Jul-2015Computational fluid dynamics was used to

Jul-2015

Computational fluid dynamics was used to analyse the performance of side entry mixers in large crude tanks
JANAKIRAMULU ADEPU, RAHUL C PATIL, AJAY GUPTA and ASIT KUMAR DAS
Reliance Technology Group
PRAVIN POTDAR, JISNA RAGHAVAN and KANNAN SRINIVASAN Central Technical Services
Reliance Industries Limited

Article Summary

In a refinery, crude oil and product storage tanks are very important factors in smooth and continuous production. Crudes are generally received in large storage tanks from ships via pipelines for further processing. The storage tanks are often equipped with side entry mixers, primarily to avoid settling/layering of sludge inside the tank. These mixers are operated to obtain a homogeneous blend of different crudes with respect to basic sediment and water (BS&W) and density of the crudes. A homogeneous crude blend is desirable for consistent operation of downstream units. The crude blend’s homogeneity is determined based on the density variation between three samples of crude taken from the top, middle and bottom of the tank and a BS&W value which is specific to an individual refinery.

The mixer can be installed in the storage tank in different configurations. The configurations can be differentiated based on number of mixers, which can vary from one to seven, and the swivel angle of the mixer which can vary about the mixer’s axis of rotation. The swivel angles of mixers in tanks are changed based on the desired results. In this study, two different configurations of mixer were simulated and the time for achieving homogeneity was computed using commercial computational fluid dynamic (CFD) software.

The CFD approach

The primary objective of the study was to understand the effect of the number of mixers in a crude storage tank on the time required to achieve homogeneity with respect to the density of the crude. CFD studies were carried out for two layouts of the tanks:

Case 1 – Tank with three mixers

Case 2 – Tank with six mixers.

Initially, steady state simulations were carried out to visualise the flow profiles inside the tank. The velocity profiles inside the storage tank were established and compared for different cases in this study. Later, unsteady state simulations were carried out to study the effect of the number of mixers on the time required for complete homogenisation. The criterion for complete homogenisation was considered as a density difference of less than 1 kg/m3 at different cross-sections of the tank.

The tank geometries for the two cases with mixers are shown in Figure 1. The mixer locations with respect to 0° location are also shown in Figure 1. The mixers are oriented towards one side by a fixed angle around their axis of rotation in both cases. If the diameter of the tank is of the order of 100m, then the tank is filled to a height of D/7.5. Analysis of flow in the domain was carried out using commercial CFD software. During the steady state simulations, the entire tank was filled with heavy crude and the mixers were rotated. The mixers were modelled in CFD using a multiple reference frame (MRF) model. The simulations were continued until the residues fell below 10-5, indicating convergence of the flow problem.

For the unsteady state simulations, equal volumes of heavy and light crudes were considered to be in the tank. Heavy crude (density = 926 kg/m3) filled the bottom half of the tank and the remaining top portion of the tank was filled with light crude (density = 859.5 kg/m3). The initial condition of the unsteady state simulation is 
shown in Figure 2, the red colour indicating heavy crude and the blue colour indicating light crude.

With this initial condition, the mixers were rotated at fixed speed and variations of density at different cross-sections of the tank were monitored to estimate the time required for complete homogenisation. An average density of 893 kg/m3 is expected after complete homogenisation in the tank. The simulations were terminated when the homogeneity criterion of a density difference at different cross-sections of less than 1 kg/m3 was met.

Results – Steady state simulations

The velocity distribution inside the crude tank for the two cases was examined. For better understanding, the velocity profiles were examined at four different cross- sections along the vertical axis of the tank.

The velocity profiles were examined at planes 1 to 4 along the vertical axis of the tank, starting from bottom to top. The maximum and average velocities for the two cases are compared in Table 1. It can be seen from Table 1 that, as expected, the maximum and average velocities at all cross-sections increases as the number of mixers increases. It can be also seen that 
as the height increases the maximum and average velocity decreases, showing that the 
effect of the jet developed due to mixer action reduces along 
the vertical axis of the tank.

The velocity vectors at the impeller level for the different cases are shown in Figure 3. The velocity is scaled in m/s with the red colour showing highest velocity magnitude and the blue colour showing lowest velocity magnitude. It is apparent from Figure 3 that the jets of the impeller intersect with each other, forming a big jet swiping the periphery of the tank. The velocity of this resultant jet increases as the number of mixers increases. The path lines of velocity for the two cases are also shown in Figure 4 for reference. It is again clear from Figure 4 that, with an increase in the number of mixers, the swiping action at the periphery of the tank increases.

Although it is not very apparent from the steady state simulations which layout, three mixers or six mixers, is better than the other, it can be seen that the number of mixers has a marked effect on the velocity profiles in the tank. An increase in the number of mixers increases the amount of swiping at the periphery of the tank and helps to achieve more velocities at each cross-section of the tank along its vertical axis. For more conclusive results, the unsteady state simulations were carried out for two layouts. The time required for achieving homogeneity was estimated from these simulations.

Unsteady state simulation

As was explained earlier, the unsteady state simulations were carried out in a tank initially filled with two crudes of equal volumes, the bottom layer heavy crude and the top layer light crude. The mixers were then rotated at fixed speed and the time required to achieve homogenisation was estimated for both layouts. The criterion for complete homogenisation was considered as a density difference of less than 1 kg/m3 at different cross-sections of the tank and simulations were stopped when this criterion was met. The densities of the mixture in the tank were compared at different planes, plane 1 to 4 from the bottom to the top of the tank.

The variation of densities at different planes of the tank with time for Case 1 is repre

sented graphically in Figure 5. The time estimated for complete homogenisation for Case 1 is 18 hours.

The variation of densities for Case 2 at different planes of the tank with time is represented graphically in Figure 6.

For Case 2, the time estimated for complete homogenisation is less, compared to Case 1. Around 11 hours are required to achieve complete homogenisation for Case 2; that is, seven hours less than for Case 1.