Apr 20, 2017 - Constraint Analysis. Designer determines a pair of thrust loading (TSL/WTO) and wing loading (WTO/S) at takeo from constraint analysis with ...
Reliability-Based Design Optimization Applied to Aircraft Initial Sizing
KSAS Conference, Spring 2017 Seul-Ki Kim
Shinseong Kang
Kyunghoon Lee
Department of Aerospace Engineering, Pusan National University
April 20, 2017
RBDO applied to the aircraft initial sizing
Outline
SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
1 Introduction
Motivation and Research Proposition
2 Formulations
Constraint Analysis Reliablity-Based Design Optimization
3 Demonstration 4 Summary and Conclusion
2 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Aircraft Initial Sizing
The overall shape, size, weight, and performance of the aircraft are determined.
Constraint Analysis
Designer determines a pair of thrust loading (TSL /WTO ) and wing loading (WTO /S) at takeo from constraint analysis with the following formula. f
TSL WTO , , T , P, CL , CD WTO S
=0
(source: Aircraft Engine Design)
(source: HowStuWorks)
Figure: Mission prole
Figure: Constraint analysis diagram
3 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Motivation and Research Proposition
Uncertainty in Flight
Uncertainty impacts the ight performance signicantly. A small variation of ight condition may cause a mission failure. 1
Aleatory uncertainty
Inherent randomness associated with a physical system or environment. Environmental factors: temperature, pressure, etc.
2
Epistemic uncertainty
Ignorance or limited data and knowledge. Aerodynamic data: CL , CD , K1 , K2 , etc.
Research Proposition Reliability-based design optimization (RBDO) in consideration of uncertainties. 4 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Constraint Analysis
Master Equation
(source: Aircraft Engine Design)
Figure: Applied forces on aircraft
A master equation for ight performance can be derived from Newton's second law of forces applied to the aircraft as " 2 q S nβ WTO nβ WTO K1 + K2 β WTO q S q S Ps . +CD0 + CDR ] + V
TSL β = WTO α
(
5 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Constraint Analysis
Air-to-Air Fighter (AAF) Example Of the various aircraft, we adopt air-to-air ghter (AAF) and selected AAF specications are delineated as follows: Mission phases and segments
Performance requirements
1
Takeo, no obstacle
2 3
Supercruise Combat turn
4
Landing
2000 ft PA, 100 °F, 0.1M tR = 3 s, sTO = 1500 ft 1.5M/30kft, no afterburning 0.9M, two 360 deg 5g sustained turn, with afterburning 2000 ft PA, 100 °F tFR = 3 s, sL = 1500 ft Drag chute diameter 15.6 ft (source: Aircraft Engine Design)
6 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Constraint Analysis
Air-to-Air Fighter (AAF) Example
Mission Phase 1: Takeo
WTO S
(
=
−b +
√
b 2 + 4ac 2a
)2 ,
where ( #) " β α TSL CL,max log 1 − ξTO − µTO , 2 ρg0 ξTO β WTO kTO q b = tR kTO 2β/(ρCL,max ), c = sTO . a=−
Mission Phase 2: Supercruise TSL β = WTO α
β K1 q
WTO S
CD0 + β/q(WTO /S)
7 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Constraint Analysis
Air-to-Air Fighter (AAF) Example
Mission Phase 3: Combat Turn
β TSL = WTO α
K1 n2
β q
WTO S
+
CD0 β/q(WTO /S)
Mission Phase 4: Landing
WTO S
(
=
−b +
√
b 2 + 4ac 2a
)2 ,
where β a= ρg0 ξL
(
log 1 + ξL
b = tFR kTD
q
"
2β/(ρCL,max ),
(−α) TSL µB + β WTO
CL,max 2 kTD
#) ,
c = sL .
8 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Constraint Analysis
Air-to-Air Fighter (AAF) Example (source: Aircraft Engine Design)
Figure: Traditional constraint analysis; oset 9 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
RBDO is a method that minimizes the objective function subject to predened probabilistic constraints.
Concept
Formulation minimize
d
f (d)
subject to P[gi (X) > 0] ≤ Pi , i = 1, · · · , Nc ,
d
lb
(source: Lecture note of Prof. Noh)
≤ d ≤ dub ,
where d = {d1 , . . . , dn } ∈ Rn is the design variable vector, X = {X1 , . . . , Xm } ∈ Rm is the random variable vector, and Pi is target probability of failure. 10 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
First-Order Reliability Method (FORM)
The probability of failure is dened by a multi-dimensional integral such that Pfailure = P[g (X) > 0] ≡
Z
Z ···
g (X)>0
fX (x)dx1 . . . dxm ,
which is the volume underneath the surface of the joint probability density function (PDF) fX (x) in the failure region g (X) > 0. In most cases, the numerical integration is so complicated that it requires considerable computation. To shorten computational burden, we approximate the performance function g (X) using rst-order reliability method (FORM).
11 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
Rosenblatt Transformation T : X → U
The rst step of FORM requires a transformation T from the original random variables X to the standard normal random variables U such that g (T(X)) ≡ g (U).
After the transformation, the probability of failure becomes Pfailure = P[g (U) > 0] ≡
Z g (U)>0
φU (u)d u.
MPP-based Method
In the second step of FORM, the performance function in U-space can be linearized at the most probable point (MPP) u ∗ that has the highest joint PDF such that g (U) ∼ = gL (U) = g (u∗ ) + ∇g T (u∗ )(U − u∗ ),
where ∇g (u∗ ) is the gradient of g (U) evaluated at u∗ .
