model based design for fuel system development · system development perimeter and interfaces 3 4th...

Model‐Based Design for Fuel

System Development

Christopher Slack, Airbus

04 October 2017

4th October 2017 MATLAB EXPO - Model‐Based Design for Fuel System Development2

17,287Aircraft sold

60Produced monthly

25,000+Daily flights

10,561Delivered

54,000Employees

€49.2billionAnnual revenue*

6,726Backlog

400Operators

System Development Perimeter and Interfaces


Fuel Pumps

Flamearrestor

Fuel Gauging& Control Computers

GaugingFuel Probes Level Sensors Temp

Sensors

Burst Discs

Pipes, Couplings

OPENL OUTR FEED 1 FEED 2 FEED 3 FEED 4 R OUTR

SHUT

REFUEL / DEFUEL VALVES

OPEN

SHUTL MID L INR TRIM CTR R INR R MID XFR

DEFUEL

MAN

REFUEL

OFFAUTO

REFUEL

MODE SELECT PRESELECT

INCREASE

DECREASE

27120 19860 13740 0 19860 27120

4000 20800 21840 21840 20800 4000 201000 kg

211000 kgAUTO REFUEL

L OUTR FEED 1 FEED 2 FEED 3 FEED 4 R OUTR

L MID L INR TRIM CTR R INR R MID

ACTUAL (FOB)

PRESELECT (PFQ) STATUS

HI LVL

HI LVL

FAULT OVERFLOW

POWER SUPPLY

BATTERY

NORMAL

SHUTOFF

TEST

APU EMERGENCY

SHUT

DOWN

Refuel Panel

Flammability Reduction

Cockpit displays

Air to Air

Refuelling

Valves

CompensatorsDensitometers

Engines

Electrical Interfaces

Functional

Requirement

Validation

Iron Bird

A300A320

A330/A340

A340-600

A380

A400M

A350

Model Use History


Textual

Requirements

Model Based

V&V

Model Based Design

V&V Auto Test

& SLDV

Use

of m

od

els

(Am

ou

nt / F

ide

lity)

Textual

RequirementsTextual

Requirements

Model Based Design

Desktop simulator

Model Based Design

Model Based System Engineering

Fuel Virtual

Integration

Platform

Specific,

Targeted V&V CofG Calcs Time

MATLAB, Matrix-X,

Easy5,

Statemate

Fortran,

HP Basic

MATLAB,

Simulink,

Statemate

MATLAB,

Simulink, RTW,

Stateflow

SysML, MATLAB, Simulink,

Stateflow, SLDV,

Simscape, …

MATLAB,

Simulink,

Stateflow, SLDV

Towards Full MBSE


System Requirement Authoring, Validation and Verification

Model Based Design Lifecycle

4th October 2017 MATLAB EXPO - Model‐Based Design for Fuel System Development

Systems

Equipments

MG5 MG7 MG9 MG11

A/C

Integration

Testing

Simulation

Multi-systems

Detailed

Design

Matlab Simulink for Fuel

Management SSRD

Req and detailed design

Desktop simulator for

Validation of requirement

and Verification of detailed

design

Desktop Simulator

Fuel Virtual Integration

Platform to support

system tests

FVIP

Suppliers s/w dev

Model translation for software

development

Architecture

Design

Fluid and

Thermal

modelling

Inerting Certification

using Modelling

Mission Distribution

0

2

4

6

8

10

12

14

16

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

1000

0

Range 1000's NM

Nu

mb

er o

f M

issio

n in

200 n

m b

lock

AMO & FSP

for Test Rig

and A/CO

Aircraft 0

Tank Modelling:

- Output for Gauging & Fuel

Management

- Output for Wing design (stress, load,

etc.)

