Presentation is loading. Please wait.

Presentation is loading. Please wait.

CHE 391 T. F. Edgar Spring 2012.

Similar presentations


Presentation on theme: "CHE 391 T. F. Edgar Spring 2012."โ€” Presentation transcript:

1 CHE 391 T. F. Edgar Spring 2012

2 Response of Linear system
๐’™ =๐‘จ๐’™ (๐’–=0 ๐‘œ๐‘Ÿ ๐’–=โˆ’๐‘ฒ๐’™, ๐น๐‘’๐‘’๐‘‘๐‘๐‘Ž๐‘๐‘˜ ๐ถ๐‘œ๐‘›๐‘ก๐‘Ÿ๐‘œ๐‘™) ๐’™ ๐‘ก 0 = ๐’™ 0 Solution: ๐’™ ๐‘ก = ๐‘’ ๐‘จ ๐‘กโˆ’ ๐‘ก 0 ๐’™ ๐‘ก 0 ๐‘’ ๐‘จ ๐‘กโˆ’ ๐‘ก 0 โ‰œ๐šฝ ๐‘ก, ๐‘ก 0 ( Transition Matrix) Usually ๐‘ก 0 =0 ๐šฝ ๐‘ก, ๐‘ก 0 = ๐‘’ ๐‘จ๐‘ก =๐‘ฐ+๐‘จ๐‘ก+๐‘จโˆ™๐‘จ ๐‘ก 2 +โ€ฆ Computationally not feasible for general use๏ƒ  need analytical solution

3 Take Laplace Transform:
๐‘ ๐‘ฟ ๐‘  โˆ’๐’™ 0 =๐‘จ๐‘ฟ ๐‘  ๐‘ ๐‘ฐโˆ’๐‘จ ๐‘ฟ ๐‘  =๐’™ 0 ๐‘ฟ ๐‘  = ๐‘ ๐‘ฐโˆ’๐‘จ โˆ’1 ๐’™ 0 Take inverse Laplace Transform ๐’™ ๐‘ก = โ„’ โˆ’1 ๐‘ ๐‘ฐโˆ’๐‘จ โˆ’1 ๐’™ 0 As ๐’™ ๐‘ก =๐œฑ ๐‘ก ๐‘ฅ 0 ๐‘’ ๐‘จ๐‘ก = โ„’ โˆ’1 ๐‘ ๐‘ฐโˆ’๐‘จ โˆ’1 (note scalar analog: ๐‘’ ๐‘Ž๐‘ก 1 ๐‘ โˆ’๐‘Ž )

4 Example ๐‘จ= 0 1 โˆ’2 โˆ’3 ๐‘ ๐‘ฐโˆ’๐‘จ = ๐‘  โˆ’1 โˆ’2 ๐‘ +3
๐‘จ= โˆ’2 โˆ’3 ๐‘ ๐‘ฐโˆ’๐‘จ = ๐‘  โˆ’1 โˆ’2 ๐‘ +3 ๐‘ ๐‘ฐโˆ’๐‘จ โˆ’1 = ๐‘ +3 ๐‘  2 +3๐‘  ๐ท โˆ’2 ๐ท ๐‘  ๐ท (Here, ๐ท= ๐‘ +2 ๐‘ +1 ) Invert, partial fraction decomposition ๐œฑ ๐‘ก = 2 ๐‘’ โˆ’๐‘ก โˆ’ ๐‘’ โˆ’2๐‘ก ๐‘’ โˆ’๐‘ก โˆ’ ๐‘’ โˆ’2๐‘ก โˆ’2 ๐‘’ โˆ’๐‘ก +2 ๐‘’ โˆ’2๐‘ก โˆ’ ๐‘’ โˆ’๐‘ก + 2๐‘’ โˆ’2๐‘ก

5 Note that ๐œฑ 0 =๐‘ฐ Also, denominator of ๐œฑ ๐‘  is ๐‘‘๐‘’๐‘ก ๐‘ ๐‘ฐโˆ’๐‘จ Roots of ๐ท are the same as the factors of characteristic equationโ‰กeigenvalues( ๐‘ ๐‘– ) of ๐‘จ Eigenvalue Equation: ๐‘‘๐‘’๐‘ก ๐‘จโˆ’๐‘๐‘ฐ = det ๐‘๐‘ฐโˆ’๐‘จ =0