12 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
MPP-based FORM Method Maximizing the joint PDF φU (u) on the limit-state function gives the model for the MPP search such that minimize
u
m X
ui2 = kuk2
i=1
subject to g (u) = 0.
The minimum distance ku∗ k from the origin to the MPP is called reliability index β . (source: Lecture note of Prof. Noh)
13 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
MPP-based FORM Method
The linearized limit-state function at the MPP (u∗ ) can be described such that g (U) ∼ = gL (U) =
n X
(Ui −
ui∗ )
i=1
∂g ∂Ui
, Ui =ui∗
since g (u∗ ) = 0. The mean and standard deviation of gL (U) is expressed as µgL =E [gL (U)] = −
n X i=1
σgL
ui∗
∂g ∂Ui
, Ui =ui∗
v u uX p u n ∂g = Var[gL (U)] = t ∂Ui i=1
2 . Ui =ui∗
14 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
MPP-based FORM Method
Consequently, the probability of failure using MPP-based FORM is given by µgL 0 − µgL = Φ . Pfailure = P[g (U) > 0] ∼ 1 − Φ = σgL
σgL
The probability of failure can be rewritten as Pfailure
where α =
∼ =Φ
µgL σgL
=Φ −
n X
! αi ui∗
= Φ −u∗T α ,
i=1
∇g (u∗ ) is the normalized gradient vector. k∇g (u∗ )k
Finally, the probability of failure can be simply expressed in terms of β such that Pfailure ∼ = Φ −u∗T α = Φ −βαT α = Φ(−β). 15 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Reliablity-Based Design Optimization (RBDO)
MPP-based FORM Method
Probabilistic constraint P[g (X) > 0] ≤ P can be replaced by Φ(−β) ≤ Φ(−β target ),
where βtarget is called the target reliability index. In conclusion, the standard formulation of RBDO using MPP-based FORM is dened such that minimize
d
f (d)
subject to βitarget ≤ βi
d
lb
i = 1, · · · , Nc ,
≤ d ≤ dub .
16 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Demonstration
Random variables for AAF Example Mission phases
Random variables K1
Takeo
CL CD 0
Supercruise Combat turn
K1 CD 0 K1 CD 0 K1
Landing
CL CD 0
Mean value 0.180 1.389 0.014 0.270 0.028 0.180 0.016 0.180 1.210 0.014
Coecient of variation
Distribution type
0.1 " " " " " " " " "
Normal " " " " " " " " "
17 / 22
SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Demonstration
Comparison of RBDO Results with Discrete Target Reliability 1.6 Solution space 1.4
Thrust Loading (TSL =WT O )
RBDO applied to the aircraft initial sizing
Takeoff Supercruise Combat turn Landing
Initial Point 99% RBDO 95% RBDO 90% RBDO
1.2
DO 1
0.8
0.6
0.4 20
30
40
50
60
70
80
90
100
110
120
Wing Loading (WT O =S)
Figure: Results of DO and RBDO1 1 DO : Deterministic optimum, RBDO : Reliability-based design optimum. 2 Non-linear optimizer: sequential quadratic programming (SQP)
18 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Demonstration
Comparison of RBDO Results with Discrete Target Reliability Methods
Costs
WTO S
∗
TSL WTO
∗
Constraints
g1 = −5.303 DO
0.365
55.911
1.055
g2 = 0 g3 = 0 g4 = −14.627 g1 = −10.985
RBDO (90%)
0.444
56.518
1.176
g2 = −0.131 g3 = −0.113 g4 = −14.020 g1 = −12.555
RBDO (95%)
0.467
56.665
1.209
g2 = −0.168 g3 = −0.146 g4 = −13.873 g1 = −17.214
RBDO (99.7%)
0.535
57.058
1.314
g2 = −0.279 g3 = −0.245 g4 = −13.480 19 / 22
RBDO applied to the aircraft initial sizing SK Kim
Demonstration
Comparison of RBDO Results with Discrete Target Reliability
Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Methods
Sref (ft2 )
DO 429.254 RBDO (90%) 424.644 (1.074% ↓) RBDO (95%) 423.542 (1.331% ↓) RBDO (99.7%) 420.625 (2.010% ↓) *W TO = 24,000 lbf
TSL (lbf )
25,320 28,224 (11.470% ↑) 29,016 (14.597% ↑) 31,536 (24.550% ↑)
Findings 1
2
Compared to the deterministic design, the reliability-based design provides higher reliable results. Design with the higher target reliability tends to be oversized than that with the lower target reliability.
20 / 22
RBDO applied to the aircraft initial sizing
Summary and Conclusion
SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Summary 1
2
Reliability-based design optimization is proposed based on MPP-based FORM. To consider uncertainty, a total of ten aerodynamic data were selected as random variables.
Conclusion 1
Reliability-based design optimization using MPP-based FORM provides a reliable optimum point for an aircraft initial sizing.
Future Work 1 2
Sensitivity analysis. MPP-based second-order reliability method (SORM).
21 / 22
RBDO applied to the aircraft initial sizing SK Kim Outline Introduction Motivation and Research Proposition Formulations Constraint Analysis Reliablity-Based Design Optimization Demonstration Summary and Conclusion
Thank you for your attention! Questions?
22 / 22