InternallyBalancedDecomposition

InternallyBalancedDecomposition[ssm]

yields the internally balanced decomposition of the state-space model ssm.

Details and Options

  • For a standard StateSpaceModel the result is a list {p,bssm}, where p is the similarity transformation matrix and bssm is the internally balanced form of ssm.
  • For a descriptor StateSpaceModel the result is a list {{p, q},bssm}, where p and q are a pair of transformation matrices.
  • InternallyBalancedDecomposition accepts a Method option with the following settings:
  • Automaticautomatically choose method
    "Eigensystem"use eigenvalue decomposition
    "SingularValues"use singular value decomposition
  • The methods "Eigensystem" and "SingularValues" call Eigensystem and SingularValueDecomposition, respectively. In each case, the additional options relevant to the corresponding function can be specified as Method->{"name",opt1-> val1,opt2-> val2,}.

Examples

open allclose all

Basic Examples  (1)

The internally balanced realization of a state-space model:

Scope  (3)

The internally balanced realization of a SISO system:

The two realizations are different forms of the same model:

The balanced realization of a MIMO system:

The balanced realization of a descriptor system:

Applications  (1)

In a balanced realization, each state is just as controllable as it is observable:

Get an approximation to the model by truncating the least controllable and observable mode:

Get submatrices of the balanced model:

Get an approximation by residualizing the least controllable and observable mode:

The truncated model better approximates the system during the transients, and the residualized model better approximates the system at steady state:

Properties & Relations  (2)

The ControllabilityGramian and ObservabilityGramian are equal for a balanced system:

The diagonal entries are given by Hankel singular values for the original system:

The original and balanced realizations are related by a similarity transformation:

Use StateSpaceTransform to transform the original system:

The system matrices from the balanced and transformed system are identical:

Possible Issues  (2)

The state-space model must be both completely controllable and observable:

The state-space model must be asymptotically stable:

It is only marginally stable:

Introduced in 2010
 (8.0)
 |
Updated in 2012
 (9.0)