6 Eigenvector/eigenvalue solution
Example: 2nd order ๐’™ ๐‘ก = ๐‘ 1 ๐’— 1 ๐‘’ ๐‘ 1 ๐‘ก + ๐‘ 2 ๐’— 2 ๐‘’ ๐‘ 2 ๐‘ก ๐’™ ๐‘ก = ๐‘ 1 ๐‘ 1 ๐’— 1 ๐‘’ ๐‘ 1 ๐‘ก + ๐‘ 2 ๐‘ 2 ๐’— 2 ๐‘’ ๐‘ 2 ๐‘ก =๐‘จ๐’™ =๐‘จ ๐‘ 1 ๐’— 1 ๐‘’ ๐‘ 1 ๐‘ก + ๐‘ 2 ๐’— 2 ๐‘’ ๐‘ 2 ๐‘ก Equating ๐‘’ ๐‘ ๐‘– ๐‘ก terms: ๐‘ 1 ๐’— 1 = ๐‘จ๐’— 1 ๐‘ 2 ๐’— 2 = ๐‘จ๐’— 2 If ๐‘ ๐‘– , ๐’— ๐‘– satisfy eigenvalue, eigenvector equations For nth order model, ๐‘จ ๐‘›ร—๐‘›

7 ๐‘จโˆ™ ๐’— 1 โ‹ฏ ๐’— ๐‘› โ‹ฎ โ‹ฑ โ‹ฎ. โ‹ฏ. = ๐‘ 1 ๐’— 1 โ‹ฏ ๐‘ ๐‘› ๐’— ๐‘› โ‹ฎ โ‹ฑ โ‹ฎ. โ‹ฏ
๐‘จโˆ™ ๐’— 1 โ‹ฏ ๐’— ๐‘› โ‹ฎ โ‹ฑ โ‹ฎ . โ‹ฏ . = ๐‘ 1 ๐’— 1 โ‹ฏ ๐‘ ๐‘› ๐’— ๐‘› โ‹ฎ โ‹ฑ โ‹ฎ . โ‹ฏ . Define ๐‘ฝ= matrix of eigenvectors (also called similarity transform, modal matrix, eigenvector assembly matrix) ๐œฆ= ๐‘ 1 0 โ€ฆ 0 0 ๐‘ โ‹ฎ 0 โ‹ฑ โ‹ฎ 0 โ‹ฏ 0 ๐‘ ๐‘› ๐‘จ๐‘ฝ=๐‘ฝ๐šฒ๏ƒจ ๐‘จ=๐‘ฝ๐šฒ ๐‘ฝ โˆ’1 ๐‘œ๐‘Ÿ ๐šฒ= ๐‘ฝ โˆ’1 ๐‘จ๐‘ฝ Relates ๐‘จ, ๐‘ฝ, and ๐šฒ

8 Example ๐‘จ= โˆ’8 2 7 โˆ’3 ๐‘ 1 =โˆ’1, ๐‘ 2 =โˆ’10, ๐’— 1 = 2 7 , ๐’— 2 = 1 โˆ’1
๐‘จ= โˆ’ โˆ’3 ๐‘ 1 =โˆ’1, ๐‘ 2 =โˆ’10, ๐’— 1 = , ๐’— 2 = โˆ’1 ๐’™ ๐‘ก = โˆ™ ๐‘ 1 ๐‘’ โˆ’๐‘ก โˆ’1 โˆ™ ๐‘ 2 ๐‘’ โˆ’10๐‘ก Note that ๐’— 1 , ๐’— 2 must be linearly independent (at ๐‘ก=0, solve for ๐‘ 1 , ๐‘ 2 ) For ๐’™ 0 = , we have ๐‘ 1 =1.11, ๐‘ 2 =7.78 Then, ๐’™ ๐‘ก = โˆ™ ๐‘ 1 ๐‘’ โˆ’๐‘ก โˆ’7.78 โˆ™ ๐‘ 2 ๐‘’ โˆ’10๐‘ก Note that when ๐‘ก is very large, neglect the 2nd term. ๐’™ ๐‘ก =๐‘ ๐‘ก ๐’— 1

9 If ๐‘จ is in canonical form,
๐‘จ= ๐‘ฐ ๐‘›โˆ’1ร—๐‘›โˆ’1 โ‹ฎ โˆ’ ๐‘Ž 0 โˆ’ ๐‘Ž 1 โ‹ฏ โˆ’ ๐‘Ž ๐‘›โˆ’1 ๐‘ฝ= 1 1 โ‹ฏ 1 ๐‘ 1 ๐‘ 2 โ‹ฏ ๐‘ ๐‘› ๐‘ ๐‘ ๐‘ ๐‘› 2 โ‹ฎ โ‹ฎ โ‹ฑ โ‹ฎ ๐‘ 1 ๐‘›โˆ’1 ๐‘ 2 ๐‘›โˆ’1 โ‹ฏ ๐‘ ๐‘› ๐‘›โˆ’1 (Vandermonde Matrix) Note, if ๐‘ ๐‘– is complex, ๐‘ฃ ๐‘– is complex.

