The Solution of Two-Phase Inverse Stefan Problem Based on a Hybrid Method with Optimization

Yu, Yang; Luo, Xiaochuan; Cui, Haijuan

doi:https://doi.org/10.1155/2015/319054

Mathematical Problems in Engineering

On this page

Abstract Introduction Conclusion Acknowledgments References Copyright Related Articles

Research Article | Open Access

Volume 2015 | Article ID 319054 | https://doi.org/10.1155/2015/319054

The Solution of Two-Phase Inverse Stefan Problem Based on a Hybrid Method with Optimization

Yang Yu,¹Xiaochuan Luo,¹and Haijuan Cui¹

Academic Editor: Evangelos J. Sapountzakis

Received20 Mar 2015

Revised30 Apr 2015

Accepted04 May 2015

Published09 Jun 2015

Abstract

The two-phase Stefan problem is widely used in industrial field. This paper focuses on solving the two-phase inverse Stefan problem when the interface moving is unknown, which is more realistic from the practical point of view. With the help of optimization method, the paper presents a hybrid method which combines the homotopy perturbation method with the improved Adomian decomposition method to solve this problem. Simulation experiment demonstrates the validity of this method. Optimization method plays a very important role in this paper, so we propose a modified spectral DY conjugate gradient method. And the convergence of this method is given. Simulation experiment illustrates the effectiveness of this modified spectral DY conjugate gradient method.

1. Introduction

The process of solidification/melting occurs in many industrial fields, such as steel-making continuous casting. In the continuous casting process [1, 2], molten steel comes into slab by solidification. And this process is shown in Figure 1. Because of the heat symmetry, we only consider 1/2 of slab thickness, and we divide this domain into two regions by the freezing front, which is shown in Figure 2. stands for the solid phase and is taken by the liquid phase. The solidification problem of continuous casting is associated with Stefan problem, which is a model of the process with a phase change.

Many researches have focused on solving Stefan problems [3–7]. Grzymkowski and Slota [8, 9] applied the Adomian decomposition method (ADM) combined with optimization for solving the direct and the inverse one-phase Stefan problem, and Hetmaniok et al. [10] used the Adomian decomposition method (ADM) and variational iteration method to solve the one-phase Stefan moving boundary problems. In comparison to the studies on the one-phase Stefan problems, researches on two-phase Stefan problems [11] are much more scarce. Słota [12] applied homotopy perturbation method (HPM) to solve the two-phase inverse Stefan problem. However, in their formulation, these two-phase inverse Stefan problems were solved when the position of the moving interface is known. While the position of the moving interface is unknown and no temperature or heat flux boundary conditions are specified on one part of the boundary, this two-phase inverse Stefan problem may be difficult to be solved. This problem is close to the problem which occurs in the continuous casting. Therefore, this paper focuses on solving this problem by a hybrid method with optimization which combines the homotopy perturbation method (HPM) with the improved Adomian decomposition method (IADM).

This paper is organized as follows. In Section 2, we give the description of the two-phase Stefan problem. And the hybrid method for solving this problem is introduced in Section 3. We give a modified spectral DY conjugate gradient method and the convergence of this method in Section 4. In Section 5, some numerical results are given to illustrate the validity of the hybrid method.

2. The Description of Two-Phase Inverse Stefan Problem

In this section, the two-phase inverse Stefan problem is considered. The two domains are described (see Figure 2):and their boundaries with unknown function are

In practice, the measurement of temperature on the boundary may not be easily obtained and the position of the moving interface described by means of function is unknown. Therefore, the designed two-phase inverse Stefan problem consists in the calculation of the temperature distribution in the domains and , respectively, as well as in the reconstruction of the functions describing moving interface . In order to solve this inverse problem, the additional information is given. In such a situation, the final time internal temperature at on the boundary is known, which can be measured in practice. The temperature distribution is determined by the functions and , which are defined in the domains and , respectively, and they satisfy the following heat conduction equations:where and refer to time and spatial location and and are the thermal diffusivity in the solid phase and the liquid phase, respectively. On the boundaries and , the initial conditions are described in the following:On the boundary , it satisfies the Dirichlet boundary or Neumann boundary condition: On the boundary , it satisfies the Dirichlet boundary conditions:On the phase change moving interface (), it satisfies the condition of temperature continuity and the Stefan condition:where is the phase change temperature. Hence,where is the latent heat of fusion per unit volume and and are the thermal conductivity in the solid and liquid phase, respectively. The final time internal temperature on the boundary satisfies

3. The Solution of Two-Phase Inverse Stefan Problem

3.1. The Homotopy Perturbation Method

The homotopy perturbation method was presented in [13, 14]; it arose as a combination of two other methods: the perturbation method and the homotopy technique from topology. Homotopy perturbation method was applied on differential equations, partial differential equations, and the direct and inverse problems of heat transfer by many authors [15–17].

