Development and Implementation of a Methodological Approach to

Support MDO by Means of Knowledge based Technologies

M. F. M. Hoogreef and G. La Rocca

Faculty of Aerospace Engineering, Delft University of Technology, Kluyverweg 1, 2629HS, Delft, The Netherlands

1 STAGE OF THE RESEARCH

The research started in mid-October 2013 with an

introduction into the European Community project

“iProd” (iProd Integrated Management of product

heterogeneous data - FP7 project of the European

Community, Grant agreement no. 257657). Over the

course of the ﬁrst year, the research has focused on

project work and becoming accustomed to the tools

and methods of the current state of the art within

multi-disciplinary design optimization (MDO) and

knowledge engineering. Currently, an initial MDO-

advisory system is under development, coupled to an

MDO architecture design system, with the end-goal

in mind to formalize MDO knowledge to achieve au-

tomation and a lower accessibility level. This ﬁrst

concept is under development together with two other

PhD candidates at the department of Flight Perfor-

mance and Propulsion, R.E.C. van Dijk and D. Steen-

huizen.

2 OUTLINE OF OBJECTIVES

Although the concept of MDO is not new at all, the

discipline is not yet fully exploited. For industrial

applications, this seems a missed chance to improve

product quality and cut design cycle time and cost.

The reasons are the currently not addressed intrinsic

complexity of MDO problems and the mathematical

background, as well as the lack in awareness and

understanding of MDO architectures. The main

purpose of this PhD research is the development

of an methodological approach, with knowledge

based technologies to support MDO, and speciﬁcally

address the two issues mentioned above. This will

include the following objectives and sub-objectives:

• Objective: Development of an MDO

Knowledge Base

A knowledge base will be developed to capture,

formalize and organize dispersed knowledge

on MDO approaches and architectures, as well

as guidelines concerning the applicability of

certain MDO strategies to problems of different

nature. Excellent surveys on MDO architectures

and their performance evaluation are available

in specialized literature such as Tedford and

Martins (Tedford and Martins, 2010) and Perez

et al. (Perez et al., 2004). Other MDO system

knowledge will be obtained from the activities

carried out in the iProd project.

• Sub-objective: Development of an MDO

Dedicated Ontology

Storing and structuring this type of knowledge

will require the investigation of existing suitable

ontologies and/or the development of an ontology

dedicated to MDO. This will build on the work

currently being ﬁnalized by R.E.C. van Dijk (PhD

candidate) and the work within the iProd project.

A categorization approach of different MDO

architectures, as well as solving algorithms, exists

in literature and will provide a suitable starting

point.

• Objective: Development of an MDO Advisory

System

An MDO advisory system will be developed

to enable the user i.e. an engineer who is not

necessarily an MDO expert to describe the

problem at hand, specify a number of selection

criteria (accuracy, speed, etc.) and obtain in

return (based on the knowledge stored in the

knowledge base) a ranked list of suitable MDO

architectures, with (links to) relative documenta-

tion, and a template to support its implementation.

• Sub-objective: Implementation of Interactive

MDO Architecture Design

The MDO advisory system will also provide

some interactive design capabilities of MDO

architectures, allowing the user to deﬁne in an

intuitive and interactive way his own MDO

F. M. Hoogreef M. and La Rocca G. (2013).

Development and Implementation of a Methodological Approach to Support MDO by Means of Knowledge based Technologies.

In Doctoral Consortium, pages 3-9

 SCITEPRESS

strategy, either from scratch or based on available

templates (MDO architecture meta-model).

• Objective: Develop MDO Work-ﬂow

Instantiations for Assembled MDO

Architectures

Finally, capabilities will be developed to trans-

late (hence automate all the software intensive op-

erations) the selected or assembled MDO archi-

tecture, into an MDO work-ﬂow instantiation by

means of a commercial work-ﬂow management

system such as, for example, Optimus by Noesis

or the open source framework RCE, developed by

DLR.

3 RESEARCH PROBLEM

Multidisciplinary Design Optimization can provide

designers with the methods to further improve the

performance of already mature solutions, and to sup-

port the exploration of innovative complex engineer-

ing products. The best design of a complex system

can only be found when the interactions between the

systems disciplines are fully considered (Tedford and

Martins, 2010). MDO provides the structured ap-

proach and the mathematical formulations to capture

such interactions.

Although the very ﬁrst MDO implementations

have been presented about 50 years ago, at date, the

discipline is not yet fully exploited at industrial level,

apart from limited application (Agte et al., 2010).

