AI through deep-learning on tomography data (DeepTomo)
The project aims to develop methods for AI through deep learning on tomographic data. This will improve and make data analysis more effective, concerning parameters such as microstructure, porosity, and surface area of ash particles, as well as distribution of phosphates and other compounds in the particles.
The project aims to develop methods to use AI through deep learning on tomography data from synchrotron experiments. The technology provides a 3D image of the sample and with the help of advanced image processing, properties of the sample can be obtained. It is for this step that AI through deep machine learning can gain great importance through improved and more efficient analyzes of the data. Our research aims to evaluate parameters such as microstructure, porosity and surface area, as well as how phosphorus compounds are placed in the ash. The project finances a post-doc and equipment.
Large amounts of phosphorus and other important nutrients are found in many of society's residual streams. By using the ash after combustion, it may be possible to recycle these nutrients. In order to be able to use the ash for nutrient recycling in an efficient and sustainable way, it is important to gain increased knowledge about how the chemical and physical properties of the ash affect the availability of nutrients when the ash is used as a fertilizer.
We currently use synchrotron-based X-ray tomography to study the porosity and pore structure of materials in 3D, as well as how nutrients are distributed in different ashes. An automated data analysis can greatly shorten the time for data analysis and the efficiency of data analysis is an important step in utilizing the capacity of the synchrotron, as each visit to synchrotron plants generates large amounts of data and manual analysis is very time-consuming.