Homotopy perturbation method was used to find the exact or the approximate solutions of the linear and nonlinear integral equations [16, 18], the two-dimensional integral equations [19–21], and the integrodifferential equations [22–25]. This method enables seeking a solution of the following operator equation:where is an operator, is a sought function, and is a known function on the domain of . The operator is presented in the form of sum:where defines the linear operator and is the remaining part of operator . So (18) can be written asWe define a new operator which is called the homotopy operator in the following:where is an embedding parameter, , and defines the initial approximation of a solution of (18); (21) is transformed into the following form:Therefore,For the solution of operator equation is equivalent to solution of problem , whereas, for , the solution of operator equation is equivalent to solution of problem (20). Thus, a monotonous change of parameter from zero to one is just that continuous change of the trivial problem to the original problem. In topology, this is called homotopy. According to the HPM, we assume that the solution of (21) and (22) can be written as a power series in :setting , the approximate solution of (20) is obtained:in most cases, the series in (25) is convergent, and in [15] an additional remark on the convergence of this series can be found.

3.2. The Improved Adomian Decomposition Method

The Adomian decomposition method was presented by Adomian [26]. This method was used to solve a wide variety of linear and nonlinear problems. Several other researchers had developed a modification to the ADM [27]. The modification usually involves a slight change and it is aimed at improving the accuracy of the series solution.

The Adomian decomposition method is able to solve a wide class of nonlinear operator equations [28] in the formwhere is a nonlinear operator, is a known element from Hilbert space and is the sought element from Hilbert space . The operator can be written as where is an invertible linear operator, is a linear operator, and is a nonlinear operator. The solution of (26) admits the decomposition into an infinite series of componentsthe nonlinear term is equated to an infinite series is the Adomian polynomials, which can be determined bySubstituting (28), (29), (30), and (31) into operator equation (26) and using the inverse operator , the following equation can be obtained:where is the function dependent on the initial and boundary conditions.

The improved Adomian decomposition method [27]: it is important to note that the improved Adomian decomposition method is based on the assumption that the function can be divided into two parts, namely, and . Under this assumption, the following equation can be obtained:Accordingly, a slight variation is proposed on the components and . The suggestion is that only the part is assigned to the component , whereas the remaining part is combined with other terms given in (33) to define . Consequently, the recursive relation is written as follows:

3.3. The Solution of the Two-Phase Inverse Stefan Problem

In this section, we introduce the solution of the two-phase inverse Stefan problem. We split the two-phase inverse Stefan problem into two problems. The first is the determination of in domain and the moving interface which is satisfying (7), (8), (10), and (13). Once the boundary and the heat flux have been obtained, the second problem for determining the temperature in domain is solved.

The solution of and moving interface based on HPM: because the heat flux on the moving position is unknown in this two-phase inverse Stefan problem, the HPM is used to compute . In this section, the homotopy perturbation method is applied to solve the inverse problem. We make the homotopy map for (22):the solution to this equation,is sought in the form of a power series in and it satisfieswhere ; by substituting (37) into (36), the following equation can be obtained:Next, by comparing the same powers of parameter in (38), the following equation can be obtained:Equations (39) and (40) must be supplemented by the boundary conditions. Therefore, for (39), we set the following boundary conditions:and for (40) we set the conditions in the following ():The above conditions are selected. So we give the -order approximate solution: because are determined by the unknown function . This function is derived in the formThus we define a cost function according to (10) and (17):Therefore, are chosen according to the minimum of (47). The moving interface and the solution of in domain can be obtained.

The boundary condition (13) on the boundary and the condition of temperature continuity (15) should be fulfilled for . In this way, while looking for the solution of this problem in domain , the initial approximation will be found by solving the system of (39) with conditions (41) and (42). Conversely, functions , , are determined by the recurrent solution of (40) with conditions (42) and (44). Consequently, the moving interface and the solution of in domain can be obtained by the homotopy perturbation method with optimization. According to the Stefan condition (16) on the phase change moving interface , the temperature in domain can be recovered by IADM with optimization method.

