Development of preconditioners for numerical simulation of two-phase flow using Krylov subspace methods

• Published in: Progress in nuclear energy (New series) 2021-09, Vol.139, p.103852, Article 103852
• Main Authors: Esmaili, H., Kazeminejad, H., Ahangari, R., Boustani, E.
• Format: Article
• Language: English
• Subjects: Central processing units; CPUs; Drift flux model; Heat transfer; Krylov solvers; Mathematical analysis; Mathematical models; Matrices (mathematics); Numerical analysis; Preconditioning; Robustness (mathematics); Simulation; Subspace methods; Two phase flow
• ISSN: 0149-1970; 1878-4224
• DOI: 10.1016/j.pnucene.2021.103852

In general, Newton-Krylov methods alone are not effective for solving non-symmetric matrices arising from the simulation of one-dimensional subcooled flow boiling inside a vertical tube with upward flow using the Drift-Flux Model (DFM); and, appropriate preconditioning is required. Because the key to the success of Newton-Krylov methods is the application of efficient and robust preconditioners, it is necessary to study the performance of the various Newton-Krylov methods preconditioned with different preconditioners. Therefore, the performance of three different preconditioners combined with four different Newton-Krylov methods, along with their implementation process, has been investigated. The results show that the preconditioning technique effectively reduces the number of iterations and the CPU time. Also, it is found that among the proposed preconditioners, the semi-implicit physics-based preconditioning (SI-PBP) has the best performance in terms of reducing the number of iterations and CPU time. Therefore, it can be used for simulation of other engineering problems. •The performance of Newton-Krylov methods with different preconditioners, along with their implementation process, is studied.•The verification and validation are performed by comparing the computational results with the relevant experimental data.•The results show that the preconditioning technique effectively reduces the number of iterations and the CPU time.•For all implemented Newton-Krylov solvers, the use of Si-PBP results in the lowest CPU time and iteration numbers.