Browsing by Author "Moller, Klaus"

Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    Open Access
    A CFD Model for a Fixed Bed Reactor for Fischer Tropsch Reaction using Ansys
    (2023) Chitranshi, Vidushi; Moller, Klaus
    Fischer-Tropsch (FT) is a process which can convert synthesis gas derived from natural gas, coal or even biomass to a variety of products including saturated and unsaturated hydrocarbon chains, while keeping the emission of greenhouse gases minimum. Among various types of reactors used for commercial FT, fixed bed tubular reactors are among the most common type of reactor. However, there is a big challenge faced by these tubular reactors. FT is a highly exothermic process and therefore, heat removal in these reactors is needed to be highly efficient to avoid a thermal runaway. To improve the heat transfer in any reactor, it is necessary to estimate the heat production correctly. Therefore, the kinetics in the FT system needs to correctly represent the heat transfer behaviour in the system. This requires an effective description of the reaction kinetics. FT is a polymerisation reaction, so the rate expressions must be able to retain the chain reaction behaviour. This is not possible with a lumped approach model which is used by most researchers in literature. Therefore, a partial equilibrium approach was employed, where thermodynamic and kinetic models were coupled, and the reaction rates depended on the concentration of reactants as well as products. The kinetic model employed in the current project was taken directly from the work of Davies and Moller. The abovementioned partial equilibrium kinetics was used to develop a CFD model for the FT reaction system. This model was reproduced using COCO simulator for a plug flow reactor for same operating conditions as Ansys to compare the results from both the softwares. The results showed a close agreement and hence, assured that the CFD model could be used for further testing. The other challenge with the FT in FBRs is the heat dissipation. To avoid the thermal runaway, some innovations in reactor design have been studied in literature. In terms of heat transfer capabilities, when shell and tube heat exchangers are compared with the plate and frame heat exchangers, various sources in literature claim that the latter is found to be more effective. However, plate type reactors have not yet been explored in detail for their heat transfer capabilities. Taking an idea from this, the CFD model developed for the tubular reactors was adapted for plate type reactors. The heat transfer capabilities of the plate type reactors were compared relative to the tubular reactors. The tube reactor and plate type reactor were compared on the basis of two criteria. One criterion was based on physical similarity between the reactors. It included having equal Reynolds Number and equal surface area available per unit volume for both type of reactors. For a plate with plate spacing t and a tube with diameter D, the latter condition resulted in the expression, D = 2t. The factors that Reynolds number for a packed bed depends on were all same for both the geometries, so by default, the Reynolds Number was identical for both cases. The other criterion was based on catalyst packing. It included having equal tube-to-particle diameter ratio for both geometries. For tube reactors, diameter is an important parameter that determines the heat dissipation behaviour, so a parametric study was carried out to study the effect of a diameter and plate spacing on heat transfer behaviour for the same set of operating conditions. It was found that the plate type reactor had a hotspot temperature which was less than the hotspot temperature of corresponding tube reactor at all plate spacings. This indicated that the heat dissipation in a plate type reactor is better than in the corresponding tube reactor. Since the tube reactors observed higher temperatures than corresponding plate reactors, the CO conversion observed in the tube reactors was higher. When the product distributions for the two geometries were compared at isothermal conditions, the results almost overlapped for the two geometries. But when they were compared for non- isothermal conditions, significant differences were observed. This showed that heat dissipation mechanisms in the system had a huge role in bringing out different performances for the two geometries. Effect of temperature and conversion on the product distribution were also studied. On the basis of tube to particle diameter ratio criterion, tube reactor was found to outperform the plate reactors in terms of temperature control when compared using the tube-to-particle diameter ratio. Therefore, the superiority of one reactor over the other was dependent on the criterion they are being compared for. The plate type reactor was then represented in PFR model by tuning the heat transfer coefficient of the tubular model in COCO. The difference between the CO conversions achieved between the plate type reactor in Ansys and the representative model in COCO was found to be very little. Hence, the plate type reactor representation could be successfully achieved in COCO. There can be a lot of further research that can be done using the current model. The areas of reaction kinetics and reactor design were highlighted in this regard. The current model can be extended to a larger number of species, for a better representation of the FT product spectrum. Formation of liquids was completely neglected in the current project. It can be taken into account as presence of liquid can affect the FT reactor system by imposing internal and external mass transfer limitations to the reactions. The current model can also be used to study the HTFT process and also to study the isomeric products in the LTFT which were assumed to be not present in the current project. In the areas of reactor design, the geometry of catalyst particles can be included in the reactor geometry. This model can also be used for plates of other shapes and sizes to study the effect of shape and size on heat transfer capabilities. Different types of corrugated plates are used in the Plate and Frame Heat exchangers nowadays. The corrugations increase the surface area available and also increase mixing. Using the current model, such modifications can be studied for their effect on the reactions in a reactive system.
