Sunteți pe pagina 1din 9

Paper for International Conference of INMARTECH2008, France

Multidisciplinary Design Optimization Method


Applied to a HOV Design
Cao Anxi1, Cui Weicheng2
1

State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai, 200030, China

China Ship Scientific Research Center, P.O.Box 116, Wuxi, Jiangsu, 214082, China

Abstract
In this paper, a Multidisciplinary Design Optimization (MDO) procedure is applied to a HOV design.
Multidisciplinary decomposition and analyses have been developed for this complex system that includes
hydrodynamics, structure, propulsion, weight & volume. The Multi-Objective Collaborative Optimization (MOCO)
method is selected to conduct the preliminary conceptual design of the HOV. This approach was able to identify
Pareto front designs. The results also demonstrate that MDO approaches are more suitable for design of the HOV
and more flexible and advanced compared with the traditional design approach.
Keywords: MDO, HOV, Conceptual Design, Multi-Objective Collaborative Optimization (MOCO)

1. Introduction
The conceptual design stage of a complex engineering design process is the most critical to its
success or failure, as at this stage 75% of the final cost and performance metrics are determined.
The design of a Human Occupied Vehicle (HOV) is a complex and multidisciplinary task, and the
task is often divided into a set of smaller and easier tractable design problems. A complete design
requires analyses of hydrodynamics, structure, propulsion, weight, control, operations, cost and
the others. It is important for each of these aspects to be addressed at the conceptual design phase.
The traditional approach for the design of a HOV is of a sequential order. As shown in Fig. 1, in
this approach, design begins with the first discipline team, where the values of certain design
variables are fixed and passed to the second disciplinary team, and so on until a complete design
emerges from the last discipline. The traditional approach may lead to non-optimal system designs
for many reasons including: [1]
(1) Analysis in upstream discipline may depend on results determined in downstream
disciplines. In the sequential order approach, upstream disciplines must assume values
which may not match the actual values when they are finally determined in the
downstream disciplines.
(2) The system objectives (cost, performance, etc.) may depend heavily on results
determined in downstream disciplines where design freedom no longer exists to make
significant changes.
(3) The design may be so fixed by time and it reaches downstream disciplines that may be
1

impossible for them to satisfy their constraints.


It is obvious that the traditional approach to a HOV design is not suitable for the development
of modern HOVs which will become more and more complex.

Hydrodynamics

Propulsion

Energy

Structure

Weight &
Volume

Fig.1 The traditional conceptual design architecture of a HOV


The field of multidisciplinary design optimization (MDO) has emerged to develop approaches
for optimizing the design of large coupled systems

[2]

. MDO is concerned with how to efficiently

analyze and optimally design a system governed by the multiple coupled disciplines or made up of
coupled components. It is a part of the concurrent engineering technology that may well be an
enabling technology for complex advanced systems

[3]

. With the rapid growth of MDO over the

past decade, MDO has been widely discussed and used not only in aerospace and aeronautical
industries[4-6], but also in other complex engineering systems such as automobile[7], underwater
vehicles[8, 9], ship[10] etc. and resulted in a more reliable and better design.
For the purpose of attaining the overall performance optimization of a HOV and improving
procedure of a HOV conceptual design, MDO technique has been employed. The purpose of this
paper is to explore how Collaborative Optimization (CO), one of the MDO methods, can be
applied in the conceptual design of a HOV.

2. Multi-Objective Collaborative Optimization (MOCO)


Collaborative Optimization (CO), one of the multidisciplinary design optimization methods,
has been developed to promote autonomy while providing a coordinating mechanism that
guaranteeing progress toward an optimum and maintaining interdisciplinary compatibility

[11, 12]

It basically consists of a two-level optimization structure. The original structure of CO is shown in


Fig. 2. Within CO, the design task is accomplished by several disciplinary teams as well as by a
system-level team. The disciplinary teams are free to define their own local designs. The task of
the disciplinary-level teams is to find a local design that satisfies local constraints and comes as
close to that specified by system-level optimizer as possible. The system-level team is in charge of
2

adjusting the system variables with the goal of minimizing or maximizing the system-level
objective. This problem is subjected to the interdisciplinary compatibility constraints equal to
zero.
System-Level optimizer
Goal: Design objective
s.t.:
Interdisciplinary
compatibility constraints

Subspace optimizer 1
Goal: Interdisciplinary
compatibility
S.t.: Analysis 1
constraints

Subspace optimizer 2
Goal: Interdisciplinary
compatibility
S.t.: Analysis 2
constraints

