Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Science

Program

Mechanical and Materials Engineering

Supervisor

Ogden, Kelly

Abstract

The presence of a stiff obstruction in the path of fluid causes the creation of a boundary layer over and around the obstruction. The flow over an idealized, two-dimensional series of blocks is numerically investigated to determine how statistical blocks height variation, such as standard deviation, mean, and skewness, influence pressure drop and heat flux. These data sets serve as a foundation for developing models for estimating the heat transfer coefficient of each block using machine learning (ML) methods. The results show that the pressure drop increased by 60% when the standard deviation of heights of blocks increased from 0.1 to 0.4 due to promoting turbulent mixing over the blocks, hence boosting pressure drop and heat flux. Furthermore, the ML model has great potential for predicting the Convective heat transfer coefficient (CHTC) of an individual block given the heights of a few nearby obstacles.

Computational fluid dynamics (CFD), Machine learning, Convective heat transfer coefficient (CHTC), Varying height

Summary for Lay Audience

In nature, the presence of obstacles in the path of a fluid, such as a group of nearby plants in the way of water flow, a forest, or a group of buildings in the way of wind, results in a distinct fluid behaviour over and around the obstacle, with the flow going over the obstacle and slowing down near it. Due to the extensive industrial applications and environmental impacts of barriers on flow behaviour, it is vital to understand the flow near obstructions. This investigation uses idealized simulations to examine the influence of height variation within a single cluster of obstacles. The obstruction might symbolize a vegetation canopy, an urban canopy, or urban structures. Previous research on this topic focuses on groups of blocks that follow a pattern. In this study, the effect of random variation of height is investigated. The height of the 14 blocks for each set of simulations is assumed to be random, and they are generated by defining the average height and the variation of heights. Each of these parameters' effect on the total heat flux and pressure drop are investigated. Additionally, the effect of the qualitative flow regimes is explored. Using simulation-generated data, a machine learning model for predicting the heat transfer coefficient of blocks has been developed. The feature importance analysis in machine learning indicates that only data for a few neighbouring blocks are required to calculate the heat transfer coefficient.

Available for download on Tuesday, August 01, 2023

Share

COinS