This seems a missed chance to improve product qual-

ity and cut design cycle time and cost. Some of the

reasons for the limited exploitation of MDO in indus-

try (in contrast with the growing interest and body

of knowledge being developed in academia) can be

found in the following:

1. Lack of adequately ﬂexible, accurate and robust

parametric models to support MDO using high-

ﬁdelity simulations.

2. Limited availability of computation resources to

solve complex problems of industrial interest.

3. Intrinsic complexity of the MDO discipline and

its mathematical foundations.

4. Lack in awareness and understanding of the many

available MDO architectures and their speciﬁc

suitability to problems of different nature.

Advances in knowledge-based parametric CAD

(Computer Aided Design) modeling, e.g. KBE

(Knowledge Based Engineering) applications

(La Rocca, 2012), and robust pre-processing tools to

support high-ﬁdelity analysis seem to successfully

address the ﬁrst issue. This PhD research will demon-

strate the beneﬁt of deploying KBE applications to

support some knowledge intensive engineering

process, as deﬁned in the EC project iProd.

3.1 Flexible Problem Formulation

In multidisciplinary design optimization (MDO), the

problem deﬁnition is considered ﬁxed. I.e. the prob-

lem formulation in terms of the number of design

variables is not ﬂexible. However, in true engineer-

ing design optimization problems, the engineer con-

tinuously varies the product model, thereby also al-

tering the mathematical problem description. Agte, et

al. (Agte et al., 2010) presented an assessment and

the direction for advancement of MDO in their pa-

per summarizing the 2006 European U.S. Multidisci-

plinary Optimization Colloquium in G

ottingen, Ger-

many. While discussing vertical growth of MDO, de-

scribed by the authors as treading new grounds, the is-

sue of implementing creativity, cognition and ﬂexibil-

ity to MDO is raised. Not only allowing for require-

ments to change over time, but also the possibility of

adding design variables, and thereby adding new di-

mensions to the design space, is considered an im-

portant, though difﬁcult direction for future research.

As Agte, et al. (Agte et al., 2010) note: Generaliza-

tion of the very concept of the design space to enable

qualitative transformation of that space by adding, re-

ﬁning, and removing design variables simultaneously

with the numerical search would simulate what to a

competent designer comes naturally and apparently is

one of the intrinsic capabilities of the human mind.

In other words, allowing the optimization to adapt the

problem formulation and design variables itself will

mimic the way the human mind works when solving

design problems. For optimization, e.g. using a ge-

netic algorithm, this would mean that every iteration,

or design mutation, has to consider a redeﬁnition of

the design space and at the same time let the math-

ematical model of the optimization manifest this re-

deﬁnition (Agte et al., 2010). These topology varia-

tions have an inﬂuence on the optimization problem,

and have to be taken into account. Implying speciﬁc

knowledge of the mathematical problem formulation

and its attributes of the problem at hand. A feat that is

not necessarily applicable to engineering optimization

problems, which are often complex, highly non-linear

and may involve changes in the problem deﬁnition.

Flexible problem formulation in MDO is also ad-

dressed by Welle, et al. (Welle et al., 2012) who

claim that no current MDO literature addresses ﬂex-

ible problem formulation. Though focusing on Ar-

IC3K2013-DoctoralConsortium

chitecture, Engineering and Construction (AEC), the

intrinsic problem is similar for any discipline. Welle,

et al. (Welle et al., 2012) achieve to link the prod-

uct model to the CAD model, allowing changes to be

made to the attributes of the product model. Their

example relates to the design of a building with 4

walls and 10 windows per wall, changing attributes

of either one window or the windows on one side of

the building or all windows individually. As Welle,

et al. (Welle et al., 2012) indicate; design deci-

sions are made upstream of any analysis modules,

requiring ﬂexible problem formulation in terms of

the product model, in contrast to conventional MDO

where ﬂexible problem formulation relates to ﬂexi-

bility in problem formulation construction in terms of

the ability to employ various optimization algorithms

or the sequencing of optimization algorithms. How-

ever, Welle, et al. (Welle et al., 2012) provide only

limited ﬂexibility to the product model (and mathe-

matical problem deﬁnition), as only the attributes of

the windows can change, yet the construction of the

building remains unchanged. A fully ﬂexible problem

formulation would also allow the building to change,

as the engineer, in his creative design phase, may con-

sider a building with either more, or fewer walls and

varying the amount of windows per wall. This re-

quires a higher level of ﬂexibility in problem formula-

tion and dynamic deﬁnition of slaves in a master-slave

MDO set-up. I.e. slaves can exist in one branch of the

optimization, but may not exist for another branch of

the optimization.

