Comparison of transport methodologies on near-field pollutant dispersion in urban environments using CFD
Until recently, the use of Computational Fluid Dynamics (CFD) appears to be gaining traction over traditional Gaussian Dispersion Modelling to predict and understand pollutant dispersion processes in urban environments. Gaussian Dispersion Models, while computationally fast, lacks in physical repres...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Final Year Project |
Language: | English |
Published: |
2019
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/78388 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | Until recently, the use of Computational Fluid Dynamics (CFD) appears to be gaining traction over traditional Gaussian Dispersion Modelling to predict and understand pollutant dispersion processes in urban environments. Gaussian Dispersion Models, while computationally fast, lacks in physical representation and accuracy but still sufficed as evidenced by its use in numerous engineering designs and research applications. The opposite holds true for CFD. In practice, the need for pollution dispersion studies is two-fold: First, for urban planning and risk analysis to pre-empt future possible disasters or industrial accidents and secondly, for analyzing the immediate impact of such events and recommendation of subsequent actions to be taken in order to minimize loss and damages. In CFD, two typical methodologies namely the passive scalar transport and the multi-species transport are used to track spatial dispersion of pollutants. The focus of this study is to understand and quantify the differences between the two models when applied to near-field dispersion of heavy gases. The two methodologies are validated by simulating the dispersion phenomena for two test cases which largely bear resemblance to urban settings: a three-dimensional street canyon setup and the Mock Urban Setting Test (MUST) field experiment. The street canyon is adapted from a series of wind-tunnel experimental measurements under controlled conditions known as CODASC which aimed to emulate dispersion of traffic exhaust in an urban setting. The MUST case study is a full-scale outdoor experiment carried out in a desert of western Utah with an array of 12 by 10 shipping containers with the intention to replicate urban dispersion. The pollutant used in the street canyon study is Sulphur Hexafluoride (SF6) while Propylene (C3H6) is used in MUST. Due care is taken to ensure accurate representation of boundary conditions, with inlet profiles obeying the power-law in the street canyon and the log-law in MUST. The standard k-ε turbulence model is chosen based on recommendations from previous research with appropriate wall functions. Computational demand is also considered when attempting to conclude which methodology works better in terms of computational efficiency. Convergence is examined through monitoring both changes in residuals of primitive variables and concentration changes in locations of interest. Statistical analysis to identify discrepancies between predicted and experimental data is subsequently undertaken to quantify the accuracy of the proposed methodologies. It is found that numerical results are highly sensitive to the value of turbulent Schmidt number (Sc_t) in both test cases. Through parametric studies, the best accuracy is attained when the Sc_t value of 0.5 is used in the street canyon case and Sc_t value of 1.5 for MUST. Generally, better agreement between numerical and experimental results is reflected in the street canyon case study compared to MUST. For the street canyon case study, the passive scalar transport model yielded better results than the multi-species transport model. For MUST, the multi-species transport model provided a more accurate prediction than the passive scalar transport model. This leads to the conclusion that the multi-species transport model does not show significant improvement over the passive scalar transport model. Possible reasons could be due to the assumption of isothermal effects or dispersion in both cases are dilute. With regards to computational efficiency, the passive scalar transport model has proven to be consistently more efficient compared to the multi-species transport model. Time taken for running simulations using the passive scalar transport model is observed to be three times shorter compared to using the multi-species model. |
---|