Journal of

Mechanical
Design Guest Editorial

Special Issue: Artificial Intelligence and Engineering Design

Downloaded from http://asmedigitalcollection.asme.org/mechanicaldesign/article-pdf/144/2/020301/6820421/md_144_2_020301.pdf by guest on 14 December 2023

Artificial intelligence (AI) has had a strong presence in engineer- and knowledge representation, support early design ideation, and
ing design for decades, and while theory, methods, and tools for even to discover design solutions to previously unsolved engineer-
engineering design have advanced significantly during this time, ing problems. In addition, the capabilities of modern AI algorithms
many grand challenges remain. Modern advancements in AI, can enhance the efficacy of later-stage system design activities, such
including new strategies for capturing, storing, and analyzing as those involving high-dimension and strongly interrelated detailed
data, have the potential to revolutionize engineering design pro- design decisions.
cesses in a variety of ways. The purpose of this special issue is to Tasks at early design stages, such as concept generation, have the
consolidate recent research activities that utilize existing or new potential to effect transformational change upon engineering system
AI methods to advance engineering design knowledge and capabilities. Yet in practice, these early-stage tasks rely heavily
capabilities. upon designer experience and intuition. Human cognitive limits
During the conception of this special issue, we identified three constrain such design processes, creating an opportunity for AI
core interfaces between the research domains of engineering models to accelerate the ability of engineering to better meet the
design and AI: (1) leveraging AI methods directly in engineering needs of humanity. Limitations of conventional methods and
design methods, (2) creating new AI capabilities that are inspired tools constrain the fidelity and scope of the design spaces that are
by unique challenges that arise in engineering design, and (3) creat- tractable. AI methods are helping to solve problems where human
ing and analyzing design methods that are tailored for the design of intuition or human processing capabilities are insufficient, making
engineering systems where the systems themselves utilize AI, such it possible to expand the richness of design spaces that can be nav-
as autonomous vehicles. The diverse body of research articles that igated successfully through AI support.
now comprise this special issue gravitate toward the first of these Some engineering functions may not even have known solutions,
themes: advancing engineering design capability through the use even via biological analogies; AI-based design strategies have the
of AI. While these articles are an exciting contribution to the potential to aid in the discovery of design solutions to previously
design research literature, significant opportunities exist for more unrealized functions through nonobvious design configurations.
fully exploring the remaining two interfaces, ideally through more Used as an identification tool, AI is enabling more efficient and
unified interdisciplinary efforts. During the process of synthesizing higher-performance design solutions during design revision at the
this editorial, we recognized a fourth interface between engineering manufacturing and assembly level by screening and identifying
design and AI: specifically, investigating how AI could be used as part consolidation opportunities and multifunctional design
an increasingly powerful tool for conducting engineering design potential.
research, such as AI tools that are used directly in research activities More comprehensive investigation of the role of AI in engineer-
(e.g., experiment planning or gathering information from human ing design methods and research appears to be reshaping how we
designers) and that are not necessarily part of the designed system think about design. AI methods that can transformatively restructure
or the design method. design spaces for various goals exist, such as more efficient naviga-
Two sets of clear themes have emerged from this special issue. tion of design alternatives or insightful interrogation to extract
The first set is expressed from the perspective of engineering design knowledge. Design spaces can be adaptable throughout
design research and design processes. The second set is organized design process stages. Less rigid formulations can produce more
in terms of AI methods. We have organized these themes in this fluid design paradigms compared to established frameworks, such
way in part to facilitate clearer communication across AI and Engi- as design optimization. AI models can also serve as the design solu-
neering Design research communities and to enable the productive tion method without needing to be linked to iterative design optimi-
collaboration that is needed to address open questions. Later, we zation. Moreover, inverse design approaches fit well with AI tools,
discuss the relationship between AI and engineering design and further bending our conception of design strategies. Publications at
then articulate the two sets of themes found in this special issue. the interface of AI and engineering design have often been led by
Finally, a vision is presented for advancing interdisciplinary design research experts rather than AI experts, resulting in potential
research in this area, including an initial outline of promising bias toward particular research questions and classes of AI methods.
research topics. Approaching this interface from additional complementary direc-
tions could help us build a more complete understanding of how
AI methods can further benefit society through advancements in
engineering design.
Artificial Intelligence and Engineering Design Articles in this issue link to a diverse set of approaches
This special issue consolidates exciting new outcomes from used in AI and utilize various neural network architectures to
recent research that focused on the application of modern AI develop surrogate models. These surrogate models typically learn
methods to the creation and analysis of methods that advance engi- abstract representations to understand the underlying processes,
neering design capabilities. Design researchers have long recog- which can be then explained to gain further insight. Several
nized that AI tools can be used in many ways to advance how approaches also assist the designer through human augmentation
engineers perform design. For example, designers can now leverage and increased automation, while also expanding design spaces
AI to support more automatic and intelligent knowledge extraction through transfer learning. These approaches are applied to many

