Abstract
We present, a collocation method based on Haar wavelet and Kronecker tensor product for solving three-dimensional partial differential equations. The method is based on approximating a sixth-order mixed derivative by a series of Haar wavelet basis functions. The present method is suitable for numerical solution of all kinds of three-dimensional Poisson and Helmholtz equations. Numerical examples are solving to establish the efficiency and accuracy of the present method. Numerical results obtained are better as compared to numerical results obtained in past.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
In many applications of engineering and science, there are various boundary value problems which involve three-dimensional partial differential equations. Only a few of these equations can be solved by analytical methods. In most cases, we depend on numerical solutions of such partial differential equations. There are several numerical methods available for solving these equations; the most common method used for solving such equations is finite difference method; but this method is slow. In the last few years, other numerical techniques were developed, which are more accurate, efficient and faster than previous numerical algorithms, such as: (a) Jacobi pseudospectral approximation for solving nonlinear complex generalized Zakharov system in [2], (b) A highly accurate collocation algorithm for solving 1 + 1 and 2 + 1 fractional percolation equations in [3], (c) Spectral-Galerkin algorithms using Jacobi polynomials for solving second- and fourth-order differential equations in [8] and [9] respectively, and (d) Legendre spectral-Galerkin method for solving multidimensional elliptic Robin boundary value problems in [11], (e) Jacobi spectral-Galerkin method for the integrated forms of fourth-order elliptic differential equations in [10]. In the last few decades, methods based on wavelet basis functions have been used abruptly. These methods are more efficient and give more accurate numerical results as compared to other well known methods. Wavelet methods are more interesting, accurate and reliable for solving integral and differential equations. Wavelets are a powerful and efficient mathematical tool that divides the data functions or operators into distinct frequency constituents and each constituent is analyzed or investigated with a resolution matching on its scale. Nowadays, wavelet methods are becoming a favorite choice of researchers for solving differential and integral equations. Haar [12] discovered a function, later known as Haar wavelet, in 1909. Such Haar functions are rectangular pair pulses and these are known as Daubechies wavelet of order 1. Also, it is a simplest orthonormal wavelet with compact support. The main disadvantage of Haar wavelets is their discontinuity and, therefore, derivatives do not exist at the points of discontinuities. Due to this, it is impossible to obtain the numerical solution of differential and integral equations. There are two possibilities for overcoming these shortcomings. First, to regularize the piecewise constant Haar functions with interpolation splines; this technique has been applied by Cattani in [4, 5]. But, by this technique, it is difficult to find the solution easily and simplicity of Haar wavelets gets lost. Another possibility, which is proposed by Chen and Hsiao in [6, 7] is that they recommended to expand the highest derivative appearing in the differential equation into the Haar series, instead of the function itself. The other derivatives (and the functions) are obtained through integrations.
Numerical solutions of differential and integral equations using Haar wavelet have been presented by Lepik in [14, 15]. Numerical solutions of two-dimensional PDEs using Haar wavelet have been presented in Lepik [16]. The fundamental idea behind the Haar wavelet method is to convert the given problem into a system of equations which involves finite number of variables. In numerical analysis, because of the property of localization, wavelet based algorithms have become an important tools for solving ordinary and partial differential equations. A new approach of the Chebyshev wavelets method for partial differential equations with boundary conditions of the telegraph type equation has been presented in [13]. Haar wavelet collocation method has been presented in [19], for solving boundary layer fluid flow problems. A numerical assessment of parabolic partial differential equations using Haar and Legendre wavelets has been presented in [20]. Numerical solution of two-dimensional elliptic PDEs with nonlocal boundary conditions has been presented in [21]. In the present paper, we use Haar wavelet collocation method for solving three-dimensional Poisson and Helmholtz equations because Haar wavelet method has sparse representation, fast transformation, and possibility of implementation of fast algorithms. The general linear partial differential equation of the second order in three independent variables is of the form
on \(K=\{(x,y,z):{a}<{x}<{b}, {c}<{y}<{d}, {e}<{z}<{f}\}\), with boundary conditions
and
where \(A_{11}, B_{11}, C_{11}, g_{0}, g_{1}, g_{2}, g_{3}, {g_{4}}, {g_{5}}\) and f are known functions. The Poisson equation in three-dimensional Cartesian coordinates system plays an important role due to its wide range of application in areas like ideal fluid flow, heat conduction, elasticity, electrostatics, gravitation and other science fields especially in physics and engineering. Poisson and Helmholtz equations are arising in different branches of science and engineering such as fluid mechanics, electricity and magnetism and torsion problems.
