GUIDANCE FOR IDENTIFYING UNSTABLE

APPROACH WITH FLIGHT DATA

April 2022
Rev 00

D4S is co-funded by
the European Union

1
DISCLAIMER ACKNOWLEDGEMENTS

The information and data presented is strictly for information The authors wish to acknowledge the contribution made by the
purposes only. It is obtained from Data4Safety (D4S) programme D4S participating members to thank them for their support in
databases comprised of data from EASA Member States and the the conduct of this work and in the preparation of this
aviation industry. It reflects knowledge that was current at the time document.
that the report was generated. Whilst every care has been taken in
preparing the content of the report to avoid errors, the
Data4Safety makes no warranty as to the accuracy, completeness
PHOTOCREDITS
or currency of the content. The D4S members, the DAP and any
other contributor shall not be liable for any kind of damages or European Union Aviation Safety Agency, 2021 Reproduction is
other claims or demands incurred as a result of incorrect, authorised provided the source is acknowledged.
insufficient or invalid data, or arising out of or in connection with
the use, copying or display of the content, to the extent permitted
by European and national laws. This document is not EASA official
recommendation or EASA official guidance material, and the
information contained should not be construed as legal advice.

2
The purpose of this document is to share the main aspects and lessons learnt
during the development of the Unstable Approach detection algorithm
Objectives of the document

Provide a complete characterisation of the logic for detection of

1 Unstable Approach events, the analysis window and their severity
classification based on FDM data

Present the different criteria and thresholds that the

2 identification of instabilities encompasses, so as to guide
industry practitioners on its implementation

Convey a set of assumptions, considerations and lessons learnt,

3 arising out of the work performed during the definition of Unstable
Approach detection algorithm, aiming to assist industry
practitioners when conducting safety analysis in this area

3
Acronyms

A/C Aircraft GPS Global Positioning System

ASR EASA Annual Safety Review HAT Height Above Touchdown

CE Consequential Event IATA International Air Transport Association

DAP Data Analytics Provider ILS Instrument Landing System

D4S Data4Safety N1 Engine 1 rotational speed of low speed spool

DH Decision Height PoC Proof of Concept

DS Directed Study SH Stabilization Height

EASA European Union Aviation Safety Agency SL Severity Level

FDM Flight Data Monitoring TAWS Terrain Avoidance and Warning System

FH Flight Hours TD Touchdown

FL Flight Level UA Unstable Approach

GA Go-Around Vref Reference speed

4
Definitions

Approaches with
Any approach with an instability condition triggered within the analysis window as per the instability criteria.
instability

Destabilization
Height at which any approach firstly encounters an instability (instability condition is triggered).
Height

Instability condition Aircraft parameter deviations from a specific nominal thresholds and within a minimum number of seconds, considered to be
/ trigger inducing instability.

Height at which any approach with an instability is fully stabilized, meaning the lowest height at which any instability criteria is no
Stabilization Height
longer triggered.

Any approach with the minimum required instability conditions triggered within the analysis window (1000ft – 0ft) as per the
Unstable Approach
instability criteria and height band.

5
 Introduction
CONTENT Unstable Approaches detection
Assumptions and lessons learnt
Conclusions
SECTION Introduction
Unstable Approaches detection

1
Assumptions and lessons learnt
Conclusions
Introduction

This document captures part of the work performed in the 2nd Directed Study
developed under the D4S programme, focused on Unstable Approaches

Context of the document

Data4Safety is a data collection and analysis programme that is delivering a set of Use Cases that will support the demonstration
supports the goal of ensuring the highest common level of safety of the D4S concept.
and environmental protection for the European aviation system.
To this end, the programme collects data from multiple sources Since the start of the programme, the aviation industry has
(including safety reports, flight data, surveillance data, and underwent significant change in a very short timeframe (e.g. COVID-
weather data) and utilises big data techniques to process it, 19 crisis). In order to underline the value of D4S in this rapidly
extracting insights on the location and nature of risks, and on the changing environment, the Steering Board decided to address a
performance of safety actions. safety topic of special concern through a Directed Study. This
Directed Study had the objective of providing insights to better
The programme, coordinated by EASA, is organised around safety understand the context of Unstable Approaches, the prevalence of
partners from the European Aviation Community: airlines, some associated risks as well as the impact of Go-Arounds as
manufacturers, ANSPs, NAAs, pilots, etc. As voluntary members of mitigation barriers to consequential events.
the programme they provide the necessary data and expertise for
the programme to achieve its stated goals.

