how to prove the transfer function of the decomposed form is the following?

55 Views Asked by At

We know that for an uncontrollable state space model of the following form:

$\begin{array}{l} \dot x\left( t \right) = Ax\left( t \right) + Bu\left( t \right)\\ y\left( t \right) = Cx\left( t \right) + Du\left( t \right) \end{array}$

it has a transfer function $G\left( s \right) = C{\left( {SI - A} \right)^{ - 1}}B + D$.

Furthermore, we can associate a controllable Kalman decomposition:

$\begin{array}{l} \left[ {\begin{array}{*{20}{c}} {\frac{{d\left( {{{\bar x}_c}} \right)}}{{dt}}}\\ {\frac{{d\left( {{{\bar x}_{\bar c}}} \right)}}{{dt}}} \end{array}} \right] = \left[ {\begin{array}{*{20}{c}} {{{\bar A}_c}}&{{{\bar A}_{12}}}\\ 0&{{{\bar A}_{\bar c}}} \end{array}} \right]\left[ {\begin{array}{*{20}{c}} {{{\bar x}_c}}\\ {{{\bar x}_{\bar c}}} \end{array}} \right] + \left[ {\begin{array}{*{20}{c}} {{{\bar B}_c}}\\ 0 \end{array}} \right]u\\ y = \left[ {\begin{array}{*{20}{c}} {{{\bar C}_c}}&{{{\bar C}_{\bar c}}} \end{array}} \right]\left[ {\begin{array}{*{20}{c}} {{{\bar x}_c}}\\ {{{\bar x}_{\bar c}}} \end{array}} \right] + Du \end{array}$

my question is: how to prove that the new transfer function is: $G\left( s \right) = {{\bar C}_c}{\left( {SI - {{\bar A}_c}} \right)^{ - 1}}{{\bar B}_c} + D$

1

There are 1 best solutions below

0
On BEST ANSWER

The controllable Kalman decomposition is a linear transformation, i.e. $\bar{x} = T\,x$ with $T$ non-singular. So the state space model in the controllable Kalman decomposition $(\bar{A}, \bar{B}, \bar{C}, \bar{D})$ is related to the initial state space model $(A,B,C,D)$ according to

\begin{align} \bar{A} &= T\,A\,T^{-1}, \\ \bar{B} &= T\,B, \\ \bar{C} &= C\,T^{-1}, \\ \bar{D} &= D. \end{align}

For any such non-singular $T$ is follows that

$$ G(s) = \bar{C}\,(s\,I - \bar{A})^{-1} \bar{B} + \bar{D} = C\,T^{-1}\,(s\,I - T\,A\,T^{-1})^{-1} T\,B + D, \tag{1} $$

which simplifies to the following when using the push-through identity

$$ G(s) = C\,(s\,I - A)^{-1} B + D. \tag{2} $$

So any non-singular $T$ does not change the transfer function $G(s)$ associated with the state space model.


If there exists such a $T$ such that

$$ \bar{A} = \begin{bmatrix} \bar{A}_c & \bar{A}_{12} \\ 0 & \bar{A}_\bar{c} \end{bmatrix}, \quad \bar{B} = \begin{bmatrix} \bar{B}_c \\ 0 \end{bmatrix}, \quad \bar{C} = \begin{bmatrix} \bar{C}_c & \bar{C}_\bar{c} \end{bmatrix}, $$

it follows that the transfer function $G(s)$ can be obtained using $(1)$. It can be noted that due to upper block triangular structure of $\bar{A}$ it holds that

\begin{align} (s\,I - \bar{A})^{-1} &= \begin{bmatrix} s\,I-\bar{A}_c & -\bar{A}_{12} \\ 0 & s\,I-\bar{A}_\bar{c} \end{bmatrix}^{-1}, \tag{3a} \\ &= \begin{bmatrix} (s\,I-\bar{A}_c)^{-1} & (s\,I-\bar{A}_c)^{-1} \bar{A}_{12} (s\,I-\bar{A}_\bar{c})^{-1} \\ 0 & (s\,I-\bar{A}_\bar{c})^{-1} \end{bmatrix}. \tag{3b} \end{align}

From this it follows that multiplying $(3)$ by $\bar{B}$ yields

$$ (s\,I - \bar{A})^{-1} \bar{B} = \begin{bmatrix} (s\,I-\bar{A}_c)^{-1} \bar{B}_c \\ 0 \end{bmatrix}, $$

after which I assume it is clear that

$$ G(s) = \bar{C}_c(s\,I-\bar{A}_c)^{-1} \bar{B}_c + D. $$