10 Another way to calculate ๐œฑ ๐‘ก using ๐‘ฝ, ๐šฒ
๐’™ =๐‘จ๐’™, let ๐’™=๐‘ฝ๐’› (transformation), then we have ๐‘ฝ ๐’› =๐‘จ๐‘ฝ๐’› ๏ƒจ ๐’› = ๐‘ฝ โˆ’1 ๐‘จ๐‘ฝ๐’›=๐šฒ๐’› The solution for the ODE is ๐’› ๐‘ก = ๐‘’ ๐šฒ๐‘ก ๐’› 0 Where ๐‘’ ๐šฒ๐‘ก = ๐‘’ ๐‘ 1 ๐‘ก ๐‘’ ๐‘ 2 ๐‘ก โ‹ฑ ๐‘’ ๐‘ ๐‘› ๐‘ก (a diagonal matrix) hence ๐‘ง ๐‘– ๐‘ก = ๐‘’ ๐‘ ๐‘– ๐‘ก ๐‘ง ๐‘– (โ€œmodesโ€ of the system) The analytical solution to ๐’™ =๐‘จ๐’™ is therefore ๐’™=๐‘ฝ๐’›=๐‘ฝ ๐‘’ ๐šฒ๐‘ก ๐’› ๐’› 0 = ๐‘ฝ โˆ’1 ๐’™ 0 ๐’™ ๐‘ก =๐‘ฝ ๐‘’ ๐šฒ๐‘ก ๐‘ฝ โˆ’1 ๐’™ 0 and ๐œฑ ๐‘ก =๐‘ฝ ๐‘’ ๐šฒ๐‘ก ๐‘ฝ โˆ’1 If ๐‘ ๐‘– are complex or repeated, modify the above procedure. (see Ogata)

11 Add controller ๐’™ =๐‘จ๐’™+๐‘ฉ๐’– Substitute ๐’™=๐‘ฝ๐’› , we have ๐‘ฝ ๐’› =๐‘จ๐‘ฝ๐’›+๐‘ฉ๐’– ๐’› = ๐‘ฝ โˆ’1 ๐‘จ๐‘ฝ๐’›+ ๐‘ฝ โˆ’1 ๐‘ฉ๐’–=๐šฒ๐’›+ ๐‘ฝ โˆ’1 ๐‘ฉ๐’– Let ๐’–=โˆ’๐‘ฒ๐’› (Modal feedback. ๐’–=โˆ’๐‘ฒ ๐‘ฝ โˆ’1 ๐’™) ๐’› = ๐šฒโˆ’ ๐‘ฝ โˆ’1 ๐‘ฉ๐‘ฒ ๐’› Multivariable root locus (some limitations in MIMO as in SISO โ€“ closed-loop response depends on eigenvectors as well as eigenvalues.) Shifts eigenvalues

12 โ€œAutonomousโ€ model ๐’™ =๐‘จ๐’™+๐‘ฉ๐’–
๐‘จ, ๐‘ฉ not functions of ๐‘ก; ๐’™, ๐’– deviation variables . Take L.T., ๐‘ ๐‘ฐโˆ’๐‘จ ๐‘ฟ ๐‘  =๐‘ฉ๐‘ผ ๐‘  ๐‘ฟ ๐‘  = ๐‘ ๐‘ฐโˆ’๐‘จ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  If ๐’€ ๐‘  =๐‘ช๐‘ฟ ๐‘  (๐’š=๐‘ช๐’™ ) (output states or controlled variables) Then the output is ๐’€ ๐‘  =๐‘ช ๐‘ ๐‘ฐโˆ’๐‘จ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  Transfer function matrix ๐‘ฎ ๐‘ ๐‘ 

13 Example: ๐’™ ๐‘ก = ๐‘ 1 ๐’— 1 ๐‘’ ๐‘ 1 ๐‘ก + ๐‘ 2 ๐’— 2 ๐‘’ ๐‘ 2 ๐‘ก ๐‘ 1 =โˆ’1, ๐‘ 2 =โˆ’10
๐’™ ๐‘ก = ๐‘ 1 ๐’— 1 ๐‘’ ๐‘ 1 ๐‘ก + ๐‘ 2 ๐’— 2 ๐‘’ ๐‘ 2 ๐‘ก ๐‘ 1 =โˆ’1, ๐‘ 2 =โˆ’10 Shift pole ๐‘ 1 to higher value However, if ๐’™ 0 is such that ๐‘ 1 =0, shifting ๐‘ 1 has no effect on response

