robust aircraft design exploration through integrated parametric … · 2017. 5. 18. · robust...

INNOVATIONS IN ENGINEERING INNOVATIONS IN ENGINEERING

Robust Aircraft Design Exploration through Integrated

Parametric Modeling and Dynamic Simulation

Nick Borer

Draper Laboratory

[email protected]

617 258 4655

25 October 2012

Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.

Outline

Motivation

Context for Robust System Design

Case Studies in Multistate Design

Twin-Engine Aircraft

Long-Endurance UAV

Fleet of UAVs for Persistent Campaign

Conclusions


Motivation

Conceptual design

methodologies typically

focus on optimizing

nominal performance,

with constraints for off-

nominal conditions

Only guarantees adequate,

not best performance in off-

nominal conditions

Yet, aerospace systems

can spend significant time

operating in off-nominal

conditions


NASA/JPL-Caltech NASA

NASA

National Museum of the USAF

If it’s Going to Spend a Lot of Time Broken, Why Not Design it to Operate That Way?

Robust System Design Context

Change the design Increased margins

Cross-strapping, redundancy

Change the performance requirements Change operations

(plan operation in favorable modes)

Remove operational requirements (e.g. do part of the mission with another system)

Change the rate of state transition Increased component

testing, qualification

Use of higher-quality components

Sizing

Performance

State

Performance

balance

System

configurations

Geometry,

Environment

Boundary

Conditions

Modes, Transition

Rates

Cost, Weight,

Dimensions, …

Range, Rates,

Limits, …

Reliability,

Extensibility,

Expandability, …

Reliability

Analysis

DR/Dl [CAME]

Multistate Design

D($,P,R) / D(x,z,l) Multidisciplinary Design

Optimization

D($,P)/D(x,z) Design

D$/Dx

System

parameters

Performance

parameters

What is the best path to design for off-nominal events?


Behavioral-

Markov

D(P,R)/D(z,l)

[PARADyM]

Use All Available Means to Increase System Robustness

Multistate Sensitivity Analysis: Twin-Engine Aircraft

Investigated the

sensitivity of

traditional aircraft

design variables on

reliability in addition

to component failure

rates for C-12

(Beech King Air)1

Results showed that

reliability was far

more sensitive to

vertical tail

parameters than to

component failure

rates

-0.1 -0.05 0 0.05 0.1 0.15 0.2

eng. failure rate

rud. failure rate

ail. failure rate

wing area

wing span

horiz. tail area

horiz. tail span

vert. tail area

vert. tail height

engine location

aileron chord

elevator chord

rudder chord

Normalized sensitivity of expected availability to design variables

5-yr

20000-hrStatic Design Variables, x

Component Failure Rates, l

1. J. S. Agte, N. K. Borer, O. de Weck, “Multistate Design Approach to the Analysis of Twin-

Engine Aircraft Performance Robustness,” Journal of Aircraft, 49(3) 781-793, 2012.


More Leverage Exists to Increase Availability through Design Variables, not Failure Rates

Multistate Tradespace Exploration: Long-Endurance UAS

Investigated the tradespace of a long-endurance UAS First considered a long-

endurance system with a notional power system and three-month endurance2

Next considered a more conventionally-fueled “squadron” of systems operated to enable persistent coverage over five years3

Both cases showed an increase in availability and reduction in cost over traditional approaches, as well as non-intuitive design features

~2100 nm

tot. coverage ~ 100000 nm

Falklands

2. J. S. Agte, N. K. Borer, O. de Weck, “Design of Long-Endurance Systems

With Inherent Robustness to Partial Failures During Operations,” Journal of

Mechanical Design, 134(10), 2012.

3. N. K. Borer, J. S. Agte, “Design of Robust Aircraft for Persistent Observation

Campaigns Using Nested Multistate Design,” AIAA-2012-5452, AIAA Aviation

Technology, Integration, and Operations Conference, September 2012.

