Dortmund Data Bank Research Articles

Prediction of parameters of group contribution models of mixtures by matrix completion.

Group contribution (GC) methods are widely used for predicting the thermodynamic properties of mixtures by dividing components into structural groups. These structural groups can be combined freely so that the applicability of a GC method is only limited by the availability of its parameters for the groups of interest. For describing mixtures, pairwise interaction parameters between the groups are of prime importance. Finding suitable numbers for these parameters is often impeded by a lack of suitable experimental data. Here, we address this problem by using matrix completion methods (MCMs) from machine learning to predict missing group-interaction parameters. This new approach is applied to UNIFAC, an established group contribution method for predicting activity coefficients in mixtures. The developed MCM yields a complete set of parameters for the first 50 main groups of UNIFAC, which substantially extends the scope and applicability of UNIFAC. The quality of the predicted parameter set is evaluated using vapor-liquid equilibrium data of binary mixtures from the Dortmund Data Bank. This evaluation reveals that our approach gives prediction accuracies comparable with UNIFAC for data sets to which UNIFAC was fitted, and only slightly lower accuracies for data sets to which UNIFAC is not applicable.

Uncertainty evaluation of thermodynamic properties estimated with online data tools: vapor pressure as case study

Online databanks are a convenient source of information whenever thermodynamic properties are required for calculations, in both scientific and engineering environment. The data listed in these repositories is usually reported along with a range of validity and their sources. The values of the thermodynamic properties can usually be retrieved from databases and handbooks, in the form of tables, curves or correlations with regression parameters. Most of the times, little to no information is given on the uncertainty of the reported estimates, let alone regression parameters uncertainty. Evaluating the sources of uncertainty in any model is fundamental to verify the significance of the results for a specific application. Knowing the uncertainty, makes the difference between trusted values and random values. Depending on the model where the property is required and its application (the use of results), the contribution to the total uncertainty of thermodynamic properties and constants can surpass the contributions of other experimental input quantities. In this work, the methods described in the ‘Guide for the expression of uncertainty in measurement’ are used to evaluate the uncertainty of a thermodynamic property of a pure substance, calculated based on the constants and references reported in two important online data banks (NIST and Dortmund Databank). The vapor pressure is considered as case study and several definitions with their relevant sources of uncertainty are presented. The aim of the work is to highlight the limitations of properties data commonly used in engineering estimations, as well as the importance of accounting for their uncertainty. Following established metrological guidelines, a full procedure has been developed, with specific considerations aimed to account for the ‘quality’ of the realization of the definition of the measurand as a source of uncertainty of the property.

An open-access database of the thermophysical properties of nanofluids

The collection, scope, utility, and development of a comprehensive database with published experimental thermophysical properties of nanofluids is described in this paper. The collected data target nanofluids composed of a base fluid, which can be a pure compound or a binary mixture, and a single type of nanoparticle. The available experimental data in published scientific literature up to date in the field of nanofluids has been collected and organized to provide easy access to knowledge on the thermophysical behavior of nanofluids. The database contains a total of 307 datasets with 8118 data records of thermophysical properties of nanofluids, covering 13 types of base fluids, and 19 nanoparticle types. The collected experimental data include five thermodynamic properties (i.e. density, isobaric heat capacity, vapor pressure, speed of sound, and vaporization enthalpy) and two transport properties (i.e. thermal conductivity, and dynamic viscosity). The first release of this database is accesible for free under the host of the Dortmund Databank. An analysis of the consistency of the data, and challenges encountered during the data collection is provided.

Problema inverso aplicado a equação de Antoine para acetona e etanol

Evolutionary Algorithms (AEs) are non-deterministic optimization methods widely used in the literature and find great relevance and wide use in problems involving non-linear, non-differentiable or multimodal functions. Differential Evolution (DE) is an algorithm that has elements of the AEs, such as the mutation, crossing and selection operator, but differs in the generation of new individuals. In this work, the DE algorithm was used to solve the inverse problem of determining the coefficients of the Antoine equation for acetone and ethanol. To this end, temperature and pressure data available in the Dortmund Data Bank (DDB) were used and 30 simulations were performed for each parameter of disturbance rate (F) and crossover probability (CR). The pressure predictions calculated using the parameters obtained by the stochastic process showed greater agreement with the experimental data when compared with the predictions obtained through the parameters consulted in the literature.