The solution of based on IADM with optimization: in this problem, we consider the operator equations (27), whereThe inverse operator is given by the following equation:The boundary condition and the Stefan condition from (14) are used to obtain the following equation:Hencethen the following recursive relation can be obtained:because and are determined by the above HPM with optimization method, can be obtained by (16). An approximation solution is expressed in the formand in this inverse Stefan problem, is unknown on the boundary and it is derived in the following form:where and are the basis function. The coefficients are selected to show minimal deviation of function (55) from conditions (11) and (15). Thus are chosen according to the minimum of the following functional:where is a nonlinear function. The optimization method plays a very important role in solving this two-phase inverse Stefan problem. Therefore, a modified spectral DY conjugate gradient method is presented. In Section 4, this optimization method is introduced.

Remark. The IADM with optimization method can be used to solve the direct and inverse one-phase Stefan problem which is designed in the literature [8, 9]. And this method is better than ADM with optimization method. In Section 5, we give two examples of one-phase Stefan problem to illustrate the effectiveness of the IADM with optimization method.

4. A Modified Spectral DY Conjugate Gradient Method

Conjugate gradient methods are very efficient for solving the unconstrained optimization problem . Their iterative formula is given as follows:where step-size is positive, , and is a scalar. In addition, is a step length which is calculated by the line search. is a very important factor in conjugate gradient methods; the FR, PRP, DY, and CD methods [29] are used for choosing this scalar. Among them, the DY method is regarded as one of the efficient conjugate gradient methods. The scalar is given bywhere and stands for Euclidean norm. Recently, Birgin and Martłnez [30] presented a spectral conjugate gradient method which is combining conjugate gradient method and spectral gradient method. And the direction is obtained bywhere is a parameter with and is regarded as spectral gradient. However, according to [31], the spectral conjugate gradient method can not guarantee producing the descent directions.

Based on the above analysis, in the following, we describe a modified spectral DY conjugate gradient method and give the convergence of this method with the Wolfe type line search rules.

4.1. The Modified Spectral DY Conjugate Gradient Method

The new method is introduced:where and . This method is called the MDY method.

Algorithm 1. Consider the following.

Step 1. The following are given: , , and the max iteration step .

Step 2. Set ; compute ; if , then stop.

Step 3. Find according to Wolfe type line search rule.

Step 4. Let and ; if , then stop.

Step 5. Compute by (59).

Step 6. Set ; go to Step 3.

4.2. Convergence Analysis

From the above, the MDY conjugate gradient method provides a descent direction for the objective function. In the following, we give the convergence of this algorithm with the Wolfe type line search rules. Firstly, we introduce some assumptions:(i)The level set is bounded.(ii)In some neighborhood of , is continuously differentiable and its gradient is Lipschitz continuous; namely, there exists a constant such thatIt is implied from the above assumptions that there exist two positive constants and such thatThe Wolfe type line search [29] isand the strong Wolfe line search iswhere .

Lemma 2. Let the sequences and be generated by the proposed algorithm (Algorithm 1) with Wolfe type line search rules; then and

Proof. According to (59), we haveIf we choose , thenaccording to (65), we haveSo . For , according to (67), we haveby induction. So we obtainthe proof is completed.

Theorem 3. Let the sequences and be generated by the proposed algorithm (Algorithm 1) with Wolfe type line search rules; thenwhere is a constant.

Proof. According to (59), we haveFrom the Wolfe type line search rules (65), we can obtainaccording to (73) and (74), we have we define . Therefore, we can obtain inequality (72).

Lemma 4. Suppose that assumptions (i) and (ii) hold; let and be generated by the proposed algorithm (Algorithm 1) with Wolfe type line search rules (62) and (63); then one has

Proof. From assumption (ii), we haveso we can obtainFrom inequality (63), we obtainAccording to inequalities (78) and (79), we haveSo we can obtainOwing to inequality (81), we haveFrom (82), we can obtain inequality (76).

Theorem 5. Suppose that assumptions (i) and (ii) hold. Let and be generated by the algorithm with Wolfe type line search rules. Then one has

Proof. By contradiction, we suppose that there exists a positive constant , such thatholds for all . From (59), we haveAccording to Lemma 4 and the above equation, we haveMultiplying the above inequality by , we haveThen from inequality (83), we havewhich contradicts (76). Therefore, (83) holds.