~20 hrs over continent@ 50k ft (~4400 nm)

avg. ingress/egress ≈2100 nm (~10 hrs)

21


Multistate Exploration Yields Reduced Cost, Increased Availability vs. Traditional Approach

Acquisition Cost vs. Availability for 3-Month UAS


0.68 0.7 0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.880.8

0.9

1

1.1

1.2

1.3

1.4

1.5

System Availability

A/C

Fly

aw

ay C

ost [%

of b

ase

lin

e c

ost]

varying x & ls

varying x

varying ls

NASA baselineOptimized Baseline

Multistate Approach Yields Lowest Cost, Highest Availability Solution

Life Cycle Cost vs. Availability for 5-year Campaign

Multistate design – cheapest LCC, highest

availability, highest O&M cost, lowest combined

acquisition cost (initial + spares)

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Baseline Combined Varying λ Varying x

Life

Cyc

le C

ost

(% b

ase

line

)

34.3%

20.8%

43.2%

36.0%

25.6%

45.4%

20.2%

53.1%

28.9%26.7%

21.3%

44.4%

Operations & Routine

Maintenance

AcquisitionCost of Spares

ContingencyRepairs* (< 1%

in all cases)

Acquisition Cost w/o Spares

LCC 100% LCC 88.8% LCC 92.8% LCC 91.7%

0.94 0.945 0.95 0.955 0.96 0.965 0.97 0.975 0.980.8

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

Campaign Availability

Life

Cycle

Co

st [%

of b

ase

lin

e c

ost]

varying x & l

varying l

varying x

baseline design

Key to above figure


Multistate Design Yields Lowest Life Cycle Cost, Least Attrition for Multiyear Persistent Campaign

Conclusions

Conceptual design of robust systems requires a

change in perspective vs. traditional “design

optimization”

Move away from constrained optimization of nominal

performance to tradespace exploration across set of

possible conditions

Integrated modeling & simulation is key to this early

exploration

Capture interactions between systems, operations,

requirements, and disciplines

Much to be gained by giving your reliability engineer

a seat at the conceptual design table

Case studies show lower-cost, higher-availability designs

are very possible


It’s Not Off-Nominal if You’ve Designed for it!

References

Multistate Sensitivity Analysis Case Study: J. S. Agte, N. K. Borer, O. de Weck, “Multistate Design Approach to the

Analysis of Twin-Engine Aircraft Performance Robustness,” Journal of Aircraft, 49(3) 781-793, 2012.

Multistate Design for Long Endurance UAV Case Study: J. S. Agte, N. K. Borer, O. de Weck, “Design of Long-Endurance

Systems With Inherent Robustness to Partial Failures During Operations,” Journal of Mechanical Design, 134(10), 2012.

Fleet of UAVs for Persistent Campaign Case Study: N. K. Borer, J. S. Agte, “Design of Robust Aircraft for Persistent

Observation Campaigns Using Nested Multistate Design,” AIAA-2012-5452, AIAA Aviation Technology, Integration, and Operations Conference, September 2012.

Behavioral-Markov Modeling for Reliability-Driven Design N. K. Borer et al., “Model-Driven Development of Reliable Avionics

Architectures for Lunar Surface Systems,” IEEE Aerospace Conference, March 2010.

Markov Modeling for Reliability-Driven Design P. S. Babcock IV, “Development the two-fault tolerant attitude control

function for the Space Station Freedom,” 20th International Symposium on Space Technology and Science, Gifu, Japan, May 1996.