Estimation of Pure Component Properties, Part 5: Estimation of the Thermal Conductivity of Nonelectrolyte Organic Liquids via Group Contributions

As an extension of the pure component property estimation techniques that we have developed for normal boiling point, critical property data, vapor pressure, and liquid viscosity, a new group contribution method for the prediction of pure component saturated liquid thermal conductivity has been developed. Predictions are solely molecular structure based. Structural group definitions for the method are a modified version of those proposed by Moller et al. for the estimation of saturated vapor pressures. Structural groups were defined in a standardized form and fragmentation of the molecular structures was performed by an automatic procedure to eliminate any arbitrary assumptions. The new method is based on liquid thermal conductivity data for about 330 components after critical evaluation of 110 000 data points for the various phases and nearly 900 components stored in the Dortmund Data Bank. Results of the new method compared favorably to several other estimation methods from the literature. A relative mean deviation in thermal conductivity of 3.87% was observed for a common set of 331 components (6264 data points). By comparison, the Nagvekar and Daubert method, the best literature method in our comparison, produced a relative mean deviation of 7.03% for 206 components. The range of the method is usually from the triple or melting point to the critical point.

Quality of Component- and Group-Interaction-Based Regression of Binary Vapor–Liquid Equilibrium Data

To describe the behavior of real liquid mixtures, binary interaction parameters of component-based models like, e.g., UNIQUAC, are fitted to experimental data or estimated values from group contribution (GC) methods like UNIFAC and mod. UNIFAC. Because parameters of GC methods are based on a large number of data sets for many similar mixtures, they are typically less precise than models fitted directly to individual data sets for the binary mixture of interest. In some cases, however, it was reported that UNIFAC with group parameters independently regressed to the same data sets represents binary mixtures more accurately than UNIQUAC, which could be due to better localization of interactions. This possible advantage was analyzed using over 3700 data sets from the Dortmund Data Bank. Advantages and disadvantages of the group-based approach are identified, which could be well-explained by the change in concentration base; the results should also hold for segment-based models like NRTL-SAC and COSMO-RS.

Present status of the group contribution equation of state VTPR and typical applications for process development

For the development and design of industrial processes the reliable knowledge of thermophysical properties, in particular phase equilibria is most important. Assuming that 1000 components are of technical interest, vapor-liquid equilibrium data for approximately 500,000 binary systems are needed to fit all required binary interaction parameters. Due to the fact that the measurement of all needed properties is nearly impossible, the process engineer depends on factual data banks.Besides the different phase equilibria, the Dortmund Data Bank as worldwide largest factual data bank for thermophysical properties contains nearly all worldwide available pure component, excess and transport properties. Although more than 66,300 binary VLE data sets for non-electrolyte systems are currently stored in the DDB, in total up to now VLE data for only 13,540 different binary systems are available. The reason is that some systems are very popular and were measured very often, e.g. ammonia – water, ethanol-water and so on. When the 1000 most important components are considered, VLE data for only 8635 binary systems are available. This means for only 1.73% of the systems the required binary interaction parameters can be fitted. Since the assumption of ideal behavior for the missing binary systems can be very erroneous and measurements are very time consuming, predictive group contribution models can be successfully applied to estimate the missing thermophysical properties.To cover sub- and supercritical conditions, group contribution equations of state have to be applied. They automatically take into account both phases and can be used up to high pressures and supercritical conditions. This allows for example the calculation of phase envelopes. Furthermore, the introduction of Henry coefficients for gaseous compounds is not required. At the same time, enthalpies, heat capacities, densities and so on can be predicted. Today the most sophisticated group contribution equation of state is the volume translated Peng-Robinson group contribution equation of state VTPR. In this paper, typical applications of VTPR for process development are shown and new parameters for 24 additional group combinations are given.

Group Contribution Methods for Phase Equilibrium Calculations.

The development and design of chemical processes are carried out by solving the balance equations of a mathematical model for sections of or the whole chemical plant with the help of process simulators. For process simulation, besides kinetic data for the chemical reaction, various pure component and mixture properties are required. Because of the great importance of separation processes for a chemical plant in particular, a reliable knowledge of the phase equilibrium behavior is required. The phase equilibrium behavior can be calculated with the help of modern equations of state or g(E)-models using only binary parameters. But unfortunately, only a very small part of the experimental data for fitting the required binary model parameters is available, so very often these models cannot be applied directly. To solve this problem, powerful predictive thermodynamic models have been developed. Group contribution methods allow the prediction of the required phase equilibrium data using only a limited number of group interaction parameters. A prerequisite for fitting the required group interaction parameters is a comprehensive database. That is why for the development of powerful group contribution methods almost all published pure component properties, phase equilibrium data, excess properties, etc., were stored in computerized form in the Dortmund Data Bank. In this review, the present status, weaknesses, advantages and disadvantages, possible applications, and typical results of the different group contribution methods for the calculation of phase equilibria are presented.

