concept of machine learning or digital twin, shape of a man combined with electronic pattern

Helmholtz AI project call showcase: Reactive Transport Experiment Digital Twin (T^6)

First-time implementation of an AI-based evaluation pipeline for microfluidic experiments to assess hydrochemical processes.

How to create a digital twin of microfluidic experiments for energy-related applications? Find out more in today's Helmholtz AI project showcase "ReacTive TransporT experimenTal digiTal Twin (T6)", a collaboration between scientists from Forschungszentrum Jülich (FZJ), the German Research Centre for Geosciences (GFZ) and the Paul Scherrer Institut (PSI).

Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?

I am Jenna Poonoosamy, a scientist in the Reactive Transport team at the Institute of Energy and Climate Research, Nuclear Waste Management and Reactor Safety (IEK-6) at Forschungszentrum Jülich (FZJ), with a background in radiochemistry and reactive transport modelling. My research deals with water-rock interactions that govern the evolution of engineered subsurface systems in energy-related applications. I am particularly interested in the understanding and the development of process-based descriptions of reactive transport phenomena in porous media, using creative and cutting-edge methods, including the development of state of the art microfluidic experiments for studying coupled flow and reactive transport processes. My ambition is to bring microfluidic experiments to the next level by creating an expanded toolbox for experimentalists integrating computer vision and AI/ML methods for ultra-fast forward hydro-geochemical simulations. The research activities within the project “ReacTive TransporT experimenTal digiTal Twin (T6)” funded by Helmholtz AI aim to provide the means for evaluating and controlling microfluidic experiments in real time.

The T6 project is a joint collaborative project with two research partners, Dr Marco De Lucia from the German Research Centre for Geosciences (GFZ) in Potsdam, Germany and Dr Nikolaos Prasianakis from the Paul Scherrer Institut (PSI) in Villigen, Switzerland. Marco De Lucia focuses on modelling water-rock interactions in natural and engineered subsurface systems. He pioneered the use of AI methods as surrogates for computationally expensive geochemical models within coupled reactive transport simulations, further explored in the Helmholtz Incubator Project “Reduced Complexity Models”. Dr Nikolaos Prasianakis and his research group focus among others on multiscale multiphysics reactive transport modelling for a large range of applications, high performance computing and digital twins.

Together, we have complementary skills, expertise and share the scientific vision in the development of digital twins of microfluidic experiments for energy-related applications.

In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?        

Microfluidic lab-on-a-chip systems with real time imaging (including spectroscopic and synchrotron-based analytical techniques) in combination with reactive transport modelling have proven to be decisive tools to generate spatio-temporal insights into processes relevant for various subsurface systems in energy-related applications. The overall goal of T⁶ is to implement a real time pipeline for an AI-assisted evaluation of microfluidic experiments assessing hydrogeochemical processes for the first time. This pipeline serves as the backbone of an automated workflow consisting of several interacting software components, many of which require simple adaptation and others further research for more efficient implementation. Once this workflow is established, it practically constitutes a Digital Twin of the experiments, allowing not only a quick evaluation of the results, but also prognosis and optimised control of the experiments while they are running.

This ambitious project is a ‘high risk, high gain’ project since several technologies have to be further developed and integrated. However, the results of this project are expected to drastically enhance the efficiency of the experimental investigations and will serve as a paradigm for several scientific disciplines and industries who rely on lab-on-a-chip technologies (e.g. pharmaceuticals, chemicals, materials).

The pipeline to be implemented requires (i) the ability to detect and monitor changes based on microscopic images, and (ii) real-time resolution of transport and chemical reactions in the microfluidic devices. AI methods are required to tackle and bypass the high computational loads required for image processing and coupled flow-transport-chemistry simulations. Being able to establish a complete pipeline will immensely benefit the experimentalist, who will then be capable of extracting maximum knowledge from long-running experiments, which is crucial to assess the consequences of subsurface utilisation (e.g., safety of underground nuclear waste repository, CO2 sequestration). Each subtopic of the project tackles aspects which in themselves are beneficial to the scientific community, enabling more efficient and less error-prone results of numerical simulations and data acquisition and evaluation.

How important has the Helmholtz AI funding and platform been to carry out this project?

The Helmholtz AI funding is specifically designed to fund ‘high risk, high gain’ methodology development based on AI. The Reactive Transport team in Jülich was able to attract a Post-doctoral researcher with expertise in computer vision. The multidisciplinary collaboration between the three research partners with complementary skills was crucial to meet our goals and would have otherwise been impossible to achieve if relying upon different funding agencies.

Figure: Concept of the Reactive Transport Digital Twin T6 towards more efficient experiments.