View
5
Download
0
Category
Preview:
Citation preview
INNOVATIONS IN ENGINEERING INNOVATIONS IN ENGINEERING
Robust Aircraft Design Exploration through Integrated
Parametric Modeling and Dynamic Simulation
Nick Borer
Draper Laboratory
nborer@draper.com
617 258 4655
25 October 2012
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 2
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 3
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
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?
Slide 4
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?
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Behavioral-
Markov
D(P,R)/D(z,l)
[PARADyM]
Use All Available Means to Increase System Robustness
Slide 5
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.
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
More Leverage Exists to Increase Availability through Design Variables, not Failure Rates
Slide 6
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Multistate Exploration Yields Reduced Cost, Increased Availability vs. Traditional Approach
Slide 7
Acquisition Cost vs. Availability for 3-Month UAS
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
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
Slide 8
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Multistate Design Yields Lowest Life Cycle Cost, Least Attrition for Multiyear Persistent Campaign
Slide 9
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
It’s Not Off-Nominal if You’ve Designed for it!
Slide 10
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.
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 11
Additional Information
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 12
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 13
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 14
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
specific excess power, Ps (fpm)
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
specific excess power, Ps (fpm)
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
specific excess power, Ps (fpm)
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
specific excess power, Ps (fpm)
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
specific excess power, Ps (fpm)
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
specific excess power, Ps (fpm)
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
specific excess power, Ps (fpm)
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
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
Slide 15
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
Reliability Cost Function, f = 0.2
Reliability Cost Function, f = 0.3
Dra
pe
r
Copyright 2012 by the Charles Stark Draper Laboratory, Inc. All rights reserved.
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.
Recommended