Degree Type

Dissertation

Date of Award

2019

Degree Name

Doctor of Philosophy

Department

Mechanical Engineering

Major

Mechanical Engineering

First Advisor

Michael G. Olsen

Abstract

The production of uniformly sized functional nanoparticles for applications including pharmaceuticals, pesticides, and cosmetics is a problem of great interest. The macroscale multi-inlet vortex reactor (macro-MIVR) was developed for nanoparticles manufacturing using flash nanoprecipitation technique due to its ability to achieve rapid mixing and high efficiency. A thorough experimental investigation was conducted to characterize and optimize the turbulent mixing of the macro-MIVR. The planar laser induced fluorescence (PLIF) was first used to acquire the instantaneous concentration fields of the passive scalar. Data were collected in three measurement planes located at different heights from the reactor bottom (¼, ½, and ¾ of the reactor height) for Reynolds numbers of 3290 and 8225 based on the reactor inlet velocity and hydraulic diameter. The instantaneous fields were used to quantify the pointwise statistics such as the mixture fraction mean, variance, and one-point concentration probability density function. These revealed that, near the reactor center, the fluid was nearly homogeneously mixed at a mixture fraction of 0.5. Away from the reactor center, the unmixed fluid formed the spiral arms structures of high concentration gradients.

The full revelation of the underlining turbulent mixing mechanisms in the macro-MIVR was achieved using the mixing parameters such as turbulent viscosity, turbulent diffusivity, turbulent Schmidt number, linear stochastic estimates, and conditional averages. Such quantities were obtained by performing the simultaneous measurements of instantaneous velocity and concentration fields using stereoscopic particle image velocimetry (stereo-PIV) and PLIF. In the free vortex region for the radial position (r) greater than 20 percent of the reactor radius (Ro), the turbulent viscosity was nearly constant. Near the center of the reactor in the forced-vortex region (r/Ro < 0.1), the turbulent viscosity significantly increased, with peak values occurring near the center. The turbulent viscosity and Reynolds shear stress were highest near the reactor exit at the ¾ plane. The dominance of high turbulent fluxes and low concentration gradients near the reactor center led to high turbulent diffusivity. Away from the center, the turbulent diffusivity was reduced because of large concentration gradients and low turbulence intensity in the spiral arm region. The turbulent Schmidt numbers were also found to correlate with concentration gradients. The turbulent Schmidt number values were found to vary from 0.1 to 1.2. The highest spatial variation in the turbulent Schmidt number was observed in the spiral arms region, where the concentration gradients are also the highest. This spatial variation in Schmidt number contrasts with the common assumption of constant turbulent Schmidt number in Reynolds-averaged CFD models.

To gain further understanding of the correlation between the concentration and velocity fields and potentially offer useful information needed to achieve a closure in the conditional moments closure methods (CMC), the conditional mean velocity and concentration profiles were extracted at various locations on a streamline passing through significant concentration gradients. Two mathematical models (i.e. the linear approximation and PDF gradient diffusion model) were also validated using experimental results. The results of the velocity conditioned on the mixture fraction proved that the linear model works well in a low turbulence region away from the reactor center. Nevertheless, near the center of the reactor, an acceptable agreement was found within ± 2Ф_rms (mixture fraction root mean square). The PDF model with an isotropic turbulent diffusivity predicted inadequately the tangential and axial conditional velocities. A modified version of the PDF model that considers all components of the turbulent diffusivity produced better agreement with experimental data, especially in regions of considerable concentration gradients. Furthermore, the mixture fraction conditioned on the velocity tensor components (〈Ф|ω_i 〉) showed a more linear behavior near the reactor center, where the probability density function (PDF) of the mixture fraction is a Gaussian distribution. As the concentration gradients became prominent away from the reactor, 〈Ф|ω_i 〉 also deviated from the linear pattern. This was especially remarkable for the mixture fraction conditioned on the tangential velocity. The overall prediction Ф|ω_i showed an improvement at higher Reynolds number as the fluid mixing was enhanced.

Moreover, the coherent structures were investigated using a technique of linear stochastic estimation (LSE). The linear stochastic estimation of the velocity fields were computed directly from the two-point spatial correlations for various basepoints located in regions of high concentration gradients. The correlations were found to be elliptical in shape, inclined, and peaked at the basepoints. The estimated instantaneous conditional velocity fields revealed obliquely oriented counter-rotating vortical structures that stir the fluid in the direction normal to the mean flow. Finally, these flow structures weakened when the Reynolds number was decreased from 8125 to 3250.

Copyright Owner

Emmanuel Hitimana

Language

en

File Format

application/pdf

File Size

135 pages

Share

COinS