Abstract
Engineering is undergoing a fundamental shift as digital and data technologies become central to the design, maintenance, and operation of physical systems. Data-Centric Engineering is a new open-access journal from Cambridge University Press that aims to bridge the gap between fundamental data science and traditional engineering disciplines. By formalizing the "new nexus" where statistical science and computing transform physical practice, the journal seeks to foster the development of systems and products that are more reliable, resilient, efficient, and safe. For more details, see cambridge.org/dce.
Introduction: The New Nexus
Engineering has always been a data-driven discipline. From Lord Kelvin’s 19th-century insistence on precise measurement to the rigorous testing protocols of the aerospace industry, data has served as the bedrock of physical design. However, the modern "confluence of digital and data technologies" has accelerated at a pace that has outstripped existing academic and professional frameworks.
Mark Girolami, in his inaugural editorial for Data-Centric Engineering, argues that we are witnessing the birth of a new discipline. This field does not merely use data as a tool but places it at the center of the engineering lifecycle. The goal of the journal is to provide a dedicated venue for high-quality, peer-reviewed research that addresses this transformation, making it accessible to both industry practitioners and academic researchers.
The Challenges of Modern Engineering
Girolami identifies three core challenges that necessitate a dedicated focus on Data-Centric Engineering (DCE):
- Lack of Definition: While "Data Science" is a ubiquitous term, its specific application and rigorous integration within engineering contexts remain broad and often ill-defined.
- Educational Lag: Current engineering training—ranging from vocational apprenticeships to doctoral programs—requires radical reconsideration. There is an urgent need to integrate data-intensive methods into the core curriculum of engineering education.
- Market Disruption: The global economy is shifting from a product-based manufacturing model to a service-based model. This transition is mediated by continuous data generation, requiring new ways to monitor, maintain, and secure infrastructure and products.
A Methodological Framework for DCE
Data-Centric Engineering is defined as the intersection of two major domains:
- Data Sciences: Encompassing Statistical Science, Applied Mathematics, and Computing Science.
- Engineering Sciences: Including Aeronautical, Civil, Mechanical, and Materials Engineering, among others.
The journal itself introduces several structural innovations to improve scientific communication and ensure the practical utility of its published research. Notably, every research article is required to include an Impact Statement—a plain-language summary explaining the significance of the work to a non-specialist audience. Furthermore, the journal has adopted the Transparency and Openness Promotion (TOP) policy, using badges to incentivize authors to share open data and code, thereby ensuring reproducibility.
Key Insights: Beyond the Hype
A central argument of the editorial is that DCE is a substantive development rather than a fleeting trend on a hype cycle. Girolami points to a three-year foundational research program at the Alan Turing Institute as evidence of the field's growing maturity.
The paper also draws historical parallels, connecting modern AI-driven engineering to innovators like William Gosset. Gosset, known for the Student’s t-distribution, developed data-driven process controls while working at the Guinness brewery, demonstrating that the marriage of statistics and engineering has deep, successful roots.
One of the most significant shifts identified is the move from "build and sell" to "monitor and maintain." In sectors like civil infrastructure and aerospace, data allows for real-time health monitoring of structures and engines, transforming how we ensure safety and longevity.
Early Impact and Results
The launch of the journal has already highlighted the breadth and practical utility of the field through its first cohort of papers:
- Yu et al. (2020): Utilized deep kernel learning to model diesel engine emissions, showcasing the power of advanced machine learning in traditional mechanical engineering.
- Webb et al. (2020): Developed predictive models for evaluating the wear of conveyor belts in global supply chains, a critical application for the mining and logistics industries.
- Dodds et al. (2020): Published a manifesto on the necessity of sharing engineering data to drive social and environmental innovation.
Limitations and Future Directions
Despite the promise of DCE, significant hurdles remain. The "education gap" is a primary concern, as the profession must adapt to handle data-intensive workflows. Additionally, while the journal promotes open data, the editorial acknowledges that proprietary or security concerns often make full transparency difficult in engineering contexts.
Looking forward, Data-Centric Engineering plans to expand its focus on the ethical, legal, and security implications of data-centric systems. The journal will also provide "Tutorial Reviews" to assist students and researchers in transitioning into this emerging, interdisciplinary field.
Summary updated