5. Numerical Simulation Experiments

In order to illustrate the validity of the IADM with optimization method, we introduce the one-phase inverse Stefan problem in Example 1. From the literature [9], we know the inverse Stefan problem consists in the calculation of the temperature distribution in the domain and in the reconstruction of the temperature distribution on the boundary, when the position of the function which describes the moving interface is known. And then, in Example 2, we solve the two-phase inverse Stefan problem based on the hybrid method which combines the Adomian decomposition method with the homotopy perturbation method.

Example 1. In this example, the one-phase inverse Stefan problem [9] is taken for the following values of parameters: , , , , , , and . Then the exact solution of such formulated one-phase Stefan problem can be found from the following functions:The following equation,is taken as basis function. The approximate solutions are compared with the exact solution and the values of the absolute errors are calculated fromwhere ( stands for the boundary condition on in the literature [9] and is the moving interface), is the domain of the temperature distribution, is the exact value of function , is the approximate value of function , is the exact distribution of the temperature in the domain, and is the reconstructed distribution of temperature. Percentage relative errors are calculated from

The comparison of the ADM and the IADM for reconstructing distribution of the temperature on boundary in the literature [9] is shown in Figure 3 for , and , . From Figure 3, the method based on the improved Adomian decomposition method (IADM) is better than the Adomian decomposition method (ADM).

(a)

(b)

(c)

Example 2. This example is introduced to illustrate the effectiveness of the hybrid method which combines the Adomian decomposition method with the homotopy perturbation method, which is presented in the previous sections. Here, we consider an example of the two-phase inverse Stefan problem [12], in which , , , , , , , , , and , , , and . The exact solution of this inverse Stefan problem is described by the following functions:

First, we use HPM with optimization method to solve and . As the initial approximation , this fulfills the initial condition:By using the function in (39) and the conditions (41) and (42), the following system of equations can be obtained:With the conditionsthis system of equations can be solved. And the solution of this system is shown in the following:

Next, we designate the functions for in the following:and the conditions are shown:By using (98) and the boundary conditions (99), we derive for ; according to (45), the approximate solution is obtained. Here, we define basis functions according to (46):We use the modified spectral DY (MDY) conjugate gradient method to solve function (47). The solution of moving interface is shown in Figure 4. So the temperature is derived in domain . Values of the error in reconstruction of the position of the moving interface and the temperature distribution are shown in Table 1.

(a)

(b)

(c)

In order to verify the validity of the modified spectral DY (MDY) conjugate gradient method, we choose , , , the initial point , and the max step number . The comparison results of DY and MDY are shown in Figure 5 and Table 2. Comparing to DY conjugate gradient method, the MDY conjugate gradient method is slightly better.

The moving interface and in domain are obtained by the homotopy perturbation method with optimization. Next, with the help of condition (16), we can determine the temperature in domain by the improved Adomian decomposition method (IADM) with optimization. Without loss of generality, we choose the basis functions

The exact and the reconstructed distribution of the temperature on boundary are shown in Figure 6 for different number of basis functions . And the errors in reconstruction of the temperature (on boundary in Figure 2) and the temperature distribution are shown in Table 3. The inversion results of the function match the exact value very well.

The measured temperature at the final time on the boundary in Figure 2 is perturbed by the random error. So in the next section we consider the influence of measurement errors on the results. We set the measured temperature with an error of and . We can obtain the moving interface and ; the results are shown in Figures 7 and 8.

In Figures 7 and 8, the exact and the reconstructed distribution of the temperature on moving interface and the boundary are shown, and the errors in reconstruction of the function describing the moving interface and boundary conditions are compiled in Table 4. The obtained reconstruction results illustrate that the functions and are reconstructed very well. From the results, the hybrid method with optimization is stable with the input errors. When the input data are contaminated by the errors, the error of the reconstructed boundary conditions does not exceed the error of the input data.

6. Conclusion

Based on the homotopy perturbation method (HPM) with the improved Adomian decomposition method (IADM), a hybrid method with optimization is proposed to solve the two-phase inverse Stefan problem. This problem is more realistic from the practical point of view. The simulation experiment is used to verify this method. Experimental results match the exact value very well. Furthermore, this investigated method can also be used for solving one-phase direct Stefan problems.

Optimization method is very crucial in this paper, so a modified spectral DY conjugate gradient method is presented. And we give the convergence of this MDY method. Simulation experiment illustrates the validity of this optimization method.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (60974091, 61333006, and 61473074) and the Fundamental Research Funds for the Central Universities (N120708001).