Journal of Mechanical Design Copyright © 2022 by ASME FEBRUARY 2022, Vol. 144 / 020301-1
aspects of the design process including systems design, conceptual computational iterations, or other resources required to obtain a
design, generative design, and several detailed design applications good design solution. Two key AI-based strategies employed to
in areas including supply chains, medical devices, and aeronautics. accelerate iterative design processes include (1) transfer and repre-
Readers may be keenly aware that the guest editors have not sentation learning (Behzadi et al., Herzog et al., and Whalen et al.)
attempted to define AI. While some excellent definitions exist, con- and (2) invertible neural networks for optimization (Ghosh et al. and
flicts in (sometimes evolving) AI definitions held by such a diverse Oddiraju et al.). While many current AI-related research efforts
set of potential authors are likely. We opted to consider submissions outside of engineering design research focus on training models
that utilized AI as defined by the authors, as long as the definition is from scratch for maximum accuracy with large datasets, many arti-
justifiable, and if the submission makes a notable contribution to the cles in this special issue (and other design research publications)
body of engineering design knowledge. As a result of this inclusive gravitate toward pretrained networks, such as used in transfer learn-
approach, some readers may not consider some methods appearing ing, to enhance accuracy with limited datasets (Ferrero et al. and
in this issue to be AI. The definition of AI will not be settled here; Yuan et al.). This observation links to the Transfer Learning AI
rather, we hope that one of the primary contributions of this special theme discussed later. In addition, AI methods that better support
issue is a helpful step toward more successful integration of AI and generation of design solutions that are substantially distinct from
engineering design research. previous designs, such as transfer learning methods that are success-

Downloaded from http://asmedigitalcollection.asme.org/mechanicaldesign/article-pdf/144/2/020301/6820421/md_144_2_020301.pdf by guest on 14 December 2023

This special issue incorporates a variety of topics that represent ful in design domains that are more distant from the training data,
the state of the art in how AI is changing how we do engineering may be particularly valuable.
design today. Articles solicited represent intellectual contributions Design Theme 3: AI-Based Direct Estimation and Tuning. In addi-
to engineering design research at the intersection of unmet engi- tion to accelerating iterative design processes, some methods proposed
neering challenges, advances in AI, and design-relevant contexts. AI-based tools that reduce or eliminate the need for iterative design.
Articles published in this issue fall largely within the general area Articles by Burge et al., Dachowicz et al., and Li et al. used neural net-
of AI being utilized as an enabling tool for advancing engineering works to design solutions for given initial conditions. The article by
design capabilities. In particular, many articles focus on enhancing Raina et al. used learning by demonstration to provide design solu-
early-stage design capabilities. Previous special issues exist for tions, and the article by Caputo et al. utilized deep reinforcement learn-
related topics, including Machine Learning for Engineering ing to achieve adaptive responses to changing environmental
Design and Data-Driven Design. This special issue has provided conditions without the need for any new design iterations.
an opportunity to publish design research linking to the broad Design Theme 4: Broader Design Process Support. A few articles
area of AI that may not have fit well with these previous special make broader contributions that cannot be strictly classified along a
issues. For example, one article (Gyory et al.) utilizes an AI specific theme mentioned earlier. The article by Wang et al. makes
agent to perform dynamic design team management based on pre- a theoretical contribution involving the use of Gaussian processes
defined logical rules. in a scalable manner to represent complex spaces. The article by
The guest editors have identified significant themes that have Kim et al. utilizes AI-based tools to perform sentiment analysis on
emerged from this collection of articles, organized first into topics product listings and how they evolve. The article by Gyory et al. pro-
that are presented from an engineering design research perspective poses and tests an AI-based team management system against a
and second into topics that are articulated from an AI method per- human-based management system; an overall increase in productiv-
spective. Please note that the distinctions between these themes can ity and content is realized through the AI system. The article by
be fuzzy; some studies lie at the intersection between two or more Caputo et al. proposes a new approach to conceptualize real
themes. The following two sections introduce these themes and options and flexibility analysis in engineering systems design, build-
point readers to relevant articles in this special issue. This editorial ing upon principles from deep reinforcement learning. This supports
then concludes with remarks on open research questions at the inter- the view that more work is needed to develop new AI and data-driven
face of AI and engineering design. methods to design complex engineered systems so that they can
better deal with uncertainty and risks—which is much needed, espe-
cially given ongoing threats from climate change and global health-
care emergencies.
Engineering Design Research Themes
When viewed through the lens of engineering design research,
four broad themes emerge from this special issue, each of which
is discussed here. AI Method Themes
Design Theme 1: Conceptual Design and Design Synthesis. When viewed from the perspective of identifying specific AI
Several articles involve advances in utilizing AI for more expansive methods that are shown to be useful for advancing engineering
and comprehensive yet practical exploration of complex design design capabilities, five themes materialize. An exposition of
configuration spaces, including topological changes, design synth- these themes is provided below.
esis, and support for enhanced human creativity. Methods are AI Theme 1: Natural Language Processing. Natural language
either automated or involve a partnership between AI and human processing has been used in several papers included in this
designers. More conventional design paradigms induce a different special issue. NLP serves as a human-interpretable way to interact
constraining of design spaces than might be possible with this with many AI-based systems, and so improvements to NLP
class of AI-based design methods. Natural language processing methods can benefit the tools adapted to engineering design prob-
(NLP) is utilized in several studies presented here as a key interface lems (T. Chen et al., Han et al., Kim et al., Lee et al., and Yuan
with the human creative process (T. Chen et al., Han et al., and Lee et al.).
et al.). Two survey papers also provide context and future directions AI Theme 2: Graph (Neural) Networks. Many design representa-
for utilizing the AI-based methods for this and other design needs tions may be represented as graphs (e.g., assemblies (Ferrero et al.)
(Han et al. and Jiang et al.). The article by Song et al. presents a and structures (Whalen et al.)). In future work, it would be benefi-
study of the influence that AI tools have on strategies followed by cial to see more studies that address remaining challenges in apply-
designers during engineering design processes. As noted earlier, ing graph networks to engineering design, such as invariance/
several articles link to two or more themes. For example, Behzadi equivariance properties, explainability, and transferability.
et al., Q. Chen et al., Nobari et al., Quigley et al., and Raina et al. AI Theme 3: Generative Models. Use of generative models con-
all link to both methods for conceptual design and Design Theme tinues as a demonstrated AI-based design strategy (Behzadi et al.,
2: Accelerating Design Processes. Q. Chen et al., Nobari et al., Oddiraju et al., Quigley et al., Yuan
Design Theme 2: Accelerating Design Processes. Multiple con- et al.); one potential benefit is the ability for these models to
tributions in this special issue utilize AI tools to reduce the time, produce solutions without iterative optimization. Within the set of