Our main aim is to develop an accurate and efficient collocation method using Haar wavelet and Kronecker product for solving three-dimensional partial differential equations such as Poisson and Helmholtz equations, by approximating a sixth-order mixed derivative by a series of Haar wavelet basis functions. In Sect. 2, Haar wavelet method has been discussed. Error analysis has been described in Sect. 3. In Sect. 4, numerical examples have been solved using the present method and compared with the exact solutions.
Kronecker product of two matrices
For saving calculation time, we use the concept of Kronecker product of matrix A with matrix B of orders \({p}\times {q}\) respectively and is defined as:
The first documented work on Kronecker products was written by Johann Georg Zehfuss between 1858 and 1868. In MATLAB, the Kronecker product of two matrices A and B is directly calculated with the command kron(A, B).
Kronecker product of three matrices
The Kronecker product of three A, B and C matrices each of orders \(p\times q\) can be calculated as:
where E is of the form:
Haar wavelet method
Consider \(x\in [\sigma _{1},\sigma _{2}]\), \(y\in [\sigma _{3},\sigma _{4}]\) and \(z\in [\sigma _{5},\sigma _{6}]\) where \(\sigma _{1}\), \(\sigma _{2}\), \(\sigma _{3}\), \(\sigma _{4}\), \(\sigma _{5}\) and \(\sigma _{6}\) are given constants. We shall define the quantities \(M_{1}=2^{J_{1}}\), \(M_{2}=2^{J_{2}}\) and \(M_{3}=2^{J_{3}}\) where \(J_{1}\), \(J_{2}\) and \(J_{3}\) are the maximal levels of resolution. Now, divide the interval \([\sigma _{1}, \sigma _{2}]\), \([\sigma _{3}, \sigma _{4}]\) and \([\sigma _{5}, \sigma _{6}]\) respectively into \(2M_{1}\), \(2M_{2}\) and \(2M_{3}\) subintervals, each of length \(\triangle {x}=(\sigma _{2}-\sigma _{1})/2M_{1}\), \(\triangle {y}=(\sigma _{4}-\sigma _{3})/2M_{2}\) and \(\triangle {z}=(\sigma _{6}-\sigma _{5})/2M_{3}\) respectively. Now, we introduce parameters : dilatation parameter \(j_{1}=0,1,2,\ldots ,J_{1}\); \(j_{2}=0,1,2,\ldots ,J_{2}\) and \(j_{3}=0,1,2,\ldots ,J_{3}\) and translation parameter \(k_{1}=0,1,2,\ldots ,m_{1}-1\); \(k_{2}=0,1,2,\ldots ,m_{2}-1\) and \(k_{3}=0,1,2,\ldots ,m_{3}-1\), where \(m_{1}=2^{j_{1}}\), \(m_{2}=2^{j_{2}}\) and \(m_{3}=2^{j_{3}}\). The wavelet numbers \(i_{1}\), \(i_{2}\) and \(i_{3}\) are calculated according the formula \(i_{1} = m_{1}+k_{1}+1\), \(i_{2} = m_{2}+k_{2}+1\) and \(i_{3} = m_{3}+k_{3}+1\) respectively. Therefore, we have
and
where \(\alpha _{1}= {\sigma _{1}}+2{k_{1}}{\omega _{1}}{\triangle {x}}\), \(\alpha _{2}= {\sigma _{1}}+(2{k_{1}}+1){\omega _{1}}{\triangle {x}}\), \(\alpha _{3}= {\sigma _{1}}+2({k_{1}}+1){\omega _{1}}{\triangle {x}}\), \({\omega _{1}}=M_{1}/m_{1}\),
\(\beta _{1}= {\sigma _{3}}+2{k_{2}}{\omega _{2}}{\triangle {y}}\), \(\beta _{2}= {\sigma _{3}}+(2{k_{2}}+1){\omega _{2}}{\triangle {y}}\), \(\beta _{3}= {\sigma _{3}}+2({k_{2}}+1){\omega _{2}}{\triangle {y}}\), \({\omega _{2}}=M_{2}/m_{2}\),
\(\gamma _{1}= {\sigma _{5}}+2{k_{3}}{\omega _{3}}{\triangle {z}}\), \(\gamma _{2}= {\sigma _{5}}+(2{k_{3}}+1){\omega _{3}}{\triangle {z}}\), \(\gamma _{3}= {\sigma _{5}}+2({k_{3}}+1){\omega _{3}}{\triangle {z}}\), \({\omega _{3}}=M_{3}/m_{3}\).
