Abstract
In this paper, the Burgers’ equation which is two-dimensional in space, time dependent parabolic differential equation was solved by b-spline collocation algorithms for solving two-dimensional parabolic partial differential equation. At first b-spline interpolation is introduced moreover, the numerical solution is represented as a bi-variate piecewise polynomial with unknown time-dependent coefficients are determined by requiring the numerical solution to satisfy the PDE at a number of points within the spatial domain i.e. we collocate simultaneously in both spatial dimensions. The accuracy of the proposed method is demonstrates by some test problems. The numerical results are found good agreement with exact solution.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Consider the two-dimensional Burgers equation
where \(\Omega =\left\{ {(x,y)\left| {a\prec x\prec b,c\prec y\prec d} \right. } \right\} \)
With exact solution
Subject to the initial condition
The Burgers equation is a fundamental pde from fluid mechanics. It occurs in various areas of applied mathematics such as modeling of gas dynamics and traffic flow.
In 1915, Bateman [1] introduced the one-dimensional Burgers equation and metioned. The steady solution is worth of the study. It was later treated by Burger [2] as a mathematical model for turbulence and after whom such an equation is widely referred to as Burgers’ equation many researchers have used various numerical methods to solve the Burgers equation [3–6].
In these methods fall into the following classes: finite difference method [7, 8], finite volume method [9], finite element method [10], boundary element method [11], and etc.
In this paper, we discuss the development of two numerical algorithms based on B-spline collocation for solving the Burgers’ equation. We consider B-spline collocation in a tensor product framework to discretize both spatial domains. We presented an algorithm that uses a fast block LU scheme with modified alternate row and column elimination.
R. D. Russell and his colleagues [13] suggested a fast algorithm upon a matrix block eigenvalue decomposition to solve spline collocation matrices. In this research, the matrices that arise with this method have two important advantages:
A similar fast algorithm based upon LU decomposition with amendment.
Alternate row and column elimination with partial pivoting to take advantage of the structure of matrices that arise and also it was used just one collocation point per subinterval, the linear system that arise are the smallest among all type of piecewise polynomial collocation method for this problem.
2 Collocation method
We consider a 2D rectangular grid based on a mesh of \(N + 1\) points
(N > 1) in [a, b] and a mesh of M + 1 points (M \(>\) 1) in [c, d] such that
We associate with the mesh on the x domain, \(C^1\)-continuous piecewise polynomials of degree p, i.e., we have a polynomial of degree p for the ith subinterval, \([x_{i-1} ,x_i ]\), \(i=1,\ldots ,N, \)with \(C^1\)- continuity imposed at the internal mesh points. Consequently the dimension of this piecewise polynomial subspace is \({KF }= {K(p-1) }+ 2.\) Similarly, in the y domain, we have a polynomial of degree q for the ith subinterval, \([y_{i-1} ,y_i ]\), \(i=1,\ldots ,N\), with \(C^1\)-continuity imposed at the internal mesh points. Consequently the dimension of this piecewise polynomial subspace is \({ LF = L(q-1)} + 2.\)
To represent the piecewise polynomials, we employ B-spline bases. Let \(\left\{ {\rho _i } \right\} _{i=1}^N , \left\{ {\eta _j } \right\} _{j=1}^M \) be the B-splines bases associated with the above meshes. Thus an approximation \(U_N (x,y,t)\) to the exact solution U(x, y, t) can be expressed as a linear combination of the tensor product of the B-spline bases functions in x and y with time-dependent coefficients, \(\delta _{ij} (t)\) as follows:
We define \(\left\{ {\alpha _i } \right\} _{i=1}^{p-1} \)and\(\left\{ {\beta _j } \right\} _{j=1}^{q-1} \) to be the canonical Gaussian points on [0,1] with \(0\prec \alpha _1 \prec \cdots \prec \alpha _{p-1} \prec 1\), and \(0\prec \beta _1 \prec \cdots \prec \beta _{q-1} \prec 1\). The collocation points in the x domain are then defined by
where \(h_i =x_i -x_{i-1}, \quad i=1,\ldots ,N\)
The collocation points in the y domain are defined to be
where \(k_i =y_i -y_{i-1}, \quad i=1,\ldots ,M\)
The Burgers’ equation is discretized over the spatial domain by simultaneously collocating at the points \(\left\{ {\lambda _i } \right\} _{i=1}^{KF-1} \) in x and the points \(\left\{ {\omega _j } \right\} _{j=1}^{LF-1} \) in y. The collocation conditions yield the following ODEs in time:
where \(i=2,\ldots ,KF-1\), and \(j=2,\ldots ,LF-1\).
The boundary conditions are
By substituting (5) into (8)–(11), we get equations in the terms of the unknowns \(\varvec{p}_{ij} (\varvec{t})\). We can then rewrite these equations in matrix Form:
In (13), \(\bar{p}(t)\) is the B-spline coefficient vector, it has the form,
The right hand side vector
where each \(F_i (t,\bar{P}(t)) \)has LF components. The expressions for \(F_i (t,\bar{P}(t)),i=1,\ldots ,KF,\)
The matrix G in, can be written as
3 2D B-spline
The B-spline coefficients characterize the projection of the approximate solution onto the B-spline tensor product basis. We can rewrite this 2D
B-spline projection in matrix form as:
where \(\bar{U}(\bar{\lambda },\bar{\omega },t)\) is the evaluation of \(\varvec{U (x, y, t)}\) and
where \(e_{LF} \) is the vector of Is of length LF, \(e_{KF} \) is the vector of Is of length KF.
A fast block matrix system solution algorithm
The algorithm we use is as follows. We assume that \(N_x \) and \(N_y \) are :
where \(L_x ,L_y \) are lower triangular matrices, and \(U_x ,U_y \) are upper triangular matrices.