Analysis 1

Analysis 2

Subspace optimizer N
Goal: Interdisciplinary
compatibility
S.t.: Analysis N
constraints

Analysis N

Fig2. The Basic Collaborative Optimization architecture


CO has been widely discussed and applied in practical engineering problems. Some
researchers have applied it to both simple test problems[11-15] and more complex engineering
design problems. These applications involve launch vehicle design[4], aircraft design[5,6,16,17],
undersea vehicles design[8,9], conceptual ship design[10]and turbine engine design[18]. CO has been
judged highly advantageous in its applications to practical engineering design problems. At the
same time, many researchers have focused on extension or modifications to CO aimed at
improving overall efficiency, permitting their use on problems with high dimensionality coupling
and simplifying their implementation. In order to relieve the numerical difficulties caused by
certain mathematical manipulations, the use of an approximation model has been proposed in
place of the disciplinary design in CO [17,19]. To resolve the convergence problem of CO, Kroo and
Manning

[12]

adopted the direct search method such as Hooke and Jeeves method or the

probabilistic search method such as genetic algorithm instead of the gradient-based method. A
variety of extensions have been made to CO including resolving the system including mixed
continual and discrete design variables [1], and introduction of multi-objective formulation[20, 21].
In this study, a MOCO has been selected to handle multiobjective systems. In the MOCO, the
goal of the system level optimizer is to minimize a system level multiobjective function of target
variables while satisfying compatibility constraints using a Pareto Genetic Algorithm based on
(PGA). PGA solves system level optimization problem with respect to system design variables.
For each generation at the system level, the disciplines are optimized for each candidate design
from the population. The system level optimization problem is described as equations (1-3)
3

Min :{f1 , f2 ,L , f N }
S .t.:

(1)

J i 0, i = 1, 2,L , N

(2)

D.V . x o = [x osh , x oaux , x1 , x 2 , , x m , ]

(3)

Equations (4-7) describe the subsystem level optimization problem for a typical subsystem, in
this case subsystem 1

Min : J 1 (x ss1 ) = (1

S .t. :

D.V .

y1 j
(x sh )1 2
(x aux ) i1 2 N
x1 2 N
)
+
(
1

)
+
(
1

)
+
(
1

) 2 (4)

o
o
o
o
(x sh )1
x1
(x aux ) i1
(x aux )1 j
i=2
i =2

g1 0

(5)

(x ss1 ) min x ss1 (x ss1 ) max

(6)

x ss1 = [(x sh )1 , (x aux ) i1 , x1 ]

(7)

The MOCO architecture for HOV design that has been developed for conceptual design is
briefly discussed, a more detailed description can be found in Ref. [21].

3. HOV design Application


A MOCO method has been completed for a deep sea HOV which is shown conceptually in
Figure 3. The shape, type of propulsion, ascent depth, pressure hull structure and material have
been identified. The design space of this vehicle has been described in Table 1. The cruise and
operation time Tco and the Ratio Rw are the objective attributes.

Dmax

2
La

Dmax
2

Lm
Fig.3 HOV configuration

Table 1 Design variables of HOV general design


variable

unit

Baseline

Lower
Bound

Upper
Bound
4

Lm

4.45

La

3.0

2.8

3.2

3.0

2.6

3.2

kn

2.5

2.5

3.0

Wpl

kg

220

200

300

kWh

110

90

140

Wbuo

kg

6374

5500

7000

Firstly, we decompose the design problem into a system module and four disciplinary
optimization modules: geometry & hydro, structure, propulsion, weight & volume. Figure 4
describes the MOCO architecture of the HOV design.

System
optimizer

J1

D o , Lo ,

D o , Lo , RTo

uo, Eo

o
o
o
W plo , Wbuo
, W pro
, V pro

RTo , V fo

Wso ,Vso ,WTo

o
o
W pro
, V pro

Wso , Vso , WTo , V fo

Discipline 1
Geometry & hydro

J2

Discipline 2
Structure

J3

Discipline 3
Propulsion

J4

Discipline 4
Weight & volume

Fig.4 MOCO architecture of the HOV design


Equations (8-10) describe the formulation of system-level optimization problem, and
Equations (11-21) describe the four subsystem optimization problems.
1 System-level stand formulation

Min : {Rw , Tco }


S .t. : J i ,

(8)

(i = 1,2,3,4)

o
o
D.V . : x o = [ D o , u 0 , E o , Lo , RTo ,V fo ,Wso , Vso , W pro
,V pro
,W plo ]