Selection of Solvents or Solvent Mixtures for Liquid–Liquid Extraction Using Predictive Thermodynamic Models or Access to the Dortmund Data Bank

A sophisticated software package for the selection of the most suitable solvents or solvent mixtures for liquid–liquid extraction was developed. For the selection in particular, phase equilibrium information such as selectivity, capacity, etc. was used. For the determination of these properties, besides group contribution methods, such as UNIFAC or modified UNIFAC (Do) and the group contribution equations of state PSRK or VTPR, experimental data collected and updated in the Dortmund Data Bank (DDB) can be applied. For the final selection, pure component and mixture properties such as the surface tension, density, flash point, viscosity, etc. are accessible via DDB also. The applicability of the program package was demonstrated by means of various extraction processes of industrial importance.

Extrapolation/interpolation of infinite dilution, activity coefficient as well as liquid and solid solubility between solvents: Part 1. Alkane solvents

Infinite dilution activity coefficients play a vital role in modeling of phase and chemical equilibrium behavior. From the dependence of the limiting activity coefficient of a solute on the solvent, important properties like liquid and solid solubility in these solvents can be deduced. Although a large amount of experimental information is available in the open literature as well as from thermophysical data banks such as the Dortmund Data Bank (DDB), the behavior of high boiling solutes in low boiling solvents is often not known, partly due to difficulties in experimental determination.A method for the extrapolation of infinite dilution activity coefficients and thus the liquid and solid solubility of any solute in different solvents within one solvent family is under development. Results of a method applicable to any solute in a large variety of different saturated hydrocarbon compounds are presented in this paper.In this method, the activity coefficient is assumed to consist of a combinatorial and residual contribution. In the case of alkane solvents, only this combinatorial contribution changes when going from one solvent to another. Several well known combinatorial expressions were evaluated and were found to produce poor extrapolations in many instances. Quite surprisingly, free-volume combinatorial expressions performed best even for some of the rather low molecular weight compounds used in this test.A new empirically modified free-volume expression is proposed which allows for accurate extrapolation. Additional activity coefficient data have been determined via GLC (gas–liquid chromatography) to validate the method. In most cases extrapolations were within 10% of the experimental findings and safely within or close to the scatter of literature data.The method allows for example the safe estimation of the infinite dilution behavior as well as liquid and solid solubility of high boiling components in hydrocarbons from data for high boiling solvents usually determined easily by gas–liquid chromatography. Additionally, the newly developed combinatorial expression for infinite dilution data should be able to greatly improve other predictive mixture models, e.g. UNIFAC, mod. UNIFAC and COSMO-RS type models in the cases when large molecules in low molecular weight solvents are considered. This was verified for a larger number of experimental limiting activity coefficient data from literature.

Activity coefficient at infinite dilution measurements for organic solutes (polar and non-polar) in fatty compounds – Part II: C18 fatty acids

In this work, activity coefficients at infinite dilution (γ13∞) have been measured for 21 solutes (subscript 1) (alkanes, cycloalkanes, alkenes, aromatic compounds, alcohols, esters, ketones, and halogenated hydrocarbons), in four solvents (subscript 3), namely one saturated fatty acid and three unsaturated fatty acids: stearic (octadecanoic) acid – C18:0, oleic (cis-9-octadecenoic) acid – C18:1 9c, linoleic (cis,cis-9,12-octadecadienoic) acid – C18:2 9c12c, and linolenic (cis,cis,cis-9,12,15-octadecatrienoic) acid – C18:3 9c12c15c, by gas–liquid chromatography. The measurements were carried out at temperatures from (303.13 to 368.19)K and the partial molar excess Gibbs free energy (ΔG1E,∞), enthalpy (ΔH1E,∞), and entropy (ΔS1E,∞), at infinite dilution were calculated from experimental γ13∞ values obtained over the temperature range. The uncertainties in determination of γ13∞ and ΔH1E,∞ are 4% and 20%, respectively. The results for stearic acid obtained in this study have been compared to those available in the Dortmund Data Bank (DDB). The real behaviour of fatty systems could be better understood through the results obtained in this work.