Let us simplify the notation by writing (4.12) as
where \(p=P(t)\) and \(c=\bar{U}(\lambda ,\omega ,t). \)The above system can then be rewritten as
Based on a property of the Kronecker product, we can rewrite the above system as
This system can then be solved in 4 steps:
-
Step 1: Solve \((L_x \otimes I_{LF} )\tilde{v}=c\) for \(\tilde{v}\)
-
Step 2: Then solve \((I_{KF} \otimes L_y )v=\tilde{v}\) for \(\varvec{v}\)
-
Step 3: Then solve \((U_x \otimes I_{LF} )\tilde{w}=v\) for \(\varvec{\tilde{w}}\)
-
Step 4: Then solve \((I_{KF} \otimes U_y )w=\tilde{w}\) for \(\varvec{w}\)
where \(\varvec{I}_{LF} \) is the \(LF\times LF\) identity matrix and \(\varvec{I}_{KF} \) is the \(KF\times KF\) identity matrix.
Based on the analysis given in [12]. The cost for step 1 is \(O(N^2p^3)\), the cost of step 2 is \(O(N^2p^3)\), the cost of step 3 \(O(N^2p^3)\) is and the cost of step 4 is \(O(N^2p^3)\). Thus the total cost of for the algorithm is \(O(N^2p^3)\).
If we were to solve (16) simply by the coefficient matrix as a blocks of size \(O(\text{ p }(\text{ LF }\times \text{ LF }))=O(\text{ NP }^2\times \text{ NP }^2)\), the cost of this almost block diagonal with block of this size is \(O(\text{ N }(\text{ NP }^2)^3)=O(\text{ N }^4\text{ P }^6)\) thus by using this algorithm, we have substantial saving.
steps (1) and (2) solve the system
for v.
substitute from step (2) into step (1); we have
Then we can rewrite this equation by using the property of the Kronker product as
A similar argument can be used to show that steps (3) and (4) solve the linear System
At certain points in the algorithm, we know the B-spline coefficients, \(\bar{P}(t)\) and We should evaluate the approximate solution at the collocation points \(\bar{U}(\lambda ,\omega ,t). \)That is, we compute\((N_x \otimes N_y )p=c\), where we know \(N_x ,N_y \)and p . We compute cwith using a matrix multiplication.
4 Numerical implementation of algorithm
We consider the 2D Burgers’ equation (1) the problem domain is \((x,y)\in (0,1)\times (0,1),t\succ 0\)
The exact solution is
We set \(\xi =0.01\), and the boundary conditions is
and initial condition is
The following notation will be used in numerical result.
-
NCOLL : The number of collocation points per subinterval
-
NUM : The number of subintervals
-
AERR : The absolute error
-
RERR : The relative error
-
TERR : The true error
-
RATE: The rate of convergence
5 Conclusion
A new numerical method, which is based on 2D B-spline collocation algorithm. This leads to an approximation of Burgers’ equation by a large system of time-dependent differential algebraic equation, we then solve using a high quality differential algebraic equation solver. The amount of computation and memory is substantially reduced. In Tables 1, 2, for the collocation solutions we compute at t = 1, the observed AERR, RERR, TERR, and corresponding approximate convergence rates. From the two tables we observe that the expected rates of convergence are indeed observed in the 2D case.
References
Bateman, H.: Some recent researches on the motion of fluids. Monthly Weather Rev. 43, 163–170 (1915)
J.M. Burger, A mathematical model illustrating the theory of turbulence, in: Adv. In App. Mech. I, Academic Press, New York, 1948, pp. 171–199
Varoglu, E., Finn, W.D.: Space-time finite element incorporating characteristics for the Burgers’ equation. Int. J. Numer. Meth. Eng. 16, 171–184 (1980)
Evans, D.J., Abdullah, A.R.: The group explicit method for the solution of Burgers’ equation equation. Computing 32, 239–253 (1984)
E.L. Miller, Predictor–corrector studies of Burgers’ model of turbulent Flow, M.S. Thesis, University of Delaware, Newark, DE, 1966
Abd-el-Malek, M.B., El-Mansi, S.M.A.: Group theoretic methods applied to Burgers’ equation. J. Comput. Appl. Math. 115, 1–12 (2000)
Gulsu, M.: A finite difference approach for solution of Burgers’ equation. Apple. Math. Compute. 175, 1245–1255 (2006)
Hassanien, I.A., Salama, A.A., Hosham, H.A.: Fourth-order finite difference method for solving Burgers’ equation. Apple. Math. Comput. 170, 781–800 (2005)
Bihari, B., Harten, A.: Multiresolution schemes for the numerical solution of 2D conservation laws. SIAM J. Sci. Comput. 18, 315–345 (1997)
Aksan, E.N.: A numerical solution of Burgers’ equation by finite element method constructed on the method of discretization in time. Apple. Math. Comput. 170, 895–904 (2005)
Chino, E., Tosaka, N.: Dual reciprocity boundary element analysis of time-independent Burgers’ equation. Eng. Anal. Boundary Elem. 21, 3–47 (1996)
Russell, R.D., Sun, W.: Spline collocation differentiation matrices. SI AM J. Numer. Anal. 34(6), 2274–2287 (1997)
Diaz, J.C., Fairweather, G., Keast, P.: FORTRAN packages for solving certain almost block diagonal linear systems by modified alternate Row and column elimination. ACM Trans. Math. Software 9(3), 358–375 (1983)
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Yousefi, M., Rashidinia, J., Yousefi, M. et al. Numerical solution of Burgers’ equation by B-spline collocation. Afr. Mat. 27, 1287–1293 (2016). https://doi.org/10.1007/s13370-016-0409-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13370-016-0409-0