Data Science (DS) bridges computer science, statistics and domain knowledge to uncover the potential concealed in Data (Wickham, 2016). It is a growing field, and it is having an impact on several domains. One of these is Engineering Design (ED), which can be defined as the process of solving technical problems within requirements and constraints to create new technical products (Pahl, 2013).
DS and ED are two fields of practice that are becoming increasingly intertwined. On the one hand, firms take advantages in having a more structured and reproducible design process thanks to data, therefore pressuring designers into streamlining the process through the reuse of past knowledge and through a more context-aware approach that takes into account customer needs and product life cycles. On the other hand, design tools are producing increasing amounts of data that can be exploited to support all the phases of ED.
DS is changing the way organizations are designing their products and is re-shaping the approaches adopted to handle data during the engineering design process (Wang, 2015). Despite that, general purpose DS need redesign in order to be properly applied in ED. Using the data-driven scientific methods to test new artefacts in medicine, engineering and other safety-critical systems is common practice yet using this in the design of consumer products is a relatively recent phenomenon.
In contrast with an increasing focus of research on DS tools and a large practical adoption of these tools in ED, there is a lack of attention on the intersection between these two fields in the academic literature. In particular, no scientific framework exists to guide researchers interested in designing DS tools specifically for ED. As a consequence, there is no theoretical foundation that allows for objective reasoning around the application of DS in the ED process.
This special issue aims to attract scientific contributions in these bridging fields, to face a number of related themes and challenges for future research. The objective is to use both publicly available (open) data and private datasets to obtain a clearer view of which information sources contain the most untapped value and which data science tool can be used to uncover it.
The main contributions are expected to highlight approaches in any data science phase (retrieval, manipulation, modelling, visualization, and communication) and product design phases (problem definition, conceptual design, preliminary design, detailed design). We welcome both theoretical and practical research with particular attention to industrial data set and case studies, fostering the interfaces between ED research and industrial practice.
This Special Issue is launched as a continuation of a state of the art paper on DS for ED (Chiarello et al, 2021, accepted for publication in March 2021in Computers in Industry, to be visible as early as April 2021), where the interested contributors can find an in-depth analysis of the gaps and bridges that exists between ED and DS. Prospective authors interested in the Special Issue are strongly advised to read this paper, which explains in more detail the boundaries of the area of research to be covered by papers to be published in the SI.
Furthermore, we published a database containing a classification of the reviewed papers to help contributors to explore and be inspired by the existing literature at the bridge between ED and DS (see figure 1 for a map of the relevant keywords).
Figure 1: Graph of ED and DS concepts and their Relationships in the Reviewed Articles
ED Related Themes
1. Develop data-driven tools for problem definition: propose systems to help in the exploratory phase of design, helping to find the right questions to answer and the right problems to solve.
2. Integrate top-down and bottom-up approaches in AI systems: to develop explicit systems using formal declarative information models that follows an imperative procedure (top-down) in standard AI systems that do not need explicit data as input in their implementations.
3. Move towards genetic algorithms for design optimization: explore other solutions for optimal design to advance the state of the art in ED and to incorporate more effective solutions coming from other fields of research.
4. Design novel ways of using CAD as a data source: use DS tools in order to identify latent features (e.g. temporal features, behavioral features) hidden in CAD data or even how to capture the design intent behind them.
5. Improve communication in design teams thanks to NLP: analyze the vast amount of text generated for communication purposes during the design process to study and improve communication in the design team, also considering the increasing amount of digital data generated for communications during COVID-19 pandemic.
6. Use of AI to assist the inventive process: study systems to help designers in the generation phase and study the social, legal and ethical issues linked to this.
7. Implement DS method in Design for X: improve Design for X using data, further developing standard knowledge-based inference methods and “What-lf’ analyses.
8. Move beyond student-generated data: strive to identify other data sources to study the application, teaching and learning of ED principles and tools. We acknowledge that this process requires trust-building and a deep connection to industrial settings that are traditionally hesitant to share design data.
9. Speeding up the prototyping process: focus on taking fast and objective decisions using prEtotypes and also rethinking the prototyping process, thanks to the opportunities offered by new technologies such as augmented reality.
10. Exploit Smart Manufacturing Data to Design Sustainable Products: study how to use data coming from sensor-enabled processes in order to design more sustainable products and obtain further data from the entire product lifecycle to feed the design process with relevant measures able to orient the design actions.
DS Related Themes
11. Map problems of ED in DS: support the problem-mapping phase, using novel data sources such as patents, papers and social media.
12. Mix Expert Systems and Machine Learning Approaches: develop hybrid systems, exploiting both the knowledge of experts and the information coming from data sources.
13. Redesign ad-hoc DS methods: redesign DS methods to work effectively in the ED domain. Scholars in the field of ED can join DS experts in order to redesign state of the art data-driven systems to properly perform in the specific context of ED and its subfields.
14. Structure design related data: focus on the development of data structures and ontologies to define fuzzy concepts that are crucial in ED phases such as users, design objectives, design problems, requirements or product failures.
15. Search for meaningful feature representation: find meaningful feature representations for different ED tasks, in order to foster the accuracy of machine learning systems.
16. Define more effective visualization tools for ED: identify which are the most effective ways of synthesizing data in the context of ED to enable better communication of data-driven analysis.
17. Scale experiments using advanced hardware: test and describe experimental set-up using state-of-the-art hardware systems (i.e. GPU, TPU) in order to make it easier for other scholars to re-use these experimental set-ups and to develop reusable data-intensive models for design.
18. Develop ad-hoc packages for ED: scholars and practitioners may develop packages especially designed for the ED context, as well as examples and datasets associated to particular methods. This may lead to the development of contextually indexed generic methods.
19. Combine textual and visual data: increase the accuracy of AI systems in ED, using the text and images together to open to the uncover of unexplored latent knowledge produced during the ED process (e.g. mixing CAD images and project reports over time).
20. Automatic data labelling: develop curated labelled data that may be extracted from multiple sources and test approaches to automatically label data, using small samples of already labelled databases or using expert systems in order to accomplish this task. For both these solutions, a greater interaction between ED researchers and practitioners has to be reached.
Dr. Filippo Chiarello, Italy, University of Pisa, [email protected]
Pr. Gualtiero Fantoni, Italy, University of Pisa, [email protected]
Dr. Paola Belingheri, Italy, University of Pisa, [email protected]
NOTE: before submitting a full paper to the journal portal, please send a short abstract (200 words) via email to [email protected] and wait until the guest editors indicate that your paper is in the scope of the SI.
Submission period: 1st of May 2021 to 30th of November 2021
Review process: On a rolling basis from May 2021 to March 2022
Papers will be published as they are accepted
Target for SI full publication and closing editorial: April 2022
Wickham, H., & Grolemund, G. (2016). “R for data science: import, tidy, transform, visualize, and model data.” O’Reilly Media, Inc.
Pahl, G., & Beitz, W. (2013). “Engineering design: a systematic approach.” Springer Science & Business Media.
Wang, L., & Alexander, C. A. (2015). “Big data in design and manufacturing engineering.” American Journal of Engineering and Applied Sciences, 8(2), 223.
Cross, N., & Roy, R. (1989). “Engineering design methods.” New York: Wiley.
Chiarello, F., Belingheri, P., Fantoni, G. (2021). “Data Science for Engineering Design: State of the Art and Future Directions.”, Computers in Industry