The Proof of Concept (PoC) phase of the D4S programme was

launched at technical level in June 2016. Among the main
outcomes expected from this phase, the Programme

8
Introduction

To enable the analysis of Unstable Approaches, it is first essential to define a

common understanding of the set of parameters that constitute ‘instability’
What do we understand by ‘Unstable Approach’?

There is a general understanding at industry level concerning

what could be considered as an Unstable Approach. It is
commonly understood as: “an undesired state by which an
aircraft arrives at the RWY threshold too high, too fast, out
of alignment with the runway centre-line or incorrectly
configured in a way that is not prepared to land”.

However, there are many variables that might be considered

within the industry to identify such undesired aircraft state,
given the wide variety of aircraft types and operational
environments at different airports. As a result, different
methodologies and criteria have been developed in the
industry which, although they are the same in essence, differ
in terms of particular parameters and thresholds analysed.

Commonly, instability conditions to be analysed include:

 Approach speed above / below the desired reference

speed

 Vertical speed too high

 Aircraft misconfiguration (landing gear or flaps)

 Engine thrust level Picture: EASA

 Approach path deviations

9
Introduction

A standardised criteria to identify ‘Unstable Approaches’ is defined at D4S level,

as a result of close cooperation between participating FDM and data experts
Defining a standardised criteria for identification of ‘Unstable Approaches’

While specific criteria might work from an organisational To that end, a multidisciplinary approach, through
perspective, a common criteria is required for the analysis of collaboration and communication between all industry
data at industry-wide level, as general as possible, to drive members participating in the D4S, was applied to develop a
conclusions that were relevant and coherent with each standardised criteria to identify Unstable Approaches based
operational environment and aircraft type. on FDM data.

Figure 1: Illustrative figures for identification of Unstable Approach analysis window and instability criteria assessed

Within the approach window (<1000ft), the following set of instability criteria
Flight path
(triggers) are evaluated:

Fast descent Low thrust High/Low airspeed

Late flap or gear Unstable attitude

TAWS alerts
extension (pitch, roll) Approach
1000ft threshold window

Height
High/Low glideslope High/Low localizer
deviation deviation

Each criterion is evaluated at different height bands and thresholds in order to Evaluation of
classify it into a severity level. Two bands are differentiated above/below 500ft criteria

An approach is considered unstable when a minimum number of criteria are Touchdown

met. Three criteria must be met above 500ft while only one instability 0
condition triggers the identification of UA below that threshold. 0 0,2 0,4 0,6 0,8 1 1,2
Time

The presented UA detection algorithm is not to be understood as the only valid definition, but as a UA metric that has been tested on a large
number of aeroplanes and could be used as a start, as it has the advantage of not being aircraft-type specific. Therefore, operators should not
hesitate to try different threshold values when found more relevant for their fleet. Furthermore, the algorithm has been validated only with
turbojet aeroplanes with a MCTOM of over 27,000 KG, not with turboprop or business jet aeroplanes.

10
SECTION Introduction
Unstable Approaches detection

2
Assumptions and lessons learnt
Conclusions
Unstable Approaches detection

The pool of FDM and data experts within the D4S programme defined and
developed a standardised algorithm for detection of Unstable Approaches
‘4’ steps for the definition of the Unstable Approach algorithm

1
A methodology and algorithm were developed, capable of harmoniously identify Unstable
Approach events on FDM data for multiple operators and aircraft types, based on the
combination of 12 different operational criteria

Multiple iterations of refinement, validation, results discussion with experts and data quality

2
analysis were conducted to ensure that Unstable Approaches detection results were reliable.
As part of this process, conditions for the identification of instabilities during approach were
reviewed and modified when deemed necessary. The consistency of its definition was tested
by using a data sample covering around 1.4 million flights from 8 different aircraft models