Also Amadori et al. (Amadori et al., 2012) present

a certain form of problem ﬂexibility their MDO prob-

lems. Their use of High Level CAD templates is sim-

ilar to the High-level Primitives as described by La

Rocca and Van Tooren (La Rocca and van Tooren,

2007). However, the implementation of ﬂexible prob-

lem formulation for these CAD templates is lim-

ited to a ﬁrst optimization where only morphological

changes are allowed and a second run where only the

topology of an optimized morphology (geometry) is

allowed to change.

3.2 Knowledge based Technologies

Implementing creativity, cognition and ﬂexibility in

MDO, requires a ﬂexible optimization work-ﬂow en-

vironment that can reconsider and reformulate the de-

sign problem at hand and adapt the optimization ar-

chitecture accordingly. Referring to the aforemen-

tioned issues 3 and 4, the intrinsic complexity of

MDO and its mathematical foundation as well as the

lack in awareness and understanding of the many

available MDO architecture makes the implementa-

tion difﬁcult. This is where a methodological ap-

proach through knowledge based technologies can

help.

Through the formalization of MDO knowledge,

for example through a dedicated ontology, this knowl-

edge can be made available for re-use and automatic

implementation, also to engineers who are not MDO-

experts. The stored and structured knowledge can

be used to automatically implement an MDO work-

ﬂow, according to user inputs for the problem at hand.

By also providing relevant documentation, or sources

of documentation, the accessibility level of MDO is

signiﬁcantly lowered. The intrinsic complexity and

mathematical background of MDO are thus not di-

rectly exposed to anyone implementing MDO and

knowledge and understanding of all available archi-

tectures and their implementation and appropriateness

is not required.

Also the implementation of KBE applications that

automate repetitive, non-creative work and allow for

rapid design iteration can boost the use and ease

of implementation of MDO. These application also

use captured design knowledge and rationale to auto-

matically create geometric engineering designs, from

high-level geometric primitives.

The combination of knowledge based technolo-

gies and these KBE applications can lower the acces-

sibility level of MDO and increase its level of imple-

mentation in engineering product development.

4 STATE OF THE ART

4.1 Engineering Knowledge

Management and Concurrent

Engineering

The engineering of complex products, such as air-

planes, automobiles or any kind of appliances is

highly knowledge intensive and requires vast amounts

of knowledge and data to be handled and managed in

a very precise way. The whole product life cycle is

built on an information model of the product with all

its aspects and related processes. Information about

the design and expertise, the know-how which resides

in the knowledge workers is combined along the engi-

neering process. The shortening timescales, widening

geographic distribution and increasing complexity of

the design task in all these aspects make the effective

management of this know-how even more important,

as observed by Wallace et al. (Wallace et al., 2005)

and Scholl et al. (Scholl et al., 2004). The Con-

current Engineering (CE) approach was proposed as

DevelopmentandImplementationofaMethodologicalApproachtoSupportMDObyMeansofKnowledgebased

Technologies

collaborative approach to support the development of

products (e.g. as described by Clarkson and Eckert

(Clarkson and Eckert, 2005)) to achieve faster, bet-

ter, cheaper products. CE emphasizes the importance

of frequent iterative interaction between traditionally

separate functional teams.

This PhD research builds on the above by access-

ing and collecting the information and design exper-

tise residing in engineering and making it available

for (automated) reuse, through a knowledge base. The

knowledge and information is formalized and stored.

4.2 Multi-Objective and

Multi-Disciplinary Design

Optimization

In the design process in highly networked and interde-

pendent industrial contexts, such as the aeronautical

industry, the effective design of components and sys-

tems and processes is essential for improving prod-

ucts cost-efﬁciency but requires the achievement of

multiple and, often conﬂicting, objectives in pres-

ence of multiple and multi-criteria constraints. To ac-

complish such tasks, appropriate design optimization

strategies using effective constraint-handling tech-

niques must be carried out. In these procedures, up-

dated design rules have to be developed to optimize

product design in order to meet the more and more

increasing requirements for performance, weight, sur-

vivability and cost reduction. Presently there are sev-

eral methods to deal with such design problems. They

can be roughly summarized into two categories:

1. Trial and error methodology (often referred to

as heuristics), supported by engineering knowl-

edge based on numerical simulations and/or ex-

periments.

