Summary

Avoiding catastrophic climate change requires dramatically decreasing greenhouse gas emissions and removing already-emitted CO2 from the atmosphere paired with permanent CO2 storage. Through carbon mineralization, CO2 can be stored as carbonates which are environmentally benign and stable, and thus make mineral carbonation a permanent and leakage free CO2 disposal method. There are many industrial wastes rich in calcium and magnesium, which are usable as feedstocks for mineral carbonation waste cement (1 Gt/yr), coal fly ash (600 Mt/yr) and steelmaking slag (400 Mt/yr). Due to the different raw material qualities (composition and particle size) of the mineral waste streams, there is a knowledge gap in terms of efficient process operation while maximizing CO2 sequestration while recovering valuable, high-purity target products (CaCO3, MgCO3). This project addresses this issue by developing an autonomous, self-learning process chain for the CO2 carbonation of mine waste considering the four step of: mineral extraction, filtration, selective precipitation, and centrifugal classification. The project is a close collaboration between the Institute of Mechanical Process Engineering (IMVM) of the Karlsruhe Institute of Technology (Dr. Marco Gleiß), the Institute of Process Engineering (IVT) of the OVGU Magdeburg (Prof. Dr. Kai Sundmacher) and Chair of Mechatronics in Mechanical and Automotive Engineering at TU Kaiserslautern (Prof. Dr. Naim Bajcinca). The autonomous process chain will be able to recognize the mentioned dynamic variations and disturbances affecting the process and accordingly steer it towards a state of maximum possible productivity, while ensuring desired purities for the carbonate products. In cases of unforeseen scenarios leading to non-optimal or undesired behavior, the self-learning feature of the controller should nonetheless enable the process, based on the observable state variables, to behave autonomously. During the first project period we investigate the filtration (step 2) and selective precipitation (step 3) in relation to the process dynamics and develop dynamic models which are integrated within the Self-Learning Robust Autonomous Controller (SLARC) for vacuum belt filtration and selective precipitation.

The second funding period covers the extension of the process chain to include mineral extraction (step 1) and centrifugal classification (step 4). We address here the self-learning autonomous systems for these two process steps as well as for the entire process chain. Furthermore, we close material cycles by integrating recirculation flows that can significantly change the process dynamics of the overall process.