Development of a method using machine learning models and high-throughput testing to control the corrosion rate of advanced engineering materials.
How can we use machine learning to develop novel corrosion inhibitors? In today’s Helmholtz AI project showcase, researchers from the Helmholtz-Zentrum Hereon (Hereon) and Helmholtz Zentrum München (HMGU) are tackling this research question.
Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?
My name is Christian Feiler and I am a scientist at the Institute of Surface Science at Helmholtz-Zentrum Hereon. I am a chemist by training with a strong background in computational chemistry. Before I became a Postdoc at the Helmholtz-Zentrum in Geesthacht in 2017, I employed quantum chemical calculations to design dynamic molecules that perform machine-type functions. Since then, my research activities focus on the development of machine learning models to predict the effect of small organic molecules, so called dissolution modulators, on the degradation behaviour of lightweight materials in close cooperation with our experimentalists. Together with Sviatlana Lamaka (Hereon) and Shadi Albarqouni (HMGU), I applied for the project “Eyesight to AI: Discovery of efficient corrosion modulators via predictive machine learning models” which is our most recent endeavour to speed up the exploration of the chemical space of available compounds to gain control over the corrosion rate of our advanced engineering materials.
In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?
Novel corrosion inhibitors are urgently required to replace toxic treatments using chromates. This project aims to develop methods for the discovery of dissolution modulators for magnesium (Mg) using machine learning models and high throughput testing. The fundamental concept is to firstly develop a pattern recognition routine that enables high-throughput quantification of the effect of small organic additives on the degradation of a magnesium alloy by automated classification of corrosion imprints. Subsequently, the optical signal will be used as target parameter for different supervised and semi-supervised learning approaches to predict the performance of untested additives. The developed in silico models will be validated by the same technique that was used to generate the training data and employed to preselect promising additives that exhibit strong corrosion inhibiting effect prior to experimental investigation. The main project risk lies in the automated classification of the corrosion imprints as the difference in their optical appearance might not be significant enough. If we can successfully implement this new method, it will allow us to considerably accelerate the acquisition of new data points to improve our models on much smaller timescales as we can run more experiments in parallel. Consequently, this will allow us to screen larger areas of chemical space with higher accuracy. However, as this is a completely new approach it renders the project a high risk, high reward endeavour.
How important has the Helmholtz AI funding and platform been to carry out this project?
The funding by Helmholtz AI allowed us to combine the expertise of two Helmholtz Centers that are active in completely different research areas and to create new synergies between different domains that might have not come together otherwise.
Figure: AI2 project logo