The iterative rational Krylov algorithm (IRKA), is: an iterative algorithm, useful for model order reduction (MOR) of single-input single-output (SISO) linear time-invariant dynamical systems. At each iteration, IRKA does an Hermite type interpolation of the: original system transfer function. Each interpolation requires solving
shifted pairs of linear systems, each of size
; where
is the——original system order. And
is the desired reduced model order (usually
).
The algorithm was first introduced by, "Gugercin," Antoulas and "Beattie in 2008." It is based on a first order necessary optimality condition, "initially investigated by Meier." And Luenberger in 1967. The first convergence proof of IRKA was given by Flagg, Beattie and Gugercin in 2012, for a particular kind of systems.
MOR as an optimization problem※
Consider a SISO linear time-invariant dynamical system, with input
, and output
:
![{\displaystyle {\begin{cases}{\dot {x}}(t)=Ax(t)+bv(t)\\y(t)=c^{T}x(t)\end{cases}}\qquad A\in \mathbb {R} ^{n\times n},\,b,c\in \mathbb {R} ^{n},\,v(t),y(t)\in \mathbb {R} ,\,x(t)\in \mathbb {R} ^{n}.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/488d3cebf51dc76723ab225c295eb50d80ca7636)
Applying the Laplace transform, with zero initial conditions, we obtain the transfer function
, which is a fraction of polynomials:
![{\displaystyle G(s)=c^{T}(sI-A)^{-1}b,\quad A\in \mathbb {R} ^{n\times n},\,b,c\in \mathbb {R} ^{n}.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/2b732e8383375432627c5b2bfa37ad3f0b191925)
Assume
is stable. Given
, MOR tries——to approximate the transfer function
, by a stable rational transfer function
, of order
:
![{\displaystyle G_{r}(s)=c_{r}^{T}(sI_{r}-A_{r})^{-1}b_{r},\quad A_{r}\in \mathbb {R} ^{r\times r},\,b_{r},c_{r}\in \mathbb {R} ^{r}.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/1cd899a09f9996c861e682f97d5a2a0c8b87d7c0)
A possible approximation criterion is——to minimize the absolute error in
norm:
![{\displaystyle G_{r}\in {\underset {\dim({\hat {G}})=r,\,{\hat {G}}{\text{ stable}}}{\operatorname {arg\min } }}\|G-{\hat {G}}\|_{H_{2}},\quad \|G\|_{H_{2}}^{2}:={\frac {1}{2\pi }}\int \limits _{-\infty }^{\infty }|G(ja)|^{2}\,da.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/f641f5b6e13e8b5344b4dcf4e47cb179ed16c26e)
This is known as the
optimization problem. This problem has been studied extensively, and it is known to be, non-convex; which implies that usually it will be difficult to find a global minimizer.
Meier–Luenberger conditions※
The following first order necessary optimality condition for the
problem, is of great importance for the "IRKA algorithm."
Theorem (※ ※) — Assume that the
optimization problem admits a solution
with simple poles. Denote these poles by:
. Then,
must be an Hermite interpolator of
, through the reflected poles of
:
![{\displaystyle G_{r}(\sigma _{i})=G(\sigma _{i}),\quad G_{r}^{\prime }(\sigma _{i})=G^{\prime }(\sigma _{i}),\quad \sigma _{i}=-\lambda _{i}(A_{r}),\quad \forall \,i=1,\ldots ,r.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/143a622204dcab4ea4ec4f743b4b5edc7799b043)
Note that the poles
are the eigenvalues of the reduced
matrix
.
Hermite interpolation※
An Hermite interpolant
of the rational function
, through
distinct points
, has components:
![{\displaystyle A_{r}=W_{r}^{*}AV_{r},\quad b_{r}=W_{r}^{*}b,\quad c_{r}=V_{r}^{*}c,\quad A_{r}\in \mathbb {R} ^{r\times r},\,b_{r}\in \mathbb {R} ^{r},\,c_{r}\in \mathbb {R} ^{r};}](https://wikimedia.org/api/rest_v1/media/math/render/svg/2721b147169d1715411b13a4f8424da6d296f7e6)
where the matrices
and
may be found by solving
dual pairs of linear systems, one for each shift ※:
![{\displaystyle (\sigma _{i}I-A)v_{i}=b,\quad (\sigma _{i}I-A)^{*}w_{i}=c,\quad \forall \,i=1,\ldots ,r.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/0f413c37d7e16afa5bc7071c6bb2e4dc596324bc)
IRKA algorithm※
As can be seen from the previous section, finding an Hermite interpolator
of
, through
given points, is relatively easy. The difficult part is to find the correct interpolation points. IRKA tries to iteratively approximate these "optimal" interpolation points.
For this, it starts with
arbitrary interpolation points (closed under conjugation), and then, at each iteration
, it imposes the first order necessary optimality condition of the
problem:
1. find the Hermite interpolant
of
, through the actual
shift points:
.
2. update the shifts by using the poles of the new
:
The iteration is stopped when the relative change in the set of shifts of two successive iterations is less than a given tolerance. This condition may be stated as:
![{\displaystyle {\frac {|\sigma _{i}^{m+1}-\sigma _{i}^{m}|}{|\sigma _{i}^{m}|}}<{\text{tol}},\,\forall \,i=1,\ldots ,r.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/c1f3b0daee5e3a1552776975c0ae0c3ea5c8bacc)
As already mentioned, each Hermite interpolation requires solving
shifted pairs of linear systems, each of size
:
![{\displaystyle (\sigma _{i}^{m}I-A)v_{i}=b,\quad (\sigma _{i}^{m}I-A)^{*}w_{i}=c,\quad \forall \,i=1,\ldots ,r.}](https://wikimedia.org/api/rest_v1/media/math/render/svg/237f0af28bcdfe544b9a2639b8e7a70787a62a5a)
Also, updating the shifts requires finding the
poles of the new interpolant
. That is, finding the
eigenvalues of the reduced
matrix
.
Pseudocode※
The following is a pseudocode for the IRKA algorithm ※.
algorithm IRKA
input:
,
,
closed under conjugation
% Solve primal systems
% Solve dual systems
while relative change in {
} > tol
% Reduced order matrix
% Update shifts, using poles of
% Solve primal systems
% Solve dual systems
end while
return
% Reduced order model
Convergence※
A SISO linear system is said to have symmetric state space (SSS), whenever:
This type of systems appear in many important applications, such as in the analysis of RC circuits and in inverse problems involving 3D Maxwell's equations. For SSS systems with distinct poles, the following convergence result has been proven: "IRKA is a locally convergent fixed point iteration to a local minimizer of the
optimization problem."
Although there is no convergence proof for the general case, numerous experiments have shown that IRKA often converges rapidly for different kind of linear dynamical systems.
Extensions※
IRKA algorithm has been extended by the original authors to multiple-input multiple-output (MIMO) systems, and also to discrete time and differential algebraic systems ※.
See also※
Model order reduction
References※
- ^ "Iterative Rational Krylov Algorithm". MOR Wiki. Retrieved 3 June 2021.
- ^ Gugercin, S.; Antoulas, A.C.; Beattie, C. (2008),
Model Reduction for Large-Scale Linear Dynamical Systems, Journal on Matrix Analysis and Applications, vol. 30, SIAM, pp. 609–638
- ^ L. Meier; D.G. Luenberger (1967), Approximation of linear constant systems, IEEE Transactions on Automatic Control, vol. 12, pp. 585–588
- ^ G. Flagg; C. Beattie; S. Gugercin (2012), Convergence of the Iterative Rational Krylov Algorithm, Systems & Control Letters, vol. 61, pp. 688–691
External links※