3
By encountering several data quality issues throughout the analysis of FDM data in the
context of the Unstable Approaches definition, relevant insights were extracted on how to deal
with flight parameters and the criteria for instabilities detection during approaches

4
Once the Unstable Approach methodology and algorithm is thoroughly defined and validated,
the identification of those events is ready to be used on the development of specific and in-
depth analysis

12
 Unstable Approaches detection

The Unstable Approach algorithm intends to identify those cases where the
approach was not stabilized or was destabilized shortly before landing
Overview of the Unstable Approach identification logic

Flight path – Within the approach window, the following

2.000
set of criteria are evaluated:

High/Low
Fast descent Low thrust
airspeed
1.500
Unstable
Late flap or
TAWS alerts attitude (pitch,
gear extension
Approach roll)
window
Height threshold High/Low High/Low
Height

1.000 glideslope localizer

deviation deviation

– Each criterion is evaluated at different height

500
bands and thresholds in order to classify it
into a severity level

– An approach is considered unstable when a

Touchdown
minimum number of criteria are met
0
0 5 10 Time 15 20 25

13
Unstable Approaches detection

A total of 12 criteria are set to be evaluated in parallel and for each approach,
within different height ranges and, in some cases, with multiple thresholds
Detailed ‘instability’ criteria and thresholds for Unstable Approach detection

Height Height
Instability conditions 1st threshold 2nd threshold Instability conditions 1st threshold 2nd threshold
range range

>(Vref + 20kt) >(Vref + 35kt) High pitch >10º and >15º and
High airspeed
[3s] [3s] attitude <-3º [3s] <-10º [3s]
Aircraft
1000ft
handling and
configuration to 50ft
Low <(Vref – 5kt) <(Vref – 10kt) High roll >15º and >30º and
airspeed [3s] [3s] attitude <-15º [3s] <-30º [3s]
1000ft
to 50ft
High
Energy Fast descent <-1200fpm <-1500fpm
glideslope >1 dot [5s] >2 dots [5s]
management (vertical speed) [5s] [3s]
deviation

Low thrust Fleet specific (1st percentile) Aircraft path Low glideslope 1000ft
<-1 dot [5s] <-2 dots [5s]
(N1) [10s] management deviation to 500ft

Excessive
Modes 1 (Alert & Warning), 2 >1 dot (left or >1.5 dots (left
TAWS alerts localizer
(Alert & Warning), 4 (Terrain) right) [5s] or right) [5s]
deviation

Late flap Any change > 2 degrees 1000ft

extension or 1 notch to 0ft
Aircraft
handling and
configuration
Late gear Any deployment
extension or not deployed

14
Unstable Approaches detection

A four-level severity classification is also proposed to categorise criteria based

on both the height at which they occurred and the severity of the occurrence
Unstable Approach identification and severity classification

2.000 Flight path Unstable Approach identification

Approach window – For an approach to be classified as Unstable, there must
- Severity thresholds + be:
Height threshold 1
 At least 3 distinct criteria triggered at severity
1000ft level 1 or 2A; OR
SL1 SL2A
1.500

 At least 1 criterion triggered at severity level 2B

or 3

– An approach is classified as the highest severity level of

Height

1.000 its criteria

Height threshold 2  For example, if an event has two criteria triggered
500ft at SL1 but one at SL2A, the UA is ultimately
classified as SL2A
SL2B SL3
! For those criteria that only have one
500

threshold (e.g. Low Thrust or Late

Flap Extension), SL2A and SL3 are
0 Touchdown used for each height band
0 5 10 15 20 25

15
SECTION Introduction
Unstable Approaches detection

3
Assumptions and lessons learnt
Conclusions
 Assumptions and lessons learnt

Lessons learnt from the implementation of

the Unstable Approach FDM logic provide
valuable insights for industry practitioners
Throughout various iterations and validations over the Unstable
Approach detection methodology – under the D4S programme –
some relevant outcomes and lessons learnt can be generally applied in
the definition of detection algorithms for organizations’ FDM programmes
and/or safety analysis

Figure 2: D4S Unstable Approach detection criteria and thresholds

The following pages provides some key points and findings derived from the implementation of the logic for the
detection of Unstable Approaches, detailing the rationale underlying the selection of different criteria and thresholds

