Engineering: Theses and Dissertationshttp://hdl.handle.net/10092/8402016-07-01T02:33:24Z2016-07-01T02:33:24ZEvaluation of Warm Mix Asphalt Performance Incorporating High Reclaimed Asphalt Pavement ContentDai, Lu Xuanhttp://hdl.handle.net/10092/124192016-06-29T15:01:02Z2016-01-01T00:00:00ZEvaluation of Warm Mix Asphalt Performance Incorporating High Reclaimed Asphalt Pavement Content
Dai, Lu Xuan
Warm mix asphalt (WMA) has gradually become more popular in the roading industry, because, compared to hot mix asphalt (HMA), WMA can bring numerous benefits, such as lower energy consumption, lower emissions, and greater ability to incorporate a high proportion of reclaimed asphalt pavement (RAP) in the mixtures. Incorporating RAP in WMA can increase the sustainability benefits and enhance the performance of WMA. This study investigated the performance of WMA by adding RAP in different proportions, from 0 up to 70% by mass of WMA. The performance of mixtures was compared with a control HMA. One type of binder, 80/100 penetration grade, and two types of additives were use: a chemical warm mix additive and a rejuvenator, namely, Evotherm and SylvaroadTM RP1000, respectively. Tests were done on the binder’s viscosity and the mechanical performance of mixtures such as moisture resistance, fatigue cracking, and rutting resistance. In this study, the semi-circular bending test was investigated to further study its applicability in asphalt pavement testing. Results from laboratory tests showed that the two additives reduced the viscosity of the binder. Mixtures with the chemical additive (Evotherm) performed better than other mixtures in terms of moisture resistance. Only the WMA mixture with the Sylvaroad rejuvenator showed a higher number of cycles to fatigue failure than the control HMA. For rutting resistance, the increase in RAP proportion greatly improved the performance of WMA mixtures. WMA without RAP had a lower number of cycles to reach maximum rut depth than the HMA. All WMA-RAP mixtures showed considerably better rutting resistance than the HMA. The study of semi-circular bending test showed that the notch depths from 5 to 15 mm are suitable for 100 mm diameter samples. The indirect tensile strengths yielded by the semi-circular test and those from the indirect tensile method could be convertible.
2016-01-01T00:00:00ZEmpirical Bayes estimation for random dot product graph representation of the stochastic blockmodelSuwan, Shakirahttp://hdl.handle.net/10092/124182016-06-29T15:00:55Z2015-01-01T00:00:00ZEmpirical Bayes estimation for random dot product graph representation of the stochastic blockmodel
Suwan, Shakira
Network models are increasingly used to model datasets that involve interacting units, particularly
random graph models where the vertices represent individual entities and the edges represent
the presence or absence of a specified interaction between these entities. Finding inherent
communities in networks (i.e. partitioning vertices with a more similar interaction pattern into
groups) is considered to be a fundamental task in network analysis, which aids in understanding
the structural properties of real-world networks. Despite a large amount of research on this task
since the emergence of graphical representation of relational data, this still remains a challenge.
In particular, within the statistical community, the use of the stochastic blockmodel for this task
is currently of immense interest.
Recent theoretical developments have shown that adjacency spectral embedding of graphs yields
tractable distributional results. Specifically, a random dot product graph formulation of the
stochastic blockmodel provides a mixture of multivariate Gaussians for the asymptotic distribution
of the latent positions estimated by adjacency spectral embedding. The first part of this
thesis seeks to employ this new theory to provide an empirical Bayes model for estimating block
memberships of vertices in a stochastic blockmodel graph. Posterior inference is conducted using
a Metropolis-within-Gibbs algorithm. Performance of the model is illustrated through Monte
Carlo simulation studies and experimental results on a Wikipedia dataset. Results show performance
gains over other alternative models that are considered.
Instead of a complete classification of vertices via community detection, one may wish to discover
whether vertices possess an attribute of interest. Given that this attribute is observed for a few
vertices, the goal is to find other vertices that possess that same attribute. As an example, if a
few employees in a company are known to have committed fraud, how can we identify others who
may be complicit? This is a special case of community detection, known as vertex nomination,
which has recently grown rapidly as a research topic. The second part of this thesis extends
the empirical Bayes model for vertex nomination based on information contained in the graph
structure. This yields promising simulation results as well as real-data results from an Enron
email dataset.
Recent studies have shown that information pertinent to vertex nomination exists not only in
the graph structure but also in the edge attributes (Coppersmith and Priebe, 2012; Suwan et al.,
2015). This motivates the third part of this thesis by further extending the model to exploit
both graph structure and edge attributes for vertex nomination. Simulation studies confirm the
benefit of doing so. However, the same benefit is not observed when the model is applied to the
Enron email dataset; further investigations suggest that this is due to the data violating one of
the model assumptions.
