised to last.

PHASE comes from flight instruction and event analysis. A first version, then called OOPS, was presented at HCI-Aero 2016, Paris, 14–16 September (the paper was not published in the proceedings; HAL hal-02468273). According to the 2016 paper, the model was used in training for two years and was employed by trainers with 240 French Air Force instructor pilots. Today’s method keeps its core: input through perception, output through action, and the question “how far ahead of the aircraft is the crew?”

paper presented at HCI-Aero 2016, Paris (not in the proceedings)
2016
paper presented at HCI-Aero 2016, Paris (not in the proceedings)
of use of the first version in training, according to the 2016 paper
2 years
of use of the first version in training, according to the 2016 paper
French Air Force instructor pilots trained with the first version
240
French Air Force instructor pilots trained with the first version
of instruction, roughly, behind the method’s designer
10,000 h
of instruction, roughly, behind the method’s designer

We say plainly what is not yet established: the effect of a coded chart on learning, and how consistently different analysts code the same case, are still to be measured. That validation is under way and open to research partners.

Founder

One of your events, read through time.

Send a message. The founder will reply personally, whether it is about analysing a case from your organisation, a training session or a research discussion.