  • No Thumbnail Available
    Item
    Open Access
    A Machine Learning Model for Octane Number Prediction
    (2023) Spencer, Victor; Moller, Klaus; Nyirenda Juwa Chiza
    Assessing the quality of gasoline blends in blending circuits is an important task in quality control. Gasoline quality however , cannot be measured directly on a process stream. Therefore a quality indicator which can be determined from the stream composition is required. Various quality indicators have been used in the existing body of literature but the indicator in this study will be the Research Octane Number (RON). This is an indicator which measures the ignition of gasoline relative to pure octane (Abdul-Gani et al. 2018). Previous research has used empirical models in the form of phenomeno-logical and machine learning models (Gonz´alez 2019). Phenomeno-logical models have been used in the past as a way of programming an engineer's thought process in the form of differential equations put together. Machine learning models are data driven with primarily regression and deep learning methods being used in literature as prediction models. This study aims to develop a parsimonious machine learning model which can be used to predict the RON from the molar composition of the gasoline product stream. Regression, ensemble learning and Artificial Neural Networks (ANN) will be used specifically in this study. The ensemble learning models which will be trained are Bayesian Additive Regression Trees (BART) and Gradient Boosting Machines (GBM). The raw data will be scraped from multiple journals online and the data frame will be comprised of volume compositions of the reference compounds and the RON of each blend. The existing data frame will be extended to include the molar composition of the structural groups present in each of the blends. The structural groups which may be referred to as functional groups are specific substituents within molecules which may be responsible for the characteristic chemical reactions of the respective molecules. This addition of structural groups adds a layer of information to differentiate between blends with different compound compositions but similar RON. It was hypothesised that the molar compositions of the additives and their substituent structural groups would rank highest and the molar composition of n-heptane would have the lowest ranking. For the Multiple Linear Regression (MLR) models, two cases were trained; one with interaction parameters and another without. Both of these cases were trained with and without the composition constraints on the compound compositions. For the ensemble learning case, a BART model with 200 trees and a GBM model with 1998 trees were trained. Four Single Layer Feed-forward Neural Network (SLFN) models were trained, each with 3, 5, 10 and 15 nodes. The choice of neural network architecture was made because the data frame was small, with only 12 input variables and 350 observations. Prior to training the models, an Explanatory Data Analysis was carried out to assess the potential dimensionality reduction, correlations and outliers. The final regression model was the interaction model with a test MSE of 7.54 and an adjusted R2 of 0.986. The BART model obtained a test MSE of 13.74 and an adjusted R2 of 0.983. The GBM model had a test MSE of 38.12 and an adjusted R2 of 0.917. Lastly the best performing ANN was the 10 node SLFN which obtained a test MSE of 11.26 and an adjusted R2 of 0.969. For each model, a variable importance was carried out and it was observed that the molar composition of n-Heptane consistently ranked high in the variable importance. In addition to these predictive statistics; the parity plots, residual plots and Analysis of Variance (ANOVA) were analysed and taken into consideration in evaluating the performance of each of the models trained. It was concluded that the MLR model performed best followed by the BART model. The ANN models ranked third and the GBM model ranked last. The hypothesis that the molar compositions of the additives and their substituent structural groups would rank highest and iv n-heptane would be the lowest ranking component was disproved as the molar composition of n-heptane and its substituent structural groups consistently ranked high . The recommendation for this study is to train the models with a more representative data set in future and to use a hybrid model which comprises of a phenomeno-logical model and a machine learning model for best results and to reduce the bias of the model in the regions with few data points. With the next step of the study being the integration of the new model into the plant-wide Advanced Process Control (APC).
