An Examination of Design Practices for Mixing Tanks Using Computational Fluid Dynamics

  • Nicodemus Muchemu

Student thesis: Doctoral Thesis

Abstract

Background and Objectives:
This thesis examines design practices of mixing tanks with baffles using the tools of computational fluid dynamics in a technique known as a multiple reference frame analysis. The goal of this thesis is to optimise the simulation process as well as to examine the effect of varying the impeller size and shape on fluid flow within the confines of a mixing tank. This was done for both a single-phase flow (water) and a multiphase flow (crude oil and water mixture).

Methods and Results:
A model tank of cylindrical shape of height to diameter ratio of 1:1 was used in the simulation. The diameter of the proposed mixing tank design was 1.08 m with the tank diameter being based on a mixing tank from an industrial supplier. AutoCAD was then used to construct this geometric shape. The model tank has a paddle blade impeller operating at a clearance of 0.36 m and diameter of 0.54 m with a moving fluid zone surrounding the agitator paddles. The tank also has two baffles on either side of the agitator with each of them having a width of 0.18 m and thickness of 0.045 m. Ansys workbench was used to further process this geometric model with Boolean operation to remove the solid regions within the mixing tank and separate the moving and stationary fluid zones. Principal component analysis based factor analysis approach was used to optimise the element size and residual to stabilise the convergence and accuracy. The optimised element size was found to be 1.34 cm and 〖1.3×10〗^(-3) was found to be the optimised residual value for the CFD simulation. Baffle dimensions were proposed based on their effect on fluid flow patterns and were labelled according to baffle length to width ratios. The 2:1 baffle (0.36 m:0.18 m) was found to be the most suitable when analysing pressure and velocity profiles at different regions within the mixing tank.With the baffle ratio of 2:1, three shapes with the same baffle length were constructed: rectangle (standard), normal distribution and error functions based were simulated. When observing average pressure and various velocity profiles at specified locations, the normal distribution curve was seen to produce the best mixing results. In particular, at the horizontal line location labelled x-axis line location, the pressure profile for the normal distribution curve had the largest maxima and minima at 208217.1 Pa and -458357 Pa respectively with an average pressure value of -115047 Pa. In contrast, the average pressure values for the error and standard baffle configurations respectively were -52345.8 Pa and -57675.5 Pa. The same pattern of the normal distribution shape producing better mixing results was consistently observed with the rest of the pressure and velocity profiles provided at different locations. To examine the viability of proposed baffle designs in multiphase mixing, 10% volume of water and 90% volume of Conroe crude oil was introduced to the mixing tank for the simulation for baffles of the proposed shapes. They were the modelled using the Eulerian multiphase model under the optimised settings. The average water volume fractions at different regions of the mixing tank were used to determine the dispersion quality of each baffle design. The normal distribution baffle had an average volume fraction of 〖2.2×10〗^(-3) at the distance furthest from the impeller paddles. This implied that of the 3 baffle configurations, for water to disperse in crude oil the normal distribution shape produced the best quality of dispersion.

Conclusions:
Of the proposed rectangle (standard) baffle designs, the 2:1 baffle configuration was found to be the most suitable from the profile analyses when operating at the optimised CFD settings. Likewise, of the proposed shapes the normal distribution baffle configuration was fluid to be the most suitable for both multiphase and single-phase mixing.
Date of Award26 Jun 2023
Original languageEnglish
SupervisorLande Liu (Main Supervisor) & John Chai (Co-Supervisor)

Cite this

'