14 Stability Requirements
Continuous: all ๐‘…๐‘’ ๐‘ ๐‘– โ‰ค0 Discrete: ๐’™ ๐‘˜+1 =๐‘ญ ๐’™ ๐‘˜ +๐‘ฎ ๐’– ๐‘˜ ๐‘ ๐‘– eigenvalues of ๐‘ญ, ๐‘ ๐‘– โ‰ค1 (for complex ๐‘ ๐‘– , all ๐‘ ๐‘– must lie within unit circle)

15 Analytical Solution: ๐’™ โˆ’๐‘จ๐’™=๐‘ฉ๐’– ๐’™ ๐‘ก ๐ŸŽ = ๐’™ 0
๐’™ โˆ’๐‘จ๐’™=๐‘ฉ๐’– ๐’™ ๐‘ก ๐ŸŽ = ๐’™ 0 Pre-multiply by ๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก =๐‘’ โˆ’๐‘จ๐‘ก (integrating factor) ๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก ๐’™ โˆ’๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก ๐‘จ๐’™=๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก ๐‘ฉ๐’– ๏ƒจ ๐‘‘ ๐‘‘๐‘ก ๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก ๐’™ =๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก ๐‘ฉ๐’– Integrating between ๐‘ก 0 and ๐‘ก 1 : ๐‘’๐‘ฅ๐‘ โˆ’๐‘จ ๐‘ก 1 ๐’™ 1 โˆ’๐‘’๐‘ฅ๐‘ โˆ’๐‘จ ๐‘ก 0 ๐’™ 0 = ๐‘ก ๐ŸŽ ๐‘ก 1 ๐‘’๐‘ฅ๐‘ โˆ’๐‘จ๐‘ก ๐‘ฉ๐’–๐‘‘๐‘ก ๐’™ ๐‘ก 1 = ๐’™ 1 =๐‘’๐‘ฅ๐‘ ๐‘จ ๐‘ก 1 โˆ’ ๐‘ก 0 ๐’™ 0 + ๐‘ก ๐ŸŽ ๐‘ก 1 ๐‘’๐‘ฅ๐‘ ๐‘จ ๐‘ก 1 โˆ’๐‘ก ๐‘ฉ๐’– ๐‘ก ๐‘‘๐‘ก Homogenous Solution, ๐šฝ ๐‘ก 1 โˆ’ ๐‘ก 0 Particular Solution

16 If ๐’™ 0 =0 ๏ƒจ convolution integral
Note that ๐’™ ๐‘ก 1 depends only on ๐‘ก 1 โˆ’ ๐‘ก 0 =โˆ†๐‘ก. Let ๐‘ก 0 =0 ๐’™ โˆ†๐‘ก =๐šฝ โˆ†๐‘ก ๐’™ โˆ†๐‘ก ๐šฝ โˆ†๐‘กโˆ’๐œ† ๐‘ฉ๐’– ๐œ† ๐‘‘๐œ† Sampled data system: ๐’– ๐œ† =๐’– 0 (0โ‰ค๐‘กโ‰คโˆ†๐‘ก) Define ๐‘ฎ= 0 โˆ†๐‘ก ๐šฝ โˆ†๐‘กโˆ’๐œ† ๐‘ฉ๐‘‘๐œ† ๐’™ โˆ†๐‘ก =๐‘ญ๐’™ 0 +๐‘ฎ๐’– 0 (๐‘ญ, ๐‘ฎ are functions of โˆ†๐‘ก only ) Generalize to any time step ๐‘˜ ๏ƒจ ๐’™ ๐‘˜+1 =๐‘ญ ๐’™ ๐‘˜ +๐‘ฎ ๐’– ๐‘˜ Calculation of ๐‘ญ, ๐‘ฎ : use infinite series expression ๐‘ญ=๐‘ฐ+๐‘จโˆ†๐‘ก+๐‘จโˆ™๐‘จโˆ™ โˆ†๐‘ก โ€ฆ