References

A. Fic, A. J. Nowak, and R. Bialecki, “Heat transfer analysis of the continuous casting process by the front tracking BEM,” Engineering Analysis with Boundary Elements, vol. 24, no. 3, pp. 215–223, 2000.
View at: Publisher Site | Google Scholar | Zentralblatt MATH
C. A. Santos, J. A. Spim, and A. Garcia, “Mathematical modeling and optimization strategies (genetic algorithm and knowledge base) applied to the continuous casting of steel,” Engineering Applications of Artificial Intelligence, vol. 16, no. 5-6, pp. 511–527, 2003.
View at: Publisher Site | Google Scholar
R. Kamal, A. Hojatollah, A. R. Jamal, and P. Kourosh, “Application of mesh-free methods for solving the inverse one dimensional Stefan problem,” Engineering Analysis with Boundary Elements, vol. 40, pp. 1–21, 2014.
View at: Google Scholar
D. Slota, “Direct and inverse one-phase Stefan problem solved by the variational iteration method,” Computers & Mathematics with Applications, vol. 54, no. 7-8, pp. 1139–1146, 2007.
View at: Publisher Site | Google Scholar | MathSciNet
B. T. Johansson, D. Lesnic, and T. Reeve, “A method of fundamental solutions for the one-dimensional inverse Stefan problem,” Applied Mathematical Modelling. Simulation and Computation for Engineering and Environmental Systems, vol. 35, no. 9, pp. 4367–4378, 2011.
View at: Publisher Site | Google Scholar | Zentralblatt MATH | MathSciNet
D. Slota, “The application of the homotopy perturbation method to one-phase inverse Stefan problem,” International Communications in Heat and Mass Transfer, vol. 37, no. 6, pp. 587–592, 2010.
View at: Publisher Site | Google Scholar
R. Grzymkowski, M. Pleszczyński, and D. Slota, “Comparing the Adomian decomposition method and the Runge-Kutta method for solutions of the Stefan problem,” International Journal of Computer Mathematics, vol. 83, no. 4, pp. 409–417, 2006.
View at: Publisher Site | Google Scholar | MathSciNet
R. Grzymkowski and D. Slota, “Stefan problem solved by Adomian decomposition method,” International Journal of Computer Mathematics, vol. 82, no. 7, pp. 851–856, 2005.
View at: Publisher Site | Google Scholar | MathSciNet
R. Grzymkowski and D. Slota, “One-phase inverse Stefan problem solved by adomian decomposition method,” Computers & Mathematics with Applications, vol. 51, no. 1, pp. 33–40, 2006.
View at: Publisher Site | Google Scholar | MathSciNet
E. Hetmaniok, D. Słota, R. Wituła, and A. Zielonka, “Comparison of the Adomian decomposition method and the variational iteration method in solving the moving boundary problem,” Computers & Mathematics with Applications, vol. 61, no. 8, pp. 1931–1934, 2011.
View at: Publisher Site | Google Scholar | MathSciNet
B. T. Johansson, D. Lesnic, and T. Reeve, “A meshless method for an inverse two-phase one-dimensional linear Stefan problem,” Inverse Problems in Science and Engineering, vol. 21, no. 1, pp. 17–33, 2013.
View at: Publisher Site | Google Scholar | Zentralblatt MATH | MathSciNet
D. Słota, “Homotopy perturbation method for solving the two-phase inverse stefan problem,” Numerical Heat Transfer Part A: Applications, vol. 59, no. 10, pp. 755–768, 2011.
View at: Publisher Site | Google Scholar
J. H. He, “Homotopy perturbation technique,” Computer Methods in Applied Mechanics and Engineering, vol. 178, no. 3-4, pp. 257–262, 1999.
View at: Publisher Site | Google Scholar | MathSciNet
J.-H. He, “An elementary introduction to the homotopy perturbation method,” Computers & Mathematics with Applications, vol. 57, no. 3, pp. 410–412, 2009.
View at: Publisher Site | Google Scholar | MathSciNet
C. Chun, H. Jafari, and Y.-I. Kim, “Numerical method for the wave and nonlinear diffusion equations with the homotopy perturbation method,” Computers & Mathematics with Applications, vol. 57, no. 7, pp. 1226–1231, 2009.
View at: Publisher Site | Google Scholar | MathSciNet
A. Yildirim, “Analytical approach to fractional partial differential equations in fluid mechanics by means of the homotopy perturbation method,” International Journal of Numerical Methods for Heat & Fluid Flow, vol. 20, no. 2, pp. 186–200, 2010.