020301-2 / Vol. 144, FEBRUARY 2022 Transactions of the ASME

articles utilizing generative models, articles by Behzadi et al. and This observation is representative of the second interface between
Nobari et al. address the challenges with small data set sizes, AI and engineering design research, as identified in the second par-
which is an important characteristic of many data-driven design agraph of this editorial, and is an opportunity for productive inter-
strategies. To enhance what can be done using small data sets, the disciplinary discussion.
article by Nobari et al. introduces a label-aware self-augmentation An interesting future direction for this effort would be to solve
training approach, and the article by Behzadi et al. uses knowledge engineering design problems using both the pretraining and from
transfer. scratch strategies for neural networks and then to compare the
AI Theme 4: Transfer Learning. Transfer learning is emerging as abstractions learned using both approaches. It would also be
an important tool for dealing effectively with the small data sets that valuable to evaluate hybrid methods, where certain layers are
are common in engineering design. The article by Whalen et al. allowed to mutate to adapt to the given problem. Alternately,
explores the transferability of a learned graph network on a networks trained from one paradigm can be used as the initiali-
variety of structural design problems. The article by Herzog et al. zation for the other, and optimal abstractions can be learned for
presents a study of the effectiveness of knowledge transfer in the the given problem sets. As observed earlier, many articles in this
context of semi-supervised learning of labels on 3D geometries. issue concentrated on new AI-enabled design methods to better
AI Theme 5: Representation Learning. The article by Raina et al. support early-stage design tasks. Research topics that are

Downloaded from http://asmedigitalcollection.asme.org/mechanicaldesign/article-pdf/144/2/020301/6820421/md_144_2_020301.pdf by guest on 14 December 2023