17 ๐‘ฎ=๐šฝ โˆ†๐‘ก 0 โˆ†๐‘ก ๐šฝ โˆ’๐œ† ๐‘‘๐œ†โˆ™๐‘ฉ Where 0 โˆ†๐‘ก ๐šฝ โˆ’๐œ† ๐‘‘๐œ† = 0 โˆ†๐‘ก ๐‘’ โˆ’๐‘จ๐œ† ๐‘‘๐œ† = ๐‘จ โˆ’1 ๐‘ฐโˆ’ ๐‘’ โˆ’๐‘จโˆ†๐‘ก = ๐‘จ โˆ’1 ๐‘ฐโˆ’ ๐‘ฐโˆ’๐‘จโˆ†๐‘ก+๐‘จโˆ™๐‘จโˆ™ โˆ†๐‘ก 2 2 โˆ’โ€ฆ =๐‘ฐโˆ†๐‘กโˆ’๐‘จโˆ™ โˆ†๐‘ก โ€ฆ Use finite number of terms ๐’™ ๐‘˜+2 =๐‘ญ๐’™ ๐‘˜+1 +๐‘ฎ๐’– ๐‘˜+1 = ๐‘ญ 2 ๐’™ ๐‘˜ +๐‘ญ๐‘ฎ๐’– ๐‘˜ +๐‘ฎ๐’– ๐‘˜+1 ๐’™ ๐‘˜+๐‘ = ๐‘ญ ๐‘ ๐’™ ๐‘˜ + ๐‘–=0 ๐‘โˆ’1 ๐‘ญ ๐‘โˆ’1โˆ’๐‘– ๐‘ฎ๐’– ๐‘˜+๐‘– No integration error for sampled data control (zero-order hold)

18 Discrete-time State Controllability
๐’™ ๐‘› โˆ’ ๐‘ญ ๐‘› ๐’™ 0 = ๐‘ญ ๐‘›โˆ’1 ๐‘ฎ๐’– 0 + ๐‘ญ ๐‘›โˆ’2 ๐‘ฎ๐’– 1 +โ€ฆ+๐‘ฎ๐’– ๐‘›โˆ’1 ๐‘› is the order of the system. We need to be able to generate any ๐’™ ๐‘› from ๐’™ 0 using ๐’– 0 , ๐’– 1 , etc. Example: ๐‘ข is a scalar, ๐‘ญ ๐‘›ร—๐‘› , ๐‘ฎ ๐‘›ร—1 , ๐‘ญ ๐‘˜ ๐‘ฎ is ๐‘›ร—1 ๐‘› equations, and ๐‘› unknowns (rank must be n, determinant โ‰  0. No rows, columns are linearly dependent.) Use Gaussian elimination to check controllability

19 Example: ๐‘ฅ 1 = ๐‘ฅ 2 +๐‘ข ๐‘ฅ 2 = โˆ’๐‘ฅ 1 โˆ’2๐‘ฅ 2 โˆ’๐‘ข ๐‘จ= โˆ’1 โˆ’2 , ๐‘ฉ= โˆ’1 ๐‘ด= ๐‘ฉ|๐‘จ๐‘ฉ = โˆ’1 โˆ’ (not linear independent) Controllability Matrix: Discrete-time: ๐‘ด ๐ท = ๐‘ฎ ๐‘ญ๐‘ฎ ๐‘ญ 2 ๐‘ฎ|โ€ฆ| ๐‘ญ ๐‘›โˆ’1 ๐‘ฎ Continuous-time: ๐‘ด ๐ถ = ๐‘ฉ ๐‘จ๐‘ฉ ๐‘จ 2 ๐‘ฉ|โ€ฆ| ๐‘จ ๐‘›โˆ’1 ๐‘ฉ

20 Observability ๐’™ ๐‘˜+1 =๐‘ญ๐’™ ๐‘˜ ๐’š=๐‘ฏ๐’™
๐’™ ๐‘˜+1 =๐‘ญ๐’™ ๐‘˜ ๐’š=๐‘ฏ๐’™ ๐’š 0 =๐‘ฏ๐’™ 0 , ๐’š 2 =๐‘ฏ ๐‘ญ 2 ๐’™ 0 , ๐’š ๐‘›โˆ’1 =๐‘ฏ ๐‘ญ ๐‘›โˆ’1 ๐’™ 0 Example: Suppose ๐’™ 0 is ๐‘›ร—1 (unknown vector), ๐’š is 1ร—1 (scalar) After ๐‘›โˆ’1 time steps, can we reconstruct ๐’™ 0 using ๐‘› values of ๐‘ฆ (๐‘ฆ 0 , ๐‘ฆ 1 , โ€ฆ, ๐‘ฆ ๐‘›โˆ’1 )? In this case ๐‘ฏ is 1ร—๐‘› ๐‘ฆ 0 ๐‘ฆ 1 โ‹ฎ ๐‘ฆ ๐‘›โˆ’1 = ๐‘ฏ ๐‘ฏ๐‘ญ โ‹ฎ ๐‘ฏ ๐‘ญ ๐‘›โˆ’1 ๐’™ 0 ๐‘› equations, ๐‘› unknowns Must have rank ๐‘›

