PhD completed 2015

Nicolin Govender, 2015. “Blaze-DEM: A GPU based large scale 3D

 discrete element particle transport framework"


Understanding the dynamical behavior of particulate materials is extremely important  to many industrial processes with a wide range of applications ranging from hopper flows  in agriculture to tumbling mills in the mining industry. Thus simulating the dynamics of particulate materials is critical in the design and optimization of such processes. The mechanical behavior of particulate materials is complex and cannot be described by a closed form solution for more than a few particles. A popular and successful numerical approach in simulating the underlying dynamics of particulate materials is the discrete element method (DEM). However, the DEM is computationally expensive and computational viable simulations are typically restricted to a few particles with realistic particle shape or a larger number of particles with an often oversimplified particle shape. It has been demonstrated for numerous applications that an accurate representation of the particle shape is essential to accurately capture the macroscopic transport of particulates. The most common approach to represent particle shape is by using a cluster of spheres to approximate the shape of a particle. This approach is computationally intensive as multiple spherical particles are required to represent a single non-spherical particle. In addition spherical particles are for certain applications a poor approximation when sharp interfaces are essential to capture the bulk transport. An advantage of this approach is that non-convex particles are handled with ease. Polyhedra represent the geometry of most convex particulate materials well and when combined with appropriate contact models exhibit realistic mechanical behavior to that of the actual system. However detecting collisions between the polyhedra is computationally expensive often limiting simulations to only a few thousand of particles.

Driven by the demand for real-time graphics, the Graphical Processor Unit (GPU) offers cluster type performance at a fraction of the computational cost. The parallel nature of the GPU allows for a large number of simple independent processes to be executed in parallel. This results in a significant speed up over conventional implementations utilizing the Central Processing Unit (CPU) architecture, when algorithms are well aligned and optimized for the threading model of the GPU. This thesis investigates the suitability of the GPU architecture to simulate particulate materials using the DEM. The focus of this thesis is to develop a computational framework for the GPU architecture that can model (i) tens of millions of spherical particles and (ii) millions of polyhedral particles in a realistic time frame on a desktop computer using a single GPU. The contribution of this thesis is the development of a novel GPU computational frame-work Blaze-DEM, that encompasses collision detection algorithms and various heuristics that are optimized for the parallel GPU architecture. This research has resulted in a new computational performance level being reached in DEM simulations for both spherical and polyhedra shaped particles.

In terms of the particle shape there are no other freely available codes that can match the geometrical fidelity in terms of accurate representation on the GPU. To the authors best knowledge there is only one study on the GPU that takes particle shape into account with a physics model of similar fidelity to Blaze-DEM. That study uses the clumped sphere method. Blaze-DEM is able to simulate 2 orders of magnitude more particles compared to other published results while being 3 times faster. The only reported implementations for polyhedra are on the CPU platform. Blaze-DEM is hundreds of times faster compared to CPU codes with physics models of a similar fidelity and 24 times faster than CPU codes with physics models of a lower fidelity. For simulations involving spherical particles Blaze-DEM is 5 times faster than other GPU based codes that have physics models of a similar fidelity.

Keywords: DEM, GPU, Parallel Processing, Polyhedra, Particle transport framework, Semi-autogenous ball mill, Particle discharge

Supervisors: Dr. D.N. Wilke, Prof. S. Kok



Mehdi Mehrabi, 2015. "Modelling and optimisation of thermophysical properties and convective heat transfer of nanofluids by using artificial intelligence methods"

Nanofluids are modern heat transfer fluids which can significantly increase the thermal performance of a thermal system. It enhances the thermal conductivity of working fluids due to adding solid nanoparticles to the base fluid. In order to use nanofluids widely in industrial applications knowing the thermophysical properties of these new heat transfer fluids are essential. In this research, the GA-PNN and FCM-ANFIS methods are employed to present models for thermophysical properties of nanofluids. Furthermore, modified NSGA-II technique has been used to optimise the convective heat transfer of nanofluids in a turbulent flow regime.

In recent years considerable correlations have been suggested by different researchers for thermophysical properties of nanofluids based on the experimental and theoretical works, which a large number of those correlations are failed to predict the thermophysical properties of nanofluids for a wide range of particle size, temperature and nanoparticle volume concentrations. In this thesis, experimental data available in literature have been used to propose models for thermophysical properties of nanofluids to overcome this problem by using artificial intelligence-based techniques. Two models based on FCM-ANFIS and GA-PNN techniques have been proposed for the thermal conductivity and viscosity of nanofluids. To show the accuracy of the proposed models, the predicted result has been compared with experimental data as well as well-cited correlations in literature. Furthermore, the convective heat transfer of nanofluids was studied and different models based on artificial intelligence techniques have been proposed to model the Nusselt number and pressure drop of nanofluids in a turbulent regime. Finally, a multi-objective optimisation technique was used to optimise the convective heat transfer characteristics and pressure drop of nanofluids to find the best design point base on the Pareto front of the results. The predictions of the models for all cases agreed with the experimental data much better than the available correlations.

