ACORN - AI-powered concrete recipe generator

Purpose and goal:

The project is aimed at developing advanced, sustainable concrete recipes using AI. By integrating data from the research literature, the project aims to use generative AI to generate optimized concrete recipes. The project will use language modeling (LLM) to consolidate datasets and AI-powered anomaly detection to ensure data quality.

The project brings about a paradigm shift for concrete development, where concrete recipes can be optimized for specific applications and produced with a reduced climate footprint. Expected results and effects:

This project aims to advance from TRL 3 to 6 and focuses on LLM data collection, AI anomaly detection, and predictive AI modeling of concrete recipes. The project goal is to build a system infrastructure for AI-based collection of concrete data, upon which predictive models are developed for simulating mechanical properties and climate impact based on constituent materials and process configuration.

The platform prioritizes modularity and openness, allowing users to integrate proprietary data and optimize simulation precision for prioritized design space in future versions. Approach and implementation:

The project begins with a design phase where user needs are defined and key data sources are identified. After the design phase, a prototype of the data collection system and the new database is developed. Finally, predictive models are trained and evaluated, with the best algorithms implemented and made available via the Ecometrix web application. The project concludes with the development of a Go-To-Market strategy.