Downstream demonstrator, downstream gate, circular feedstock - SLAMD

This study evaluates the SLAMD (Sequential Learning App for Materials Discovery) platform, which integrates a digital laboratory environment with sequential learning algorithms to optimize mortar formulations containing recycled glass and concrete fines. The objective was to achieve at least 30% cement substitution while maintaining a 28-day compressive strength of ≥30 MPa. SLAMD iteratively proposed and refined formulations across six optimization cycles, guiding experimental testing under real laboratory conditions. In total, 32 mortar recipes were tested, demonstrating that SLAMD can identify technically feasible low-cement formulations and improve predictive accuracy as experimental data accumulates. Despite early experimental variability, model performance stabilized in later cycles, confirming SLAMD’s learning capability and its potential to significantly shorten experimental development time. The results provide scientific evidence for AI-enabled circular construction practices and contribute to Reincarnate’s impacts on data-driven material optimization, CO₂ reduction through waste utilization, urban mining, and the creation of new value chains for alternative binders.