Additional Information


Integrated Parametric Flight Dynamics Model

Built around MATLAB® and Simulink® for ease of integration with other analyses MATLAB® calls to open-

source aerodynamics module, custom-built analyses

Flight dynamics model built as Simulink S-function from open-source flight dynamics engine Allows for evaluation of

control architectures, failure models in Simulink

Outputs can be ported to quasi-real-time flight simulator for visualization and “pilot” operation

Mass

Properties

Aerodynamic

Forces &

Moments

Propulsion

6-DoF

Simulation

Geometry

Engine deck: T, FF f (h, M)

Drag polar

Stability derivatives

Fuel tank location

CGs

Inertias

Gross weight

CGs

Inertias

Dimensions Dimensions

Conf iguration type

Parameters

Group weights Flight conditions

Test data

Flight conditions

Test data

Controls/

Displays

Position

Attitude

Health

Control inputs

Location

Environment

MATLAB scripts

for preprocessing

MATLAB scripts for

piecewise inertia estimation, mass scaling

MATLAB script for parasite drag,

AVL for induced drag & derivatives

Lookup table

(bootstrap)

JSBSim S-function in

Simulink

FlightGear

Parametric Analysis

Dynamic

“Real”-

Time Sim

4

outputs

3

propulsion

2

controls

1

states

JSBSim_SFunction

JSBSim3

sleep_usec

2

dt_sec

1

inputs

double

double

double

double

double

double

double


Behavioral-Markov Modeling

Behavioral-Markov modeling offers an opportunity to predict and capture system behavior in an exhaustive combination of off-nominal events

Legacy: CAME (Computer Aided Markov Evaluator) used for fault-tolerant system development for high-reliability space and ground systems

– Able to capture cascading failures through integrated system model

PARADyM (Performance and Reliability Analysis via Dynamic Modeling) enhances CAME’s capabilities by directly simulating every Markov state with a behavioral model

– System performance is captured for every Markov state

– Status is determined directly from system performance

tbPtaPdt

dP

tPbadt

dP

212

11

tbabt

tba

eetP

etP

2

1

or ?

time

failure

occurs

perturbation


Effect of Geometry of Performance in Failed States

-500 0 500 1000 1500 2000 250020

25

30

35

40

specific excess power, Ps (fpm)

ba

nk a

ng

le (

de

g)

Nominal State

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 1: Left Engine Failed

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 2: Rudder Failed

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 3: Ailerons Failed

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 4: Left Engine, Ailerons Failed

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 5: Left Engine, Rudder Failed

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 6: Rudder, Ailerons Failed

unsafe region

safe region

-500 0 500 1000 1500 2000 250020

25

30

35

40


ba

nk a

ng

le (

de

g)

State 7: All Failed

unsafe region

safe region

Actual (non-weighted) performance in each Markov state for the 23 geometry cases

Loss if Ps < 200 fpm, bank angle not held within 10 degrees


Aircraft Performance and Cost Model

Aircraft sizing and

performance

Added loop to optimize

cruise operation for

maximum endurance in

each failure configuration

Cost

Development and

Procurement Cost of

Aircraft model (DAPCA

IV) updated for UAVs

Component reliability

cost based on

Draper/NASA study of

cost vs. parts grades4

0

5000

10000

15000

20000

25000

0 1000 2000 3000 4000 5000 6000

No

rma

lize

d C

ost

Normalized Mean Time Between Failure

Large Unit (~1400 parts)

Medium Unit (~750 parts)

Small Unit (~100 parts)

Reliability Cost Function, f = 0.1



Dra

pe

r


Geometry

Mass

Properties

Engine

Aero

Throttle

Controls/

TrimControls/

TrimControls/

TrimControls/

TrimControls/

TrimControl/

TrimK

Range/

Endur-

ance

failures

Itera

te o

n d

efl

ectio

ns

to m

inim

ize t

rim

dra

g

Iterate on airspeed/altitude for best endurance

opt. alt/

velocity

Performance

c = vector of operational parameters

x = vector of static design variables

λs = vector of transition rates, size n

1 2 3 … … n

opt. drag

polar

Gk

4. N. K. Borer et al., “Model-Driven Development of Reliable Avionics Architectures

for Lunar Surface Systems,” IEEE Aerospace Conference, March 2010.

robust aircraft design exploration through integrated parametric … · 2017. 5. 18. · robust...

Documents