17
Assumptions and lessons learnt

The establishment of a maximum height range in which to consider potential

Unstable Approaches is required to define the rest of algorithm criteria

1 Height ranges

1st 2nd Height 1st 2nd Height

threshold threshold range threshold threshold range

>(Vref + 20kt) >(Vref + 35kt) Late gear Any deployment 1000ft

High airspeed
[3s] [3s] extension or not deployed to 0ft

<(Vref – 5kt) <(Vref – 10kt) High pitch >10º and <- >15º and
Low airspeed
[3s] [3s] attitude 3º [3s] <-10º [3s]
1000ft 1000ft
to 50ft to 50ft
Fast descent <-1200fpm <-1500fpm High roll >15º and >30º and
(vertical speed) [5s] [3s] attitude <-15º [3s] <-30º [3s]

KEY POINTS
Low thrust Fleet specific (1st percentile) High glideslope
>1 dot [5s] >2 dots [5s]
(N1) [10s] dev.
The upper limit of 1,000ft above airfield elevation for instability conditions is set as a mean of the
definition of the rest of criteria and a common standard consideration
Modes 1 (Alert & Warning), 2 Low glideslope
among industry experts. 1000ft
TAWS alerts <-1 dot [5s] to
<-2 dots [5s]
(Alert & Warning), 4 (Terrain) deviation
– Findings: even though approaches may1000ft be identified above this threshold (e.g. below 2,000ft), 500ft the
to 0ft
detection of Unstable Approaches with the standardisedExcessive
criteria presented in this document is defined
Late flap Any change > 2 degrees >1 dot (left >1.5 dots (left
for a window height
extension or below
1 notch1,000ft. Otherwise, an overhaul of dev.
localizer many oforthe criteria
right) [5s] thresholds would be
or right) [5s]
necessary to adapt for the extended flight period

18
Assumptions and lessons learnt

The criteria and thresholds of airspeed conditions are generalised, but the
computing method may depend on the data available or specific aircraft types

2 Airspeed condition

1st 2nd Height 1st 2nd Height

threshold threshold range threshold threshold range

>(Vref + 20kt) >(Vref + 35kt) Late gear Any deployment 1000ft

High airspeed
[3s] [3s] extension or not deployed to 0ft

<(Vref – 5kt) <(Vref – 10kt) High pitch >10º and <- >15º and
Low airspeed
[3s] [3s] attitude 3º [3s] <-10º [3s]
1000ft 1000ft
to 50ft to 50ft
Fast descent <-1200fpm <-1500fpm High roll >15º and >30º and
(vertical speed) [5s] [3s] attitude <-15º [3s] <-30º [3s]

KEY POINTS
Low thrust Fleet specific (1st percentile) High glideslope
>1 dot [5s] >2 dots [5s]
(N1) [10s] dev.
The high / low airspeed ‘instability’ condition compares the aircraft airspeed during approach against the
reference speed (Vref), plus some operational margins. Low glideslope
Modes 1 (Alert & Warning), 2
1000ft
TAWS alerts <-1 dot [5s] <-2 dots [5s] to
(Alert & Warning), 4 (Terrain) deviation
– Findings: some aircraft may not have the Vref available or not sufficiently reliable to be used in 500ft
1000ft the
to 0ft
Unstable Approach detection. In that case and considering the importance of >1.5
this dots
criteria in the
Late flap Any change > 2 degrees Excessive >1 dot (left (left
evaluation of instabilities,
extension or 1 notcha value can be estimated bylocalizer
using landing
dev. charts
or right) of
[5s]specific aircraft
or right) [5s] models,
as a function of the flap position and the Gross Weight

19
Assumptions and lessons learnt

The Vref value could be estimated by using landing charts of specific aircraft
models, as a function of the flap position and the Gross Weight

2 Airspeed condition – Three (3) potential ways of using the ‘Vref’ depending on the data available

A Use decoded Vref parameter Figure 3: Aircraft reference speed table (Vref) example