2. Computer-assisted optimization.

Trial and error methodology has been the standard de-

sign procedure in industry for a long time and still

is regarded as very important. This is an experi-

ence based method that is used to reduce the need

for calculations pertaining to general equipment size,

performance, or operating conditions. Heuristics are

therefore fallible and they do not guarantee an opti-

mal solution although they may be of value because

they offer time saving approximations in the prelimi-

nary design process. Examples of heuristics in aero-

nautics are almost uncountable: in particular, most

of the experimental work for product development

makes to some extent substantial use of it. On the

other hand, the multidisciplinary nature of aeronau-

tical engineering problems has led researchers to in-

vestigate formal optimization methods (i.e. featuring

a solid mathematical basis alternative to heuristics)

for the component design process. Sobieszczanski-

Sobieski and Haftka (Sobieszczanski-Sobieski and

Haftka, 1997) give a review of developments in

multidisciplinary design optimization for aerospace

problems and Giesing and Barthelemy (Giesing and

Barthelemy, 1998) provide an industrial perspective

of multidisciplinary optimization research. The com-

mon basis joining the works above is that consid-

erable effort is spent building efﬁcient optimization

method coupled with numerical solvers (which led to

the so-called computer-assisted optimization), partic-

ularly aerodynamic and aero-elastic ones. Computer-

assisted optimization methods applied to aircraft

are mainly of two types: gradient-based and non-

gradient-based. Moreover, the theoretical approach

to multi-objective optimization is somewhat weak in

that multiple objectives are always resembled into one

objective function using arbitrary weights. To some

extend, computer-assisted optimization makes use of

KBE applications to rapidly create and vary geome-

tries according to a set of design rules. However, the

implementation of KBE applications is still limited.

This research aims at going one step further and to

develop a methodology for choosing the most appli-

cable solver, optimization architecture and algorithm

for the problem at hand. This will be done through an

MDO advisory system, that suggests an MDO strat-

egy and implementation to the engineer. Formalizing

this knowledge can make MDO more accessible.

5 METHODOLOGY

To address the objectives stated in Section 2 and de-

velop, and implement a methodological approach to

support MDO by means of knowledge based tech-

nologies, the following top level tasks have been iden-

tiﬁed:

1. Review of the current state of art in the ﬁeld

of MDO (and supporting technologies, such as

KBE, integration frameworks, etc.) and Knowl-

edge Technologies (Knowledge bases, reasoning

mechanisms, ontology engineering, etc.).

2. Deﬁnition of knowledge base(s) and ontologies to

capture MDO architectures/strategies.

3. Development of an MDO advisory system to sug-

gest suitable MDO strategies and support the def-

inition and implementation of the actual MDO

framework.

4. Veriﬁcation and validation of the developed tools

and methods.

