Subject matter

The concrete industry is increasingly in need of more accurate and reliable materials models and intelligent tools to further develop and use concrete in important structures critical to the functioning of society. It also needs more advanced simulation tools for concrete performance validation and uncertainty quantification. Such requirements are well aligned with recent developments in data-driven models based on artificial intelligence (AI). Although theory-driven models can help building a more fundamental understanding and extrapolating in unknown areas, data-driven models can help bring accuracy in areas where "real life" varies significantly from "model or idealized", thus bridging gaps in basic understanding. These also allow closed-loop optimizations by integrating the simulation of manufacturing processes of all application-relevant phases and on all scales. AI techniques do not impose specific models on the data and can excel at capturing highly complex and nonlinear relationships between the input parameters and performance. However, they require robust and comprehensive data sets for training and validation. Hence, a better use of existing data, as well as the availability of more structured and validated information of the materials and components, are essential for the ability to reliably simulate options and make sound decisions. The value of data lies in its use and reuse. However, open data sharing among the concrete research community is still in its infancy. To ensure that it matures into widespread practice, benefiting the scientific community, and by extension the entire society, actions at various levels are required.

The primary objective of this new TC is to gather, analyse and present the state-of-the-art in a report on the use of AI algorithms (machine learning and deep learning) in concrete structures; as well as to look for unified views on the most suitable tools to extract information and/or knowledge from data for different application types, incorporating the developments made in the last 10 years. The work of this committee is based on surveying the pertinent literature and experiences gained by the committee members. The following topics are intended to be included:

• Latest developments on data-science and AI algorithms
  • Machine learning types
  • Deep learning
  • Methods for explainability and interpretability (target: more transparency)
• Application of machine learning and deep learning in concrete structures
  • Concrete engineering properties and mix-design
  • Structural health monitoring
  • Durability and service-life assessment
  • Life-cycle assessment
• Hybrid approaches (combination of theory-driven and data-driven material modelling)
• Recent advances in sensor technologies (target: to collect data)
  • at the material level (constituent materials’ characterization; mixing process; rheological properties)
  • at the structural level (strength evolution, deformations, concrete chemistry, moisture and temperature state, rebar corrosion...)
• Internet of things (IoT) for remote monitoring and communication
• Open data
  • Small data/big data in concrete engineering
  • Data repositories
  • Standardisation in data formats

Terms of reference

• The TC is anticipated to start at the RILEM Spring Convention in 2022 and has intended to take five full years.
• Membership: potential members are located in Europe (The Netherlands, Belgium, Germany, Finland…..), North-America (USA, Canada, …) and Asia (China, Japan….), based on the majority of the research performed over the last decade. The main composition of the group will be academic, but efforts will be made to include pertinent industry (concrete producers, materials suppliers, companies producing sensors,…) and governmental institutions (research centers, standard issuing entities, etc.).
• An objective of the TC is to perform bibliographical research. A document with recommendations on how to foster more data sharing among the concrete research community can also be envisioned, dependent on the progress rate of the TC.
• No experimental work will be done, and nor will new equipment be developed. The report and related recommendations will be based on the information found in the literature and the experience gained by the committee members.
• A closer look at the current literature reveals that publications utilizing AI methods continue to rise in the domain of concrete structures, which indicates that the concrete materials and structures community is interested in these approaches. As such, facilitating the integration of these techniques in the concrete technology and concrete structures field is of utmost importance. By doing so, we can further promote the digitalization and industrialization of the construction industry.

Detailed working programme

The mission of this committee is three-fold:

• Create a STAR report on the use of AI in concrete structures, for which a first draft is anticipated to be ready in 2027.
• Further educate the civil engineering community, both from academia and practice, on the potentialities of data-science tools for prediction of concrete’s engineering properties and structural performance.
• Foster increased (open) data sharing among the concrete research community and propose best practices and recommended formats for data. Based on the STAR, the recommendations could be created in 2028.

The development of the state-of-the-art report will continue during the committee's tenure. The amount of information available in the literature is extensive and is expected to increase significantly over the coming years. Information will be gathered on the use of AI (machine learning and deep learning) to model different concrete properties, whether at the stage of material or component selection or the mix-proportioning stage, during concrete production, or at the structural level during its lifetime. For each set of properties, information on existing instruments (sensors) to collect data will be gathered and described. Different measuring instruments and communication systems will be discussed, along with outlining how they could contribute to generating relevant data. While gathering information for the state-of-the-art report, best practices or recommended procedures for increased (open) data sharing will be discussed. Upon consensus of the committee, these recommendations can be summarized in a separate document. The goal of this document is to provide practical guidelines on sharing data as of today (short-term) while ensuring reproducibility and not becoming excessively complex. Based on this document, the information will be disseminated to the entire concrete research community. Another important outreach task will be (co)organizing the First International RILEM Symposium on Data-Driven Concrete Science, in Europe, near the closure of the committee. A joint session with ACI can be requested on recent developments on data-driven models in concrete materials and structures.