– Use the Vref recorded in the

FDM database FLAPS (º)
WEIGHT (1,000 KG)
40 30
B Both GW & flap parameters 85 160 168
available
80 155 163
– Estimate Vref as a function of
Gross Weight and flap 75 151 158
configuration 70 146 153
65 141 148
C Only GW parameter available 60 135 142
– Estimate Vref as a function of 55 128 136
Gross Weight but assuming 50 122 129
constant flap configuration (e.g.
Flap 40), in case this parameter 45 115 122
is not available or presents any 40 108 115
other quality issue

20
Assumptions and lessons learnt

The thresholds established in the fast descent condition are important so as

not to end with excessive false positive triggers in UA detection

3 Fast descent condition

1st 2nd Height 1st 2nd Height

threshold threshold range threshold threshold range

>(Vref + 20kt) >(Vref + 35kt) Late gear Any deployment 1000ft

High airspeed
[3s] [3s] extension or not deployed to 0ft

<(Vref – 5kt) <(Vref – 10kt) High pitch >10º and <- >15º and
Low airspeed
[3s] [3s] attitude 3º [3s] <-10º [3s]
1000ft 1000ft
to 50ft to 50ft
Fast descent <-1200fpm <-1500fpm High roll >15º and >30º and
(vertical speed) [5s] [3s] attitude <-15º [3s] <-30º [3s]

Low
KEY thrust
POINTS Fleet specific (1st percentile) High glideslope
>1 dot [5s] >2 dots [5s]
(N1) [10s] dev.

Setting the limiting value and the duration of a fast descent exceedance, as it is proposed, is found 1000ft
essential Modes 1 (Alert & Warning), 2 Low glideslope
TAWS alertsto conclude with a valid instabilities detection criterion. <-1 dot [5s] <-2 dots [5s] to
(Alert & Warning), 4 (Terrain) deviation
1000ft 500ft
– Findings: when using lower values (e.g. <-1000fpm)
to 0ft and shorter time periods (e.g. 3 seconds) in the 1st
Latethreshold
flap forAny
thechange
fast > 2 degrees
descent Excessive
criteria, the proposed UA algorithm >1 dot (left high
captures >1.5 dots (leftof false
amounts
extension or 1 notch localizer dev. or right) [5s] or right) [5s]
positive triggers which may compromise the overall accuracy of the methodology

21
Assumptions and lessons learnt

Low thrust criterion for Unstable Approaches detection is found more accurate
and reliable when defined at a product-specific level (1/2)

4 Low thrust condition

KEY POINTS
1st 2nd Height 1st 2nd Height
threshold threshold range threshold threshold range
In the context of the UA detection algorithm, a
High airspeed
>(Vref + 20kt) >(Vref + 35kt) generalised
Late gearcondition isAnyimplemented
deployment to standardise
1000ft
[3s] [3s] N1 threshold. Thus,or this
the extension not deployed to 0ft at
criteria is fine-tuned
product-specific level (aircraft and engine), by using
Low airspeed
<(Vref – 5kt) <(Vref – 10kt) the High pitch
1st percentile >10º
of N1and <-
values for>15º
all and
approaches and
[3s] [3s] attitude 3º [3s] <-10º [3s]
1000ft engines of each product as the threshold. 1000ft
to 50ft to 50ft
Fast descent <-1200fpm <-1500fpm – Findings:
High roll >15º and >30º and
(vertical speed) [5s] [3s] attitude <-15º [3s] <-30º [3s]
 Better accuracy is achieved by fine-tuning this
parameter at product-specific level based on
Low thrust Fleet specific (1st percentile) High glideslope
the distribution of N1>2values,
>1 dot [5s] dots [5s] instead of
(N1) [10s] dev.
establishing the same threshold across all
products (e.g. N1<35% / N1<30%) 1000ft
Modes 1 (Alert & Warning), 2 Low glideslope
TAWS alerts Additionally,<-1the
deviation dot [5s] <-2of
duration dotsthe to
[5s] low thrust
(Alert & Warning), 4 (Terrain)
1000ft condition is defined at 10 seconds as seen 500ftthe
to 0ft most representative time period under this
Late flap Any change > 2 degrees Excessive >1 dot (left >1.5 dots (left
extension or 1 notch criterion
localizer dev. or right) [5s] or right) [5s]

22
Assumptions and lessons learnt

Low thrust criterion for Unstable Approaches detection is found more accurate
and reliable when defined at a product-specific level (2/2)