Keywords:     Nanofluids, thermophysical properties of nanofluids, Brownian motion, nanolayering, artificial intelligence, multi-objective optimisation, GA-PNN (genetic algorithm-polynomial neural network) method, FCM-ANFIS (fuzzy C-means clustering- adaptive neuro-fuzzy inference system) technique, convection heat transfer, NSGA-II (modified non-dominated sorting genetic algorithm) multi-objective optimisation




Dr. Mohsen Sharifpur and Prof. Josua P. Meyer


Johnathan John Vadasz, 2015. "Vibration effects on Natural convection in a porous layer heated from below with application to solidification of binary alloys"

Directional solidification has a wide interest due to its importance to the iron and steel industry. Examples of further application can be found in the aerospace industry regarding the manufacture of turbine blades and the semiconductor industry regarding single-crystal growth applications. Solute convection in the solidification process results in channel formation, which has a freckle-like appearance in cross-section and has a critical effect on the mechanical strength of a casting. For a solidification process that occurs via planar solidification from a solid boundary, one may consider the presence of three distinct regions often identified as horizontal layers, i.e. a fluid binary mixture (the melt), the solid layer and a two-phase (fluid-solid) mushy layer, separating the other two. The mushy layer is practically a porous medium consisting of an interconnected solid phase having its voids filled with the melt binary fluid. Channelling in the mushy layer and the creating of freckles are being considered the main reasons for non-homogeneous solidification and production of defects in the resulting solid product. The production of defects adversely affects the mechanical properties of the solid product leading to undesirable constraints on its industrial use.

The purpose of this study is to evaluate the effect the vibrations have on the heat transfer during the solidification process as well as on the average density of the solid product and void formation. Experimental work on solidification of paraffin was performed. Theoretical results of heat convection in a porous layer heated from below and subject to vibrations are presented by using a truncated spectral method in space. The results show that the heat convection subject to vibration is generally reduced when compared with the corresponding convection without vibrations. The exception for a certain frequency range shows about a 10% enhancement in the weak turbulent regime of convection, however, a 10% enhancement is still lower than the heat transfer prior to the transition to weak turbulence. Therefore, the heat transfer mechanism can be excluded as the main reason behind the improvement in solidification when vibrations are applied.

KEYWORDS: vibration, solidification, mushy layer, porous media, natural convection.

Supervisors:  Professor Josua P. Meyer, Dr. Saneshan Govender




The small-scale dish-mounted open solar thermal Brayton cycle (1-20 kW) with recuperator has an advantage in terms of cost and mobility and can offer an off-grid electricity solution to the people of the water-scarce southern Africa. South Africa has an advantage in terms of solar resource, but this solar resource is not used extensively due to high-cost and low-efficiency solar-to-electricity systems. The dish-mounted solar thermal Brayton cycle with recuperator offers a solution. However, heat losses and pressure losses in the cycle components can decrease the net power output of the system tremendously. In addition, the costs due to solar tracking and perfect dish optics can be high. The purpose of the study was to develop the small-scale (1-20 kW) dish-mounted open solar thermal Brayton cycle by optimising an open-cavity tubular solar receiver and counterflow plate-type recuperator with the method of total entropy generation minimisation. The optimised receiver was also tested in an experimental dish collector set-up. Modelling methods to predict the performance of the cycle and to optimise the solar receiver and recuperator were developed and tested so that the small-scale open solar thermal Brayton cycle could be developed further. SolTrace was used as ray-tracing method to determine the effects of inaccurate dish optics. An optimum concentration ratio of 0.0035 was identified for a collector with a maximum tracking error of 1° and an optical error of 10 mrad. It was shown that the open-cavity tubular solar receiver surface temperature and net heat transfer rate for heating air depended on the receiver size, mass flow rate through the receiver, receiver tube diameter, receiver inlet temperature and dish errors. Receiver efficiencies of between 43% and 70% were found for a receiver with mass flow rates of between 0.06 kg/s and 0.08 kg/s, tube diameters of between 0.05 m and 0.0833 m, air inlet temperatures of between 900 K and 1 070 K operating on a dish with 10 mrad optical error and maximum solar tracking error of 1°. With the use of Matlab and Flownex, it was shown that the small-scale open solar thermal Brayton cycle could generate a positive net power output with solar-to-mechanical efficiencies in the range of 10-20% with much room for improvement. The maximum receiver surface temperature was restricted to 1 200 K and the recuperator weight was restricted to 500 kg. An experimental set-up with a 4.8 m diameter parabolic dish with rim angle of 45° on a two-axis tracking system was constructed to test the receiver. An optimised open-cavity stainless steel tubular receiver with tube diameter of 88.9 mm was tested in the experiment. The experimental results showed the challenges regarding the design and construction of a solar thermal Brayton cycle collector. It was found that the insulation arrangement around the large receiver tube diameter influenced the heat loss due to convection and conduction. Results showed that with further research, the small-scale open solar thermal Brayton cycle could be a competitive small-scale solar energy solution to the people of South Africa.

