Systematic computer aided methods and tools for lipids process technology


Lipids are naturally occurring compounds which cover a wide applications range: from biofuels to food, health and personal care products, and which involve different industries, such as biodiesel, edible oils and oleochemical industry. Expansion of lipids related industries led to new challenges regarding the design and development of better performing processes and products. Despite the advances in property modelling and process design techniques available in different computer-aided methods and tools for the petrochemical and chemical industries, the lipids related industries were not able to exploit this knowledge. The main reason is the lack of experimental data and lack of available property models within commercial software tools. An essential aspect of process synthesis, modelling and simulation is represented by phase equilibria prediction, which is highly dependent on data used for regression of model parameters. The objectives of this PhD project are to further extent SPEED lipids data base, to develop and apply a systematic identification method for thermodynamic property modelling for improving the phase equilibria prediction for lipids systems, and then apply the new thermodynamic models for process design.


Process synthesis, design, optimization, control as well as energy, economic and environmental impact analysis are performed through model-based tools, where for a class of systems, accurate and consistent prediction of phase equilibrium is very important. While experimental determination of phase equilibrium is accurate, it is time consuming, expensive and not suitable for multiple requirements. In this case, it is better to use collected experimental data to fit the parameters for a suitable thermodynamic model. The important issues are however, the amount and consistency of data, the selection of the appropriate model and its parameters. For the scope of the model-based system, the models need to be predictive and it should be possible to extrapolate from regressed model parameters. Also, it should be possible to obtain the best identified models with a minimum of the available data, since there may be limitations on data availability.
The demand of oils and fats has grown from 40.8 million tons in 1980 to more than 210 million tons 2016 [1]. The consumption of 17 vegetable oils for bio fuels and food and other applications are presented in Figure 1. Limited availability of physical and thermodynamic properties for lipids compounds and their mixtures lead to difficulties in providing models to accurately describe their phase behaviour. VLE predictions for lipid systems with original UNIFAC model have been found to be poor [2]. Also, in addition to phase equilibrium models, pure component property models are also needed. Therefore development of accurate and consistent thermodynamic models is needed for oleochemical systems. In this work, development and application of a systematic method for identification of thermodynamic models is presented. The aim of the method is to improve the phase equilibria predictions by providing reliable and correctly identified parameter sets for predictive group contribution models. The main steps of the method are: data collection and organisation, data analysis-selection, and parameter estimation and validation, Figure 2. The method is applied for improvement of VLE predictions involving lipid systems with the original UNIFAC model. First, all available binary VLE data sets are collected from SPEED Lipids Database, and organised into groups according to the binary interaction parameters that need to be identified. Next, the datasets with the best consistency test performances are selected for parameter estimation, which is performed according to an established sequence as presented in Figure 3. The calculation sequence of the parameters results from the first part of the methodology.The new set of the parameters is analysed by comparing the model performance for all available VLE datasets as well as their extrapolation to SLE data. The results show significant improvement in the performance of original UNIFAC model. The same procedure is applied to fit the model parameters for LLE data as well as group contribution models for pure component properties. In principle, the systematic model identification method should be applicable for all thermodynamic model development.

Conclusions and future work

Development of lipids property modelling with emphasis on process simulation, presents a big evolution in the last years [3]. However, there are still a lot of research areas to cover within the lipids modelling and technology development. The future steps of this project will try to cover some of these aspects: extension of available lipids data base, lipids property models improving, generation of a systematic framework for oils and fats able to improve the research within lipids related industries.


Prof. John M. Woodley (Principal supervisor)
Bent Sarup
Prof. Georgios M. Kontogeorgis

PhD Study started: September 2015 to be completed: September 2018

To see figures and references for this text, please find the original in the Graduate Schools Yearbook 2017


Olivia Ana Perederic
DTU Kemiteknik

To see figures and references for this PhD study, please download the Graduate School Yearbook.

The yearbook provides an overview of most of the PhD theses currently being written at DTU Chemical Engineering and is updated every year in February.