Chapter 2

PROBLEM STATEMENT

The phenomenology that occurs in the PEM membranes constitutes an important factor in
the operation of the fuel cells and in view of the growth that has had this technology and
the great amount of commercial uses that it is given to it, to investigate in the improvement
of new materials And improving the components of the fuel cells, it plays an important role
in [22], shows that the new materials are the key of the fundamental advances in this type of
technology, in addition this brief review demonstrates that the change of materials in bulk
or at nanoscale can change signicantly the properties of electrodes and electrolytes. While
[23] describes the design and manufacturing alternatives for PEM fuel cells in the context
of vehicular applications, it also seeks to facilitate the selection of materials and processes
through the consideration of the design and manufacturing variety before the investment
of capital for large-scale production. It is for this reason that this research work becomes
more important as it will allow us to optimize the operation, improve the characteristics and
properties of the materials and extend the understanding of the PEM and its eect on fuel
cells.

In this work we are interested in the solution and simulation of the Stokes ow problem
between the interface of two nano channels of a proton exchange membrane. In previous
research, it has been found that when connecting two PEMs and their nanoscopic pores,
Resistance at the interface cause it is still unexplained to see [5]. The use of this type of fuel
cell has grown transcendentally in recent decades, being used mainly in small electronic and
automotive devices, thus the development of a numerical method to simulate these situations
becomes more relevant.

The model to be studied includes the Stokes uid equations, the Poisson equation for the
electric potential and the Nernst-Planck equation for the concentration of protons, which are
Found in a system of nonlinear and coupled partial dierential equations. The solution of
this model will be achieved by the new meshless method known as MAPS. The solution of
this model is a great contribution to the eld of computational analysis in phenomena related
to fuel cells, improving the knowledge of the materials of the PEM and its design.

Therefore, we formulate the following hypothesis: Is it possible to develop a meshless nu-

merical method based on radial basis functions with individual approximate solutions in two
dimensions and using it to solve the coupled system of partial dierential equations modeling
CHAPTER 2. PROBLEM STATEMENT 5

ow problem Stoke on and between the interface of two nano channels of a proton exchange
membrane (PEM) of a fuel cell?

The meshless numerical method of radial basis functions with two-dimensional approximate
particular solutions can be used to solve the coupled system of non-linear partial dierential
equations, which models the Stokes ow problem between the interface of two nano channels
of a proton exchange membrane (PEM) of a fuel cell.

Next, we describe the set of dierential equations that govern the phenomena related to fuel
cells.

2.1 Governing equations

The work presented by [5], is a reference conguration. We consider the case of two straight
cylindrical channel segments with the same radius and parallel axes. These channels are
moved by a set of x-coordinates between axes (see gure 2.1). The oset is the distance
between the center of the circle in the plane interface. The reference pressure p̄ in solid
corresponding to the situation of two channels with a displacement, is varying in space (in
the interface). That is, having p̄ = pR = p∗ − 2γd
in Part I that belongs to the body of the
cylinder, while p̄ = p in Part of I near the part belonging to the covered cylinder.
∗

Figure 2.1: Domains: L, the liquid, S the solid, I the interface, O1 inlet (anode), O2 outlet
(cathode). Unknowns: in L: u, p, φ, c; in S : U . Dashed part of the interface: initial pore
shape, part of the interface with continuous lines: equilibrium pore shape ([5]).

Let C = L ∪ S ∪ I denote the whole domain with the open domains L and S . I = ∂L ∩ ∂S
denotes the interior free boundaries (or interface) between S and L and ∂C are all outer
boundaries of C . We denote the inlet by O1 and the outlet by O2 . ν denotes the outer nor-
mal on I ∪ ∂C , pointing on ∂C outside C and pointing on I \ ∂C always form the liquid into
the solid. Consequently ∂n f = ∇f · n denotes the derivative in direction of the outer normal
for a dierentiable function f . τ1 and τ2 denote the two orthonormal tangential vectors on I .
CHAPTER 2. PROBLEM STATEMENT 6

Table 2.1: General problem data

PARAMETER Symbol Unit
Temperature T K
The proton concentration →
−c mol/m3
The velocity eld →
−u m/sg
(i)
The outer pressure at in-/outlet (i=1,2) p0 N/m2
Typical pressure solid p̄ N/m2
Orthonormal tangential vectors on I (i=1,2) →
−τi
The pressure →
−p N/m2
The dynamic viscosity µdyn
¯ N s/m2
The outer normal on I ∪ ∂C ν
The Cauchy stress tensor σS (∇U )
The surface tension γ N/m
Surface charge density σc C/m2
The electric potential in the liquid L φ V
Electric bulk permitivity r I

Diameter of a channel d m
Length of 2 channels l = 10d m

The problem to be studied is made up of the following equations:

The ow is modeled by the Stokes equation for u and p.

 
T
−∇ · (µdyn ∇u + (∇u) ) +∇p = −F c∇φ, in L (2.1)
∇ · u = 0, in L (2.2)

Where F the Faraday constant. Modeling conservation of momentum and incompressibility

of the uid, this is complemented by the boundary conditions (b.c.) on in-/outlet and on the
interface.

− µdyn (∇u + (∇u)T ) ν = pi0 ν, on Oi , i, j = 1, 2 (2.3)

u · τj = 0 on Oi , i, j = 1, 2 (2.4)
u = ∂t U, on I (2.5)

The electric potential φ obeys the Poisson equation

− 0 ∇ · (r ∇φ) = F c, in L (2.6)

completed with Neumann boundary conditions

−∇φ · ν = (−1)i Eext , on Oi , i = 1, 2 (2.7)

0 r ∇φ · ν = σc , on I. (2.8)
CHAPTER 2. PROBLEM STATEMENT 7

Where 0 is the vacuum permittivity. The proton concentration c fulls the Nernst-Planck
(NP) equation
 F 
− ∇ · j+ = ∇· D∇c + D c∇φ = 0, in I (2.9)
RT
with homogeneous Neumann boundary conditions for the chemical potetial RT ln (c/c̄) + F φ,
R denoting the universal gas constant and T the temperature.
F
∂ν c + (−1)i+1 Eext c = 0, on Oi , i = 1, 2 (2.10)
RT
F
∂ν c + σc c = 0, on I (2.11)
RT 0 r
where the ux of protons j+ is
D
j+ = − c∇µ + cu
RT
Where D is the Diusivity of H + in H2 O.

In the literature seek the stationary solution of the equation system discussed above. There-
fore, they are neglecting the time derivatives thereof.
Chapter 3
STATE OF THE ART

The state of art is divided into two main topic to be addressed. in the rst place the modeling
and simulation of the proton exchange membranes in fuel cells, are addressed by dierent
investigations carried out in reference to the membranes. Secondly, models of the transport of
protons in nano channels of the proton exchange membranes are presented and the solutions
of dierent authors.