The collocation points are obtained as:
and
Consider the approximate wavelet solution of the form
Integrating (13), twice with respect to x, from 0 to x, we obtain
Putting \(x=1\) in (14), we obtain
Again, integrating (16), twice with respect to y, from 0 to y, we obtain
where
Putting \(y=1\) in (17), we obtain
From (17), using (21), we obtain
Now, integrating (13) twice with respect to z from 0 to z, we obtain
Putting \(z=1\) in (23), we obtain
From (23), using (24), we obtain
Again, integrating (25) twice with respect to x, from 0 to x, we obtain
where
Putting \(x=1\) in (26), we obtain:
Substituting (30) in (26), we obtain
Now, integrating (13) twice with respect to y, from 0 to y, we obtain:
Putting \(y=1\) in (32), we obtain
Again, integrating (34), twice with respect z, from 0 to z, we obtain
where
Putting \(z=1\) in (35), we obtain
Again, integrating (40), twice with respect to x, from 0 to x, we obtain
where
Putting \(x=1\) in (41), we obtain
Substituting the values from (22), (31) and (40) in (1), we obtain
where
and
Expressions for \(\{h_{i_{1}}(x)\}\), \(\{h_{i_{2}}(y)\}\) and \(\{h_{i_{3}}(z)\}\) are given below:
Discretising (49)–(53) using (10)–(12), we obtain the following system of equations in matrix form
where W represents the wavelet coefficient matrix. The value of matrix R can be calculated as:
where
and
The value of matrix S can be calculated as:
where
and
The value of matrix T can be calculated as:
where
and
Each component of F can be evaluated as:
where
The numerical solution of given problem is obtained by substituting the values of wavelet coefficients into (48).
Error analysis for three-dimensional PDEs
In this section we present the error analysis for our proposed scheme. In order to analyze the convergence of our method, we state and prove the following convergence theorem:
Theorem
Suppose that u(x, y, z) satisfies a Lipschitz condition on \(D=[0,1)\times [0,1)\times [0,1)\), that is there exist a positive constant \(L_{1}, L_{2}, L_{3}\) and \(L_{4}\), such that for all \((x_{1},y,z)\), \((x_{2},y,z)\), \((x_{3},y,z)\), \((x_{4},y,z)\), \((x_{5},y,z)\), \((x_{6},y,z)\), \((x_{7},y,z)\) and \((x_{8},y,z)\) in D, we have
Then, the error bound \(\parallel {E_{m}}\parallel _{2}\) obtained from above is
Here, the order of convergence is of the order 4.
Proof
Consider \(M_{1}=M_{2}=M_{3}=M\). Let \(u_{exact}(x,y,z)\) and \(u_{approximate}(x,y,z)\) be the exact and approximate solutions of the partial differential equation. The error at the Jth level of resolution is defined as:
where
and the wavelet coefficients are calculated as:
Here \(\Big <.\Big>\) shows the inner product. Define \({\parallel {.}\parallel }_{2}\) as:
Using definition of inner product, from (80), we obtain
Using orthogonality conditions, from (81), we obtain
According to (9), we can write
Applying mean value theorem, that is there exist \(z_{1}\in [{\frac{k}{m}}, {\frac{k+0.5}{m}}]\) and \(z_{2}\in [{\frac{k+0.5}{m}}, {\frac{k+1}{m}}]\), such that
Again,
From (85), using the definition of inner product, we obtain
Using (8), from (86), we obtain
Again, applying mean value theorem, we obtain:
After simplifications, from (88), we obtain
Hence,
From (90), using (7), we obtain
Applying mean value theorem, from (91), we obtain
After simplifications, from (92), we obtain
Using (74), from (93), we obtain
where \(L=max\{L_{1}, L_{2}, L_{3}, L_{4}\}\). After simplifications, from (94), we obtain
Squaring both sides, from (95), we obtain
By substituting (96) in (82), we obtain
After simplifications, from (97), we obtain
Expanding (98), we obtain
From (99), after simplification, we obtain
From (100), after series summation, we obtain
After taking square root, we obtain
This shows that the convergence is of the order 4. \(\square\)
Numerical examples and discussion
We have applied our method on some numerical examples, to observe the accuracy and efficiency of the present method for solving three-dimensional Poisson equations.