IC3K2013-DoctoralConsortium

The methodology of the research requires becoming

accustomed with current tools and technologies used

in the ﬁeld of MDO and knowledge technologies,

for example during the work performed in the iProd

project. The actual basis for the framework, acting as

an MDO advisory system, that is to be developed and

implemented is a knowledge base, which is structured

on an ontology and can capture MDO architectures

and MDO problem descriptions. As such, this ontol-

ogy will also be linked to an optimization ontology

which is currently being developed in iProd. Using

this semantic description for the knowledge base, a

graph-database stored in Allegro Graph, it is possible

to reason on this structured knowledge and the knowl-

edge can be easily reused and extended. The develop-

ment of such a framework consists partly of the devel-

opment of an advisory system for multidisciplinary

design optimization (MDO) that can enable the user

to specify an MDO problem, suggest optimization

strategy and algorithm and guide the user during the

implementation of the suggested approach. The MDO

advisory system that is to be developed, will enable

any user, i.e. engineers who are not necessarily MDO

experts, to describe the optimization problem and

specify certain selection criteria. These selection cri-

teria could e.g. be accuracy, speed, etcetera. In return,

the system will provide a ranked list of suitable MDO

architectures, based on the knowledge that is stored

in the knowledge base. This also requires a speciﬁca-

tion of the mathematical problem formulation of the

optimization problem at hand. Based on mathemati-

cal speciﬁcations, e.g. the type of design variables or

the decomposition into coupled or non-coupled sub-

problems, the system will be able to return a list of

MDO architectures that are most suitable to the spe-

ciﬁc problem. In addition to the ranked list of archi-

tecture (or strategies), (links to) related documenta-

tion will be provided. After selection of an architec-

ture by the user, the advisory system will provide a

template (from the knowledge base) that supports the

implementation in simulation workﬂow management

software. This interaction allows the user to still in-

ﬂuence the optimization strategy to his desire, for ex-

ample for user with more expert knowledge on MDO

that are looking for a speciﬁc implementation. The

interactive design capabilities of an MDO architec-

ture allow the user to deﬁne, in an intuitive and in-

teractive way, his own MDO strategy. This could be

from scratch of based on the availabel templates (or

MDO architecture metamodels). This “new” knowl-

edge will then be added to the knowledge base for fu-

ture use. The possibility to select a desired algorithm,

which is part of an MDO strategy, from the ones

stored in the knowledge base. The knowledge base

will be structured through a dedicated MDO ontology.

This ontology will contain the different MDO strate-

gies and algorithms, structured and linked through

the different classes, properties and restrictions. The

development of an MDO Advisory System, imple-

menting an an interactive MDO Architecture Design,

building on the knowledge bases and reasoning mech-

anisms developed within the iProd project, is focusing

on the following questions:

1. Can a methodological approach, to support MDO

by means of knowledge based technologies, help

non-MDO-experts in industry and academia in the

implementation of MDO in engineering?

2. Can this methodology reduce the number of itera-

tions and the time consumed for optimization and

improve the designs of coupled KBE applications

and analysis tools? (through workﬂow manage-

ment software)

3. Can the design framework lower the accessibility

level of MDO, allowing also non-experts and stu-

dents to use proper and applicable MDO architec-

tures and optimization algorithms in engineering

optimization problems?

The developed methodologies and implementa-

tions will be tested on use cases from the different

research projects, such as the Pininfarina and Fokker

Aerostructures use cases from the iProd project, in-

volving KBE applications and optimization. Valida-

tion will focus on achieving better solutions (in terms

of design objectives) for the projects in equal or less

time, or having similar solutions in the less time.

To summarize, an MDO advisory system will be

developed that is able to suggest a suitable MDO strat-

egy, based on the knowledge and rules that are col-

lected from literature and projects, such as the iProd

project. In addition, the choices made by users of

this framework will be collected and stored for future

research when implementing other MDO problems.

The advisory systems will also be able to assist in

the deﬁnition and implementation of the actual MDO

framework and, where possible refer to relevant liter-

ature and information sources during the implemen-

tation of an MDO problem. The automatic work-ﬂow

instantiation of an assembled MDO architecture re-

lates to the latest work within the group of Flight Per-

formance and Propulsion, such as the new methodol-

ogy for the development of simulation work-ﬂows in

the thesis of Chan (Chan, 2013).

DevelopmentandImplementationofaMethodologicalApproachtoSupportMDObyMeansofKnowledgebased

Technologies

6 EXPECTED OUTCOME

This PhD research will deliver a knowledge base, an

MDO advisory system and an interactive MDO archi-

tecture designer system. These will enable - also non

MDO experts - to ﬁnd knowledge related to MDO

system architecture, get advice on the most suitable

method for the problem at hand, get assistance in the

actual deﬁnition of a MDO architecture and ﬁnally

obtain an MDO framework implementation that is

ready for use, without the need to go through (all)

the software speciﬁc operations required to operate

the given MDO framework system (e.g. NOESIS Op-

timus). Figures 1 and 2 show the example interface

for a ﬁrst prototype of an MDO architecture design

system.

Figure 1 shows that problem deﬁnition tab, where

it is possible to specify the objective function, select

design variables from the product tree, specify con-

straints and select the desired responses. These are all

directly related to the KBE application that holds the

parametric geometrical model to be optimized. This

allows for the suggestion of e.g. design variables from

the code of the KBE application, but also gives room

to adapt the mathematical model interactively, which

may also be constructed to be reﬂected in the code.

The idea for this interactive approach and the vari-

ability is the ﬂexible problem formulation. E.g. an

actuator mechanism, where not only the size of ac-

tuator arms and location of actuation points and ac-

tuators is variable, but also the number of arms and

actuation points. This means that the number of de-

sign variables is variable throughout the optimization

and also that the product tree that is shown on the left,

is variable.

Figure 1: Example interface for the ﬁrst prototype architec-

ture design system, showing the problem deﬁnition tab.

(

⃝Van Dijk, Steenhuizen and Hoogreef, 2013).