4 Low thrust condition

– Despite the fact that some operators could establish Criteria

their own N1 thresholds for their products (some may
not), there is a need to define a generalised approach For each product (aircraft & engine):
across all product types
– Threshold: maximum value of N1(%) for the
1st percentile
– As such, a method using the 1st percentile could be
implemented, demonstrated to be consistent and – Condition: N1 < Threshold for 10 seconds
reliable

Figure 4: Illustrative example of N1 value distribution during approach (1,000ft to 50ft)

Threshold Under 1st percentile

Over 1st percentile

N1(%)
registers

30% 40% 50% 60%

N1(%) values

23
Assumptions and lessons learnt

Terrain Avoidance and Warning System (TAWS) criterion considers only Modes
one (1), two (2) and four (4) to prevent false UA triggers detection

5 TAWS condition
KEY POINTS
1st 2nd Height 1st 2nd Height
threshold threshold range threshold threshold range
When evaluating TAWS condition, including only
High airspeed
>(Vref + 20kt) >(Vref + 35kt) triggers over Mode 1 Any
Late gear (Alert & Warning), Mode
deployment 1000ft 2
[3s] [3s] (Alertextension or not 4
& Warning) and Mode (Terrain) is foundtoto
deployed 0ftbe
the most representative in terms of Unstable
<(Vref – 5kt) <(Vref – 10kt) High pitch
Approaches events>10º and <-
identification >15º and
Low airspeed
[3s] [3s] attitude 3º [3s] <-10º [3s]
1000ft 1000ft
– Findings: to 50ft
to 50ft
Fast descent <-1200fpm <-1500fpm High roll >15º and >30º and
(vertical speed) [5s] [3s] Mode 5 (Glideslope)
attitude <-15º [3s] is not found
<-30º [3s] appropriate
to be included due to the generation of too
Low thrust Fleet specific (1st percentile)
many false positives and the overlap made with
High glideslope
(N1) [10s] the Low/High
dev.
>1 dotGlideslope
[5s] >2 dots [5s]
criterion in the
present UA logic
1000ft
Modes 1 (Alert & Warning), 2 Low glideslope
TAWS alerts <-1 dot [5s] <-2 dots [5s] to
(Alert & Warning), 4 (Terrain) deviation
1000ft 500ft
to 0ft
Late flap Any change > 2 degrees Excessive >1 dot (left >1.5 dots (left
extension or 1 notch localizer dev. or right) [5s] or right) [5s]

24
Assumptions and lessons learnt

Late flap extension criterion evaluates significant flap changes during an

approach window, yet excludes non-deployment occurrences

6 Late flap extension condition

KEY POINTS
1st 2nd Height 1st 2nd Height
threshold threshold range threshold threshold range
Late flap extension condition would be triggered when
High airspeed
>(Vref + 20kt) >(Vref + 35kt) thereLate
is any
gearchange greater than two (2) degrees
Any deployment (or 1
1000ft
[3s] [3s] notch) below 1,000 feet, as a single threshold for0ftthe
extension or not deployed to
criterion.
<(Vref – 5kt) <(Vref – 10kt) High pitch >10º and <- >15º and
Low airspeed
[3s] [3s] – attitude
Findings: 3º [3s] <-10º [3s]
1000ft 1000ft

Fast descent <-1200fpm <-1500fpm

to 50ft A roll
High ‘not deployed’ flaps is >30º
>15º and includedtoin50ft
not and the
(vertical speed) [5s] [3s] logic, as it <-15º
attitude would[3s]change<-30º
the [3s]
nature of the
condition and may be mostly capturing a
Low thrust Fleet specific (1st percentile)
different operational case; principally a
High glideslope
(N1) [10s] technical >1
dev.
dot [5s]
flaps problem >2 dots [5s]
rather than a
procedural issue, thereby jeopardising the
actual instabilities identification 1000ft
Modes 1 (Alert & Warning), 2 Low glideslope
TAWS alerts <-1 dot [5s] <-2 dots [5s] to
(Alert & Warning), 4 (Terrain) deviation
1000ft 500ft
to 0ft
Late flap Any change > 2 degrees Excessive >1 dot (left >1.5 dots (left
extension or 1 notch localizer dev. or right) [5s] or right) [5s]