(9)
(10)
5

2 Discipline 1, Geometry and hydro subspace stand formulation:

Min : J 1 = (1

Vf 2
RT 2
D 2
L 2
)
+
(
1

)
+
(
1

)
+
(
1

)
Do
Lo
RTo
V fo

(11)

S .t. : 4 L 5m 2.8 L 3.2m 2.6 D 3.2m


m
a

(12)

D.V . : x ss1 = [ La , Lm , D]

(13)

3 Discipline 2, structure subspace stand formulation:

W
V
W
D 2
L
) + (1 o ) 2 + (1 so ) 2 + (1 so ) 2 + (1 To ) 2
o
D
L
Ws
Vs
WT

Min : J 2 = (1

(14)

S .t. : 2.6 D 3.2m

(15)

D.V . : x ss1 = [ D, L, WT ]

(16)

4 Discipline 3, propulsion subspace stand formulation:

W pro 2
V pro 2
R
u 2
E 2
)
+
(
1

)
+
(
1

)
+
(
1

) + (1 To ) 2
o
o
o
o
u
E
W pro
V pro
RT

Min : J 3 = (1

(17)

S .t. : 2.5 u 3.0kn 90 E 140kWh

(18)

D.V . : x ss1 = [u , E , RT ]

(19)

5 Discipline 4, weight & volume subspace stand formulation:

Min : J 4 = (1

W pl
W

+ (1

o
pl

) 2 + (1

Vf
Wbuo 2
W
V
) + (1 o ) 2 + (1 so ) 2 + (1 so ) 2
o
Wbuo
Vf
Ws
Vs

W pro 2
V pro 2
WT 2
)
+
(
1

)
+
(
1

)
o
o
WTo
W pro
V pro

(20)

S .t. : 200 W 300kg 5500 W 7000kg 40 V W 80kN W 22t (21)


pl
bou
sw
T
T
D.V . : x ss1 = [W pl , Wbuo , V f , Ws , Vs , W pro , V pro ]

(22)

5.1. Results and Discussion


A Multi-Objective Collaborative Optimization has been run for 328 generation with a
population of 50 HOVs. In system-level optimization problem, the relaxation factor of compatible
constraints is set to 0.0001, the crossover probability, the mutation probability and the maximum
generation are set to 0.9 , 0.1 and 350. For the sub-space optimization problems, the sequence
quadratic programming (SQP) is used to attain the discipline optimization solution. The different
6

sub-space optimizations are solved in-parallel.


Results are presented in Figure 5. None of these HOV can be identified as "the best". Selection
of the preferred design is up to the designer. Figure 5 provides the designer with important
information to make this selection. Table 2 list four designs which is selected from the Pareto
optimal solutions

Parato front
7.0

HOV2

HOV1
6.5

Tco (h)

6.0

5.5

HOV3
5.0

HOV4

4.5

4.0
0.0120

0.0125

0.0130

0.0135

0.0140

0.0145

0.0150

Rw

Fig. 5 Pareto optimal solutions of HOVs multi-objective multidisciplinary design

Tab. 1 The four designs in Pareto front.


desig
n

D /m

Lm /m

La /m

HOV1

2.93

4.60

2.95

HOV2

2.87

4.63

HOV3

2.81

HOV4

2.83

u /k

E / kW h

Wbuo /k

W pl /k

Tco /

Rw

2.50

131.99

g
7000.00

g
265.70

h
7.00

0.0123

2.93

2.50

119.91

6845.00

300.00

6.46

0.0142

5.07

3.08

2.51

110.47

6656.00

300.00

5.29

0.0146

4.60

2.93

2.51

98.83

6482.00

300.00

4.39

0.0148

HOV1 and HOV4 are located at the ends of the Pareto front, and HOV2 and HOV3 are located
at the middle of the Pareto front. HOV1 has the longest cruise and operation time which is up to 7
hours, but the alternative has the minimum ratio which is just 0.0123. In contrast, HOV4 has the
maximum ratio and the shortest cruise and operation time. If the time is the most important
performance for a HOV, the HOV between HOV1 and HOV2 are excellent choices.

4. Summary and Conclusions


An application of Multidisciplinary Design Optimization to a HOV conceptual design has
been presented. In this application, the MOCO architecture was used. The method integrates the
multi-objective optimization methods within the collaborative optimization framework, which
remains the main metrics of CO architecture and ability of PGA to seeking non-inferior solution
set. So the method is effective in that it makes a chance to execute in-parallel for disciplinary
design, and it is more flexible in that it enables the designer to select the fittest solution among the
Pareto optimal set in according with their preference and the nature of the design problem. These
practical advantages make the architecture well-suited for the design of HOVs.
References:
[1]

Balling R J, Gale D L. Collaborative Optimization of Systems Involving Discrete Design at the

Discipline Level [J]. Journal of Mechanical Design, 1998(120):32-39.


