Open Access

For mix design of Alkali-Activated Concrete (AAC) accurate compressive strength predictions are required with limited availability of experimental data. Traditional machine learning models base their reliability on the amount of data used to build them. Therefore, the generalization across diverse data of AAC mixtures utilizing different precursors and activators is an issue for these models. This study introduces a meta-learning approach, leveraging Model-Agnostic Meta-Learning (MAML) and Reptile, to enable the rapid development of a predictive model for new AAC mix design. Using initially only a limited amount of data and training the model on a variety of tasks defined by AAC mix properties and curing conditions, MAML learns an optimal initialization, facilitating few-shot learning. This allows efficient fine-tuning with available data, offering enhanced adaptability and generalization across new AAC formulations. Our results demonstrate the model’s potential for real-time compressive strength predictions, optimizing both research and industry applications by reducing experimental workload and costs.

The research conducted at the Federal Institute for Material Research and Testing (BAM) focuses on key challenges of the energy transition, spanning hydrogen technologies, electrical energy storage, and renewable energy systems. In the field of energy storage, our primary areas of interest include the safety of electrical energy storage systems, sustainable energy materials, and advanced battery diagnostics. One of our central objectives is to deepen our understanding of the processes contributing to lithium-ion cell degradation, an essential step toward improving next-generation systems and meeting the rapidly growing demand for lithium-ion battery technology. The complexity of these systems, which comprise organic and inorganic compounds in multiple aggregation states, presents significant analytical challenges. To address these challenges, we are developing novel analytical methods to further expand our insight into battery degradation mechanisms. Using GD-MS for depth-resolved lithium isotope analysis, we have recently established a correlation between lithium isotope fractionation and the growth of electrode–electrolyte interphases at electrode surfaces. In addition, we are developing GD-OES and LIBS methods for depth-resolved and lateral fluorine analysis, respectively, of lithium-ion battery electrodes to monitor electrolyte and additive degradation. These approaches might also provide valuable analytical tools for assessing the homogeneity of fluorinated active materials.

This dataset accompanies the study on sequential learning–based optimisation of bio-ash–cement binder formulations under seasonally varying material availability. It provides a fully synthetic but chemically inspired benchmark design space for evaluating data-driven optimisation strategies in cementitious materials research. The dataset comprises 5,006 unique binder formulations, each defined by the mass fractions of cement and five bio-based ash components (A1–A5). Ash components represent generic bio-ash types derived from agricultural residues (e.g. rice husk ash, cassava peel ash), and their internal proportions are systematically varied under mass-balance constraints. Cement content ranges from 0 to 100 wt% in discrete steps. To reflect dynamic supply conditions, the dataset includes season-specific ash usage metrics for four seasons (S1–S4), expressing the fraction of available ash resources consumed by each formulation. A synthetic compressive strength value is assigned to every formulation using a nonlinear scoring function based on chemically inspired descriptors, with added noise to generate a structured yet non-trivial optimisation landscape. These strength values do not represent calibrated physical predictions and are intended solely as a hidden objective function for benchmarking sequential learning algorithms. The dataset is designed for in silico benchmarking, reproducibility studies, and methodological comparisons of optimisation and active learning strategies. It enables systematic evaluation of algorithmic performance without the need for physical experiments.

As part of the EU Reincarnate project, BAM has developed an AI-based tool that signifi-cantly accelerates the development of materials in an iterative process of laboratory work, experimental validation, and data-driven optimization. This approach has already been suc-cessfully tested and validated. In Reincarnate SLAMD is currently being applied in demon-stration projects together with two industrial partners – a recycling company specializing in construction and demolition waste and a leading supplier of cement-based building materi-als. The recycling company aims to process recycled concrete fines and glass waste into composite cements with improved performance characteristics. The building materials' manufacturer, on the other hand, aims to reuse processed concrete waste primarily as aggre-gate and to develop suitable materials for five different exposure classes. The integration of SLAMD is envisioned to enable the rapid identification of optimal material compositions that consider the performance and ecological and economic aspects. In addition, we currently further develop in the Circular B-I/O project, a joint effort with African partners to develop alternative, more biobased building materials. The goal is to integrate agricultural residues – such as corn cobs, rice hulls, and sugarcane bagasse – into large-scale, robust supply chains for the construction industry. AI-driven decision-making will be used to create a sustainable framework to transform these seasonally variable raw materials into reliable, high-performance concrete components. Building stable ecosystems and supply chains is crucial to ensure the continuous availability of sustainable materials for a resilient and circular construction industry.

The use of supplementary cementitious materials (SCM) is an important part of the roadmap for reducing CO2 emissions and extending the service life of reinforced concrete structures. To accelerate the adoption of SCMs, the RILEM Technical Committee 298-EBD evaluates scaled-down cement paste test methods to assess the effect of SCM on resistance to chloride and sulfate ingress and reactivity, which are critical to concrete durability. This review focuses on methods for measuring chloride diffusivity and is divided into four sections: diffusivity models and parameters, diffusion test methods (including NMR and chloride measurements), migration test methods and implications for future research. Key insights highlight the complexities of multi-species ionic and molecular diffusion/migration, including various binding interactions, and compares the different measurement methodologies. The review also addresses the test scale and aggregate effects, noting the pros and cons of testing at the paste, mortar, and concrete scales. The review underscores the need for further investigation into testing protocols and the influence of SCM on chloride diffusion, emphasizing that comprehensive testing across different scales provides complementary information for assessing durability performance.