PhD student in machine learning, hybrid models and chemical engineering

About the project
What is a hybrid model in chemistry?

The kinetics of a given set of chemical reactions is a (typically) non-linear dynamical system. Under favourable circumstances exact models based on first principles can be formulated. However, setting up such a model requires a detailed understanding of the underlying physical mechanisms of the process such as the kinetics of a given set of reactions. For complex processes such as biofuel production, such knowledge is often not available and is expensive to produce. The idea of hybrid models is to incorporate some prior knowledge into the model, such as physical conservation laws or partial knowledge of the reaction kinetics, and model the missing pieces using data and machine learning.

The main goal of the project is to develop hybrid models for complex processes where only partial first principle knowledge is available, focusing on the production of renewable hydrocarbon fuels.

Examples include unknown reaction kinetics, complex relationships between process parameters and hydrogen mass transfer rate, catalyst suspension and catalyst deactivation. Accurate and predictive hybrid models can be used for process optimization as well as planning new experimental setups. We expect the use of hybrid models to benefit all development of real-world complex processes by providing enhanced predictive capability. To achieve these aims in the case of biofuel production, we need to push the boundaries of the current state-of-the-art of hybrid modelling to handle the complexities of industrial processes.

Why use hybrid models?
Models based on first principles are desirable since they require relatively little experimental data to set up and they extrapolate well outside the data domain. In contrast to models based on first principles, a purely data driven model as commonly used in classical machine learning, requires no prior knowledge of the physical mechanisms involved. However, such black-box models do require large amounts of data to be trained and they typically extrapolate poorly outside the domain of the training data. Hybrid models occupy a middle ground between these two extremes and thereby alleviate some of their respective drawbacks.

We are looking for a PhD candidate in a multidisciplinary project that combines machine learning, chemical engineering and mathematical models. The aim of the project is to use modern techniques in physics informed machine learning (hybrid models) to tackle hard problems in the area of biofuel reaction engineering. This is a joint project between the Research Institutes of Sweden RISE and Chalmers University of Technology.

Who are you?
The candidate must have a Masters’ degree in one of the following areas: computer science, machine learning, chemical engineering, physics, engineering physics or mathematics. Experience and strong interest in computation and mathematical modelling is required. Interdisciplinary experience within the mentioned scientific areas or experience in for example CFD modelling will be considered a plus.

The successful candidate will be employed by RISE for the duration of the project and will be enrolled at the graduate school Chemical Engineering at Chalmers University of Technology. This is a four-year project during which the candidate will conduct research and take PhD courses at Chalmers University of Technology. University teaching is not required.

We have an open position in Kista/Stockholm. Some travel to Gothenburg will be required.

Welcome with your application!
Include the following in your application: CV, short letter of interest, copy of Masters’ thesis, copy of Bachelor thesis (if applicable), attested copies of Masters’ degree diploma and transcript of academic records (classes taken, grades and other certificates such as TOEFL test results). Directions for the letter of interest: introduce yourself, describe your previous experience of relevance for the position (for example education, thesis work and, if applicable, any other research activities), and describe your future goals and desired long term research focus.

International applications are welcome.

Applications without the required documents will be discarded.

If you have any questions, please contact Director – Applied digitalization: Peter Söderman, +46 10 228 42 99. The deadline for applications is January 16 2023. Applications will be continuously
reviewed throughout the application period.

Our union representatives are Lazaros Tsantaridis, SACO, 010 516 62 21 and Bertil Svensson, Unionen, 010-516 53 56.

PHD student, machine learning, hybrid models, chemical engineering, biofuels, physics informed
machine learning

PhD student in machine learning, hybrid models and chemical engineering | RISE

URL to this page
https://www.chalmers.se/en/about-chalmers/Working-at-Chalmers/Vacancies/Pages/default.aspx?rmpage=job&rmjob=11311&rmlang=UK