Nanostructure evolution in oxide materials (NanOX-ML)

In the vast majority of cases, refractories are coarse ceramic materials and are indispensable for many things in everyday life. For example, they are required for the production of steel, glass and cement to make industrial high-temperature processes possible in the first place. Refractories are used for the lining of furnaces. In use, temperatures of 1600C and sometimes even higher are reached. The basic components for refractories are almost exclusively thermally stable oxides, typical examples are periclase (MgO) and corundum (Al 2O3). The durability of refractories is essential for environmental and climate protection and for the economic profitability of the processes due to the considerable amount of CO2 emitted during their production (often in the order of 1.5kg CO2 per kg of product). In addition, the refractories play an important role in the quality of the products manufactured. During use, i.e. at high temperatures, diffusion processes, structural changes and different thermal expansion of the components occur in the refractories. This causes mechanical stresses in the structure, which influence their macroscopic behaviour. The project quantifies the changes in the microstructure and their effects as well as the macroscopic mechanical behaviour at temperatures of up to 1600C. To this end, X-Ray Diffraction (XRD) experiments will be carried out at the D2AM beamline of the ESRF (European Synchrotron Radiation Facility) in Grenoble. However, Synchrotron beam time is an expensive resource and, in typical setups, a large amount of beam time is spent on irradiating uninformative parts of the crystal. We will devise novel Machine Learning (ML) algorithms that are sensitive to exploration costs and, hence, minimize the exploration of uninformative parts of the crystal. This will further have a direct impact on resource usage, wasting less time and cost. Real-time optimization of XRD experiments presents new challenges for ML algorithms, partially due to data characteristics: The observed data (i.e., the diffraction patterns of the crystal) is represented in the Fourier space, while the experimental parameters guide the exploration in the real space (i.e., crystal coordinates). The macroscopic material behaviour will be investigated in a specialised laboratory at Montanuniversität Leoben. Here, in addition to the time-dependent deformation at high temperatures under compressive loads, the fracture energy will also be determined. The insight into the development of the nanostructure should provide information on the further development of refractory building materials. The aim is to develop recommendations for reducing the CO2 footprint of refractories. The findings of this project will also impact the run time and cost of thousands of experiments at the European Synchrotron Radiation Facility and other XRD facilities worldwide.