Structural materials and meta-data for thermochemical electrolysis (METALYSIS)

Royce’s Materials Challenge Accelerator Programme (MCAP)  awarded £43,145.17 to this collaboration from the University of Aberdeen. The project developed database and informatics (data classification, machine learning algorithm, relationship between materials and information) for high temperature thermochemical electrolysis. The informatics could be used as part of future research and development in relation to electrolysers material selection, alongside the modular construction of high temperature thermochemical electrolyser. The research output also helped us forming a consortium and developing a proposal focusing on UK’s overall Net Zero Strategy (in an area of thermochemical electrolysis). 

One of the important challenges in materials science is to develop coating materials for thermochemical containment vessels and pipes that encounters the highly corrosive and harsh environment produced by the molten salt at high temperature. The aim of this research was to summarise structural and coating materials that can withstand thermochemical cycle corrosive environment.

Through data analytics approach (machine learning or ML), this research presents findings published in the scientific literature related to high temperature aggressive corrosion of materials, specifically geared towards nuclear thermochemical cycles leading to hydrogen production. Data related to materials, composition, synthesis were gathered. Corrosion environment data such as environment and test time were analysed.  

In this research, readily accessible datasets on substrate corrosion caused by thermochemical processes for a variety of substrates were gathered. A machine-learning approach was used for estimating the corrosion rate. The random forest approach was chosen as the best model because it considers both substrates and characteristics that pertain to substrates as input characteristics.

The rate of corrosion was accurately predicted using the machine learning-based corrosion rate estimation model, with good results. The corrosion rate prediction model’s capacity to generalise was strengthened by this research, which also demonstrated that machine learning is a possible way to assess corrosion resistance. The results of this research demonstrated the value of machine learning’s sophisticated regression and data mining capabilities for the analysis of corrosion data associated with thermochemical cycle electrolysis and other high temperature processes. 

So far in this project no Royce capabilities were utilised, however their expertise and facilities may be needed in future.  

We can say that this is First-Of-A-Kind work where we developed database and informatics (data classification, machine learning algorithm, relationship between materials and information) for high temperature thermochemical electrolysis. The informatics could be used as part of future research and development in relation to electrolysers material selection, alongside the modular construction of high temperature thermochemical electrolyser. The research output also helped us forming a consortium and developing a proposal focusing on UK’s overall Net Zero Strategy (in an area of thermochemical electrolysis).  

 Prof Nadimul Faisal, Professor of Surface Engineering & Micromechanics 

Further Information

Total Funding amount received: £43,145.17