Researchers Map Fragmented Engineering Datasets to Speed Design Research

Across the product lifecycle, engineering teams generate vast quantities of data—yet much of it remains scattered, proprietary, or impossible to discover. A new research effort tackles this fragmentation by proposing a systematic way to catalog and navigate engineering datasets.

The Data Landscape Problem#

Engineering design and systems engineering have long suffered from a reproducibility crisis unlike other AI-heavy fields. While computer vision and natural language processing boast mature benchmark ecosystems that enable broad collaboration, engineering research typically relies on small, closed, or one-off datasets cobbled together for individual projects. This "digital thread" of data across the system lifecycle could drive genuine innovation, but its inaccessible nature hampers method validation and slows progress.

Framework and Discovery Tool#

The proposed solution is a multi-dimensional Map of Datasets in EDSE (Engineering Design and Systems Engineering)—a taxonomy that classifies datasets by domain, lifecycle stage, data type, and format to enable faceted search and discovery. Researchers have built a working prototype with an interactive discovery interface powered by a knowledge graph data model. This approach captures semantic relationships not just between datasets themselves, but also between datasets, tools, and publications, creating a richer context for researchers.

Identifying Data Gaps and Clusters#

An analysis of the current landscape revealed telling patterns: certain areas are data deserts, including early-stage design and system architecture phases where research is particularly starved for reference material. Conversely, data oases exist in predictive maintenance and autonomous systems, where datasets are relatively abundant. These insights point to where the community should prioritize new data collection efforts.

Challenges and Next Steps#

The authors acknowledge that curation and long-term sustainability remain thorny challenges. Maintaining a dynamic, community-driven resource of this scale requires clear governance, incentive structures, and technical infrastructure. The paper lays out mitigation strategies and establishes a foundation for what could become a shared research asset to accelerate data-centric engineering innovation.