View at: Publisher Site | Google Scholar
E. Hetmaniok, I. Nowak, D. Slota, R. Witula, and A. Zielonka, “Solution of the inverse heat conduction problem with neumann boundary condition by using the homotopy perturbation method,” Thermal Science, vol. 17, no. 3, pp. 643–650, 2013.
View at: Publisher Site | Google Scholar
J. Biazar and H. Ghazvini, “Numerical solution for special non-linear Fredholm integral equation by HPM,” Applied Mathematics and Computation, vol. 195, no. 2, pp. 681–687, 2008.
View at: Publisher Site | Google Scholar | MathSciNet
X. Zhang, J. Zhao, J. Liu, and B. Tang, “Homotopy perturbation method for two dimensional time-fractional wave equation,” Applied Mathematical Modelling, vol. 38, no. 23, pp. 5545–5552, 2014.
View at: Publisher Site | Google Scholar | MathSciNet
M. Ghasemi, M. Tavassoli Kajani, and E. Babolian, “Application of He's homotopy perturbation method to nonlinear integro-differential equations,” Applied Mathematics and Computation, vol. 188, no. 1, pp. 538–548, 2007.
View at: Publisher Site | Google Scholar | MathSciNet
A. Golbabai and M. Javidi, “Application of He's homotopy perturbation method for $n$ th-order integro-differential equations,” Applied Mathematics and Computation, vol. 190, no. 2, pp. 1409–1416, 2007.
View at: Publisher Site | Google Scholar | MathSciNet
J. Biazar, Z. Ayati, and M. R. Yaghouti, “Homotopy perturbation method for homogeneous Smoluchowski's equation,” Numerical Methods for Partial Differential Equations, vol. 26, no. 5, pp. 1146–1153, 2010.
View at: Publisher Site | Google Scholar | MathSciNet
A. Alawneh, K. Al-Khaled, and M. Al-Towaiq, “Reliable algorithms for solving integro-differential equations with applications,” International Journal of Computer Mathematics, vol. 87, no. 7, pp. 1538–1554, 2010.
View at: Publisher Site | Google Scholar | Zentralblatt MATH | MathSciNet
H. Aminikhah and J. Biazar, “A new analytical method for solving systems of Volterra integral equations,” International Journal of Computer Mathematics, vol. 87, no. 5, pp. 1142–1157, 2010.
View at: Publisher Site | Google Scholar
J. Biazar, B. Ghanbari, M. G. Porshokouhi, and M. G. Porshokouhi, “He's homotopy perturbation method: a strongly promising method for solving non-linear systems of the mixed Volterra-Fredholm integral equations,” Computers & Mathematics with Applications, vol. 61, no. 4, pp. 1016–1023, 2011.
View at: Publisher Site | Google Scholar | MathSciNet
G. Adomian, Stochastic Systems, vol. 169 of Mathematics in Science and Engineering, Academic Press, New York, NY, USA, 1983.
View at: MathSciNet
M. Almazmumy, F. A. Hendi, H. O. Bakodah, and H. Alzumi, “Recent modifications of Adomian decomposition method for initial value problem in ordinary differential equations,” American Journal of Computational Mathematics, vol. 2, no. 3, pp. 228–234, 2012.
View at: Publisher Site | Google Scholar
D. Lesnic, “Convergence of Adomian's decomposition method: periodic temperatures,” Computers & Mathematics with Applications, vol. 44, no. 1-2, pp. 13–24, 2002.
View at: Publisher Site | Google Scholar
W. Cao, K. R. Wang, and Y. L. Wang, “Global convergence of a modified spectral CD conjugate gradient method,” Journal of Mathematical Research & Exposition, vol. 31, pp. 261–268, 2011.
View at: Google Scholar
E. G. Birgin and J. M. Martłnez, “A spectral conjugate gradient method for unconstrained optimization,” Applied Mathematics and Optimization, vol. 43, no. 2, pp. 117–128, 2001.
View at: Publisher Site | Google Scholar | MathSciNet
L. Zhang, W. J. Zhou, and D. H. Li, “Global convergence of a modified Fletcher-Reeves conjugate gradient method with Armijo-type line search,” Numerische Mathematik, vol. 104, no. 4, pp. 561–572, 2006.
View at: Publisher Site | Google Scholar | MathSciNet

Copyright

Copyright © 2015 Yang Yu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

1784

Downloads

953

Citations