[2]

Sobieszczanski-Sobieski J, Haftka R T. Multidisciplinary Aerospace Design Optimization: Survey

of Recent Developments [J]. Structural Optimization, 1997, 14(1):1-23.


[3]

Balling R J, SobieszczanskiSobieski J. Optimization of coupled systems: A critical overview of

approaches [J]. AIAA Journal, 1996, 34(1):617.


[4]

Braun R D, Moore A A, Kroo I M. Collaborative architecture for launch vehicle design [J].

Journal of Spacecraft and Rockets, 1997, 34(4):478-486.


[5]

Sobieski I P, Kroo I M. Collaborative Optimization Applied to an Aircraft Design Problem[C].

AIAA Paper 96-0715, the 34th AIAA Aerospace Sciences Meeting and Exhibit. Reno, Nevada, January
1518, 1996.
[6]

Batill S M, Stelmark M A, Yu X Q. Multidisciplinary design optimization of an electric-powered

unmanned air vehicle [J]. Aircraft Design, 1999(2):1-18.


[7]

Kodiyalam S. Evaluation of methods for multidisciplinary design optimization, Phase I [R]. Tech.

Report, NASA/CR-2000-210313, National Aeronautics and Space Administration, 1998.


[8]

Belegundu A D, Halber E, Yukish M A, Simpson T W. Attribute-based multidisciplinary

optimization of undersea vehicles [C]. AIAA Paper 2000-4865, the 8th AIAA/USAF/NASA/ISSMO
Symposium on Multi- disciplinary Analysis and Optimization. Long Beach, CA, Sept. 6-8, 2000.
[9]

McAllister C D, Simpson T W, Kurtz P H, Yukish M. Multidisciplinary design optimization test

based on autonomous underwater vehicle design[C]. The 9th AIAA/ISSOM symposium on


Multidisciplinary Analysis and Optimization. Atlanta, Georgia, Sept. 4-6, 2002.
[10]

Kodiyalam S, Sobieszczanski-Sobieski J. Bi-level integrated system synthesis with response

surfaces [J]. AIAA Journal, 2000, 38(8): 1485-1497.


[11]

Braun, R. D. Collaborative Optimization: An architecture for large-scale distributed design [D].

Ph.D. thesis, Stanford University, Department of Aeronautics and Astronautics, 1996.


[12]

Kroo I, and Manning V, Collaborative Optimization Status and Directions [C]. AIAA Paper

2000-4721, the 8th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and


Optimization. Long Beach, CA, Sept. 6-8, 2000.
[13]

DeMiguel A, Murray W. An Analysis of Collaborative Optimization Methods[C]. AIAA Paper

2000-4720, the 8th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and


Optimization. Long Beach, CA, Sept. 6-8, 2000.
8

[14]

Braun R D, Gage P, Kroo I M. Implementation and performance issues in collaborative

optimization[C]. AIAA-96-4017, the 6th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary


Analysis and Optimization. Washington, September, 1996.
[15]

Alexandrov N M, Lewis R M. Analytical and Computational Aspects of Collaborative

Optimization for Multidisciplinary Design [J]. AIAA Journal, 2002, 40(2):301-309.


[16]

Manning V. High Speed Civil Transport Design via Collaborative Optimization [D]. Ph.D. thesis,

Stanford University. 1999.


[17]

Jun S, Jeon Y, Rho J, D Lee. Application of Collaborative Optimization Using Response Surface

Methodology to an Aircraft Wing Design [C]. AIAA-2004-4442, the 10th AIAA/ISSMO


Multidisciplinary Analysis and Optimization Conference. Albany, New York, Aug. 30-1, 2004.
[18] Rohl P J, He B, Finnigan P M. A collaborative optimization environment for turbine engine
development[R]. AIAA Paper, No. 98-4734, 1998.
[19] Sobieski I P, Kroo I M. Collaborative Optimization using Response Surface Estimation [J]. AIAA
Journal, 2000, 38 (10): 1931-1939.
[20]

Tappeta R V, Renaud J E. Multiobjective collaborative optimization [J]. Journal of Mechanical

Design, 1997, 119(3): 403-411.

[21]

Cao An-xi, Cui Wei-cheng. Multi-Objective collaborative optimization in multidisciplinary

design for submersible [J]. Journal of Ship Mechanics, Vol.12, No.2, 2008.4.