Skip to main content

Table 4 Averages of accumulated iteration counts and time measurements per time step for comparison of classic, physics-based block preconditioners and the newly proposed hybrid preconditioner in the pressure wave example

From: A hybrid interface preconditioner for monolithic fluid–structure interaction solvers

Mesh ID

pw1

pw2

pw3

pw4

pw5

pw6

M

16

32

64

128

192

256

(a) Average accumulated number of linear iterations per time step

   BGS(AMG)

68.6

65.5

56.6

57.0

59.7

64.1

   H-BGS(AMG)

46.1

49.7

42.8

44.8

47.8

50.7

   Savings (%)

32.8

24.1

24.4

21.4

19.9

20.9

   AMG(BGS)

52.6

54.3

41.9

42.4

44.7

49.1

   H-AMG(BGS)

37.0

42.0

33.4

35.0

37.4

41.4

   Savings (%)

29.7

22.7

20.3

17.5

16.3

15.7

(b) Average accumulated linear solver time per time step

   BGS(AMG) (s)

6.4

12.4

17.4

20.8

27.8

31.6

   H-BGS(AMG) (s)

5.7

11.0

15.5

18.2

21.1

24.4

   Savings(%)

10.9

11.3

10.9

12.5

24.1

22.8

AMG(BGS) (s)

5.6

11.4

16.5

19.1

28.8

28.0

   H-AMG(BGS) (s)

5.0

9.6

13.8

15.8

19.4

22.2

   Savings (%)

10.7

15.8

16.4

17.3

32.6

20.7

  1. Iteration counts of the linear solver are not sensitive w.r.t. mesh refinement for all types of preconditioners.
  2. Timings of the linear solver exhibit only slight increases with mesh refinement due to the implementation. Relative savings due to the hybrid preconditioner are reported