Towards Tomography-Controlled Multiphase Flows

Abstract

This thesis investigated, for the first time, the possibilities and limitations of tomography-based real-time control of multiphase flows. The investigation was performed in a gas-liquid axial cyclone separator, chosen as representative of quasi-1D multiphase flow processes. The analysis performed has a multidisciplinary aspect, combining swirling gas-liquid flow physics, tomography, system dynamics and real-time control.
In relation to the swirling gas-liquid flow physics, the vertical upward swirling gas-liquid pipe-flow patterns were experimentally mapped in the axial cyclone for different swirl intensities. Mechanistic models were proposed based on the identified physics to predict the upward vertical swirling gas-liquid flow pattern transitions for pipe diameters in the order of centimeters and swirl numbers in the order of 1. The proposed models were able to successfully predict the flow pattern transitions observed in the axial cyclone and reported in literature.
Tomographic measurements are usually too slow for real-time control, due to the computationally demanding inverse problem of the image reconstruction step. To overcome this limitation, a fast real-time application-specific Electrical Resistance Tomography (ERT) algorithm was proposed to measure the phase distribution in axial cyclones. The proposed algorithm reconstructs the cyclone phase distribution via simple correlations based on raw ERT data instead of solving the inverse problem, resulting in measurements three orders of magnitude faster and up to five times more precise than general non-iterative traditional ERT algorithms based on the inverse problem.
Regarding system dynamics and control, this thesis showed that the phase distribution dynamics in axial cyclones can be split into two components: (i) the intrinsic dynamics due to the swirling gas-liquid flow patterns, and (ii) the phase distribution response to external process disturbances, e.g., changes in the flow rates in the cyclone inlet. The intrinsic dynamics are too fast compared to typical flow actuators, such as control valves, and cannot be controlled. A system dynamics model of the phase distribution in the cyclone was proposed, and used to design a tomography-based controller to suppress external process disturbances in the phase distribution much slower than the intrinsic dynamics. Despite the successful suppression of slow external disturbances, the controller did not significantly improve the efficiency of separation of the process due to the high impact of the (uncontrolled) intrinsic dynamics on the cyclone performance.
The results obtained with the axial cyclone suggest that suppressing external disturbances in the phase distribution of multiphase flows is insufficient to control the efficiency of the process. Therefore, the way forward towards tomography-controlled multiphase flows relies on the development of fast sensors and actuators to attempt to control the intrinsic dynamics, something challenging due to its chaotic nature.