Fuel cell modeling is very useful for fuel cell developers, as it can lead to improvements in
their design. In order to improve the operation and operation of the fuel cell, studies and
research must be directed to the design, materials and optimization of its processes. There
are many published models for PEM fuel cells in the literature, so it is necessary to begin
to understand the complexity of current models see [24], [25]. According to [19] multi-scale
models claim too much precision compared to those that focus on a specic component. In
addition, they assert that the "Models must be provided along with algorithms that explain
how to use them in the design of materials, manufacture and characterization of properties,
as well as assembly, testing and optimization of fuel cells", see also [20].

3.1 Modeling and simulation of proton exchange mem-

branes in fuel cells

Dierent types of approaches have been used for the modeling and simulating of PEM mem-
branes, it is the case of the work done by [26] where by a technique to technique is proposed
to identify the parameters of the proton exchange membrane fuel cell (PEMFC) using an
optimization approach, an algorithm based on particle swarm optimization (PSO), known
as local optima avoidance particle swarm optimization comparing their results with experi-
mental data. Achieving almost identical results to the results obtained from experiments. In
another research approach performed by [27], there is applied a combined model approach
using Monte Carlo kinetic simulations and molecular dynamics to investigate both chemical
degradation and mechanical deformation of Naon membranes at a molecular level. They
concluded that the Naon membrane has a the same direction of linear deformation, leav-
ing thin brils of a few Naon strands deployed between the surfaces of craze and that its
modeling approach demonstrates a promising way to follow exploring changing the physical
CHAPTER 3. STATE OF THE ART 9

and chemical properties of Naon membranes during the operation of fuel cells. In [28],
they investigate the membranes through a study of molecular dynamics, looking at proton
transport and local structures. They discover that the sizes of the water groups are larger in
the Nation at the highest hydration levels. Because the water is more widespread and there
is less water in the rst layer of solvation around the sulfonate in the nation, the amount
of hectares hydrologically around a sulfonate is higher. They present results from several
computational papers that use reactive molecular dynamics, which explicitly describes bond
breaking and formation, to study proton transport in several polymers commonly used in
PEMs. They found that increased side chain length and actual exibility proton transport.
We can not fail to propose the primary mechanism for proton transport in PEMs, which
are passed between sulfonate groups, rather than diusing in a bulk-water region like the
center of a water pore. This proposed mechanism is further strengthened by the observation
that protons spend very little time away from sulfonate groups, so that free diusion can
not have a substantial contribution to the overall transport. Similarly [29], They present
the results of several computational documents that use reactive molecular dynamics, which
explicitly describe the rupture of bonds and formation, to study the transport of protons
in several polymers commonly used in PEM. They discovered that increasing the length of
the side chain and exibility actually improves proton transport. Then, we could propose
that the primary mechanism for proton transport in the PEMs involves passing between the
sulfonate groups, rather than diusing freely in a region of bulk water, such as the center of
a water pore. This proposed mechanism is reinforced by the observation that protons spend
very little time away from sulfonate groups, so that free diusion can not make a substantial
contribution to general transport. In the research carried out by [30], the eects of polymer
morphology on solvation and transport of protons on a hydrated Naon membrane using a
reactive molecular dynamics approach and three types of models is studied by [17]. They
stated that protons are transported mainly along the surfaces of the sulphonate group and a
larger radius leads to a faster diusion of the proton, similar procedures and approaches were
obtained [31], [32], [33] y [34]. So similar in [35] which used an approach for the modeling of
the conduction of protons through PEM of Naon to the mesoscale with a unique combina-
tion of molecular dynamics of coarse grain. The comparison of their experiments was found
that in both cases the inclusion of local electrostatic within the membrane plays an important
role in aecting the conductivity as a function of the hydration behavior of the membrane.
See also [36], who discuss the recent modeling of PFSA proton exchange membranes, with
special attention to the proper treatment of the proton transport mechanism at the angstrom
level, modeling the morphology of the mesoscale hydrated membrane, and multiscale emerg-
ing methods. With their work they establish the bases to perform a multiscale modeling of
the PFSA membranes with some condence in their predictive capacity for the membrane
design eorts, see also [6].

3.2 Modeling of proton transport in nano channels of

proton exchange membranes

In this paper we focus on electrodynamic models in order to study Proton ow problem in
the membrane of a fuel cell. According to literature, there are very few authors, who have
been interested in studying this type of phenomenology, ow of cannulated uid in the PEM
CHAPTER 3. STATE OF THE ART 10

membrane. In the research conducted by Schmidt and Chen [1], they states that a Naon
membrane is composed mainly of parallel cylindrical channels surrounded by hydrophilic side
chains; they also explain important features of Naon, including diusion of water and pro-
tons through Naon and its persistence at low temperatures.

Peighambardoust et al [37], show a detailed review of the evolution in research done on

proton exchange membranes and the importance they have for the operation of a fuel cell;
In [38], They consider that high-temperature proton exchange membrane fuel cells (HT-
PEMFC) oer several advantages, such as high proton conductivity, low fuel permeability,
low electroosmotic drag coecient, good chemical / thermal stability, good mechanical prop-
erties and low cost, although it concludes that the operation at high temperature often leads
to dehydration, which deteriorates the performance of the membrane, consequently, several
approaches have been adopted to retain water at high temperatures such as: inorganic addi-
tives, solvents not aqueous and low volatile and utilization of a solid state proton conductor.
Similarly, Dupuis [40] shows the dierent materials that are currently believed to be potential
membrane materials, see also [39], [42] and [41]
It should be noted that the protons in a fuel cell are transported through a PEM membrane
which is composed of hydrophilic elastic pores lled with water. There are several works at
dierent scales that are interested in studying the charged ow or the passage of these protons
and water through these pores. [3] Berg and Ladipo [3] nd the exact solution of an electro-
osmotic ow problem in a cylindrical channel of polymeric electrolytic membranes, and they
present it as a solution of the Poisson-Boltzmann equation, although it was previously solved
by [10], however Tsao within its approach do not take into account the mechanical displace-
ment of the membrane . On the other hand [2] studies a mesoscopic model that describes
the velocity of the dynamic uid based on a Navier-Stokes-Nernst-Planck-Poisson coupled
system. The authors states that transport properties of charged particles are aected to
some extent by the charge at the liquid solid interface and that everything depends on an
externally applied electric eld.

