No label defined (Q3337291)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 3869139
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 3869139 |
Statements
1984
0 references
Model reduction
0 references
stochastic balancing
0 references
0.9304731488227844
0 references
0.8898729681968689
0 references
0.8727576732635498
0 references
0.8710538148880005
0 references
0.8552532196044922
0 references
[For the entire collection see Zbl 0527.00041.] Model reduction is an important problem in system identification, the concept of stochastic balancing was introduced by \textit{U. B. Desai} and \textit{D. Pal} [IEEE Trans. Autom. Control AC-29, 1097-1100 (1984)]. Model reduction, in the discrete-time case, was considered by them too. This paper first formulates the problem in continuous-time; the Markovian representation of a zero mean, stationary, rational, stochastic process with covariance \(\Lambda(\tau)\) can be shown. Let \(N_ F\underline\triangle E X(t)X^ T(t)\) be the state covariance of the FIR for a forward Markov model, and \(N_ b\) be that of the BIR for the backward Markov model. It follows that there exists a transformation T under which \(N_ F\to^{T}N\) and \(N_ b\to^{T}N\) with \(N=diag(\nu_ i)\) where \(\nu^ 2_ j=\lambda_ j(N_ FN_ b)\) denotes the eigenvalues of \((N_ FN_ b)\). This is the balancing transformation which yields the balanced stochastic realization. Once the balanced stochastic realization is obtained, model reduction consists simply of checking to see if \(\nu_ r\gg \nu_{r+1}\) for some \(r<n\). If such an r exists, then the reduced order model is the first \(r\times r\) subsystem of the FIR in balanced coordinates. It is also shown that in both the continuous time and discrete-time cases the reduced order model is asymptotically stable and dissipative, and in the continuous-time case the reduced order model is minimal. The method proposed in this paper is interesting and practical.
0 references