[This linear system can be written in matrix form as
(6.16)
where the 2N x 2N coefficient matrix A = [A1A2] is defined by
(6.17)
and]
(6.18)
The 2N vector of unknowns x is defined by
(6.19)
and the known right-hand side is
(6.20)
As shown below, with the DQMOM approximation, the right-hand side of Eq. (6.10) is closed in terms of N weights and abscissas. As N increases, the quadrature approximation will approach the exact value, although at a higher computational cost.
If the abscissas dq are unique, then A will be full rank. For this case, the source terms for the transport equations of the weights ωq and weighted diameters δq can be found by inverting A in Eq. (6.16):
There are cases for which the matrix A is not full rank (the matrix is singular). These cases can occur when one or more of the abscisses dq are non-distinct.
|
[page 8]
This linear system can be written in matrix form as:
(13)
where the 2N x 2N coefficient matrix A = [A1A2] is defined by
(14)
and
(15)
The 2N vector of unknowns α is defined by
(16)
and the known right-hand side is
(17)
[page 9]
As shown below, with the DQMOM approximation the right-hand side of Eq. 18 is closed in terms of the N weights and abscissas. [...] As N increases, the quadrature approximation will approach the exact value, albeit at a higher computational cost.
If the abscissas Lα are unique, then A will be full rank. For this case, the source terms for the transport equations of the weights ωα and weighted lengths can be found simply by inverting A in Eq. 13:
If at any point in the computational domain two abscissas are equal, then the matrix A is not full rank (or the matrix is singular), and therefore it is impossible to invert it.
|