One of the most interesting studies in the investigation of the phenomenology of the transport
of charged uid is the one proposed by [5] who in their work propose a hydrodynamic model
that allows them to describe the ow of charged uid and change the pore morphology due to
mechanical deformations, solve a set of numerically coupled nonlinear equations and examine
the shape of the pores of equilibrium, their model suggests that this interface resistance
depends mainly on two factors: (a) the oset value (the distance between the pore centres
at the intersection plane) and its stochastic distribution and (b) on the deformed pore shape
due to the balance of elastic and electrohydrodynamic forces. In the work done by [7],
the equilibrium form of the interface representing the lateral boundary of a channel of the
pore embedded in an elastomer is studied. Their model includes a system of PDEs, which
comprises a linear equation of elasticity for displacements within the elastomer and a non-
linear Poisson equation for the electric potential inside the channel (lled with protons and
water). From the thermodynamic models of [5], [8] they state that an ohmic resistance
is observed between the interfaces of two electrolyte membranes of polymer joint, without a
theoretical explanation. They propose a new model for The uid ow charged in and between
the nanocanals and investigated numerically, solve it for velocity, pressure, electric potential
and proton concentration, as in [5], they concluded that the interfacial resistance between
CHAPTER 3. STATE OF THE ART 11

two circular cylindrical pores, and therefore, two PEM, depends mainly on two factors: the
distance between the pore centers in the intersection plane and the deformed pore shape
due to the balance of elastic and electrohydrodynamic forces. Finally, the work of [9] by a
continuous decoupled model study the distribution and ow of protons within a cylindrical
nanopore lled with water of a PEM, in which the Poisson-Nernst-Planck equations are
modied. They conclude that for the conductivity the most realistic values were obtained
at large pore radii, while for the waterdrag it compares better the experimental data with
small radii. Small pores can dominate water entrainment, while larger pores can determine
conductivity.

3.3 RBF meshless collocation method

Radial Basis Functions (RBFs) have been widely used in global and continuous interpolation
of scattered data sets. This interpolation strategy has become the foundation of the col-
location meshless methods to solve PDEs through the development of the Kansa's method
[60]. Kansa used the Multiquadric (MQ) function to obtain an accurate meshless solution
to the advection-diusion and Poisson equations without employing any special treatment
for the advection term (upwinding), due to the high order of the resultant scheme and the
intrinsic relationship between governing equations and the interpolation. This strategy in-
volved all the scattered nodes that cover the domain and therefore it produced a global fully
populated matrix. Although the Kansa's method was used by many authors to solve the
Laplace, the Poisson, the Helmotz and Parabolic equations showing better accuracy com-
pared to traditional methods ([63], [64], [61], [62]), Kansa and Hon [44] concluded that the
matrix ill-conditionality worsens as the number of nodes increases and, in the case of using
MQ function, a singular matrix is obtained for some values of the shape parameter. They
presented three strategies to overcome the ill-conditioned matrix which were block partition-
ing, truncating radial basis functions and multizone strategy.

In [51], it is proposed the method of approximate particular solutions (MAPS) to solving

partial dierential equations. Then the MAPS has been successfully applied to various phys-
ical and engineering problems, such as anisotropic problems [52], nonlinear Poisson problems
[53], wave problems [54], elasticity problems [55], Stokes ow problems [12] and [16], and
convection-diusion problems [56] and [15], (see also [57],[58] and [11]).

Chen et al. [59] present an alternative formulation of the MAPS to solve elliptic linear PDEs
with variable coecients. The PDE is rewritten as a Poisson equation in which all of equation
terms dierent of the laplacian term are located at the right side. Unlike the rst case with
constant coecients, the elliptic equation contains derivatives of rst order of u and they are
approximated by direct derivation of the linear combination of the particular solutions. The
MAPS, however, outperforms the method of Kansa in the evaluation of the derivatives when
MQ RBFs are used in both methods, some other authors have used the method proposed by
[59], were [11], [13] and [14].
CHAPTER 3. STATE OF THE ART 12

3.3.1 EDP solution through MAPS

This section aims to show the work carried out to solve EDPs of vector variables through
the Approximate particular solutions method and to show the recent interest of the scien-
tic community to apply this method for the solution of problems. For example [51], they
propose an approach using the method of particular solutions to solve linear inhomogeneous
dierential equations without the need to nd a homogeneous solution. Which results in a
much simpler numerical scheme with precision similar to the traditional approach.

On the other hand Zhu [52], who studies MAPS to solve anisotropic problems of elliptical
type, introducing the design of an anisotropic radial base function. They present results for
a two-dimensional anisotropic diusion problem, stating that this method allows the choice
of location points regardless of the magnitude of the anisotropy. In [66], they develop the
MAPS using trigonometric functions as base functions, which is proposed to solve elliptical
EDPs in two dimensions, concludes that in comparison with the RBF used by MPS, Cheby-
shev polynomials and high order polynomials, The proposed MPS based on trigonometric
functions can not choose the shape parameter in the RBFs, locating the placement points in
some special positions and dealing with highly conditioned matrices. On the other hand, in
[67] a free mesh method with radial basis functions is developed to approximate the eigen-
values of the Stokes equations in primitive variables in a square domain, which arm that in
comparison with nite element methods , the method used results are more accurate without
any mesh generation. Similarly [68], they develop a meshless method of particular and fun-
damental solutions (MPS-MFS) which is implemented to solve time-dependent Navier-Stokes
equations. They found that the meshless method based on MFS-MPS is simple and eective,
and easily applied to the non-linear system of time-dependent equations, see also ([11],[12]).
Finally, in [69] they use the localized method of approximate particular solutions (LMAPS)
to solve the systems of two-dimensional transient Navier-Stokes equations. They conclude
that the numerical experiments show that the newly developed LMAPS is suitable for solving
incompressible Navier-Stokes equations with high precision and eciency.
Chapter 4
THEORETICAL BACKGROUND

The theoretical foundations of the approximate particular solutions (MAPS) method are
presented below. The method was developed by [51] to solve two-dimensional linear problems.
The objective of the procedure is to nd the coecients that t the approximate function
given to the given information, that is, the RBFs are mainly multivariable functions and
their values depend only on the distance from the origin so φ(x) = φ(r)R, xR, rR; or
alternatively on the distance from a point of a given set ξj , and φ(x − ξj ) = φ(rj )R. Then,
given a function f (x) can be approximated by a linear combination of RBFs, namely.
N
f (x) ≈ αφ(kx − ξj k2 ), xΩ
X
(4.1)
j=1

For review of the traditional RBFs see [43]. The RBFs are used in the MAPS to dene the
particular solutions integrating the auxiliary equations. Unlike the Kansa method, RBFs
are used indirectly in MAPS. When solving Stokes equations, No pressure-velocity coupling
strategies are required because the velocity satises the continuity equation and the partic-
ular pressure solution lies directly from the expression obtained from the particular velocity
solution.