21 Observability Matrix Discrete: ๐‘ต ๐ท = ๐‘ฏ ๐‘‡ ๐‘ญ ๐‘‡ ๐‘ฏ ๐‘‡ โ€ฆ| ๐‘ญ ๐‘‡ ๐‘›โˆ’1 ๐‘ฏ ๐‘‡
๐‘ต ๐ท = ๐‘ฏ ๐‘‡ ๐‘ญ ๐‘‡ ๐‘ฏ ๐‘‡ โ€ฆ| ๐‘ญ ๐‘‡ ๐‘›โˆ’1 ๐‘ฏ ๐‘‡ must have rank ๐‘› Continuous: ๐’™ =๐‘จ๐’™, ๐’š=๐‘ช๐’™ ๐‘ต ๐ถ = ๐‘ช ๐‘‡ ๐‘จ ๐‘‡ ๐‘ช ๐‘‡ โ€ฆ| ๐‘จ ๐‘‡ ๐‘›โˆ’1 ๐‘ช ๐‘‡

22 ๐ต+๐ถ ๐‘˜ 3 ๐‘ˆ๐‘›๐‘˜๐‘›๐‘œ๐‘ค๐‘› ๐ถ๐‘œ๐‘š๐‘๐‘œ๐‘ข๐‘›๐‘‘๐‘  (๐‘›๐‘œ๐‘ก ๐ต, ๐ถ, ๐‘œ๐‘Ÿ ๐ท)
Example ๐ด ๐‘˜ 1 ๐ต ๐ด+๐ถ ๐‘˜ 2 ๐ท ๐ต+๐ถ ๐‘˜ 3 ๐‘ˆ๐‘›๐‘˜๐‘›๐‘œ๐‘ค๐‘› ๐ถ๐‘œ๐‘š๐‘๐‘œ๐‘ข๐‘›๐‘‘๐‘  (๐‘›๐‘œ๐‘ก ๐ต, ๐ถ, ๐‘œ๐‘Ÿ ๐ท) Linearize Equations -> deviation variables ๐‘‘ ๐ถ ๐ด ๐‘‘๐‘ก =โˆ’ ๐‘˜ 1 ๐ถ ๐ด โˆ’ ๐‘˜ 2 ๐ถ ๐ด ๐ถ ๐ถ ๐‘‘ ๐ถ ๐ต ๐‘‘๐‘ก = ๐‘˜ 1 ๐ถ ๐ด โˆ’ ๐‘˜ 3 ๐ถ ๐ต ๐ถ ๐ถ ๐‘‘ ๐ถ ๐ถ ๐‘‘๐‘ก =โˆ’ ๐‘˜ 2 ๐ถ ๐ด ๐ถ ๐ถ โˆ’ ๐‘˜ 3 ๐ถ ๐ต ๐ถ ๐ถ ๐‘‘ ๐ถ ๐ท ๐‘‘๐‘ก = ๐‘˜ 2 ๐ถ ๐ด ๐ถ ๐ถ ๐’™ =๐‘จ๐’™ ๐’™= ๐ถ ๐ด โˆ’ ๐ถ ๐ด ๐‘  ๐ถ ๐ต โˆ’ ๐ถ ๐ต ๐‘  ๐ถ ๐ถ โˆ’ ๐ถ ๐ถ ๐‘  ๐ถ ๐ท โˆ’ ๐ถ ๐ท ๐‘ 

23 4 linear independent columns ๏ƒ  Observable
๐‘จ= ๐›ผ ๐›ผ ๐›ผ 21 ๐›ผ 22 ๐›ผ ๐›ผ 31 ๐›ผ 32 ๐›ผ ๐›ผ ๐›ผ 43 0 Case 1: Measure A, C, D (๐’š=๐‘ช๐’™) ๐‘ช= ๐‘ต ๐ถ = ๐‘ช ๐‘‡ | ๐‘จ ๐‘‡ ๐‘ช ๐‘‡ = โˆ— 0 โˆ— โˆ— โˆ— โˆ— 0 โˆ— โˆ— 4 linear independent columns ๏ƒ  Observable

24 Case 2, 3: Measure (A,B,D) , (B,C,D) ๏ƒจObservable
Case 4: Measure A, B, C ๏ƒจ Not observable Physical interpretation: since D does not affect any other outputs (does not appear in any rate equations), we donโ€™t know D(0) unless it is measured directly Note: if adding ๐‘จ ๐‘‡ ๐‘— ๐‘ช ๐‘‡ doesnโ€™t increase the rank of the sequence ๐‘ช ๐‘‡ , ๐‘จ ๐‘‡ ๐‘ช ๐‘‡ , ๐‘จ ๐‘‡ 2 ๐‘ช ๐‘‡ , โ€ฆ , you donโ€™t need to check any further. (Same for controllability).

