# Control of Multiple-Input, MultipleOutput (MIMO) Processes Chapter 18

18.1 Process Interactions and Control Loop Interactions 18.2 Pairing of Controlled and Manipulated Variables 18.3 Singular Value Analysis 18.4 Tuning of Multiloop PID Control Systems 18.5 Decoupling and Multivariable Control Strategies 18.6 Strategies for Reducing Control Loop Interactions

1

Control of Multivariable Processes

Chapter 18

In practical control problems there typically are a number of process variables which must be controlled and a number which can be manipulated. Example: product quality and throughput must usually be controlled.

• Several simple physical examples are shown in Fig. 18.1. Note the "process interactions" between controlled and manipulated variables. 2

Chapter 18

SEE FIGURE 18.1 in text.

3

Chapter 18 4 .

B. xB .• Controlled Variables: xD . and hB • Manipulated Variables: D. P. and QB Note: Possible multiloop control strategies = 5! = 120 Chapter 18 5 . QD . R. hD .

. Multivariable Control: Each manipulated variable can depend on two or more of the controlled variables. • If process interactions are significant. Definitions: • Multiloop control: Each manipulated variable depends on only a single controlled variable.e. Chapter 18 • In these situations there are incentives for considering multivariable control strategies. Examples: decoupling control. i. a set of conventional feedback controllers. model predictive control 6 • . even the best multiloop control system may not provide satisfactory control.• In this chapter we will be concerned with characterizing process interactions and selecting an appropriate multiloop control configuration.

n! possible pairing configurations. Select controlled and manipulated variables. 7 .g.. Specify types of FB controllers. one for each controlled variable. Example: 2 x 2 system Two possible controller pairings: (1-1/2-2 pairing) U1 with Y1. 3. 2. U2 with Y1 (1-2/2-1 pairing) Note: For n x n system. • Control system design Chapter 18 1.Multiloop Control Strategy • • Typical industrial approach Consists of using n standard FB controllers (e. U2 with Y2 or U1 with Y2. Select pairing of controlled and manipulated variables. PID).

= GP 22 ( s ) U1 ( s ) U2( s ) (18 − 1) Thus. the input-output relations for the process can be written as: Y1( s ) = GP11( s )U1( s ) + GP12 ( s )U 2 ( s ) Y2 ( s ) = GP 21( s )U1( s ) + GP 22 ( s )U 2 ( s ) (18 − 2 ) (18 − 3) 8 . = GP12 ( s ) U1 ( s ) U2( s ) Y2 ( s ) Y2 ( s ) = GP 21( s ).Transfer Function Model (2 x 2 system) Two controlled variables and two manipulated variables (4 transfer functions required) Chapter 18 Y1( s ) Y1( s ) = GP11( s ).

Or in vector-matrix notation as. ⎡ Y1 ( s ) ⎤ ⎡U1 ( s ) ⎤ Y(s )= ⎢ U( s ) = ⎢ ⎥ ⎥ Y s U s ( ) ( ) ⎣ 2 ⎦ ⎣ 2 ⎦ ⎡ GP11( s ) GP12 ( s ) ⎤ G p( s ) = ⎢ ⎥ G ( s ) G ( s ) ⎢ ⎥ P 22 ⎣ P 21 ⎦ And Gp(s) is the transfer function matrix for the process (18 − 6) 9 . Y ( s ) = G p ( s )U ( s ) (18 − 4) (18 − 5) Chapter 18 where Y(s) and U(s) are vectors.

Chapter 18 10 .

Controller tuning becomes more difficult. Example: 2 x 2 system Control loop interactions are due to the presence of a third feedback loop. Closed-loop system may become destabilized. 11 . ii. Chapter 18 • Problems arising from control loop interactions i.Control-loop Interactions • Process interactions may induce undesirable interactions between two or more control loops.

Chapter 18 12 .

Chapter 18 13 .

1-1/2 -2 pairing From block diagram algebra we can show Chapter 18 GP12GP 21GC 2 Y1( s ) = GP11 − (second loop closed) U1( s ) 1 + GC 2GP 22 Note that the last expression contains GC2. Example: 2 x 2 system. Y1( s ) = GP11( s ).Block Diagram Analysis For the multiloop control configuration. the transfer function between a controlled and a manipulated variable depends on whether the other feedback control loops are open or closed. U1( s ) (second loop open) (18-7) (18-11) 14 .

Chapter 18 15 .

Chapter 18 16 .

Chapter 18 17 .

Chapter 18 18 .