4.1 Boundary Value Problem Solution by RBF Collo-

cation Methods

The PDE solution is one of the most common and recently studied applications of interpo-
lation of RBFs in the eld of engineering and mathematical applications [45]. The Kansa
method is the basis of the meshless collocation methods for PDE solution [44].

A problem of value boundary linear and non-homogeneous is dened by equations (4.2) and
(4.3), where L and B are linear partial dierential operators which are applied in the problem
domain Ω and its boundary ∂Ω, respectively, and φ is the dependent variable. In the method
of placement without mesh with RBFs are two generic procedures to obtain the solution:
The method of Kansa and the symmetrical method see [46].
CHAPTER 4. THEORETICAL BACKGROUND 14

L[φ(~x)] = f (~x) (4.2)

B[φ(~x)] = g(~x) (4.3)
Both procedures require a set of N nodes located in the domain Ω and its boundary ∂Ω
known as test points ξj .Also a set of locations xi located as set points, usually coincides with
the test points.

From the set of N scattered points, n are placed at the border and N − n at the (internal)
domain. Therefore, it is necessary to know the value of B[φ] in n locations and the value of
L[φ] in N − n as the data needed to achieve a solution, see [47].

4.2 Method of Aproxímate Particular Solutions

Recently, [51] proposed an integrated RBF method for the case of a linear PDE, which
dierential operator L(·(~x)), or only part of it, can be written in terms of the radial component
of a polar or spherical coordinate systems, as:

L(u(~x)) = L1r (u(r)) + L2~x (u(~x)) = f (~x) (4.4)

In this case, these authors proposed to approximate the radial component of the PDE, i.e.
1r (·(r)), in terms of a RBF interpolation and by integrating the resulting non-homogeneous
ordinary dierential equation, it is possible to obtain an approximate representation of the
eld variable by a superposition of the corresponding particular solution. This is similar to the
integrated radial basis function approach ([48] indirect scheme), but instead of approximating
the higher-order derivatives in terms of a RBF and integrating them in each direction along
a Cartesian grid to obtain an approximation of the eld variable, the approximated eld
variable is found in terms of the complete integration of the operator L1r independently of
any Cartesian grid. In this way, the following approximation is dened:
N
X
L1r = αk φ(rk ) (4.5)
k=1

with corresponding eld variable as:

N
X
u(~x) = αk û(rk ) (4.6)
k=1

where the particular solution û(r) is given by the solution of 4.7, in terms of the RBF non-
homogeneous term φ(r),

L1r (û(r)) = φ(r) (4.7)

By substituting the obtained approximation of u(~x) into the full expression of L(u(~x)) and
into the boundary conditions of the problem, taking into consideration (2), a meshless inte-
grated RBF solution of the boundary value problem is obtained, which is free of the denition
CHAPTER 4. THEORETICAL BACKGROUND 15

of a direct or auxiliary Cartesian grid. This integrated RBF meshless approach has been re-
ferred by [51] as the Method of Approximate Particular Solutions (MAPS)
In the case of a linear boundary value problem 4.2 and 4.3, the implementation of the MAPS
reduces to the solution of the following linear system of algebraic equations:

B[û(~x1 , ξ~1 )] . . . B[û(~x1 , ξ~N )]

    
α1 g(x~1 )
 .. . . ..   ..   ..
 . . .  .   .


 B[û(~xN , ξ~1 )] ~
 
. . . B[û(~xNb , ξN )]   αNb   g(x~Nb ) 
   
 b
~ =
Nb +1 − ξ1 |) . . . φ(|~xNb +1 − ξ~N |)  αNb +1   f (xN~b +1 ) 
 φ(|~x  

 .. . . ..   ..   ...
    
 . . .  . 
φ(|~xN − ξ~1 |) . . . φ(|~xN − ξ~N |) αN f (x~N )
for Nb boundary points and Ni internal points, with N = Nb +Ni (see gure 4.1). The solution
of the above boundary value problem is achieved after solving the resulting algebraic system
for the coecients α.

Figure 4.1: Nodal distribution in the solution domain: internal (·) and boundary (◦) points.

The particular solution û of (4.7), has been used before in connection with the numerical
solution of a linear boundary value problem using the classical decomposition in terms of its
particular and homogeneous solutions, combined with Method of Fundamental solution (see
[49]) and with the Boundary Element method (see [50]) to nd the corresponding homoge-
neous solution.

4.3 Global Stokes MAPS

Bustamante et al. [55] use the procedure of the MAPS, as was presented before, to solve
elasticity problems modeled by the linear equations of Navier. A linear combination of the
displacement particular solutions is employed to approximate the variables of the problem,
i.e., the displacement components in this case. The displacement particular solutions solve a
non-homogeneous Navier's equations in which the MQ RBF is used as source term. In order
CHAPTER 4. THEORETICAL BACKGROUND 16

to obtain a closed expression, the displacement particular solutions are expressed in terms
of the Galerkin vector and this in turns is expressed in terms of an auxiliar potential. This
specic procedure to obtain the particular solutions diers of the procedure explained before
due to the vectorial character of the two-dimensional problems considered in [55] These ideas
are used by the authors in later papers to solve linear and nonlinear uid dynamics problems
so the procedure for the obtainment of the particular solutions is detailed below for the
Navier-Stokes equations case.
The global Stokes MAPS is proposed in [12] and is used to solve two-dimensional Stokes and
Navier-Stokes problems in [12] and [11], respectively. The procedure proposed by Chen et al.
[51] is used in the global Stokes MAPS but the particular solutions solve a Stokes system of
equations. In addition, the particular solution denition, in terms of an auxiliary potential,
leads to satisfy the continuity equation automatically.
Consider an arbitrary two-dimensional region Ω with boundary Γ. The steady-state Navier-
Stokes equations in its primitive variable formulation is given by
∂uj (~x)
= 0, (4.8)
∂xj
∂ui (~x) ∂ 2 ui (~x) ∂p(~x)
ρuj (~x) =µ − , (4.9)
∂xj ∂xj ∂xj ∂xi
where ~x = (x1 , x2 ) ∈ Ω, ρ is the density and µ the dynamic viscosity. ~u = (u1 , u2 ) and p are
the velocity and pressure elds, respectively. The boundary conditions for ~x ∈ Γ are dened
in terms of given values of the velocity and/or surface traction, i.e.,

ui (~x) = uib , for ~x ∈ Γu , (4.10)

