Laurent PELTIERS - AIRBUS Operations · 2015-06-30 · They are based on the mentioned assumptions...

Exploitation des données d’essais en vol un défi scientifique et technologique

24th June 2015 AG MaDICS - Lyon

AG MaDICS - Lyon Laurent PELTIERS - AIRBUS Operations

© AIRBUS Operations S.A.S. All rights reserved. Confidential and proprietary document.

Some figures about AIRBUS

• 74000 employees over mainly 4 countries (more than 1000 in the US)

• Revenues (2013) € 42,000 million / EBIT € 1,710 million

• 8121 aircrafts in service operated by more

than 400 companies

• Largest civil aircraft: the double deck A380

• 629 aircrafts delivered in 2014




MG3 MG13 MG7 MG9 MG11 MG4.1 MG5 MG6

Data for Manufacturing

Integration Tests Flight Tests

Production Assembly line Ramp Up

Concept

Definition

6 Year Development lead time

~24 months

~24 months

MG4.2

Flight & Integration test center : our place in A/C development

Test

Phase

A/C certified!


Flight &Integration test center

Design office & program domains

Flight & Integration test center: our mission


Tests definition

Tests preparation

Ground or flight tests Tests analysis

Tests reports

Aircraft certification


Evolution of data collected


150 TB archived 320 000 parameters

12 000 parameters

12,8 TB archived

Parameters #

14 000 parameters

670 000 parameters

Data archived

x50

x50

8,5 TB archived

450 TB archived

AMPEX 28 tracks

8 GB

SONY AIT 2/3

50/100 GB

SCSI hard disk

300 GB

SSD hard disk

700 GB


Why Big data ? Current way of working


Mainly

“read only”

Post-processing

(Event detection,

protocol checking…)

Primary data

Computed data Databases with

consolidated data

Archiving

(up to 20 years)

Online storage

(at least 3 months)

Testers

EV subcontractors

Analysis tools

Capacity ~200 TB ~300 users

internal

~300 users external

Up to 2,5 TB per day

30 000 batches per

month

6000 sensors + Avionic buses +

Video + log

Storage ~850 TB

Sequential approach


AG MaDICS - Lyon

Why Big data ? - Volume Forecast

24th June 2015

A lot of restore up to 10 years after

0

10 000

20 000

30 000

40 000

50 000

60 000

Q1-1

3Q

2-1

3Q

3-1

3Q

4-1

3Q

1-1

4Q

2-1

4Q

3-1

4Q

4-1

4Q

1-1

5Q

2-1

5Q

3-1

5Q

4-1

5Q

1-1

6Q

2-1

6Q

3-1

6Q

4-1

6Q

1-1

7Q

2-1

7Q

3-1

7Q

4-1

7Q

1-1

8Q

2-1

8Q

3-1

8Q

4-1

8

FT data volume production forecast (w/o video) - GigaBytes per quarter -

Simu NEO

Simu A350

Simu A400M

A350-1000

NEO

A350-900

A340

A320

A400M

A380

Concurrent access

High volume management

http://www.iconarchive.com/show/colobrush-icons-by-gianni-polito/system-attention-icon.html


Big Data: fit to our need



Main conclusions of RFI Big Data Flight Tests (from IT)


Flight Tests Use Case is well appropriated to Big Data concept & technologies.

POC results have demonstrated that Big Data technologies provide better performances (taking into account extraction & rendering steps).

POCs have been realized in a very short time: huge optimizations & techninal design improvements are possible.

Transition to Big Data can be done in several steps.

First step can be realized with limited impact on legacy clients (Visage, …)

All vendors propose Hadoop/Map-Reduce as target solution for extraction step

Different solutions have been proposed for storage/rendering steps (HDFS, NoSQL, MPP, …) Results are difficult to compare. NOSQL Design has an important impact on performances

Big Data technologies provide new business opportunities (combining new data sources & existing ones, real time analysis, ….)