Measurement of vapor-liquid equilibria for diethylamine + methanol system represented double azeotropy on elevated pressure

Accurate phase equilibrium data can be used for the design and operation for separation processes based on phase behavior. An azeotrope is the mixture consists of two or more substances, the equilibrium vapor composition is equivalent to the liquid composition, and the equilibrium temperature (or pressure) indicates extreme value. If the mixture to be separated is an azeotropic system, no separation into pure components can be achieved by simple distillation. Knowledge of the acurate azeotropic data is necessity for chemical process synthesis. According to the Dortmund Data Bank [1], approximately 47% of the stored vapor-liquid equilibrium (VLE) data show azeotropic behavior. There are two types of azeotrope for binary systems, which are positive and negative azeotropes, generally. The binary system of diethylamine + methanol is known for the polyazeotropy at given temperature or pressure. The system forms a homogeneous maximum boiling azeotrope at 101.3 kPa. The second minimum boiling azeotropic point appears in the VLE at the condition of elevated pressure, around 300 kPa. Aucejo et al. [2] has measured the VLE at 101.3 and 300 kPa and discussed the detail of these phenomena thermodynamically. The purpose of this work is describing the trajectory of two different of azeotropic points on elevated pressure from experimental data.

A revised LIQUAC and LIFAC model (LIQUAC*/LIFAC*) for the prediction of properties of electrolyte containing solutions

The reliable knowledge of the thermodynamic behavior plays a major role for the design of different industrial processes. A lot of thermodynamic models do not incorporate electrolyte solutions. For these systems special models have been published. However a lot of these models suffer from the fact that highly concentrated electrolyte solutions cannot be calculated with the desired accuracy which leads to problems when calculating, e.g. salt solubilities. In this article, a revised LIQUAC and LIFAC model is presented, which enables the user to describe mean activity coefficients, osmotic coefficients and also vapor–liquid equilibria reliably. Furthermore the prediction of salt solubilities in aqueous solutions and mixed solvents can be performed successfully. In this paper the predicted data are compared with published data stored in the Dortmund Data Bank. The improvements achieved are additionally shown by a comparison with the results of the original LIQUAC and LIFAC model.

Comparison of the a Priori COSMO-RS Models and Group Contribution Methods: Original UNIFAC, Modified UNIFAC(Do), and Modified UNIFAC(Do) Consortium

A comparison of the performances of the COSMO-SAC, COSMO-RS(Ol), original UNIFAC, modified UNIFAC(Do), and modified UNIFAC(Do) Consortium for activity coefficients at infinite dilution and binary VLE data is presented. The σ-profiles used in performing COSMO-SAC and COSMO-RS(Ol) calculations were taken from the published σ-profile database VT 2005. The predicted results were compared with the experimental data stored in the Dortmund Data Bank and analyzed with respect to the types of components in the mixture. The results show that the UNIFAC models based on experimental data are superior to the a priori COSMO-RS models.

Thermodynamic models and factual data banks - ideal tools for the development of chemical processes

The synthesis, design and simulation of chemical processes, in particular thermal separation processes is today carried out by solving the resulting balance equations of a mathematical model of the considered unit operation or the whole chemical plant using sophisticated commercial process simulation software. The reliability and correctness of the simulation results is mainly influenced by the reliability and correctness of the thermophysical property parameters used for the pure compounds and their mixtures. For the description of the required phase equilibria GE-models and equations of state were developed. These models allow the prediction of the phase equilibrium behavior of multicomponent systems using binary experimental data only. However, since the number of experimental binary data is limited, these methods often cannot be applied. Therefore in particular for process development reliable predictive models with a large range of applicability are most important. With the help of the worldwide largest factual data bank (Dortmund Data Bank (DDB)) for pure component properties, phase equilibria and excess properties, in the last 35 years powerful group contribution methods (UNIFAC, modified UNIFAC) have been developed in my research group. By combination of cubic equations of state with the group contribution concept, now even the phase equilibrium behavior with supercritical compounds can be predicted. At the same time other important properties (e.g. densities, enthalpies, heat capacities, etc.) can directly be predicted. With the development of an adequate electrolyte model (LIQUAC, LIFAC) the equation of state approach was even extended to systems with strong electrolytes. Today ideal predictive thermodynamic tools are available, which can be used in combination with factual data banks for the development of chemical processes. In the lecture the status of the factual data bank and the different predictive models will be shown. Furthermore important applications of industrial interest of the predictive thermodynamic models and the Dortmund Data Bank will be presented.