25
 Assumptions and lessons learnt

Any deployment or non-deployment of the landing gear within the height

range window is considered as a UA trigger

7 Late gear extension condition

KEY POINTS 1st 2nd Height 1st 2nd Height

threshold threshold range threshold threshold range

Either a landing>(Vref
High airspeed gear deployment
+ 20kt) or a non-
>(Vref + 35kt) Late gear Any deployment 1000ft
[3s] [3s]
deployment during the 1,000ft to 0ft approach extension or not deployed to 0ft
period is considered as an Unstable Approach trigger.
Furthermore, <(Vref – 5kt) <(Vref – 10kt) High pitch >10º and <- >15º and
Low airspeeda two consecutive conditions without
[3s] attitude 3º [3s] <-10º [3s]
value changes for this[3s]
sensor parameter is included
1000ft 1000ft
to mitigate potential quality issues to 50ft to 50ft
Fast descent <-1200fpm <-1500fpm High roll >15º and >30º and
(vertical speed) [5s] [3s] attitude <-15º [3s] <-30º [3s]

Low thrust Fleet specific (1st percentile) High glideslope

>1 dot [5s] >2 dots [5s]
(N1) [10s] dev.

1000ft
Modes 1 (Alert & Warning), 2 Low glideslope
TAWS alerts <-1 dot [5s] <-2 dots [5s] to
(Alert & Warning), 4 (Terrain) deviation
1000ft 500ft
to 0ft
Late flap Any change > 2 degrees Excessive >1 dot (left >1.5 dots (left
extension or 1 notch localizer dev. or right) [5s] or right) [5s]

26
 Assumptions and lessons learnt

Glideslope and localizer conditions detect prolonged deviations in aircraft

approach path, evaluated within a thoroughly selected height range window

8 Glideslope and Localizer condition

KEY POINTS 1st 2nd Height 1st 2nd Height

threshold threshold range threshold threshold range

TheHigh
defined thresholds
airspeed
>(Vref +for glideslope
20kt) >(Vref + (high
35kt) or low) Late gear Any deployment 1000ft
deviations are [3s]
and excessive localizer [3s] set to capture a extension or not deployed to 0ft
prolonged (5 seconds) erroneous approach path,
within <(Vref – 5kt) <(Vref – 10kt) High pitch >10º and <- >15º and
Low the 1,000ft to
airspeed 500ft analysis window.
[3s] [3s] attitude 3º [3s] <-10º [3s]
1000ft 1000ft
– Findings: to 50ft
to 50ft
Fast descent <-1200fpm <-1500fpm High roll >15º and >30º and
 The
(vertical trigger of[5s]
speed) this criterion [3s]
is substantially attitude <-15º [3s] <-30º [3s]
affected by the height range window,
dependent Fleet
Low thrust
on operational conditions and
specific (1st percentile) High glideslope
the actual airfield, together with potential >1 dot [5s] >2 dots [5s]
(N1) [10s] dev.
sensor data issues that may be present
1000ft
enable Modes
 Toalerts 1 (Alert & Warning),
a generalisation of this2 criterion, Low glideslope
TAWS <-1 dot [5s] <-2 dots [5s] to
(Alert & Warning), 4 (Terrain) deviation
the lower end of the window range is limited 1000ft 500ft
to 500ft. Nevertheless, the height range could to 0ft
Late flap Any change > 2 degrees
be modified for specific operations and Excessive >1 dot (left >1.5 dots (left
extension or 1 notch localizer dev. or right) [5s] or right) [5s]
airfields, if deemed necessary (e.g. 200ft for
glideslope and 50ft for localizer)

27
Assumptions and lessons learnt

The glideslope and localizer deviation parameters may generate UA events at a

higher rate than expected, thus needing in-depth analysis and considerations

8 Glideslope and Localizer condition – Issue diagnostic

Main contributors to high rates

Parameter deviation at low height

01. Parameters usually behave as expected up to the end of the approach (∼200ft for
glideslope, ∼50ft for localizer), where fast growth of deviation occurred, thereby
triggering unusual instabilities events