 
∂ui (~x) ∂uj (~x)
τi (~x) = −p(~x)ni + µ + nj = τib , for ~x ∈ Γτ , (4.11)
∂xj ∂xi
where Γu and Γτ are the subsets of Γ in which velocity and surface traction conditions,
respectively, are imposed.
The global Stokes MAPS approximates the velocity and pressure elds as a linear combination
of a set of functions u
bli and pbl , i, l = 1, 2,

N h
X i N
X
ui (~x) := αk1 b1i (~x, ξ~k )
u + αk2 b2i (~x, ξ~k )
u = bli (~x, ξ~k ),
αkl u
k=1 k=1
N h N
(4.12)
X i X
p(~x) := αk1 pb1 (~x, ξ~k ) + αk2 pb2 (~x, ξ~k ) = αkl pbl (~x, ξ~k )
k=1 k=1

which depends on the Euclidean distance, rk , between the eld point ~x and the test point ξ~k .
The equations in (4.12) can be written in vector notation as:
 1 2 T [α1 ] T [α1 ]
and p = pb1 , pb2 (4.13)

bi , u
ui = u .
[α2 ] [α2 ]
bi

In the present formulation of the MAPS, the ow elds (u

bli , pbl ) are particular solutions of the
Stokes problem (4.14)-(4.15) where the non-homogeneous term is dened as the MQ-RBF,
CHAPTER 4. THEORETICAL BACKGROUND 17
√
φ(r) = ~.
r2 + c2 , with r = |~x − ξ|
~
ulj (~x, ξ)
∂b
= 0, (4.14)
∂xj
∂ 2u ~
bli (~x, ξ) ~
∂ pbl (~x, ξ)
− = φ(r)δil . (4.15)
∂xj ∂xj ∂xi
bli and pbl are referred as Stokes' particular
For clarity in the presentation, the eld variables u
velocities and pressures, respectively.
In order to obtain a closed expression for the particular solutions, the Oseen's decomposi-
tion formula (see [65]) is used to nd the corresponding fundamental solution of the Stokes
equations. First of all, we dene the particular velocities in terms of an auxiliary potential
ψ , as
2 2
~ = ∂ ψ(r) δil − ∂ ψ(r) .
bli (~x, ξ)
u (4.16)
∂xm ∂xm ∂xi ∂xl
The continuity equation (4.8) is satised as consequence of the last denition for the particular
solutions, since:

" N # N  2
∂ 2 ψ(rk )

∂uj ∂ X l l X
l ∂ ∂ ψ(rk )
= α u (rk ) = αk δjl −
∂xj ∂xj k=1 k j k=1
∂x j ∂xm ∂xm ∂xj ∂xl
N N
∂ 3 ψ(rk ) ∂ 3 ψ(rk ) ∂ 3 ψ(rk ) ∂ 3 ψ(rk )
X   X  
l l
= αk δjl − = αk −
k=1
∂x j ∂xm ∂xm ∂x j ∂xj ∂xl
k=1
∂x l ∂xm ∂xm ∂xl ∂xj ∂xj
= 0.

By substituting (4.16) into (4.15), expressions for the auxiliary potential in terms of the MQ
RBF and for the particular pressures as a function of the auxiliary potential are obtained in
(4.17) and (4.18), respectively.

∂ 4 ψ(r)
µ = φ(r) (4.17)
∂xm ∂xm ∂xk ∂xk

∂ 4 ψ(r) ~
∂ pbl (~x, ξ)
µ + = 0. (4.18)
∂xm ∂xm ∂xi ∂xl ∂xi
By direct integration of the rst of these two previous equations, the following expression for
ψ in terms of r is obtained

 
1 1 4 2 2 4

3 2 1 2 2

3
ψ(r) = φ(r) 4r + 48r c − 61c − c ln(c)r − 5r − 2c c ln (c + φ(r)) .
12µ 75 5
(4.19)
Substituting (4.19) into (4.16) and (4.18), expressions for u
bli and pbl , in terms of r are obtained
to be used in (4.12) to approximate the ow eld (~u, p). Replacing expressions (4.12) into
the momentum equations (4.9) leads to the following nonlinear equation in terms of the
particular solutions
CHAPTER 4. THEORETICAL BACKGROUND 18

N N
ul (rk )
 2 l
bi (rk ) ∂ pbl (rk )

X ∂b X ∂ u
ρuj (~x) αkl i − αkl µ − = 0. (4.20)
k=1
∂xj k=1
∂xj ∂xj ∂xi
From direct substitution of (4.15) in the previous equation and using matrix notation, the
following momentum equation is attained

u1i (r1 ) u1i (rN ) ∂bu2i (r1 ) u2i (rN ) [α1 ]

 
∂b ∂b ∂b
ρuj (~x) ,··· , , ,··· , −
∂xj ∂xj ∂xj ∂xj [α2 ]
[α1 ]
[φ(r1 )δi1 , · · · , φ(rN )δi1 , φ(r1 )δi2 , · · · , φ(rN )δi2 ] 2 = 0,
[α ]

or using a compact notation, it can be written as

T !
u2
u1i ∂b

∂b [α1 ]
ρ uj (~x) , i − [φδi1 , φδi2 ]T = 0. (4.21)
∂xj ∂xj [α2 ]
Similarly the surface traction boundary conditions can be expressed in terms of the coecients
αkl as:
" T  1 T ! # 1
u2j
 1
 1 2 T ∂b
ui ∂bu2i ∂b
uj ∂b [α ]
− pb , pb ni + , + , nj 2 = τi b. (4.22)
∂xj ∂xj ∂xi ∂xi [α ]

Collocating the boundary conditions (4.10) at those nodal points in Γu , the boundary con-
ditions (4.22) at those points in Γτ and the momentum equation (4.21) at the Ni internal
nodes, a system of nonlinear algebraic equations is obtained, that can be solved with an
appropriate scheme.
Chapter 5
OBJETIVES AND METHODOLOGY

After analyzing the current state of research in the eld of numerical simulation, the following
objectives for the thesis proposal and the proposed means for achieving these objectives are
presented.

5.1 Ob jetives

5.1.1 Main objetive

The overall objective of this work is:
To develop a meshless numerical method of radial basis functions in two-dimensional with
approximate particular solutions to solve the Stokes ow problem in and between the interface
of two nanocanales of a proton exchange membrane (PEM) in a fuel cell. Allowing the
identication and characterization of the factors that generate the electro-ohmic resistance
in the transport of protons in a nanochannel

5.1.2 Specic objectives

For the development of the overall objective the following specic objectives are proposed:

• Study the characteristics of the numerical method meshless of radial basis functions
with approximate particular solutions.

• Solve the linear and non-linear scalar problems for the Poisson equations in the electric
potential and Nernst-Planck equations in the proton concentration of the proposed
model.