introduces new ways of representing design policies (e.g., notably absent from a design methods research perspective
sequences of design decisions) when the action space structurally include those that investigate AI-based methods intended for
changes during the design process. Gaussian process (GP) models more comprehensive late-stage design (e.g., AI agents that aid
that learn latent categorical representations are presented in the automation of fully detailed design specifications required for
article by Wang et al., and the scalability issue of GP models is the production).
investigated. In addition to the outputs of AI research benefitting design
research, the converse is also possible. For example, tools from
current engineering design research, such as design automation
methods, could be used to rapidly generate rich and tailored data
Future Opportunities for AI and Engineering Design sets that amplify the impact of AI methods. Furthermore, engineer-
ing design challenges and knowledge from design research could
Research help inform new advancements in AI, presenting a new opportu-
This special issue offers a partial snapshot of the current research nity for interdisciplinary collaboration and motivation for
results in AI and engineering design. The previous sections sum- enhanced communication across research disciplines. Another
marize some aspects that were addressed in the special issue; class of open research questions involves the creation of design
here, we articulate what appears to be missing, and what could methods that are tailored to meet the needs that are unique to
serve as the basis for rich future collaborations in interdisciplinary design systems where AI is part of the system being designed
research between AI and engineering design researchers. We iden- and not necessarily part of the design method. Finally, AI
tify both AI methods that need further study as elements in engi- methods could be used as a tool in engineering design research,
neering design methods, as well as ways AI could be addressed which is distinct from use as part of a design method. For
differently in the context of engineering design (e.g., new AI con- example, many design research studies involve human compo-
tributions motivated by design, design of AI-based systems, and nents; AI could play a significant role as a research tool in inter-
AI methods as tools in engineering design research). facing with human designers, users of designed systems, and
The collection of articles utilizes several well-established model- other stakeholders. It could also transform planning of research
free AI methods for the purpose of accelerating engineering design experiments and data collection.
processes, but leaves open the opportunity to investigate the use of Addressing these questions is distinct from fundamental
model-based learning techniques in these contexts. This may be advances in AI theory and algorithms alone; generation of new
because most of the applications studied here do not have clear knowledge that is relevant to engineering design and that links
associated dynamics. The solutions obtained from these methods to AI requires scientific rigor in both domains and motivates
can be used not only as solutions to engineering design problems deeper collaboration between design research and AI research
but also model-based strategies that leverage assumptions about experts. Completely new AI concepts may take engineering
the form of the system being modeled (e.g., physics-based phe- design capabilities in unexpected directions, and vice versa.
nomena or distributions) could lead to a deeper qualitative design Many important open questions remain, and only a few have
understanding. Model- and dynamics-based learning methods been highlighted here. We encourage readers to apply their own
could be exciting and insightful avenues for engineering design unique perspective and expertise to distill and identify additional
research. scientific questions at the intersection of AI and engineering
Several articles in this issue utilize pretrained networks, which design research, especially those that may benefit from deep inter-
are either adapted to identify/classify design models or to develop disciplinary collaboration. We hope to work collectively toward
a framework for explaining engineering design related processes. more impactful research at the interface of AI and engineering
While the pretraining approach has been discussed in AI under design. Achieving this will require extensive future efforts. Exist-
the names of transfer and curriculum learning, its deeper under- ing work does involve independent forays into adjacent disciplines
standing is particularly important to engineering design due to fre- (e.g., design researchers gaining AI expertise), but, on the whole,
quent scenarios where data are limited or expensive, e.g., during the studies at this interface are still performed in a largely siloed
exploration of new material systems. In such scenarios, it would be manner. Perhaps the fundamental differences in language, perspec-
valuable to be able to learn effective representations, rules, or solu- tives, and goals between the AI and engineering design research
tion search strategies from one set of design problems and apply that communities add to the underlying difficulties. A significant
knowledge to a different (or more complicated) set of problems. investment across disciplines would be required to realize this
While empirical evidence has been reported where such transfer vision of intrinsic interdisciplinarity for research at the interface
has potential, we expect deeper investigation into the theoretical of AI and engineering design.
conditions for certifiable performance of pretrained models, and
more comprehensive ontologies that define knowledge transfer
within the broad context of engineering design. The tendency James T. Allison
toward pretraining approaches in engineering design methods University of Illinois at Urbana-Champaign,
exemplifies the deeper underlying needs that are unique to engineer- Urbana, IL 61801
ing design and motivate additional fundamental AI investigations. e-mail: [email protected]

Journal of Mechanical Design FEBRUARY 2022, Vol. 144 / 020301-3

 Michel-Alexandre Cardin Daniel Selva
Imperial College London, Texas A&M University,
South Kensington, College Station, TX 77845
London SW7 2DB, UK e-mail: [email protected]
e-mail: [email protected]
Conrad Tucker
Chris McComb Carnegie Mellon University,
Carnegie Mellon University, Pittsburgh, PA 15213
Pittsburgh, PA 15213 e-mail: [email protected]
e-mail: [email protected]
Paul Witherell
Max Yi Ren National Institute of Standards and Technology,
Arizona State University, Gaithersburg, MD 20899
Tempe, AZ 85226 e-mail: [email protected]
e-mail: [email protected]

Downloaded from http://asmedigitalcollection.asme.org/mechanicaldesign/article-pdf/144/2/020301/6820421/md_144_2_020301.pdf by guest on 14 December 2023

Yaoyao Fiona Zhao
McGill University,
Montreal, QC, H3A 0C3, Canada
e-mail: [email protected]

020301-4 / Vol. 144, FEBRUARY 2022 Transactions of the ASME