Contemporary Engineering Sciences, Vol. 11, 2018, no.

91, 4517 - 4524

HIKARI Ltd, www.m-hikari.com
https://doi.org/10.12988/ces.2018.89488

Numerical Solution of the Problem of Distribution

of Electrical Charges by Means of Finite Differences
and Finite Elements
Diego Fernando Devia Narvaez

Department of Mathematics and GREDYA

Universidad Tecnológica de Pereira
Pereira, Colombia

Diana Marcela Devia Narvaez

Department of Mathematics and GIMAE

Universidad Tecnológica de Pereira
Pereira, Colombia

Germán Correa Vélez

Department of Mathematics and GIMAE

Universidad Tecnológica de Pereira
Pereira, Colombia

Copyright
c 2018 Diego Fernando Devia Narvaez et al. This article is distributed under
the Creative Commons Attribution License, which permits unrestricted use, distribution,
and reproduction in any medium, provided the original work is properly cited.

Abstract
The theory around the approximate techniques to solve differential
equations is widely accepted in diverse areas of knowledge. To solve
the differential equations it is necessary to discretize the domain in
fragments depending on the space. We present the methodology of
finding a recurrence equation that allows to traverse the domain of a
problem in order to find a linear system of equations whose solution is
the potential at each point of the domain.

Keywords: Problem of distribution, electrical charges, finite differences

4518 Diego Fernando Devia Narvaez et al.

1 Introduction
We consider an equation of the form [1]

d2 f (x) df (x)
p (x) + q (x) + r (x) f (x) = z(x), (1)
dx2 dx
where x0 ≤ x ≤ xn , f (x0 ) = f x0 and f (xn ) = f xn . Now, to use the general
numerical solution algorithm using the finite difference method, we have to:

• Discretize the domain. Divide the domain into a finite mesh of points in
which the approximate solution of the differential equation (1) is sought.

• Rewrite the differential equation (1) in terms of centered divide differ-

ences (first and second order). For each point of the mesh or discretiza-
tion of the domain, the solution of f (x) of the differential equation (1)
is approximated as a set of linear equations of the form AX = B.

• Solve the system of linear equations of the form AX = B to obtain the

approximate solution in each of the points of the discretized domain.

2 Dirichlet problem in two dimensions

Some physical phenomena can be modeled with partial differential equations
by means of the Laplace equation [2]:

∇2 u = 0 (2)
From Eq.(2) we have the Laplacian scalar operator. This equation seeks to
model the distribution of potential u in areas that accumulate loads. In [7]
it is proposed to calculate the distribution of the potential in a bar of known
dimensions of a dielectric material with regular cross section. On the other
hand, Eq.(2) can be rewritten as partial differential equations of the form:

∂ 2 u(x, y) ∂ 2 u(x, y)
+ = 0, (3)
∂x2 ∂y 2
with u(x, 0) = u(x, yn ) = 0, 0 ≤ x ≤ a, u(a, y) = 0, u(0, y) = u0 , 0 ≤ y ≤ b.
To find the solution of the potential u of the Eq.(3) it is necessary to formulate
for each point of the domain Ω the Eqs.(4) and (5) and obtain a system of
scattered linear equations. The results obtained for the distribution of load in
a medium other than vacuum are found in appendex of this paper [3].

∂ 2 f (xi, yi) fi−1,j − 2fi,j + fi+1,j

= (4)
∂x 2
∆x2
Numerical solution of ... 4519

∂ 2 f (xi, yi) fi,j−1 − 2fi,j + fi,j+1

= (5)
∂y 2
∆y 2

Figure 1: Load distribution in the Dielectric.

Now, we want to solve the following problem

∂ 2 u(x, y) ∂ 2 u(x, y)
+ = 0, (6)
∂x2 ∂y 2
with the conditions u(x, 0) = u(x, yn ) = 0, 0 ≤ x ≤ a u(a, y) = 0, 0 ≤ y ≤ b,
u(0, y) = u and 0 ≤ y ≤ b. Now, the functions of each triangular element lead
to ZZ Ω  
∂ui (x, y) ∂uj (x, y) ∂ui (x, y) ∂uj (x, y)
αij = + dxdy (7)
Ω ∂x ∂x ∂y ∂y

βi = f (x, y)ϕi (x, y) (8)

Coordinates of the first triangle

V ertex1 = (0, 0), V ertex2 = (0.1, 0), V ertex3 = (0, 0.1)

         
u1 1 (x1 , y1 ) 1 0 0 a 1 1 0 0 a
u1 1 (x2 , y2 ) = 1 0.1 0   b  1 = 1 0.1 0   b 
u1 1 (x3 , y3 ) 1 0 0.1 c 0 1 0 0.1 c
   
1 a
−10 =  b 
0 c
4520 Diego Fernando Devia Narvaez et al.

u1 (x, y) = 1 − 10x, β1 = u(x, y) (1 − 10x)