MG3

6

Model Based Design - In Practice

Develop models to specify system functionality

–Describes behavioural & functional aspects

Details become the System (and Sub-System)

Requirements

–Exercise the model to Validate

Requirements

Delivered to Fuel System Supplier

–Model contains Requirements and intent

–Model execution provides system

understanding

–Minimal Work to turn into Code

–Separate layer for independent validation

4th October 2017 MATLAB EXPO - Model‐Based Design for Fuel System Development

Page

7

Radar network

HF / VHFSATCOM RadarAir Traffic Control

SITA/ARINC network

3D positions

voice

&data

voice

&data

Environment TOP LEVEL REQ

FUNCTIONAL/

SYSTEM REQ

LOW LEVEL

EQUIP REQ

LOW LEVEL

EQUIP DEV

MBSE – Functional System Requirements

• MATLAB/Simulink/Stateflow Application

• Development of Control System Reqts

–Normal and Failure Operating Modes

–Crew Procedures

• Control Logic separated from Aircraft Environ

–System Designers focus on

– Control Functions

– HMI

– Robustness & Validation

–Specialist Modellers focus on:

– Aircraft & Environmental Simulation

– Physics (Fuel, Thermal)

– Auto-Test Capabilities


MBSE – Stateflow for Requirements Authoring

• Aircraft Fuel System Statecharts:

• Linked Requirements

– System Requirements Documents Cascade

– Requirements Database (DOORS)

• Separate Chart for each Major A/C Function

– Allows for collaborative development

• Transition booleans calculated externally

– Input from Simulink

– Stateflow graphical function

• Driven behaviour of Stateflow logic separated from

driving conditions

– Allows easier readability and testing


Model Based Design - Reuse

• Integrated Desktop Simulator

– Requirements & Environment Model

– AutoCode using Simulink Coder

– Optional Interfaces to Cockpit Display & Flight Warning

• OCASIME, VIP & Aircraft -1

– Entire Software Simulation

– Interfaces Identical to Full Flight Simulator

• Aircraft-0 (Iron Bird)

– Cockpit Avionics & Displays

– Integrated of Real & Simulated Systems

– Virtual Hosting of Supplier’s Code

• Full Flight Simulator

– Single model for all platforms

– Training Flight and Ground Crews

4th October 2017 MATLAB EXPO - Model‐Based Design for Fuel System Development10 4th October 2017 MATLAB EXPO - Model‐Based Design for Fuel System Development

• Model V&V has to go through several loops

– When the model is the requirements, the distinction

between “Model Verification” and “Requirements

Validation” is somewhat blurred

If a test fails, what is at fault?

– the requirement?

– the model?

– the test?

Model Development Test/Verification/Validation Cycle


Requirements MODEL

TestObjectives

TestScripts

ModelValidation

ModelVerification

Test Script

Validation

Requirement

Validation

Model

Specification

Results &

Problems

Test Objective

Validation

Environment

Using Simscape to Model A350 Refuel System


Component export, parameter estimation

And model simplification for Real Time performance

Use of Simscape

• Fuel Design Model Developed in Flowmaster

– Architecture and Component Performance

– Spec Model Only - not real-time

– Cannot produce C-Code or embedded simulations

• Exploiting new SimHydraulics Toolbox

• Mathworks Consultancy

– Airbus provision of core models and perf data

– Majority of development by Mathworks


Flowmaster Diagram

Component Development

• Mapping of Flowmaster components to

Simscape/ SimHydraulics equivalents

– Most 1:1 equivalents

– Some required customisation from base library

• Curve Fitting Toolbox

– Fit source data to SimHydraulics block equations

– Saved as Matlab Script for re-use


Library Construction and Parameterisation


Component LibrarySystem Library

System Palette

• Component Library to customise standard

Hydraulic Library Components

• System Library contains “System Level”

Components

• Each System Palette contains Multiple

Components

– E.g. There are several different type of pumps

• System Palette constructed using MATLAB

scripts

– Self Documenting

– Re-run if design model updated

Model Simplification

• Design Model has ~900 individual Components

– A reduction of the number of blocks by a factor of 10 can potentially yield a simulation speed improvement by a factor of 1,000.

• Reduction Strategies

– Reduce multiple serially Connected Pipes/Bends/Losses to a single Equivalent pipe/loss combination

• Design Optimisation toolbox

– Established Equivalent Parameters

• Reduction in the number of components

– Pipe components reduced from 290 to 60

– Total Components reduced from 900 to 170

– So would expect ~120 x speed-up


Model Reduction

• During Refuel or Defuel, certain valves are not in use

– Fluid network behind those closed valves do not contribute to pressure/flow calculations

– Therefore can be removed

• This can be repeated for each combination of tank that needs to be studied.

– The reduced model can be constructed automatically with MATLAB scripts that analysis network topology.