  • Thumbnail Image
    Item
    Open Access
    A study of configurational alternatives of a gas-to-liquids process based on Fischer-Tropsch technology
    (2018) Khazali, Ashcaan Tendo; Moller, Klaus; van Steen, Eric
    Environmental concerns, associated legislation, limited oil reserves and fluctuating crude oil prices are some of the factors that highlight the need for alternative and environmentally friendly routes to fuels. One alternative is to use Fischer-Tropsch Synthesis (FTS) as the major technology in conversion of carbon containing feedstock to transportation fuels. The FTS product, called syncrude, can be refined to high quality transportation fuels in Coal, Gas or Biomass to liquid plants (denoted as CTL, GTL, and BTL, respectively, and collectively referred to as XTL). The economic viability of XTL processes is generally subject to the present price of crude oil and past studies show that traditional refining is generally more economically viable. However, XTL processes have been shown to be more economical and in some cases more environmentally friendly than conventional options when legislative measures aiming to curb traditional fossil fuel usage are considered. This study explores XTL process configurations that can improve plant carbon efficiency to diesel and liquids. The configuration encompasses technologies used, operating conditions, and layout of unit operations. A basic GTL process configuration consists of an Air Separation Unit (ASU), Auto-Thermal Reforming (ATR), syngas cleaning, full conversion Low Temperature Fischer-Tropsch (LTFT) and wax hydrocracking (WHC). These operations are modeled individually and combined to produce a plant model for study with the aim of determining the effects of configurational alternatives on the process efficiency to liquids and diesel. Furthermore, given that the ASU is a major contributor to costs the effect of using oxygen-enriched or pure air is investigated. Since production of heavy wax is prioritized, FTS represents the use of cobalt catalyst in LTFT operation. Where air is used, FTS is run to high conversion in once through mode to avoid the unfavorable economics of recycling nitrogen. After separation of the syncrude, the light fraction is reformed back to syngas in order to maximize carbon efficiency. The heavy wax is hydrocracked to maximize distillate range material. The light products from the WHC are combined with the lights from FTS and the heavy wax is recycled. Carbon efficiency, liquid selectivity and diesel yield are the means of assessing performance. The Scilab programming language is used along with physical properties estimated using the COCO/ChemSep pure component database as a starting point. Estimation of properties for alkanes and olefins of carbon chain length up to C200 has been carried out. The presence of 25% nitrogen in the ATR was found to beneficial to the H2 : CO ratio in the resulting syngas. Furthermore, in FTS the presence of 10-20% nitrogen produced the lowest reduction in carbon monoxide conversion and _FTS. In general, the introduction of nitrogen resulted in decreased conversion of methane in the ATR and both decreased _FTS and conversion in FTS. WHC performance was found to benefit from alpha being as high as possible; however, when the heavy wax recycle was inactive the optimal value was 0.92. The OOT80 configuration was found to have the highest liquid selectivity, while the efficiency to diesel was maximized for the OIRC40 configuration.
  • Thumbnail Image
    Item
    Open Access
    Development of a Kinetic Model for Low Temperature Fischer-Tropsch Synthesis
    (2021) Davies, Imaad; Moller, Klaus
    Globally, there is a need to replace our dependence on fossil fuels as the main source of energy. This requires a shift towards renewable and sustainable alternatives. The well-established FischerTropsch (FT) synthesis is a potential process route to produce liquid fuels and speciality chemicals and address this challenge. FT synthesis is a polymerisation reaction in which syngas, a mixture of CO and H2, is converted to hydrocarbon products ranging from methane to wax when low temperature conditions are used. Subsequent product upgrading steps allow high quality liquid fuels to be obtained which are clean burning. This will help to mitigate the impact of human activity on the environment. The versatility of this process route is attributed to the ability of syngas to be generated from any carbon-containing feed such as coal, natural gas or biomass. The latter is attractive to enable a shift to a more sustainable way of living. Particularly for biomass-to-liquid plants, the high cost of syngas generation means that FT synthesis should use syngas as efficiently as possible. This requires an effective description of the FT reaction kinetics. This study therefore focuses on the development of a kinetic model for low temperature FT (LTFT) synthesis to improve understanding of the reaction behaviour and aid in the development of a biomass-to-liquid process route. Although the kinetics of the FT reactions under the low temperature conditions of 180-260 ◦C and 20-30 bar(a) have been extensively studied, the challenge to kinetic model development is the large number of possible reaction products. A common simplification is to consider the formation of the main products only, which are linear n-paraffins and 1-olefins. The polymerisation character of FT synthesis means that its product distribution could ideally be described using models based on probability theory. Deviations from probability theory distribution, however, occur especially at the conditions of LTFT synthesis. These deviations are a high methane yield, low ethene yield and the change from mainly 1-olefins at low carbon number to mainly n-paraffins at high carbon number. Comprehensive kinetic models in literature focus on finding a kinetic explanation for these deviations. These kinetic models, however, cannot easily be used with few being extended to include the formation of products of higher carbon number. An aspect ignored in current kinetic model development is that FT synthesis shares many aspects of an equilibrium-controlled process. This is since CO hydrogenation which leads to monomer formation is the rate-determining step for the FT reactions. Consequently, the rate of chain growth is rapid in comparison. This leads to the distribution of n-paraffins and 1-olefins being controlled by equilibrium. By modelling FT synthesis as an equilibrium-controlled process, the kinetic model formulation could be simplified, consist of fewer rate