Example 1
Consider the following linear three-dimensional Poisson equation
on \(K=\{(x,y,z):{0}{<}{x}{<} {1}, {0}{<}{y}{<}{1}, {0}{<}{z}{<}{1}\}\), with boundary conditions:
where
The exact solution is \(u(x,y,z)=-{\frac{1}{3{\pi ^2}}}.{\sin (\pi {x})}.{\sin (\pi {y})}.{\sin (\pi {z})}\). Table 1 shows the maximum absolute errors of Example 1.
The proposed method is more simplest and different from the method presented in [23]. For \(J=3\), the maximum absolute error obtained by [23] is \(1.3730E-004\), where as in our research paper maximum absolute error is \(5.3267E-005\).
Example 2
Consider the following linear three-dimensional Poisson equation
on \(K=\{(x,y,z):{0}{<}{x}{<}{1}, {0}{<}{y}{<}{1}, {0}{<}{z}{<}{1}\}\), with boundary conditions:
where
The exact solution is \(u(x,y,z)={x^3}{y^3}{z^3}{(1-x)}{(1-y)}{(1-z)}\). Table 2 shows the maximum absolute errors of Example 2.
Example 3
Consider the following linear three-dimensional Helmholtz equation
on \(K=\{(x,y,z):{0}{<}{x}{<}{1}, {0}{<}{y}{<}{1}, {0}{<}{z}{<}{1}\}\), with boundary conditions:
where
The exact solution is \(u(x,y,z)={\sin (\pi {x})}.{\sin (\pi {y})}.{\sin (\pi {z})}\). Table 3 shows the maximum absolute errors of Example 3.
Example 4
Consider the following linear three-dimensional Helmholtz equation
on \(K=\{(x,y,z):{0}{<}{x}{<}{1}, {0}{<}{y}{<}{1}, {0}{<}{z}{<}{1}\}\), with boundary conditions:
where
The exact solution is \(u(x,y,z)={x^3}{(1-x)}.{\sin (\pi {y})}.{\sin (\pi {z})}\). Table 4 shows the maximum absolute errors of Example 4.
Conclusion
It is concluded from here that the Haar wavelet method is a powerful mathematical tool for solving three-dimensional partial differential equations. As we increase the values of \(2M_{1}\), \(2M_{2}\) and \(2M_{3}\), absolute errors decrease rapidly and the numerical solutions are much closer to the exact solutions. Also, proposed method gives better results as compared to numerical results presented in [23]. Also, proposed method is applicable to many types of three-dimensional Poisson equations (see, for example, in Sect. 4, Example 2) whereas method presented in [23] is applicable to only one type of three-dimensional Poisson equation. In view of numerical results, it is concluded that proposed method based on Haar wavelet is more efficient and accurate for solving three-dimensional partial differential equations.