Figure 2 shows the architecture design tab, where

the ﬂow through the problem architecture can be spec-

iﬁed. Building on the problem description made in the

previous tab, where the objective function and sub-

problems of the optimization can be speciﬁed, the de-

sign variables, desired outputs and constraints can be

selected and the product structure (or topology) of the

model to analysed, an optimization strategy is sug-

gested. The user can then, interactively, relate in-

puts and outputs of the different modules (or sub-

problems) to each other, constructing a ﬂow. This

ﬂow can later be translated to an actual simulation

workﬂow in simalution workﬂow management soft-

ware.

Figure 2: Example interface for the ﬁrst prototype architec-

ture design system, showing the architecture design tab.

(

⃝Van Dijk, Steenhuizen and Hoogreef, 2013).

The knowledge base will provide the functionali-

ties to store also knowledge relative to business pro-

cess, hence it will enable engineers to leverage their

experience both in human-oriented tasks as well as in

simulation and optimization work-ﬂows. In this case

the human-oriented tasks can provide the rationale be-

hind a certain simulation or technical work-ﬂow im-

plementation.

While the proposed approach is innovative in na-

ture and goes behind the state of the art (no MDO

advisory systems are discussed in literature), it will

leverage on the experience and demonstration tools

developed by Delft University of Technology in iProd

and other currently running research initiatives.

REFERENCES

Agte, J., De Weck, O., Sobieszczanski-Sobieski, J., Arend-

sen, P., Morris, A., and Spieck, M. (2010). Mdo: As-

sessment and direction for advancement-an opinion of

one international group. Structural and Multidisci-

plinary Optimization, 40(1-6):17–33.

Amadori, K., Tarkian, M.,

Olvander, J., and Krus, P. (2012).

Flexible and robust CAD models for design automa-

tion. Advanced Engineering Informatics, 26(2):180–

195.

IC3K2013-DoctoralConsortium

Chan, P. K. M. (2013). A new methodology for the develop-

ment of simulation workﬂows: Moving beyond moka.

Master’s thesis, Delft University of Technology - Fac-

ulty of Aerospace Engineering.

Clarkson, J. and Eckert, C. (2005). Design process improve-

ment: a review of current practice. Springer.

Giesing, J. and Barthelemy, J.-F. (1998). A summary of

industry mdo applications and needs. In Multidisci-

plinary Analysis Optimization Conferences, pages –.

American Institute of Aeronautics and Astronautics.

La Rocca, G. (2012). Knowledge based engineering: Be-

tween ai and cad. review of a language based technol-

ogy to support engineering design. Advanced Engi-

neering Informatics, 26(2):159–179.

La Rocca, G. and van Tooren, M. J. (2007). A knowledge

based engineering approach to support automatic gen-

eration of fe models in aircraft design. In 45th AIAA

Aerospace Sciences Meeting and Exhibit, volume 130.

Perez, R. E., Liu, H. H. T., and Behdinan, K. (2004). Evalu-

ation of multidisciplinary optimization approaches for

aircraft conceptual design. In Proceedings of the 10th

AIAA/ISSMO multidisciplinary analysis and optimiza-

tion conference, AIAA paper, volume 4537.

Scholl, W., K

onig, C., Meyer, B., and Heisig, P. (2004).

The future of knowledge management: an interna-

tional delphi study. Journal of knowledge manage-

ment, 8(2):19–35.

Sobieszczanski-Sobieski, J. and Haftka, R. (1997). Mul-

tidisciplinary aerospace design optimization: Survey

of recent developments. Structural Optimization,

14(1):1–23. cited By (since 1996)255.

Tedford, N. P. and Martins, J. R. R. A. (2010). Benchmark-

ing multidisciplinary design optimization algorithms.

Optimization and Engineering, 11(1):159–183.

Wallace, K., Ahmed, S., and Bracewell, R. (2005). En-

gineering knowledge management. In Clarkson, J.P.

Eckert, C.: Design process improvement - a review of

current practice, chapter 13, pages 326–343. Springer.

Welle, B., Fischer, M., Haymaker, J., and Bazjanac, V.

(2012). Cad-centric attribution methodology for mul-

tidisciplinary optimization (cammo): Enabling de-

signers to efﬁciently formulate and evaluate large de-

sign spaces. Technical report, CIFE Technical Report

- TR195, Stanford University, Stanford, CA.

DevelopmentandImplementationofaMethodologicalApproachtoSupportMDObyMeansofKnowledgebased

Technologies