• Solve the coupled Stokes ow problem: Stokes ow equations, electric potential (Poisson
equation) and proton concentration (Nernst-Planck equation).

• Validate the results obtained and the eciency of the numerical algorithm without
mesh of radial basis functions with approximate particular solutions identifying the
eects generated by the electro-ohmic resistance in the transport of protons in a PEM
membrane.
CHAPTER 5. OBJETIVES AND METHODOLOGY 20

5.2 Methodology

For the development of this work several main stages have been arranged: rst a study
and understanding of the mathematical and numerical behavior of the model and the exist-
ing formulation of the system, through a review of the existing literature; second modeling
and simulation of the uid ow charged in and between the interface of two Cylindrical
nanochannels,working through a two-dimensional geometry and third, the validation of the
results obtained and the eciency of the method, this with the purpose of identifying and
characterizing the factors that generate resistance in the transport of protons. The proposed
method is implemented in computer routines using the Visual Studio editor and the FOR-
TRAN compiler.

In the rst stage we intend to make an analysis of the dierential equations that make up
the model to be solved, to understand the inuences of each one of the variables in the model
and to carry out a study of the solutions with similar methods that have had the equations
In addition, an analysis of the characteristics of the numerical method without mesh of pro-
posed radial basis functions is carried out.

In the second stage the MAPS will be applied to solve the linear and nonlinear scalar prob-
lems for the Poisson equations in the electric potential and Nernst-Planck equations in the
proton concentration of the proposed model, MAPS will be applied in the Poisson equations,
Stokes ow to make generic codes, which allow to analyze the behavior of the method. After
selecting the mathematical model to be solved it is necessary to apply the method for the
solution of the model of the problem to obtain rst generic results of linear and nonlinear
Fortran 90 codes of MAPS to solve the equations of uid ow loaded in and between two
Cylindrical nanochannels. The strategy consists of implementing in MAPS the Stokes ow
equations, the Poisson equation and the Nernst-Planck equation and selecting a coupling
algorithm from of the literature. Once programmed; the general code validation and veri-
cation of the procedures will be done with the experimental results, numerical and analytical
found in the literature.

Finally, athe factors and characteristic of the resistance that is generated in the transport of
ow of loaded uid are identied in the nanochannels of the PEM membrane. In this stage
the approximate particular solutions method will be applied even to systems of coupled
dierential equations, MAPS will be applied to each of the equations involved in the model
and then will be coupled as proposed in the literature see [5]. Once the code is working
correctly, and the results have been validated, the physical behavior of the nanopore will be
analyzed in dierent contexts. It is worth noting that the development of the codes has to be
complemented by the selection of an ecient and accurate solution procedure of the system
of linear and non-linear equations. The resources that will be used are personal computer,
FORTRAN compiler and preprocessor and postprocessor software.

5.3 Schedule

For the development of activities of this research is to be carried out as follows:

CHAPTER 5. OBJETIVES AND METHODOLOGY 21

Table 5.1: Schedule

ACTIVITIES ACTIVITIES PER SEMESTER
S1 S2 S3 S4 S5 S6 S7 S8 S9
Review of the literature
Development of the proposal
Review of the mathematical model
Initial selection of parameters
Characteristics of the numerical method
Denition of the mathematical model
Solution of the Stokes ow problem
Stokes ow solution by MAPS
Code Implementation
Code validation
Application problem
Solution of the charged uid ow
Development of the formulation by MAPS
Code Implementation
Code validation
Application problem
Solution of the system of coupled equations
and validation of the obtained results
Code Implementation
Code validation
Application problem
Written report
Writing articles for journals
CHAPTER 5. OBJETIVES AND METHODOLOGY 22

5.4 Scope, impact and contribution

The realization of this research generates great interest in the extent to which society seeks the
generation of energy through the non-conventional exploitation of hydrocarbons as sources of
energy. This work intends on the one hand to propose new methods to solve the dierential
equations in the models made to analize the fuel cells, with the study of the proton exchange
membrane (PEM) will allow to optimize the performance, improve the characteristics And
properties of the materials and extend the understanding of the PEM and its eect on
fuel cells. On the other hand it will be of great interest for the scientic community, the
application of new numerical methods for the solution of PDEs more precise, ecient and
easy to implement. This work is limited to an implementation of the proposed solution
method in two dimensions, which is done to facilitate the solution of the model to solve, in
addition, the nanochannel are considered with a cylindrical geometry and only two channels
are taken in front and this generalizes Behavior for the other nanochannels.

5.5 Expected products

After completing this research we expect to have the following products:

• A research paper that shows the results of the proposed solution method applied to
solve the linear and nonlinear scalar problems of the proposed model.

• Research article that shows the solution of the model to be solved using MAPS.

• Research article that shows the solution of the model to be solved using MAPS compar-
ing the results with those of the literature, emphasizing the identication of the factors
that generate resistance in proton transport.

• The creation and registration of software generated in the solution of the problem model
by applying the proposed solution method.
Bibliography

[1] Schmidt-Rohr, K., Chen, Q. (2008). Parallel cylindrical water nanochannels in Naon
fuel-cell membranes. Nature materials, 7(1), 75-83.

[2] Schmuck, M. (2009). Analysis of the Navier-Stokes-Nernst-Planck-Poisso system. Math-

ematical Models and Methods in Applied Sciences, 19 (06), 993-1014

[3] Berg, P., Ladipo, K. (2009, June). Exact solution of an electro-osmotic ow problem in a
cylindrical channel of polymer electrolyte membranes. In Proceedings of the Royal Society
of London A: Mathematical, Physical and Engineering Sciences. The Royal Society.

[4] Ladipo, K. O., Berg, P., Kimmerle, S. J., Novruzi, A. (2011). Eects of radially depen-
dent parameters on proton transport in polymer electrolyte membrane nanopores. The
Journal of chemical physics, 134(7), 074103.

[5] Kimmerle, S. J., Berg, P., Novruzi, A. (2013). An electrohydrodynamic equilibrium

shape problem for polymer electrolyte membranes in fuel cells. In System Modeling and
Optimization (pp. 387-396). Springer Berlin Heidelberg.

[6] Schmuck, M., Berg, P. (2014). Eective Macroscopic Equations for Species Transport
and Reactions in Porous Catalyst Layers. The Electrochemical Society (pp. E3323-3327).
Journal of The Electrochemical Society.

[7] Berg, P., Kimmerle, S. J., Novruzi, A. (2014). Modeling, shape analysis and computation
of the equilibrium pore shape near an PEM-PEM intersection. Journal of Mathematical
Analysis and applications, 410 (1), 241-256.

