Taylor flow dynamics under Fischer Tropsch synthesis conditions: Experimental visualization and simulation study

This study includes a comprehensive investigation on the taylor flow of hydrogen and carbon monoxide gases with gasoil in L and T channels. The aim of this thesis is to investigate the bubble, slug lengths and bubble velocities formed in a 1 mm diameter square channel during the flow, to model the obtained bubble and slug lengths with machine learning and to demonstrate the simulation performance of the L and T channel geometry with Ansys Fluent software. In Taylor flow, bubble and slug lengths are critical factors in relation with many variables such as mass transfer, retention rate and film thickness. Investigation of the variation of these variables according to position and time in the flow is very necessary for the use of this flow type in industrial and chemical processes. The machine learning models created within the scope of the study are very important both in terms of revealing the complex behaviour between variables and revealing the applicability of machine learning based control systems, which are among the new technology studies, in taylor flow. In addition, simulation studies are very important to understand the L and T channel geometry. There are various approaches to the bubble and slug lengths in Taylor flow. Relationships between channel geometry, mass transfer, bubble formation frequency and bubble and slug lengths with dimensionless numbers such as Ca, Re are included in the literature. In this context, flow stability and flow homogeneity and bubble velocity are important measures for modelling bubble and slug lengths. In addition, the need for a machine learning study, which stands out with its validation feature for a flexible model proposal within the experimental study limits of the modelling studies carried out in classical ways so far, has been identified. In all this process, simulation studies have gained an important place by providing many data such as film thickness, slug velocity, gas liquid mixture properties in taylor flow. In simulation studies, large differences were observed between experimental and simulation results in L and T channels under conditions where Re>1 in Taylor flow, and strategies such as periodic flow and special functions were developed to solve this problem. In the first part of the study, in order to analyse this relation ball, the significant effects such as channel type, gas type, gas and liquid velocities on the formation of bubbles and slugs and the changes of bubbles and slugs in the channel were investigated. In the second part of the study, the obtained bubble and slug lengths are used to propose a model that covers the bubble and slug length models in the literature. Finally, the hydrogen gasoil performances in the channels are analysed to reveal the simulation problem in L and T channels used in the study. In this study, bubble and slug lengths and bubble velocity in the channel were obtained with the help of image processing from the flow patterns obtained by imaging the gas and liquid phases sent to the T and L shaped channels formed in the plexiglass channel with the help of a high speed camera. In these data, as the bubble and slug length increased, deviations were observed between the bubble and slug formation sizes at constant gas and liquid velocity. This was attributed to the pressure drop created by the long liquid slug and the rapid movement of the long bubbles. It is also observed that the bubble and slug lengths follow the Burr distribution instead of the normal distribution and this distribution behaviour gradually changes from the bubble and slug formation. In the modelling of the bubble and slug lengths obtained from the first part with the help of machine learning, Re and Ca numbers, gas and liquid flow rates, gas and channel type, non-parametric Gaussian process regression were used in the modelling and a much more successful model was obtained than the models in the literature.The individual graphs of the variables in the proposed models were analysed and it was found that the proposed model was quite successful in terms of the bubble and slug lengths of the variables and their relationship with each other. In the simulation studies, it is concluded that the gas and liquid feeds from separate branches to the 3D channel geometry are not affected by variables such as inlet and outlet pressures and flow viscosity. It is concluded that the experimental and theoretical bubble velocities obtained in the channel are the same under flow conditions where the flow pattern obtained experimentally in the channel is deviated. Under these conditions, it was revealed that the gas and liquid flow rates and retention rates are restricted within a certain range by obeying the Fourier transform of the retention rate. No correlation between experimental and theoretical data was found in L channel. This study reveals the behaviour of the bubble and slug in the channel in Taylor flow and will be useful in determining the conditions that will provide a smooth mass and heat transfer in studies using hydrogen and carbon monoxide gases. In addition, the non-parametric models proposed for the bubble and slug length were found to be quite successful, revealing the preferability of these models in flow control applications with new technology. In the simulation studies, the causes of the T and L channel problem were revealed and it was concluded that different strategies are required for the solution of the problem.