Phase Equilibria Prediction for Systems Containing Lipids

In recent years, utilization of fats and oils has started to shift towards the bioindustrymakingthem one of the most important renewable materials for the future chemical industry, and leadingto oils and fats industry expansion along with a scale reorientation from local to largescaleindustry [1]. Such developments have led to new challenges regarding the design anddevelopment of better performing processes and products. Despite the advances in propertymodelling and process design techniques available via different computeraidedmethods andtools for the chemical and petrochemical industries, the oleochemical industry is not able toexploit this knowledge due to a lack of experimental data and property models within commercialsoftware applications and the ability to describe accurately the phase behaviour of systems withlipids. Over the past years, new methods and models for predicting single properties andtemperature dependent properties (e.g. critical properties [2], viscosity [3], heat capacity [4], heatof vaporization [4], [5], vapour pressure [5]) have been reported. Likewise, another important modelling task is phase equilibria prediction which is directly related to process synthesis,modelling and simulation. An important aspect in phase equilibria prediction is represented byquality of the data used for regression of model parameters. In previous work, Cunico et al. [5]applied several consistency tests for VLE data sets involving lipids that are available in openliterature and their results show that only 3% of the analysed data sets have quality factors over0.5 (where the quality factor varies between 0 – minimum, and 1 – maximum) [5].In this work, our available extended CAPEC Lipids Database and CAPEC Lipids MixturesDatabase is used for revising the Original UNIFAC model group contribution parameters for lipidsby proposing new values, aimed to offer a better prediction of phase equilibria calculation(vapourliquidequilibrium VLE, solidliquidequilibrium, SLE). The regression of the newparameters is done using carefully selected VLE data sets, screened out for possible erroneousdata. VLE data selection is performed based on the quality factor given by the differentconsistency tests available in ThermoData Engine (TDE) from NIST. More than 60 VLE data setsconsisting of over 600 data points, available in CAPEC LIPIDS Mixture Database, are used for theregression of the 54 binary interaction parameters corresponding to 10 groups for OriginalUNIFAC model. Note that only 10 groups are needed to represent all the lipids data sets.However, to allow a better performance of the model for this type of systems, two new groupswere introduced: one group is describing the behaviour of hydroxyl within mono and diglycerides(OHacyl) and another one is describing the glycerol molecule (GLY). The parameters are testedand evaluated on VLE and SLE data and by using a cross validation method. Compared tooriginal UNIFAC, the performance of the new parameters for the lipids systems present asubstantial improvement in phase equilibria predictions.