2015-01-01T00:00:00ZEffect of Edge Beam Deformations on the Slab Panel MethodGu, Taidihttp://hdl.handle.net/10092/124002016-06-28T15:01:30Z2016-01-01T00:00:00ZEffect of Edge Beam Deformations on the Slab Panel Method
Gu, Taidi
Previous accidental fires in multi-storey buildings and large scale fire tests have shown
composite floors with unprotected steel beams had large deformations but collapse did not
occur due to a mechanism known as tensile membrane action. The Slab Panel Method (SPM)
was developed as a simplified performance-based design method including tensile membrane
action for everyday design.
As SPM has some provisions of edge beam deformations in its analysis, this research was
carried out to assess how realistic these provisions are. The study was performed on SPM and
VUCLAN, which is a non-linear finite element program that includes both thermal and structural
analysis. The effect of various edge continuity conditions on the basic slab panel was assessed
by VULCAN models and compared with SPM. 9 m x 9 m slab panel were firstly investigated,
then 9 m x 6 m and 9 m x 12 m slab panels were modelled to investigate the effect of aspect
ratio on SPM’s provisions.
The VULCAN analyses showed that the slab panels perform better with more internal edges and
edge beam deformations are always critical to the fire resistance of the slab panels. The SPM
and VULCAN deflections were in good agreement in the 9 m x 9 m slab panel, but it
underestimates the deflections of 9 m × 6 m slab panel and was slightly conservative on the 9 m
× 12 m slab panel.
It is concluded that slab continuity over the edge beams significantly enhanced the fire
resistance of the slab panel. SPM provided good estimates on the fire resistance of slab panel
when the edge beams are designed to be very strong or well protected, but underestimated the
external edge beam deformations. It is recommended that when using the SPM in design, the
external edge beams need to be designed for higher loadings than just from the yield line
tributary area and that for irregular panels or high variable loading Finite Element Analysis (FEA)
software is recommended to check the results.
2016-01-01T00:00:00ZMathematical modelling of self-propelling, coalescing dropletsKhodabocus, M. I.http://hdl.handle.net/10092/123812016-06-27T15:01:22Z2016-01-01T00:00:00ZMathematical modelling of self-propelling, coalescing droplets
Khodabocus, M. I.
When two Newtonian droplets touch, this can lead to either droplets self-propulsion
or coalescence phenomenon. This subject does not appear to have hitherto formed a
prime subject of attention for combined theoretical and numerical treatment, and is
undertaken in this thesis.
The first part analyses the response of an interface subject to a volatile sol-
vent source driven by an air-blow effect. The volatile solvent effect is taken into
account through the full definition of the surface tension gradient and the air-blow
effect through a vapour pressure gradient model equation. Employing the long-wave
approximation of the coupled Navier-Stokes and advection-difusion equations, the
mathematical description reduces to a degenerate fourth order nonlinear parabolic
h-evolution equation coupled with one Poisson equation and one non-homogeneous
Lagrangian derivative equation. Computing these equations using the COMSOL Mul-
tiphysics software, the results are presented and contrasted with those which would
present themselves had the surface tension gradient been expressed in truncated form
and the vapour pressure gradient disregarded.
The second part treats of the self-propulsion of a miscible bi-droplet system in a
capillary tube. The mathematical framework consists of the two-phase
flow, phase field equation set, an advection-diffusion chemical concentration equation, and clo-
sure relationships relating the surface tension to the chemical concentration. The
numerical experiments are carried out using the COMSOL Multiphysics software.
The dynamical response of the bi-droplet reveals a rectilinear motion of the sys-
tem at early-times and an exponential at late-times. A parametric study shows that
the motion obeys Poiseuille flow at early-times. The results are compared success-
fully with available experimental data, thereby establishing a general mathematical
description of the phenomenon.
The third part proposes a versatile framework to study droplets coalescence phe-
nomenon in an unconfined environment. The framework uses the laminar two-phase flow
moving mesh method coupled with an advection-diffusion equation, and is con-
structed in such a way that upon variations of a single parameter, the computational
domain yields geometries ranging from a mono to a bi-droplet system. Taking advan-
tage of its geometrical properties, a theory is developed which establishes a generic
equation describing the growth of that highly curved meniscus neck in the power-law
regimes. Using the COMSOL Multiphysics software, the model is tested by proving
the leading order laws numerically and illustrating the corresponding coalescence flows.
Finally, the thesis discusses how more complex situations can be derived out of
those four parts for further scientific exploration.
2016-01-01T00:00:00Z