Coordinates of the second triangle

V ertex1 = (0, 0.1), V ertex2 = (0.1, 0), V ertex3 = (0.13, 0.08)

         
u2 2 (x1 , y1 ) 1 0 0 a 1 1 0 0 a
u2 2 (x2 , y2 ) = 1 0.1 0   b  1 = 1 0.1
  0   b
2
u2 (x3 , y3 ) 1 0.13 0.08 c 0 1 0.13 0.08 c
   
0.72731 a
−7.2727 =  b 
2.7273 c

u2 (x, y) = 0.7273−7.2727x+2.7273y, β 2 = u(x, y) (0.7273−7.2727x+2.7273y)

Coordinates of the third triangle

V ertex1 = (0, 0.1), V ertex2 = (0.08, 0.16), V ertex3 = (0.13, 0.08)

 3         
u3 (x1 , y1 ) 1 0 0 a 1 1 0 0 a
u3 3 (x2 , y2 ) = 1 0.08 0.16  b  1 = 1 0.08 0.16  b 
u3 3 (x3 , y3 ) 1 0.13 0.08 c 0 1 0.13 0.08 c
   
1.5319 a
−8.5106 =  b 
−5.3191 c

u3 (x, y) = 1.5319−8.5106x−5.3191y β3 = u(x, y) 1.5319−8.5106x−5.3191y

The first approach consists of discretizing the values of β1 , β2 , β3 equal to

1 by means of v(x, y) = 1. On the other hand, the coefficients αij are given by
the Eq.(7)

        
1 1 1 γ1 1 γ1 −4.71
0.35 0.68 0.76 γ2  = 1 ; γ2  =  23.26  ,
0.81 0.82 0.81 γ3 1 γ3 −17.54

where the results obtained are (Figure 3 shows the load distribution surface in
the material [4-5]).
Numerical solution of ... 4521

Figure 2: Discretization triangular finite elements.

Figure 3: Finite Elements Solution.

3 Conclusion
Figure 4 shows the behavior of the error obtained when comparing the results
of the numerical analysis. The maximum absolute error as shown in Figure 4
is 5 times compared to the results obtained with the finite element method,
this is because the finite element method discretizes the domain in a finer mesh
compared to the method of Finite differences [6].

As the divisions of the mesh (discretization) of each of the methods de-

scribed and applied in this document are increased, the response of the vari-
ables of the problem modeled by the Laplace equation can be improved, a
4522 Diego Fernando Devia Narvaez et al.

Figure 4: Finite Elements Solution.

consequence of discretizing the domain of the problem in a finer mesh it is the

computational cost that increases, since the quantity of potentials to calculate.

Acknowledgements. We would like to thank the referee for his valuable

suggestions that improved the presentation of this paper and our gratitude
to the Department of Mathematics of the Universidad Tecnológica de Pereira
(Colombia), the GEDNOL Research Group and GIMAE (Grupo de Investi-
gación en Matemática Aplicada y Educación).
Numerical solution of ... 4523