25 Other Controllability Considerations
Modal equation: ๐’› =๐šฒ๐’›+ ๐‘ฝ โˆ’1 ๐‘ฉ๐’– ๐‘ฝ โˆ’1 ๐‘ฉ must not have a null row (must be able to control all modes) (2) ๐‘Ž๐‘‘๐‘— ๐‘ ๐‘ฐโˆ’๐‘จ must have no common factor with det ๐‘ ๐‘ฐโˆ’๐‘จ (3) For non-distinct eigenvalues, rules must be modified; (4) Output controllability of system ๐’™ =๐‘จ๐’™+๐‘ฉ๐’– , ๐’š=๐‘ช๐’™+๐‘ซ๐’– (usually ๐‘ซ=๐‘ถ, ๐‘ช ๐’Žร—๐’ , ๐’š ๐‘šร—1 ) ๐‘Ÿ๐‘Ž๐‘›๐‘˜ ๐‘ช๐‘ฉ ๐‘ช๐‘จ๐‘ฉ ๐‘ช ๐‘จ 2 ๐‘ฉ|โ€ฆ =๐‘š

26 MIMO Model Linearization
Nonlinear model: ๐’› =๐’‡ ๐’›,๐’— ๐’›: state variable (๐‘›ร—1) ๐’—: decision variable (๐‘Ÿร—1) Linearization๏ƒ  define deviation variables: ๐’™=๐’›โˆ’ ๐’› ๐‘ ๐‘  ๐’–=๐’—โˆ’ ๐’— ๐‘ ๐‘  (Requires iterative solution of ๐’‡ ๐’› ๐‘ ๐‘  , ๐’— ๐‘ ๐‘  =0)

27 Analytical vs. numerical
Use 1st order Taylor Series (invalid for large ๐’™, ๐’–) Scalar case: ๐‘ง = ๐‘ฅ =๐‘“ ๐‘ง ๐‘ ๐‘  , ๐‘ฃ ๐‘ ๐‘  + ๐œ•๐‘“ ๐œ•๐‘ง ๐‘ ๐‘  ๐‘งโˆ’ ๐‘ง ๐‘ ๐‘  + ๐œ•๐‘“ ๐œ•๐‘ฃ ๐‘ ๐‘  ๐‘ฃโˆ’ ๐‘ฃ ๐‘ ๐‘  = ๐œ•๐‘“ ๐œ•๐‘ง ๐‘ ๐‘  ๐‘ฅ + ๐œ•๐‘“ ๐œ•๐‘ฃ ๐‘ ๐‘  ๐‘ข Vector case: ๐’™ =๐‘จ๐’™+๐‘ฉ๐’– ๐‘จ ๐‘›ร—๐‘› = ๐œ• ๐‘“ ๐‘– ๐œ• ๐‘ง ๐‘— ๐‘ ๐‘  (๐ฝ๐‘Ž๐‘๐‘œ๐‘๐‘–๐‘Ž๐‘›), ๐‘ฉ ๐‘›ร—๐‘Ÿ = ๐œ• ๐‘“ ๐‘– ๐œ• ๐‘ฃ ๐‘— ๐‘ ๐‘  Analytical vs. numerical