Big data project: the scope of the first step


Acquire • Raw data

Organize

• Current tools compatibility Data triggering

Analyse • Machine learning…

Decide • Discovery tools

No functional impact for the user

Expected enhancements :

• Less overhead to get the data

• Full campaign on-line


Big data project: OBS



Big data project: One year project



Technical choice


SW

HW

https://www.google.fr/url?url=https://cloud.google.com/hadoop/&rct=j&frm=1&q=&esrc=s&sa=U&ei=Re1dVePJK4neU-GzgbAH&ved=0CDUQ9QEwBA&usg=AFQjCNGGLOUW39-UHlKdZHulClsmyl9KPA

https://www.google.fr/url?url=https://vimeo.com/cloudera&rct=j&frm=1&q=&esrc=s&sa=U&ei=jwJ2VZDWIIGsU5ylgegM&ved=0CDYQ9QEwEA&usg=AFQjCNEngzFsR43bB1pJThcmpV38qOjRjg


Technical choice vs IT organisation



Main challenges ?


Big data appliance

• noSQL database

• HADOOP cluster

Up to 4 flights 2 times a day

http://en.wikipedia.org/wiki/File:Random-data-plus-trend-r2.png

https://www.google.fr/url?url=https://cloud.google.com/hadoop/&rct=j&frm=1&q=&esrc=s&sa=U&ei=Re1dVePJK4neU-GzgbAH&ved=0CDUQ9QEwBA&usg=AFQjCNGGLOUW39-UHlKdZHulClsmyl9KPA


Next steps: processing


Predictive analysis:

• Sensor failure

• Anomalies database

Descriptive analysis:

• Complex event gathering

• Wide correlation between parameters

https://spark.apache.org/

http://www.r-project.org/index.html


Next steps: Vizualisation



Big Data application « Data retrieval»

Save flight hours

• Easy search will be generalized

« search autopilot on & altitude > 30000 ft & Mn>0.8 & bank > 5° then

and plot the vertical load factor »

• Opportunities to bridge with other database (configuration, logs,…)

Compare versions in the frame of incremental development


A320 :1988 Sharklet 2012 NEO 2015


Big Data application « Data re-use»

Increase system design maturity

• use the data for design verification prior implementation in avionic

software (robustness)

Continuous detection of anomalies on the whole campaign

• Tuning and re-launch surveillance algorithms like comparison

between sensors or COM-MON lanes

• Patterns recognition out of the « usual » enveloppe



Big Data application « Test Analysis»

Enhance the analysis through statistical tools

• Correlation analysis:

eg : longitudinal oscillation A380 specific flight needed to understand

the root cause


Mn=0.85->0.8 Mn=0.85 Mn=0.85

No stimulation stimulation No stimulation

• « Self » learning and clustering capability

Are some test points beyond criteria ?

Classical view

Statistical view


Conclusion: A first step


Technology mature

Lot of perspectives for the business

Easy to put in place even in a complex environment

but

A big change for

• IT departments

• The users due to a new way of thinking



© AIRBUS Operations S.A.S. All rights reserved. Confidential and proprietary document. This document and all information contained herein is the sole property of AIRBUS Operations S.A.S. No intellectual property rights are granted by the delivery of this document

or the disclosure of its content. This document shall not be reproduced or disclosed to a third party without the express written consent of AIRBUS Operations S.A.S. This document and its content shall not be used for any purpose other than that for which it is

supplied. The statements made herein do not constitute an offer. They are based on the mentioned assumptions and are expressed in good faith. Where the supporting grounds for these statements are not shown, AIRBUS Operations S.A.S will be pleased to

explain the basis thereof. AIRBUS, its logo, A300, A310, A318, A319, A320, A321, A330, A340, A350, A380, A400M are registered trademarks.

Thank you

Laurent PELTIERS - AIRBUS Operations · 2015-06-30 · They are based on the mentioned assumptions...

Documents

Transcript of Laurent PELTIERS - AIRBUS Operations · 2015-06-30 · They are based on the mentioned assumptions...