Practical design of a lab-scale simplified multiple dividing wall column

Abstract

Dividing wall columns have proven to be a beneficial technology over many years and are already widely used in the industry. With the possibility of receiving three pure product fractions in one column, they achieve up to 30 % savings in investment and operating costs, thus contributing to energy savings as well as to CO2 reduction. Nevertheless, dividing wall columns have a higher complexity than 2-product columns, which results from the installation of the partitioning wall. The higher complexity is associated with an increase in the degrees of freedom, such as an internal liquid split above the dividing wall or an internal vapor split below the dividing wall. Due to consequent further development, by placing more than one partitioning wall in one column shell, a multiple dividing wall column is obtained, which offers the possibility to produce four or more pure product fractions. With the multiple dividing wall column with three partitioning walls, savings of up to 50 % can be achieved, while on the other hand the complexity also increases significantly. The high complexity, which affects all subareas from simulation, design and control structures to operation, is the main reason why there is no real existing column so far. Of particular concern are the additional vapor splits under the partitioning walls, which cannot be controlled during operation. A solution to reduce the complexitiy of multiple dividing wall columns is to simplify them by reducing the number of partitioning walls inside the column. This leads to fewer vapor and liquid splits and therefore to less degrees of freedom. Multiple dividing wall columns gained more interest over the past 10 years and have been studied in more detail on a theoretical basis, which led to more publications in this field. However, some areas, such as the operation of a multiple dividing wall column, which can be investigated using dynamic simulations, are still less researched. This work is intended to explain in detail the entire design process of a multiple dividing wall column. The aim is to contribute to a better understanding of the individual areas such as initialization, optimization, dynamic behavior, design as well as the detailed engineering and operation of a multiple dividing wall column. Furthermore, the investigations are carried out with easily accessible programs, such as commercial process simulators in order to make the work more easily reproducible and thus accelerate an industrial implementation of a multiple dividing wall column. In this work a simplified multiple dividing wall column with two partitioning walls has been designed, built and first operational results have been obtained from the pilot plant. The investigated simplified version is called the 2-2-4-a configuration. To the best of the authors knowledge the realized multiple dividing wall column is the first application of its kind world wide. First of all, a risk analysis is performed to create possible feed mixtures that are as non-hazardous as possible in order to make the plant as safe as possible to operate. In the next step, the V̇min-method is used to check which simplified multiple dividing wall column is more suitable in terms of lower energy requirement for the separation of all previously determined feed mixtures into their pure products. For the investigations Aspen Hysys© and Aspen Plus© are applied as the main simulation programs. The simulation is created via a substitute model, since there is no unit-operation model for dynamic simulations available in these process simulators. In order to achieve convergence in the steady-state simulation, good starting values must be provided for all internal streams. For this purpose, the V̇min-method will be used again, which generates initial values for different feed mixtures that robustly lead to convergence in the simulation. In the next step, the built-in optimization tool is used to obtain an optimized operating point by minimizing the reboiler duty, while adjusting the gas and liquid splits. Product purities of at least 99 mol% are set as boundary conditions. Basically, there are different calculation methods available in process simulators such as the sequential modular or the equation-based approach. For the optimization of multiple dividing wall columns a clear recommendation can be made for the use of the equation oriented mode as it is more robust and more reliable compared to the sequential modular mode. With the results of the optimization the thermodynamic design of the plant can be carried out and therefore the determination of the internal vapor and liquid loads as well as the required theoretical stages for the separation task. Based on this information the fluid dynamic design, which determines the diameter of the column sections is performed. Last but not least the detailed engineering of the plant is carried out. A main feature of the detailed engineering is the modularity, which means that each separation section is implemented as an individual column, eliminating heat transfer through the partitioning walls and individual separation sections can be easily replaced in the future by others. After the design of the column is finalized it is built and installed in the laboratory of Ulm University. Therefore, the column is limited in height to 9.8 m. Each separation section is equipped with packings with approximately 20 theoretical stages, resulting in a total of 220 theoretical stages for the entire column. The diameters of the individual separation sections are either 50 mm or 80 mm. A water-driven glass condenser is installed at the top of the column as well as a vacuum pump to allow the column to operate under vacuum conditions. A stainless steel reboiler with electric heating coils is installed in the bottom of the column. Numerous temperature and pressure measurements in the column allow a flexible adjustment of the control structure to be suitable for separating different feed mixtures. A comprehensive safety concept, consisting for example of a Makrolon enclosure, a gas warning system as well as several collecting trays with leakage sensors, was designed to ensure a safe operation of the plant. Since steady-state simulations do not provide insight into the dynamic operation of the column, nor are they capable of investigating the performance of control structures, it is necessary to create a dynamic simulation. In the dynamic simulation, significantly more information has to be provided, such as the type of internals, additional valves, or the sizing of the periphery, as this information is required for the calculation, which further increases the complexity. Standard PI controllers are used to design the control structures for the column. Temperature controllers are installed in the column to ensure product purities. To select the sensitive temperature stages for temperature control, a sensitivity analysis is performed with all manipulated variables available for control. In the next step, a singular value decomposition is carried out to include the dependence of the manipulated variables on each other. With the results from the singular value decomposition the sensitive temperature stages can be determined and thus finalize the design of the control structure. The interactions between the control loops are subsequently determined using the relative gain array analysis. After the control structures are designed, the dynamic behavior of the column can be studied. For this purpose, different disturbances in the feed mass flow as well as in the feed composition are introduced to the column and the performance of the control structures are investigated. Thereby especially two parameters are important. On the one hand it should be proven that a stable operation is possible after the disturbances have been introduced and on the other hand it is necessary that the required product purities can be met. Different control structures were designed and compared to investigate the dynamic behaviour of the multiple dividing wall column. In order to accomplish a comparison between the control structures, the same inventory control loops, e.g. for pressure and liquid levels, which are essential for stable operation, are applied in all control structures. The differences between the control structures are the temperature control loops. Since five different separations are performed in the 2-2-4-a configuration, five temperature or temperature-difference controllers are needed. The control structure that could best handle the introduced disturbances in the feed mass flow and feed composition has been shown to be the one in which temperature-differences were controlled in the prefractionator and the middle column, while temperature controllers were installed in the main column. To investigate the flexibility of the multiple dividing wall column, three additional feed mixtures were investigated by means of dynamic simulations with the previously determined control structure. The results show that for all investigated mixtures, stable operation is possible a short time after the disturbances have been introduced, and that the product purities can still be guaranteed. This proved the flexibility of the plant and showed that a flexibly designed multiple dividing wall column can be operated in different services. Although only two test runs were carried out at the manufacturer´s premises, the functionality of the pilot plant could be tested and confirmed. Even though no high-purity products were obtained during the test runs, the respective main component was in the corresponding product stream, proving that separation took place. Furthermore, it was shown that a stable operating point of the column can be achieved. Based on the findings of this work, a design guideline for multiple dividing wall columns was developed. It contains all necessary design steps to separate a zeotropic quaternary mixture into its pure products in a simplified multiple dividing wall column. For this purpose, the most suitable version of a simplified multiple dividing wall column must first be selected in order to fulfill the planned separation task. In the next step, the thermodynamic design and thus the number of theoretical stages and the internal gas and liquid loads are determined. Subsequently, the fluid dynamic design of the column can be carried out, by which the internals as well as the diameter of the column are determined. In the last step, the control structure of the column is designed and the dynamic behavior of the column is investigated by means of dynamic process simulation.