[8] Kimmerle, S. J., Ladipo, K., Novruzi, A., Berg, P. (2014). Contact Resistance at PEM-
PEM Interfaces: Charged Fluid Flow Between Nanochannels. ECS Transactions, 59(1),
145-159.

[9] Berg, P., Benjaminsen, B. E. (2014). Eects of Finite-size Ions and Relative Permittivity
in a NanoporeModel of a Polymer Electrolyte Membrane. ECS Transactions, 59(1), 145-
159.

[10] Tsao, H. K. (1998). Counterion distribution enclosed in a cylinder and a sphere. The

Journal of Physical Chemistry B, 102(50), 10243-10247.

[11] Bustamante, C. A., Power, H.,Florez, W. F. (2013). A global meshless particular solution
method for solving the two-dimensional Navier-Stokes system of equations. Computers
and Mathematics with Applications, 65(12), 1939-1955.
Bibliografía 24

[12] Bustamante, C. A., Power, H.,Sua, Y. H., Florez, W. F. (2013). A global meshless collo-
cation particular solution method (integrated Radial Basis Function) for two-dimensional
Stokes ow problems. Applied Mathematical Modelling, 37(6), 4538-4547.

[13] Yao, G. (2015). An improved localized method of approximate particular solutions for
solving elliptic PDEs. Computers and Mathematics with Applications.

[14] Yao, G., Chen, C. S., Li, W., Young, D. L. (2015). The localized method of approximated
particular solutions for near-singular two- and three-dimensional problems. Computers
and Mathematics with Applications, 70(12), 2883-2894.

[15] Yan, L., Yang, F. (2015). The method of approximate particular solutions for the time-
fractional diusion equation with a non-local boundary condition. Computers and Math-
ematics with Applications, 70(3), 254-264.

[16] Granados, J. M., Bustamante, C. A., Power, H., and Florez, W. F. (2017). A global
Stokes method of approximated particular solutions for unsteady two-dimensional
NavierStokes system of equations. International Journal of Computer Mathematics,
94(8), 1515-1541.

[17] Karniadakis, G., Beskok, A., Aluru, N. (2006). Microows and nanoows: fundamentals
and simulation (Vol. 29). Springer Science and Business Media.

[18] Spiegel, C. (2011). PEM fuel cell modeling and simulation using MATLAB. Academic
press.

[19] Eikerling, M. (2016). Theory and Modeling of Fuel Cells: Extravaganza to Panacea?.
Fuel Cells, 16(6), 663-668.

[20] Kompis, C., Malek, K. (2016). Fuel Cell Modeling Strategic Roadmap: A Systematic
Approach. Fuel Cells, 16(6), 760-770.

[21] Arce, A., Ramírez, D. R., Del Real, A. J., Bordons, C. (2007, December). Constrained
explicit predictive control strategies for pem fuel cell systems. In Decision and Control,
2007 46th IEEE Conference on (pp. 6088-6093). IEEE.

[22] Aricò, A. S., Bruce, P., Scrosati, B., Tarascon, J. M., Van Schalkwijk, W. (2005). Nanos-
tructured materials for advanced energy conversion and storage devices. Nature materi-
als, 4(5), 366-377.

[23] Mehta, V., Cooper, J. S. (2003). Review and analysis of PEM fuel cell design and
manufacturing. Journal of Power Sources, 114(1), 32-53.

[24] Grujicic, M., Chittajallu, K. M. (2004). Design and optimization of polymer electrolyte
membrane (PEM) fuel cells. Applied surface science, 227(1), 56-72.

[25] Balaji, V., Jeevanantham, P., Krishnan, M. M. (2013). Analysis of Flow Parameters in
Fuel Cell for Ecient Power Generation, 2(9), 160164.
Bibliografía 25

[26] Isa, Z. M., Rahim, N. A. (2013, November). PEM fuel cell model parameters optimization
using modied particle swarm optimization algorithm. In Clean Energy and Technology
(CEAT), 2013 IEEE Conference on (pp. 442-445). IEEE.

[27] Xie, J., Ban, S., Liu, B., Zhou, H. (2016). A molecular simulation study of chemical
degradation and mechanical deformation of hydrated Naon membranes. Applied Sur-
face Science, 362, 441-447.

[28] Tse, Y. L. S., Herring, A. M., Kim, K., Voth, G. A. (2013). Molecular dynamics sim-
ulations of proton transport in 3M and naon peruorosulfonic acid membranes. The
Journal of Physical Chemistry C, 117(16), 8079-8091.

[29] Arntsen, C., Savage, J., Tse, Y. L., Voth, G. A. (2016). Simulation of proton transport
in proton exchange membranes with reactive molecular dynamics. Fuel Cells, 16(6),
695-703.

[30] Feng, S., Savage, J., Voth, G. A. (2012). Eects of polymer morphology on proton solva-
tion and transport in proton-exchange membranes. The Journal of Physical Chemistry
C, 116(36), 19104-19116.

[31] Liu, S., Savage, J., Voth, G. A. (2015). Mesoscale study of proton transport in proton
exchange membranes: Role of morphology. The Journal of Physical Chemistry C, 119(4),
1753-1762.

[32] Awatani, T., Midorikawa, H., Kojima, N., Ye, J., Marcott, C. (2013). Morphology of
water transport channels and hydrophobic clusters in Naon from high spatial resolution
AFM-IR spectroscopy and imaging. Electrochemistry Communications, 30, 5-8.

[33] Gruger, A., Régis, A., Schmatko, T., Colomban, P. (2001). Nanostructure of Naon R
membranes at dierent states of hydration: An IR and Raman study. Vibrational Spec-
troscopy, 26(2), 215-225.

[34] Takimoto, N., Ohira, A., Takeoka, Y., Rikukawa, M. (2008). Surface morphology and
proton conduction imaging of Naon membrane. Chemistry letters, 37(2), 164-165.

[35] Jorn, R., Voth, G. A. (2012). Mesoscale simulation of proton transport in proton ex-
change membranes. The Journal of Physical Chemistry C, 116(19), 10476-10489.

[36] Jorn, R., Savage, J., Voth, G. A. (2012). Proton conduction in exchange membranes
across multiple length scales. Accounts of chemical research, 45(11), 2002-2010.

[37] Peighambardoust, S. J., Rowshanzamir, S., Amjadi, M. (2010). Review of the proton
exchange membranes for fuel cell applications. International journal of hydrogen energy,
35(17), 9349-9384.

[38] Bose, S., Kuila, T., Nguyen, T. X. H., Kim, N. H., Lau, K. T., Lee, J. H. (2011). Polymer
membranes for high temperature proton exchange membrane fuel cell: recent advances
and challenges. Progress in Polymer Science, 36(6), 813-843.
Bibliografía 26

[39] Wang, Y., Chen, K. S., Mishler, J., Cho, S. C., Adroher, X. C. (2011). A review of poly-
mer electrolyte membrane fuel cells: technology, applications, and needs on fundamental
research. Applied energy, 88(4), 981-1007.

