New machine learning research could transform healthcare and energy storage.
Researchers at Sheffield are developing powerful machine learning algorithms that predict the structure of materials as they are created.
This two year Engineering and Physical Sciences Research Council (EPSRC) funded project will use real-time monitoring and control during self-assembly to facilitate the continuous manufacture of large quantities of micron and nano-scale patterned material. Some material can self-assemble, in the same way that salt crystallises or snowflakes form.
This process can currently create small amounts of material, but imperfections prevent the manufacture of large quantities. The small quantities made in this way demonstrate remarkable optical, electronic and structural properties which makes these materials valuable in a range of applications. A more detailed control of this process (at a nano-scale level) could allow manufacturing in larger quantities and lead to improvements in sensors for medical diagnostic equipment, solar cells and other technologies dependent on these materials.
Currently there is no reliable method to manufacture the large quantities needed in industrial applications. Adding an automated control system that makes micron and nano-scale observations of the material (as it forms) will help researchers develop algorithms that can adjust the manufacturing process.
The aim is to continuously produce large quantities of material with well-defined structures, that will be benefit as energy production/storage and healthcare. This could lead to advances in the manufacture of technology such as high-performance batteries and medical sensors.
The project will be led by Dr Stephen Ebbens, from the Department of Chemical and Biological Engineering, with Professor George Panoutsos from the Department of Automatic Control and Systems Engineering (acting as co-investigator) and Dr Alan Dunbar providing expertise in instrument development.
Dr Stephen Ebbens says “Achieving control over a material’s structure at small scales is crucial for the development of new technologies and improved performance in areas such as energy storage, energy generation and healthcare. For example a solar cell would benefit from this type of nanoscale structure building.”
Professor George Panoutsos says "In the manufacture of self-assembled materials, there are not many feasible opportunities to monitor the evolution of underlying material behaviours in appropriate resolution and time-scales. One aspect of the proposed underpinning research is to create accurate data-driven Machine Learning models that are capable of forecasting the complex self-assembling behaviours using limited amount of information."