Complete ModelReduced Model

Refuel from

Left Wing

Only

http://www-test1.mathworks.com/cmsimages/98594_wl_Modeling_Airbus_fig7_wl.jpg

http://www-test1.mathworks.com/cmsimages/98595_wl_Modeling_Airbus_fig8_wl.jpg

Simscape Summary of Results

• Two system-level models of the Defuel system created in

SimHydraulics

– One complete: all components required to model the system

behaviour included

– One simplified: all “isolated” components located behind closed

valves removed

• Performance of the simplified model sufficient for real-time

– Tested with Simulink Real Time on industrial PC

• Performance of the complete model not sufficient for real-time

implementation, despite simplifications made.

– Depends on the solver chosen to a large extent

– Improves substantially from with later Simscape versions

– Near real-time performance in exploiting Simscape local solver

• New blocks and demos added to SimHydraulics as a direct result of this

work


Code Efficiencies and Performance Enhancement


Fuel Temperature Prediction Software

Fuel Temperature Prediction for AirlinesAn Exercise in Code Efficiencies

• Low outside temperatures with long exposure times

• Fuel temperature may drop close to or below freezing point

– Software written in MATLAB

– Predict fuel temperatures given Flight Profiles & Global Air

Temperatures.4th October 2017 MATLAB EXPO - Model‐Based Design for Fuel System Development20

Inter-tank

Fuel Transfer

Solar radiation

Heating from earth

T∞

Emissivity from skin

Emissivity from skin

Earth

Sky Night/Day

Radiation network

external heatsource / sink

T∞

Convection

Conduction

Buoyancy

Comparison of Predicted and Measured Fuel Temperature

-40

-30

-20

-10

0

10

20

30

0 1000 2000 3000 4000 5000 6000 7000 8000

Distance (nm)

Tem

per

atu

re (

°C)

OUTER Meas

OUTER Pred

INNER 1/4 Meas

INNER 1/4 Pred

INNER 2/3 Meas

INNER 2/3 Pred

TRIM Meas

TRIM Pred

Fuel Temperature Prediction for AirlinesAn Exercise in Code Efficiencies


Intended Usage

Flight Plan

Fuel Temperature

Prediction

Cold Fuel?

Y

Run-Time ~ 40 seconds (reasonable)

NFly

Route

Actual Usage

Flight PlanFlight PlanFlight PlanFlight PlanFlight PlanFlight PlanFlight PlanFlight Plann

Calculate

“Best”

Route

Up to 50 potential Routes Run-Time ~ 50 * 40 seconds = Half Hour…BAD!

Other Safety & Economic Factors

Fuel Temperature

Prediction

Fly Route

Using MATLAB Profiler to Identify Code Efficiency Bottlenecks

• Exploit MATLAB Profiler

• Built into MATLAB

profile on ; run program ; profile viewer

• Creates timing profiles of every function called

• Look at the “Self Time” for time spent within function

• Profile Report highlights most expensive L.O.C.

• Iterative process to increase code efficiencies.


Code Optimisation Strategies

• Equation Vectorisation

• Loop Unrolling

• Switch…case statements

– Reduce volume of code inside each “case”

• Use c-mex for time-critical functions

– Check target platforms

• Minimise Globals

– Very slow in MATLAB

• Reduce calculations inside for loops

– Pre-calculate invariant parts of equations


May Jun Jul Aug Sep Oct0

5

10

15

20

25

30

35

40

Ave

rag

e S

imu

lati

on

Tim

e p

er

Fli

gh

t (s

)

Run-Time Improvements of Fuel Temperature Prediction Module

Target Time

Keeping Track of Mathworks Release Cycles


Industry Model Testing

Industry Model & Code Testing

• Aircraft Development 3-5 years, Mathworks upgrades every 6 months– One solution to reduce cost of (continuing) upgrade cycles

• Testing infrastructure utilising customer models and MATLAB scripts

– Release Compatibility

– Performance

Win-Win Situation:

• Value to Customers

– Reduced product upgrade cost

– Increased productivity

– Early knowledge of regression

• Value to Mathworks

– Compatibility testing

– Performance testing

– Increased tool adoption


Establish

NDA

Package

Models/Scripts

Send

Package

Review and

Act on Results

Simple 4-step process

AFSME

Airbus Fuel System Modelling Environment

10

Flow Rates

9

Temp

8Tot Weight

7Mom Y

6Mom X

5

Mass

4Targ Cg

3XCg

2Altitude

1Jettison

Density

Vol 2 Mass

Engine Burn

APU Burn

Recirc Rate

Valve States

Transfer Mode

Tank Empty

Flow Rates

Valves_2_Flowrates

Mass

Temperature

Flow

Valves

Modes

Flight Profile

CG

Aux

Time Plotting

Volumes

Mach

Air Temp

Temperature

Temperature

Flow Rate

Leak Rate

Volume

Tank Empty

Tanks

Program AFSME 3.2 STR 465Date: Fri Jan 18 11:24:10 2002Ref: SCF/SYS/GEN/61/2623 Iss. 6

User Guide: SCF/SYS/GEN/61/2909

Leak

Leaks

[M]

[V]

[F]

[M]

[V]

[F]

Mass

Recirc_Inhibit

Engine Burn

APU Burn

Recirc Flow

Jettison

Altitude

Pitch

Roll

Mach

Air Temp

FP Mux

Flight_Profile

Simulation.Temperature

Enable Temperature

Simulation.CG

Enable CG Calc

Pitch

Roll

Mass

XCg

Targ_Cg

Mom X

Mom Y

Tot Weight

CG Mux

Cg_Calc

4Recirc Inhibit

3Auxiliary Plot

2Modes

1Valves

AFSME

Airbus Fuel System Modelling Environment

10

Flow Rates

9

Temp

8Tot Weight

7Mom Y

6Mom X

5

Mass

4Targ Cg

3XCg

2Altitude

1Jettison

Density

Vol 2 Mass

Engine Burn

APU Burn

Recirc Rate

Valve States

Transfer Mode

Tank Empty

Flow Rates

Valves_2_Flowrates

Mass

Temperature

Flow

Valves

Modes

Flight Profile

CG

Aux

Time Plotting

Volumes

Mach

Air Temp

Temperature

Temperature

Flow Rate

Leak Rate

Volume

Tank Empty

Tanks

Program AFSME 3.2 STR 465Date: Fri Jan 18 11:24:10 2002Ref: SCF/SYS/GEN/61/2623 Iss. 6

User Guide: SCF/SYS/GEN/61/2909

Leak

Leaks

[M]

[V]

[F]

[M]

[V]

[F]

Mass

Recirc_Inhibit

Engine Burn

APU Burn

Recirc Flow

Jettison

Altitude

Pitch

Roll

Mach

Air Temp

FP Mux

Flight_Profile

Simulation.Temperature

Enable Temperature

Simulation.CG

Enable CG Calc

Pitch

Roll

Mass

XCg

Targ_Cg

Mom X

Mom Y

Tot Weight

CG Mux

Cg_Calc

4Recirc Inhibit

3Auxiliary Plot

2Modes

1Valves

IRP_MODE_SELECTION

5.1.7(1)

GROUND_OPS/ 1

SURGE2

5.1.(1)

AUTO_REFUEL TRANSFER

5.1.4.(1)5.1.3.(1) 5.1.2.(1)

MANUAL_REFUEL

AR_

CONF

5.1.1.(1)

OFF_

CONFDF_

CONF

DEFUEL SOT OFF

5.1.5.(1) 5.1.6.(1)

GO_D = DELAY(d_t)

function

TR_

CONF

MR_

CONF

[XFER&...DELAY(D_GOS)]

[~XFER ...| SOT_INITIATED]

[AR &...DELAY(D_GOS)] [(~GND_SURGE_RELIEF_ACTIVE &...

~AR) | SOT_INITIATED][DELAY(D_GOS)][((AR | XFER |...

DF | IDLE) &...~GND_SURGE_RELIEF_ACTIVE) |...