Operational-specific cases
02. Erratic parameter evolution may be observed due to late interception or non-
interception of glideslope / localizer signals, generating false positive UA events

28
Assumptions and lessons learnt

Glideslope and localizer may not behave as expected near the low height
threshold, presenting abnormal deviations and triggering false UA events

8 Glideslope and Localizer condition –01.

01. Parameter deviations at low heights

Example of glideslope deviations Example of localizer deviations

50ft HAT
LEMD LEBL 200ft HAT
200ft HAT

Pressure altitude Glideslope Localizer

FINDINGS

 Deviations usually take place near the end of the approach, but just before a theoretically reasonable
height threshold (∼200ft for glideslope, ∼50ft for localizer) for triggering the criterion

 These deviations generate false UA events even when the parameters had actually behaved as expected
throughout the approach window

29
Assumptions and lessons learnt

Glideslope and localizer may show erratic behaviour due to late or non-
interception of signals when approaching specific runways

8 Glideslope and Localizer condition –02.

02. Operational-specific cases

Scenarios related to interception of glideslope / localizer

A
 Intercepted glideslope / localizer with good behaviour of the parameter along the approach phase

B
 Late interception of the glideslope / localizer with erratic behaviour of the parameter at the beginning of the approach


C Not intercepted glideslope / localizer, with erratic behaviour of the parameter along the approach phase (visual approach /
instrumental guidance not available)

A Intercepted B Late interception C Not intercepted

30
Assumptions and lessons learnt

Late interceptions may capture operations that took a long period to align with
the runway, triggering glideslope / localizer conditions in the approach window

8 Glideslope and Localizer condition –02.

02. Operational-specific cases

A Intercepted B Late interception C Not intercepted

Late interceptions

 Glideslope and localizer may have late interceptions in some flights,

stabilising at the end of the approach near the touchdown point
 Late interceptions depend on the aircraft operation and mainly
occurs when the approach takes too long to align with the runway
 Specific filtering conditions could be applied by practitioners over
individual cases, considering concrete operations manoeuvres

1000ft (HAT) 500ft (HAT) 200ft (HAT)

Glideslope
Localizer

31
Assumptions and lessons learnt

Glideslope / localizer signals may not intercept at any moment during the entire
approach, due to the lack of ILS systems at specific landing runways

8 Glideslope and Localizer condition –02.

02. Operational-specific cases

A Intercepted B Late interception C Not intercepted

No intercepted

 Glideslope and localizer may present random values throughout the

entire approach
 No interception of glideslope and localizer are mainly linked to a
lack of guidance systems at the destination runway
 Specific filtering conditions could be applied by practitioners over
individual cases, based on landing runways and the presence of ILS
systems

200ft (HAT) Glideslope

Localizer

32
SECTION Introduction
Unstable Approaches detection

4
Assumptions and lessons learnt
Conclusions
Conclusions

The D4S programme defined a generalised Unstable Approach

detection methodology, providing valuable insights for the industry
Key conclusions from the document

As part of the work performed in the Data4Safety programme, an algorithm and a comprehensive
methodology to harmoniously identify Unstable Approach events in FDM data were developed,
which is based on the combination of 12 operational criteria and valid for multiple operators and
aircraft types

Due to the provenance of the FDM data source used in the Data4Safety programme, the Unstable
Approach logic presented in this document is accordingly generalised, enabling its applicability for
any operator or product type under analysis

Outcomes from the performed iterations, validations and results discussions over the presented
Unstable Approach criteria, provide valuable insights over how to deal with different flight
parameters to extract instability conditions during approaches

The methodology described in this document allows practitioners to systematically identify

Unstable Approach events whilst enabling further in-depth and specific analysis for industry
practitioners or at an organisational level

The presented UA detection algorithm is not to be understood as the only valid definition, but as a UA metric that has been tested
on a large number of aeroplanes and could be used as a start, as it has the advantage of not being aircraft-type specific. Therefore,
operators should not hesitate to try different threshold values when found more relevant for their fleet.

34
Conclusions

The work presented in this document was performed through a close

cooperation between D4S participants and powered by ALG
Task Team composition

Powered by

35
 D4S is co-funded by
the European Union

36