x y u x y u x y u x y u x y u
0 0 0.5 0 0.3 1.0 0.221 0.311 0.529 0.189 0.664 0.6 0.158 0.287 0.626
1.0 0 0 0 0.2 1.0 0.543 0.0756 0.0528 0.0762 0.732 0.801 0.0637 0.353 0.858
1.0 1.0 0 0 0.1 1.0 0.443 0.912 0.0892 0.733 0.924 0.0252 0.638 0.0748 0.0361
0 1.0 0.5 0.48 0.513 0.268 0.0874 0.444 0.824 0.671 0.806 0.0787 0.353 0.936 0.0925
0.1 0 0 0.0481 0.0439 0.485 0.719 0.664 0.0961 0.925 0.0923 0.00764 0.43 0.833 0.169
0.2 0 0 0.956 0.048 0.00232 0.911 0.644 0.0281 0.0756 0.907 0.548 0.166 0.431 0.673
0.3 0 0 0.0439 0.952 0.51 0.91 0.745 0.0227 0.908 0.926 0.00764 0.775 0.631 0.0776
0.4 0 0 0.952 0.956 0.00233 0.176 0.104 0.319 0.0926 0.0755 0.437 0.321 0.125 0.194
0.5 0 0 0.755 0.356 0.0839 0.826 0.898 0.0207 0.568 0.333 0.173 0.126 0.678 0.715
0.6 0 0 0.361 0.265 0.304 0.105 0.824 0.639 0.7 0.443 0.118 0.681 0.872 0.0518
0.7 0 0 0.268 0.637 0.485 0.899 0.174 0.0189 0.917 0.544 0.0288 0.857 0.418 0.0494
0.8 0 0 0.657 0.728 0.109 0.605 0.423 0.168 0.738 0.754 0.0711 0.142 0.0449 0.188
0.9 0 0 0.0898 0.161 0.657 0.397 0.624 0.328 0.389 0.71 0.295 0.956 0.142 0.0067
1.0 0.1 0 0.161 0.91 0.311 0.372 0.388 0.359 0.288 0.385 0.464 0.858 0.956 0.00714
1.0 0.2 0 0.911 0.844 0.0148 0.824 0.585 0.0617 0.63 0.654 0.14 0.045 0.858 0.797
1.0 0.3 0 0.841 0.0903 0.0163 0.825 0.374 0.0585 0.901 0.347 0.0311 0.517 0.291 0.191
1.0 0.4 0 0.452 0.769 0.202 0.525 0.742 0.172 0.411 0.457 0.33 0.791 0.431 0.0755
1.0 0.5 0 0.558 0.237 0.142 0.46 0.259 0.214 0.465 0.593 0.27 0.643 0.274 0.115
1.0 0.6 0 0.23 0.453 0.569 0.262 0.536 0.521 0.515 0.831 0.126 0.663 0.354 0.125
1.0 0.7 0 0.687 0.534 0.125 0.607 0.571 0.169 0.166 0.517 0.679 0.677 0.602 0.125
1.0 0.8 0 0.815 0.688 0.0566 0.663 0.135 0.058 0.484 0.161 0.134 0.309 0.701 0.39
1.0 0.9 0 0.259 0.158 0.304 0.142 0.354 0.696 0.765 0.511 0.0886 0.298 0.307 0.41
0.9 1.0 0 0.852 0.259 0.039 0.354 0.858 0.194 0.773 0.274 0.0649 0.632 0.493 0.157
0.8 1.0 0 0.159 0.741 0.605 0.824 0.795 0.0388 0.242 0.724 0.47 0.405 0.546 0.338
0.7 1.0 0 0.746 0.842 0.0472 0.181 0.218 0.514 0.275 0.24 0.385 0.436 0.386 0.291
0.6 1.0 0 0.734 0.0928 0.0301 0.218 0.819 0.4 0.47 0.706 0.226 0.741 0.588 0.096
0.5 1.0 0 0.267 0.909 0.184 0.788 0.184 0.0433 0.335 0.584 0.411 0.209 0.386 0.586
0.4 1.0 0 0.0903 0.267 0.764 0.92 0.45 0.0276 0.409 0.328 0.295 0.387 0.79 0.237
0.3 1.0 0 0.846 0.49 0.0553 0.445 0.0731 0.0739 0.506 0.42 0.238 0.821 0.321 0.0551
0.2 1.0 0 0.34 0.505 0.414 0.552 0.922 0.0531 0.292 0.464 0.478 0.339 0.652 0.381
0.1 1.0 0 0.483 0.336 0.234 0.0765 0.553 0.846 0.56 0.497 0.204 0.45 0.657 0.264
0 0.9 1.0 0.533 0.65 0.202 0.646 0.929 0.0335 0.406 0.201 0.213 0.34 0.44 0.409
0 0.8 1.0 0.61 0.857 0.0763 0.0691 0.645 0.847 0.204 0.593 0.602 0.345 0.334 0.367
0 0.7 1.0 0.391 0.135 0.161 0.354 0.068 0.0969 0.594 0.779 0.118 0.588 0.146 0.0837
0 0.6 1.0 0.138 0.609 0.717 0.335 0.187 0.259 0.708 0.168 0.0583 0.587 0.711 0.148
0 0.5 1.0 0.711 0.232 0.076 0.268 0.0757 0.153 0.287 0.842 0.274 0.529 0.565 0.223
0 0.4 1.0 0.312 0.778 0.321 0.93 0.256 0.0177 0.645 0.198 0.0879 0.709 0.297 0.0928

References
[1] P. Hartman, On the local linearization of differential equations, Proceed-
ings of the American Mathematical Society, 14 (1963), no. 4, 568-573.
https://doi.org/10.2307/2034276

[2] N. Adabala, M. Varma, K. Toyama, Computer aided generation of stylized

maps, Computer Animation and Virtual Worlds, 12 (2007), no. 2, 133-
140. https://doi.org/10.1002/cav.168

[3] A. Escobar, A. Molina, V. Correa, Método Iterativo para el Cálculo de

Potenciales Electrostáticos, Scientia et Technica, 32 (2006), 133-138.

[4] P. Francisco, Control de un Sistema de Levitación Magnética con Compen-

sación de Redes Neuronales, Diss., Centro de investigación y de estudios
avanzados del instituto politécnico nacional, 2009.

[5] Diana Marcela Devia Narvaez, Fernando Mesa, Pedro Pablo Cardenas
Alzate, A Note on the Implementation of the Finite Element to Solve
Differential Equations, Contemporary Engineering Sciences, 11 (2018),
no. 50, 2457-2464. https://doi.org/10.12988/ces.2018.84149
4524 Diego Fernando Devia Narvaez et al.

[6] F. Mesa, P. Cardenas Alzate, C. Rodriguez Varela, Numerical Solution

of the Laplace Equation: Electrostatic Potential, Advanced Studies in
Theoretical Physics, 11 (2017), no. 12, 717-723.
https://doi.org/10.12988/astp.2017.71155

Received: September 24, 2018; Published: October 30, 2018