The universal group contribution equation of state VTPR present status and potential for process development

For the development and design of the various processes in the chemical, oil and gas processing, pharmaceutical, food and environmental industry, reliable thermophysical properties and particular phase equilibrium data are required. Because of the cost-intensive and time consuming measurement of the required data for the synthesis of unit operations or whole plants, the process engineer depends on models which are able to predict the thermophysical properties of the investigated systems reliably. Equations of state in contrast to g E-models show the advantage, that they can be applied to supercritical compounds and that they can be used to calculate enthalpies, heat capacities, densities (e.g. excess volumes), entropies, etc. of the pure compounds and their mixtures. After the development of g E-mixing rules, the group contribution concept could be applied for equations of state. This leads to group contribution equations of state like LCVM, PSRK, etc. In the meantime these group contribution equations of state were further revised by using an improved equation of state and better mixing rules. At the same time all kind of thermodynamic data are used to fit the required reliable temperature dependent group interaction parameters simultaneously to all kind of phase equilibrium data (VLE, γ ∞, LLE, SLE of eutectic systems, gas solubilities, azeotropic data) and excess properties ( h E, c P E) stored in the Dortmund Data Bank. The result is the universal group contribution equation of state VTPR which was first published in 2001 by Ahlers and Gmehling.

Solubilities of NaCl, KCl, LiCl, and LiBr in Methanol, Ethanol, Acetone, and Mixed Solvents and Correlation Using the LIQUAC Model

The solubilities of NaCl, KCl, LiCl, and LiBr in pure methanol, ethanol, and acetone were measured over a temperature range from 293.15 to 333.15 K. Furthermore salt solubilities in the mixed solvents (water + methanol, water + ethanol, water + acetone, methanol + ethanol, methanol + acetone, ethanol + acetone) were determined at 313.15 K. For a few systems solubility data are reported for the first time. In a few cases a comparison with published data stored in the Dortmund Data Bank (DDB)(1) showed disagreement. The LIQUAC model was used to correlate the experimental data. The calculated salt solubilities are in good agreement with the experimental results for the systems NaCl + water + methanol and KCl + water + methanol.

Effect of surface area parameters on the vapor–liquid equilibrium calculations using a lattice fluid equation of state with hydrogen bonding

Vapor–liquid equilibrium calculations at ambient and elevated pressures were performed using the lattice fluid equation of state with hydrogen bonding (NLF-HB EoS) proposed by You et al. [7,8] and Lee et al. [17]. Vapor–liquid equilibrium calculations composed of typical pure components were compared with electronic experimental database (Dortmund Data Bank). Special attention has been paid to the correction of surface area parameters. Bulkiness factors can be used to modify theoretical surface area parameters for molecules with nonlinear shapes. Empirical bulkiness factors were obtained from liquid density data and VLE data sets with n-hexane chosen as a reference component. Using empirical bulkiness factors, overall prediction performances without binary interaction parameters have been significantly improved. It is shown that the NLF-HB EoS have comparable prediction capability with UNIQUAC method or Peng–Robinson Equation of State only with pure component parameters and component-type-dependent hydrogen-bonding energy parameters for most systems considered in this study.

Factual data banks and their application to the synthesis and design of chemical processes and the development and testing of thermophysical property estimation methods

Abstract Factual data banks nowadays play an important role as a source for thermophysical property data for use in chemical process simulation, environmental models, and many other computer-based applications. In this work, the historical developments leading to modern factual data banks, the differences compared to other more bibliographically oriented data banks, and their most important applications and future potential will be discussed by using the example of the Dortmund Data Bank (DDB). As the development of the different predictive models for mixtures is covered in a separate publication, this paper focuses on pure-component property estimation, regression of model parameters, test and verification of model parameters prior to process simulation, and advanced topics in process synthesis such as selection of entrainers and data-mining applications.

Analysis of a concept for predicting missing group interaction parameters of the UNIFAC model using connectivity indices

The reliability of a method for the determination of missing group interaction parameters for the UNIFAC model1 proposed by Gani et al.2 was carefully examined. In this method, called CI-UNIFAC, the missing group interaction parameters are determined by means of connectivity indices. Such a method would be very desirable since it would save a lot of time and money spent for measuring VLE data to fill the gaps in the UNIFAC parameter matrix. Because of the great importance of distillation processes for the chemical industry mainly the results for vapor-liquid equilibria (VLE) were investigated using the data stored in the Dortmund Data Bank and VLE data measured. There are examples in which the method works quite well, but often the procedure leads to poor results. © 2009 American Institute of Chemical Engineers AIChE J, 2009