expressions and contain the minimum number of model parameters without compromising on prediction quality. At the conditions of LTFT synthesis, both the vapour and liquid phases exist during reaction. The formation of liquid and its effect on the kinetics of FT synthesis has, however, often been neglected with most kinetic models considering the vapour phase only. This is despite of the effect that liquid formation has on product selectivity. The kinetic model developed in this thesis therefore aimed to combine the interaction between chemical equilibrium, kinetics and liquid formation and account for the formation of products of high carbon number. This will assist in providing a comprehensive description of the observed FT reaction behaviour in a simple and tractable manner. A pre-requisite to kinetic model development is the creation of a physical property database for nparaffins and 1-olefins which is extendable to high carbon numbers. This is since only the physical properties of low carbon number n-paraffins and 1-olefins are known because they are common in most industrial processes. For chemical and phase equilibrium calculations, the critical and ideal gas properties needed to be estimated. The Constantinou Gani group contribution method, with modification to the group contributions, proved to be an effective strategy to predict the physical property data for low carbon number n-paraffins and 1-olefins. The correlations developed should therefore provide an adequate approximation of the properties of their higher carbon number relatives. To model the phase behaviour of FT synthesis, the Peng-Robinson equation of state is used. Modifications were made to the alpha function of this equation of state to ensure it remained valid when the describing the behaviour of heavy hydrocarbons. The kinetic model development which relies on the equilibrium aspects of the reactions involved to describe the formation of n-paraffins and 1-olefins. The reaction pathway implemented is based on the alkyl mechanism and assumes that FT synthesis can be viewed as a methylene (CH2) polymerisation. In addition, the water gas shift (WGS) reaction is also considered. Methylene is taken as the monomer and enabled the reactions in FT synthesis to be represented using an equilibrium approach. Each rate expression is formulated as an equilibrium-controlled process, using species activity as the kinetic driving force. This proved to be an effective strategy to account for the observed reaction behaviour, namely a high methane yield, low ethene yield and the change from mainly 1-olefins at low carbon number to mainly n-paraffins at higher carbon number. This approach also allowed the model to effectively capture changes in product selectivity and the product distribution as a function of process conditions (CO conversion, temperature, pressure and H2/CO feed ratio). These changes could be explained by considering the equilibrium aspects of the reactions involved. The model only requires six adjustable parameters i.e. rate constants. An important part of model development is knowing how the model rate constants determine the model output. This provides insight regarding which rate constants can be determined from the regression of data. For this purpose, a sensitivity analysis was performed on the selectivity to C1, C2, C3, C4, C5+ and CO2 as a function of CO conversion. This analysis revealed that CO hydrogenation is rate-determining which agrees with findings in literature. This analysis also revealed that model rate constants are cross-correlated when product selectivity as a function of CO conversion data is studied. This means that meaningful estimation of all rate constants using data of this form is not possible. However, it was found that when product distribution data at constant CO conversion is used instead, then meaningful estimates of the rate constants could be determined. This is if CO conversion is below 60%. Over this CO conversion range, the product distribution is independent of the WGS reaction. The WGS rate constant should thus be approximated using data in literature. As such, five rate constants need to be determined from the regression of product distribution data. Model validation occurred by regression of product distribution data at constant CO conversion. Emphasis was placed on the ability of the model to predict changes in the product distribution with temperature. A quantitative measure of the model fit is the precision with which the rate constants were estimated. A good fit to experimental product distributions in both fixed-bed and slurry reactors is obtained. The kinetic model has therefore been shown to be independent of reactor type. The good fit to data is quantified by the small error in the estimated rate constants, particularly for CO conversions up to approximately 30%. Higher variability in the estimated rate constants was obtained for higher CO conversions. This emphasised the importance of estimating rate constants at conditions where the product distribution is most sensitive. The temperaturedependence of the rate constant could be described by an Arrhenius expression. The effect of liquid formation on the kinetic behaviour of FT synthesis was modelled by assuming that the vapour and liquid phases are in equilibrium. The choice of species activity as the kinetic driving force allowed the kinetic model to be applied in both the vapour and liquid phases. Although the system was mainly in the vapour phase, liquid formation alters selectivity. Single-phase simulations are valid up to a CO conversion of 20% and predict a higher selectivity to products of carbon number in the diesel product grade (C10-C20). Between a CO conversion of 20-90%, it becomes essential to account for liquid formation to ensure that the favourable selectivity to wax products (C21+) in FT synthesis is adequately captured. The predictions of the single- and two-phase simulations were assessed by comparison to the wax product from LTFT reactors in Sasol processes. Both simulations were found to be useful in describing these wax product distributions. The kinetic model developed in this thesis therefore effectively describes the behaviour of FT synthesis. The ability of the model to predict changes in product selectivity and the product distribution as a function of process conditions will make it a powerful tool involved in the design of FT processes. It is recommended that the approach taken to develop the model be used to study the kinetics of other gas-to-liquid processes, for reactor design and flowsheet development.