[40] Dupuis, A. C. (2011). Proton exchange membranes for fuel cells operated at medium
temperatures: Materials and experimental techniques. Progress in Materials Science,
56(3), 289-327.

[41] Laberty-Robert, C., Valle, K., Pereira, F., Sanchez, C. (2011). Design and properties of
functional hybrid organicinorganic membranes for fuel cells. Chemical Society Reviews,
40(2), 961-1005.

[42] Tripathi, B. P., Shahi, V. K. (2011). Organicinorganic nanocomposite polymer elec-

trolyte membranes for fuel cell applications. Progress in Polymer Science, 36(7), 945-979.

[43] Chen, W., Fu, Z. J., Chen, C. S. (2014). Recent advances in radial basis function collo-
cation methods. Springer Berlin Heidelberg.

[44] Kansa, E. J., Hon, Y. C. (2000). Circumventing the ill-conditioning problem with mul-
tiquadric radial basis functions: applications to elliptic partial dierential equations.
Computers and Mathematics with applications, 39(7-8), 123-137.

[45] Franke, C., Schaback, R. (1998). Solving partial dierential equations by collocation
using radial basis functions. Applied Mathematics and Computation, 93(1), 73-82.

[46] Jumarhon, B., Amini, S., Chen, K. (2000). The Hermite collocation method using radial
basis functions. Engineering analysis with boundary elements, 24(7), 607-611.

[47] Chen, C. S., Fan, C. M., Wen, P. H. (2012). The method of approximate particular
solutions for solving certain partial dierential equations. Numerical Methods for Partial
Dierential Equations, 28(2), 506-522.

[48] Mai-Duy, N., Tran-Cong, T. (2001). Numerical solution of dierential equations using
multiquadric radial basis function networks. Neural Networks, 14(2), 185-199.

[49] Golberg, M. A., Chen, C. S. (1998). The method of fundamental solutions for potential,
Helmholtz and diusion problems. Boundary Integral Methods-Numerical and Mathe-
matical Aspects, 103-176.

[50] Florez, W. F., Power, H. (2002). DRM multidomain mass conservative interpolation ap-
proach for the BEM solution of the two-dimensional navier-stokes equations. Computers
and Mathematics with Applications, 43(3-5), 457-472.

[51] Chen, C. S., Fan, C. M., Wen, P. H. (2012). The method of approximate particular
solutions for solving certain partial dierential equations. Numerical Methods for Partial
Dierential Equations, 28(2), 506-522.

[52] Zhu, H. (2014). The method of approximate particular solutions for solving anisotropic
elliptic problems. Engineering Analysis with Boundary Elements, 40, 123-127.
Bibliografía 27

[53] Reutskiy, S. Y. (2013). Method of particular solutions for nonlinear Poisson-type equa-
tions in irregular domains. Engineering Analysis with Boundary Elements, 37(2), 401-
408.

[54] Kuo, L. H., Gu, M. H., Young, D. L., Lin, C. Y. (2013). Domain type kernel-based
meshless methods for solving wave equations. CMC: Computers, Materials and Continua,
33(3), 213-228.

[55] Bustamante, C. A., Power, H., Florez, W. F., Hang, C. Y. (2013). The global approxi-
mate particular solution meshless method for two-dimensional linear elasticity problems.
International Journal of Computer Mathematics, 90(5), 978-993.

[56] Jiang, T., Li, M., Chen, C. S. (2012). The method of particular solutions for solving
inverse problems of a nonhomogeneous convection-diusion equation with variable coef-
cients. Numerical Heat Transfer, Part A: Applications, 61(5), 338-352.

[57] Fu, Z. J., Chen, W., Ling, L. (2015). Method of approximate particular solutions
for constant-and variable-order fractional diusion models. Engineering Analysis with
Boundary Elements, 57, 37-46.

[58] Chen, C. S., Muleshkov, A. S., Golberg, M. A. (1999). The numerical evaluation of
particular solutions for Poisson's equation-a revisit. WIT Transactions on Modelling
and Simulation, 25.

[59] Chen, C. S., Fan, C. M., Wen, P. H. (2011). The method of approximate particular
solutions for solving elliptic problems with variable coecients. International Journal of
Computational Methods, 8(03), 545-559.

[60] Kansa, E. J. (1990). MultiquadricsA scattered data approximation scheme with ap-
plications to computational uid-dynamicsII solutions to parabolic, hyperbolic and
elliptic partial dierential equations. Computers and mathematics with applications,
19(8), 147-161.

[61] Cheng, A. D., Cabral, J. J. S. P. (2005). Direct solution of ill posed boundary value prob-
lems by radial basis function collocation method. International Journal for Numerical
Methods in Engineering, 64(1), 45-64.

[62] Chantasiriwan, S. (2004). Cartesian grid methods using radial basis functions for solv-
ing Poisson, Helmholtz, and diusionconvection equations. Engineering Analysis with
Boundary Elements, 28(12), 1417-1425.

[63] Zerroukat, M., Power, H., Chen, C. S. (1998). A numerical method for heat transfer
problems using collocation and radial basis functions. International Journal for Numer-
ical Methods in Engineering, 42(7), 1263-1278.

[64] Li, J., Cheng, A. H. D., Chen, C. S. (2003). A comparison of eciency and error con-
vergence of multiquadric collocation method and nite element method. Engineering
Analysis with Boundary Elements, 27(3), 251-257.
Bibliografía 28

[65] H. Power, L. Wrobel (1995), Boundary Integral Methods in Fluid Mechanics, Comput.
Mech. Publ, Southampton.

[66] Tian, Z., Li, X., Fan, C. M., Chen, C. S. (2018). The method of particular solutions using
trigonometric basis functions. Journal of Computational and Applied Mathematics, 335,
20-32.

[67] Golbabai, A., Rabiei, H. (2012). A meshfree method based on radial basis functions
for the eigenvalues of transient Stokes equations. Engineering Analysis with Boundary
Elements, 36(11), 1555-1559.

[68] Nath, D., Kalra, M. S., Munshi, P. (2016). Numerical simulation of time-dependent
NavierStokes and MHD equations using a meshless method based on fundamental and
particular solutions. Engineering Analysis with Boundary Elements, 67, 81-95.

[69] Zhang, X., Chen, M., Chen, C. S., Li, Z. (2016). Localized method of approximate
particular solutions for solving unsteady NavierStokes problem. Applied Mathematical
Modelling, 40(3), 2265-2273.