SOT_INITIATED]

[MR | FAULT]/d_i=0; [AR]/d_i=0; [~AR][~MR] [~XFER] [XFER]/d_i=0;

[DF]/d_i=0; [IDLE]/d_i=0;[~IDLE][~DF]

[~IDLE ...| FAULT | SOT_INITIATED ][~DF | SOT_INITIATED]

[SOT_INITIATED]

[Mode_Status[TR_SOT_EXIT] &...~GND_SURGE_RELIEF_ACTIVE]

[IDLE &...DELAY(D_GOS)]

[DF &...DELAY(D_GOS)]

evaluate_conditions()function 5.1.(2)

{

AR= ...IRP_MODE[IRP_AUTO_REFUEL] & ...~IRP_MODE[IRP_MANUAL_REFUEL] & ...~IRP_MODE[IRP_TRANSFER] & ...~IRP_MODE[IRP_DEFUEL] & ...~IRP_MODE[IRP_OFF] &...~SOT_INITIATED;

XFER= ...IRP_MODE[IRP_TRANSFER] & ...~IRP_MODE[IRP_MANUAL_REFUEL] & ...~IRP_MODE[IRP_AUTO_REFUEL] & ...~IRP_MODE[IRP_DEFUEL] & ...~IRP_MODE[IRP_OFF] &...~SOT_INITIATED &...~GND_SURGE_RELIEF_ACTIVE;

DF= ...IRP_MODE[IRP_DEFUEL] & ...~IRP_MODE[IRP_MANUAL_REFUEL] & ...~IRP_MODE[IRP_AUTO_REFUEL] & ...~IRP_MODE[IRP_TRANSFER] & ...~IRP_MODE[IRP_OFF] &...~SOT_INITIATED &...~GND_SURGE_RELIEF_ACTIVE;

IDLE= ...IRP_MODE[IRP_OFF] & ...~IRP_MODE[IRP_MANUAL_REFUEL] & ...~IRP_MODE[IRP_AUTO_REFUEL] & ...~IRP_MODE[IRP_TRANSFER] & ...~IRP_MODE[IRP_DEFUEL] &...~SOT_INITIATED &...~GND_SURGE_RELIEF_ACTIVE;

MR= ...IRP_MODE[IRP_MANUAL_REFUEL] & ...~IRP_MODE[IRP_DEFUEL] & ...~IRP_MODE[IRP_AUTO_REFUEL] & ...~IRP_MODE[IRP_TRANSFER] & ...~IRP_MODE[IRP_OFF] &...~SOT_INITIATED;

FAULT= ...~AR & ~XFER & ~DF & ~IDLE & ~MR;}

IRP_MODE_SELECTION

5.1.7(1)

GROUND_OPS/ 1

SURGE2

5.1.(1)


5.1.4.(1)5.1.3.(1) 5.1.2.(1)

MANUAL_REFUEL

AR_

CONF

5.1.1.(1)

OFF_

CONFDF_

CONF

DEFUEL SOT OFF

5.1.5.(1) 5.1.6.(1)

GO_D = DELAY(d_t)

function

TR_

CONF

MR_

CONF






SOT_INITIATED]




[SOT_INITIATED]




IRP_MODE_SELECTION

5.1.7(1)

GROUND_OPS/ 1

SURGE2

5.1.(1)


5.1.4.(1)5.1.3.(1) 5.1.2.(1)

MANUAL_REFUEL

AR_

CONF

5.1.1.(1)

OFF_

CONFDF_

CONF

DEFUEL SOT OFF

5.1.5.(1) 5.1.6.(1)

GO_D = DELAY(d_t)

function

TR_

CONF

MR_

CONF






SOT_INITIATED]




[SOT_INITIATED]





{








{







Models

GUIs

Tests & Results

Models

Secure Models

Server

ITAR Rated Firewall

Summaries and Lessons Learnt


Lessons Learnt

• Deployment of MBSE

– As much about Competences as Technologies

– Skillsets & Mindsets

– Integration of Functional & Non-Functional models

• Model Build Reveals Emergent Properties

– Validation for free

– System difficult to model will be difficult to build/test

• Validation/Verification Testing

– A test that is more complex than that being tested is

probably wrong

– Easy to be caught in the trap of “Test for Success”

– Testing for intentional but not unintentional

behaviour

– Automated Test/Analysis allows regression testing

– Formal Proof more thorough than test scripts


• System Designers Focus on Designing the System

– The System Model is the System Requirements

– Extra functionality required to exercise the model are not

requirements

– Need to clearly identify what are requirements and what

are the extras

• Model Architecture

– Must match System Architecture

– Also conducive to multi-team development

• Easy for Designers can be Difficult for Simulators

– Engineers can be very “ingenious”

– Break downstream processes

– Model exchange with suppliers

– Automatic code generators

– Require adherence to Style Guidelines and Design Patterns