ESR 2 - Development of advanced mathematical data interpretation methods

Host: DTU, Supervisor: Krist V. Gernaey

This PhD project has the focus on extracting as much information as possible from the obtained experimental data. In practice, this can be achieved by coupling advanced mathematical data interpretation to the MBR experiments. This project is therefore devoted to technologies that support MBR development.

Tasks and methodology

Performing experiments at microscale will generate a considerable amount of data, and in the end a person operating such a microscale system will have difficulties to interpret all generated data using traditional tools such as a spreadsheet program. Therefore, ESR2 will focus on streamlining data interpretation. First, ESR2 will work on describing cultivations (WP6) and biocatalysis processes at microscale (WP7) by mechanistic models of reactor systems, either based on ordinary differential equations (ODEs) or partial differential equations (PDEs), and will specifically work on the modeling of S. cerevisiae batch and continuous cultivations (Yr 1). Based on an existing MatlabTM toolbox developed at DTU, ESR2 will also work on the application of uncertainty and sensitivity analysis for this model with the aim to use the analysis results for proposing targeted new experiments in order to collect informative experiments with experimental set-ups. ESR2 will furthermore develop software sensors for the MBR platforms (Yr2), to extend the number of variables for which on-line information is available. The two chemometric methods principal component analysis (PCA) and partial least squares (PLS) regression are commonly applied together with spectroscopic data (WP8) and process data (WP6), and will be used to predict variables such as biomass concentration and substrate concentration which are usually difficult to measure on-line. From Yr2 onwards, DTU will also give support to the consortium for development of software sensors. Finally, ESR2 will work on automating experiments (Yr3), coupling simulations with a model with techniques like Design of Experiments, Monte Carlo simulation or optimisation methods on the one hand, and then transforming results into new experiments that are to be performed in a microbioreactor. Results for the latter experiment(s) are then fed back to a new cycle of parameter estimation/simulations until a user-defined optimality or performance criterion is reached.

  • Use of uncertainty/sensitivity analysis to MBR platform fermentation case study
  • Software: Method for Design of Experiments (DoE) linked to microbioreactor platform
  • Mechanistic model of biocatalytic case study completed
  • Software: software sensors implemented and in operation on MBR platform fermentation case study