References
Aziz, I., Sarler, B.: Wavelets collocation methods for the numerical solution of elliptic BV problems. Appl. Math. Model. 37, 676–697 (2013)
Bhrawy, A.H.: An efficient Jacobi pseudospectral approximation for nonlinear complex generalized Zakharov system. Appl. Math. Comput. 247, 30–46 (2014)
Bhrawy, A.H.: A highly accurate collocation algorithm for 1+1 and 2+1 fractional percolation equations. J. Vibrat. Control 22, 2288–2310 (2016)
Cattani, C.: Haar wavelet splines, J. Interdiscipl. Math., 4, 35–47 (2001)
Cattani, C.: Haar wavelets based technique in evolution problems, Proc. Estonian Acad. Sci. Phys. Math., 53, 45–65 (2004)
Chen, C.F., Hsiao, C.H.: Haar wavelet method for solving lumped and distributed- parameter systems, IEE Proc. Control Theory Appl. 144, 87–94 (1997)
Chen, C. F. Hsiao, C. H.: Wavelet approach to optimising dynamic systems, IEE Proc. Control Theory Appl., 146, 213–219 (1997)
Doha, E.H., Bhrawy, A.H.: Efficient spectral- Galerkin algorithms for direct solution of second-order differential equations using Jacobi polynomials. Num. Algorith. 42, 137–164 (2006)
Doha, E.H., Bhrawy, A.H.: Efficient spectral- Galerkin algorithms for direct solution of fourth-order differential equations using Jacobi polynomials. Appl. Num. Math. 58, 1224–1244 (2008)
Doha, E.H., Bhrawy, A.H.: A Jacobi spectral-Galerkin method for the integrated forms of fourth-order elliptic differential equations. Num. Methods Par. Diff. Equ. 25, 712–739 (2009)
Doha, E.H., Bhrawy, A.H.: An efficient direct solver for multidimensional elliptic Robin boundary value problems using a Legendre spectral-Galerkin method. Comp. Math. Appl. 64, 558–571 (2012)
Haar, A.: Zur theorie der orthogonalen Funktionsysteme. Math. Annal. 69, 331–371 (1910)
Heydari, M.H., Hooshmandasl, M.R., Maalek, F.M., Ghaini, : A new approach of the Chebyshev wavelets method for partial differential equations with boundary conditions of the telegraph type. Appl. Math. Model. 38, 1597–1606 (2014)
Lepik, Ü.: Application of Haar wavelet transform to solving integral and differential equations. Proc. Estonian Acad. Sci. Phys. Math. 56(1), 28–46 (2007)
Lepik, Ü.: Solving differential and integral equations by Haar wavelet method, revisted. Int. J. Math. Comput. 1, 43–52 (2008)
Lepik, Ü.: Solving PDEs with the aid of two-dimensional Haar wavelets. Comp. Math. Appl. 61(7), 1873–1879 (2011)
Mittal, R.C., Bhatia, R.: A numerical study of two dimensional hyperbolic telegraph equation by modified B-spline differential quadrature method. Appl. Math. Comput. 244, 976–997 (2014)
Mittal, R.C., Gahlaut, S.: High-order finite-differences schemes to solve Poissons equation in polar coordinates, IMA J. Numer. Anal., 11, 261–270 (1987)
Siraj ul Islam, Arler, B., Aziz, I., Haq, F.: Haar wavelet collocation method for the numerical solution of boundary layer fluid flow problems, Int. J. Therm. Sci., 50, 686–697 (2011)
Siraj-ul-Islam, Aziz, I., Al-Fhaid, A.S., Shah, A.: A numerical assessment of parabolic partial differential equations using Haar and Legendre wavelets, Applied Mathematical Modelling, 37, 9455–9481 (2013)
Siraj-ul-Islam, I., Aziz, M., Ahmad, : Numerical solution of two-dimensional elliptic PDEs with nonlocal boundary conditions. Comp. Math. Appl. 69, 180–205 (2015)
Sutmann, G., Steffen, B.: High–order compact solvers for the three–dimensional Poisson equation, J. Comput. Appl. Math., 187, 142–170 (2006)
Shi, Zhi, Cao, Yong-yan, Chen, Qing-jiang: Solving 2D and 3D Poisson equations and biharmonic equations by the Haar wavelet method. Appl. Math. Modell. 36, 5143–5161 (2012)
Acknowledgements
We are grateful to the anonymous reviewers for their valuable comments which let to the improvement of the manuscript.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Singh, I., Kumar, S. Wavelet methods for solving three-dimensional partial differential equations. Math Sci 11, 145–154 (2017). https://doi.org/10.1007/s40096-017-0220-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40096-017-0220-6
Keywords
- Haar wavelet method
- Three-dimensional partial differential equations
- Linear systems
- Kronecker tensor product