Keywords:             solar, Brayton, receiver, optimisation, entropy

Supervisors:         Prof. T. Bello-Ochende, Prof. J. P. Meyer.

Aggrey Mwesigye, 2015. "Thermal Performance and Heat Transfer Enhancement of Parabolic Trough Receivers – Numerical Investigation, Thermodynamic and Multi-Objective Optimisation"


Parabolic trough systems are one of the most commercially and technically developed technologies for concentrated solar power. With the current research and development efforts, the cost of electricity from these systems is approaching the cost of electricity from medium-sized coal-fired power plants. Some of the cost-cutting options for parabolic trough systems include: (i) increasing the sizes of the concentrators to improve the system’s concentration ratio and to reduce the number of drives and controls and (ii) improving the system’s optical efficiency. However, the increase in the concentration ratios of these systems requires improved performance of receiver tubes to minimise the absorber tube circumferential temperature difference, receiver thermal loss and entropy generation rates in the receiver. As such, the prediction of the absorber tube’s circumferential temperature difference, receiver thermal performance and entropy generation rates in parabolic trough receivers therefore, becomes very important as concentration ratios increase.

In this study, the thermal and thermodynamic performance of parabolic trough receivers at different Reynolds numbers, inlet temperatures and rim angles as concentration ratios increase are investigated. The potential for improved receiver thermal and thermodynamic performance with heat transfer enhancement using wall-detached twisted tape inserts, perforated plate inserts and perforated conical inserts is also evaluated.

In this work, the heat transfer, fluid flow and thermodynamic performance of a parabolic trough receiver were analysed numerically by solving the governing equations using a general purpose computational fluid dynamics code. SolTrace, an optical modelling tool that uses Monte-Carlo ray tracing techniques was used to obtain the heat flux profiles on the receiver’s absorber tube. These heat flux profiles were then coupled to the CFD code by means of user-defined functions for the subsequent analysis of the thermal and thermodynamic performance of the receiver. With this approach, actual non-uniform heat flux profiles and actual non-uniform temperature distribution in the receiver different from constant heat flux profiles and constant temperature distribution often used in other studies were obtained.

Both thermodynamic and multi-objective optimisation approaches were used to obtain optimal configurations of the proposed heat transfer enhancement techniques. For thermodynamic optimisation, the entropy generation minimisation method was used. Whereas, the multi-objective optimisation approach was implemented in ANSYS DesignXplorer to obtain Pareto solutions for maximum heat transfer and minimum fluid friction for each of the heat transfer enhancement techniques.

Results showed that rim angles lower than 60o gave high absorber tube circumferential temperature differences, higher receiver thermal loss and higher entropy generation rates, especially for flow rates lower than 43 m3/h. The entropy generation rates reduced as the inlet temperature increased, increased as the rim angles reduced and as concentration ratios increased. Existence of an optimal Reynolds number at which entropy generation is a minimum for any given inlet temperature, rim angle and concentration ratio is demonstrated. In addition, for the heat transfer enhancement techniques considered, correlations for the Nusselt number and fluid friction were obtained and presented. With heat transfer enhancement, the thermal efficiency of the receiver increased in the range 5% – 10%,         3% – 8% and 1.2% – 8% with twisted tape inserts, perforated conical inserts and perforated plate inserts respectively. Results also show that with heat transfer enhancement, the absorber tube’s circumferential temperature differences reduce in the range 4% – 68%, 3.4 – 56% and up to 67% with twisted tape inserts, perforated conical inserts and perforated plate inserts respectively. Furthermore, the entropy generation rates were reduced by up to 59%, 45% and 53% with twisted tape inserts, perforated conical inserts and perforated plate inserts respectively. Moreover, using multi-objective optimisation, Pareto optimal solutions were obtained and presented for each heat transfer enhancement technique.

In summary, results from this study demonstrate that for a parabolic trough system, rim angles, concentration ratios, flow rates and inlet temperatures have a strong influence on the thermal and thermodynamic performance of the parabolic trough receiver. The potential for improved receiver thermal and thermodynamic performance with heat transfer enhancement has also been demonstrated. Overall, this study provides useful knowledge for improved design and efficient operation of parabolic trough systems.

Key words: Absorber tube; Computational fluid dynamics; Concentration ratio; Entropy generation rate; Heat transfer enhancement; Monte-Carlo ray tracing; Multi-Objective optimisation; Parabolic trough receiver; Perforated plate inserts; Perforated conical inserts; Rim angle, Twisted tape inserts. 

Supervisors:         Prof. T. Bello-Ochende, Prof. J. P. Meyer.


Published by Barbara Huyssen

Copyright © University of Pretoria 2021. All rights reserved.

FAQ's Email Us Virtual Campus Share Cookie Preferences