28 Minimal State Vector Representation (MIMO Systems)
๐’™ =๐‘จ๐’™+๐‘ฉ๐’– ๐’™=๐‘ฝ๐’› ๏ƒ  ๐‘ฝ ๐’› =๐‘จ๐‘ฝ๐’›+๐‘ฉ๐’– ๐’› = ๐‘ฝ โˆ’1 ๐‘จ๐‘ฝ๐’›+ ๐‘ฝ โˆ’1 ๐‘ฉ๐’–=๐šฒ๐’›+ ๐‘ฝ โˆ’1 ๐‘ฉ๐’– ๐‘ ๐‘ฐโˆ’๐šฒ ๐’ ๐‘  = ๐‘ฝ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  ๐’ ๐‘  = ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฝ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  , ๐‘ฝ โˆ’1 ๐‘ฟ ๐‘  =๐’ ๐‘  ๐‘ฝ โˆ’1 ๐‘ฟ ๐‘  = ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฝ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  ๐‘ฟ ๐‘  =๐‘ฝ ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฝ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  ๐’€ ๐‘  =๐‘ช๐‘ฟ ๐‘  =๐‘ช๐‘ฝ ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฝ โˆ’1 ๐‘ฉ๐‘ผ ๐‘  Make ๐‘ช๐‘ฝ=๐œธ, ๐‘ฝ โˆ’1 ๐‘ฉ=๐‘ฌ, then the transfer function is ๐‘ฎ ๐‘ ๐‘  =๐‘ช๐‘ฝ ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฝ โˆ’1 ๐‘ฉ=๐œธ ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฌ Question: Given ๐‘ฎ ๐‘ ๐‘  , how do we find ๐‘จ, ๐‘ฉ, ๐‘ช which gives a model of minimal order? (or given step response for all inputs/outputs)

29 Example: 2ร—2 (Second order)
๐‘ฎ ๐‘ ๐‘  = 2๐‘ +11 ๐‘ +4 ๐‘ +2 ๐‘ +6 ๐‘ +4 ๐‘ +2 ๐‘ โˆ’5 ๐‘ +4 ๐‘ +2 ๐‘ โˆ’2 ๐‘ +4 ๐‘ +2 =๐œธ ๐‘ ๐‘ฐโˆ’๐šฒ โˆ’1 ๐‘ฌ= ๐›พ 11 ๐›พ 12 ๐›พ 21 ๐›พ ๐‘ โˆ’ ๐‘ ๐‘ โˆ’ ๐‘ ๐ธ 11 ๐ธ 12 ๐ธ 21 ๐ธ 22 = ๐ธ 11 ๐›พ 11 ๐‘ โˆ’ ๐‘ ๐ธ 21 ๐›พ 12 ๐‘ โˆ’ ๐‘ ๐ธ 12 ๐›พ 11 ๐‘ โˆ’ ๐‘ ๐ธ 22 ๐›พ 12 ๐‘ โˆ’ ๐‘ ๐ธ 11 ๐›พ 21 ๐‘ โˆ’ ๐‘ ๐ธ 21 ๐›พ 22 ๐‘ โˆ’ ๐‘ ๐ธ 12 ๐›พ 21 ๐‘ โˆ’ ๐‘ ๐ธ 22 ๐›พ 22 ๐‘ โˆ’ ๐‘ 2 (2 poles => 2nd order, 2ร—2)

30 Expand ๐‘ฎ ๐‘ ๐‘  by partial fraction expansion ๐‘ฎ ๐‘ ๐‘  = โˆ’ 3 2 ๐‘  ๐‘ +2 โˆ’1 ๐‘  ๐‘  ๐‘ +4 โˆ’ 7 2 ๐‘ +2 3 ๐‘ +4 โˆ’ 2 ๐‘ +2 Matching coefficients, ๐ธ 11 ๐›พ 11 ๐ธ 12 ๐›พ 11 ๐ธ 11 ๐›พ 21 ๐ธ 12 ๐›พ 21 = โˆ’3 2 โˆ’ Let ๐›พ 11 =1, then ๐ธ 11 = โˆ’3 2 , ๐ธ 12 =โˆ’1, ๐›พ 21 =โˆ’3 ๐ธ 21 ๐›พ 12 ๐ธ 22 ๐›พ 12 ๐ธ 21 ๐›พ 22 ๐ธ 22 ๐›พ 22 = โˆ’7 2 โˆ’2 Let ๐›พ 12 =1, then ๐ธ 21 = 7 2 , ๐ธ 22 =2, ๐›พ 22 =โˆ’1

31 Therefore ๐‘ฌ= ๐‘ฝ โˆ’1 ๐‘ฉ= โˆ’3 2 โˆ’ โˆ’ (1) ๐œธ=๐‘ช๐‘ฝ= โˆ’3 โˆ’ (2) Assume ๐‘ช, calculate ๐‘ฝ in Eq. (2); Using ๐‘ฝ โˆ’1 , find ๐‘ฉ in Eq.(1); ๐‘จ=๐‘ฝ๐šฒ ๐‘ฝ โˆ’1 . Note that there is an infinite number of realizations to yield ๐‘ฎ ๐‘ ๐‘ 

32

33

34

35

36

37


Download ppt "CHE 391 T. F. Edgar Spring 2012."

Similar presentations


Ads by Google