Chapter 18 • Requires knowledge of steady-state gains but not process dynamics. Recommendation about best pairing of controlled and manipulated variables. 19 . Measure of process interactions 2.Relative Gain Array • Provides two types of useful information: 1.

Example of RGA Analysis: 2 x 2 system • Steady-state process model. λij. relates the ith controlled variable and the jth manipulated variable ( ∂yi / ∂u j )u open-loop gain λij = ( ∂yi / ∂u j ) y closed-loop gain (18 − 24 ) 20 . is defined as: ⎡ λ11 λ12 ⎤ Λ = ⎢ ⎥ λ λ ⎣ 21 22 ⎦ where the relative gain. Λ. Chapter 18 y1 = K11u1 + K12u2 y2 = K 21u1 + K 22u2 The RGA.

Scaling Properties: i. 1 . λ11 = λ12 = 1 − λ11 = λ21 K12 K 21 − 1 K11K 22 (18-34) Recommended Controller Pairing It corresponds to the λij which have the largest positive values that are closest to one. λij is dimensionless ii. Chapter 18 ∑λ = ∑λ ij i j ij = 1.0 For a 2 x 2 system. 21 .

Otherwise. 2. Pairings which correspond to negative pairings should not be selected. Chapter 18 Examples: Process Gain Matrix. K : ⎡ K11 ⎢ 0 ⎣ ⎡ 0 ⎢K ⎣ 21 ⎡ K11 ⎢ 0 ⎣ ⎡ K11 ⎢K ⎣ 21 Relative Gain Array. Λ : 0 ⎤ K 22 ⎥ ⎦ K12 ⎤ 0 ⎥ ⎦ K12 ⎤ K 22 ⎥ ⎦ 0 ⎤ K 22 ⎥ ⎦ ⇒ ⇒ ⇒ ⇒ ⎡1 0 ⎤ ⎢0 1 ⎥ ⎣ ⎦ ⎡0 1 ⎤ ⎢1 0 ⎥ ⎣ ⎦ ⎡1 0⎤ ⎢0 1 ⎥ ⎦ ⎣ ⎡1 0⎤ ⎢0 1 ⎥ ⎣ ⎦ 22 . choose the pairing which has λij closest to one.In general: 1.

K12 K 21 K11K 22 λ12 = 1 − λ11 = λ21 Chapter 18 ∴ Recommended pairing is Y1 and U1.5⎤ =⎢ ⎥ K 22 ⎥ . 1 5 2 ⎣ ⎦ ⎦ λ11 = 1− 1 .29 −1.For 2 x 2 systems: y1 = K11u1 + K12u2 y2 = K 21u1 + K 22u2 Example 1: ⎡ K11 K =⎢ ⎣ K 21 K12 ⎤ ⎡ 2 1.64 0. 36 0 .5⎤ K =⎢ ⇒ ⎥ ⎣1. 1 29 2 29 − ⎣ ⎦ Example 2: ∴ ⎡ −2 1.5 2 ⎦ ⎡ 0.36 ⎤ Λ= ⎢ ⎥ 0 . . Y2 and U2. 64 ⎣ ⎦ Recommended pairing is Y1 with U1 and Y2 with U2.29 ⎤ ∴ Λ= ⎢ ⎥ . ⎡ 2. 23 .

and Tc. Th.EXAMPLE: Thermal Mixing System Chapter 18 The RGA can be expressed in two equivalent forms: Wh ⎡ T − Tc W⎢ Th − Tc ⎢ K= ⎢ Th − T T ⎢ ⎣ Th − Tc Wc Th − T ⎤ Th − Tc ⎥ ⎥ T − Tc ⎥ Th − Tc ⎥ ⎦ and Wh ⎡ Wc W⎢ Wc + Wh ⎢ Λ= ⎢ Wh T ⎢ ⎣Wc + Wh Wc Wh ⎤ Wc + Wh ⎥ ⎥ Wc ⎥ Wc + Wh ⎥ ⎦ Note that each relative gain is between 0 and 1. 24 . The recommended controller pairing depends on nominal values of T.

y = Ku Hij is the (i. Chapter 18 u1 u2 y1 ⎡ λ11 λ12 λ21 λ22 y2 ⎢ ⎢ Λ = ⎢ ⎢ yn ⎢ ⎣ λn1 λn1 un λ1n ⎤ λ2 n ⎥ ⎥ ⎥ ⎥ λnn ⎥ ⎦ (18 − 25) Each λij can be calculated from the relation.RGA for Higher-Order Systems For and n x n system. -1 T Λ ≠ KH .j) -element of the H = K Note : ( ).j) -element of the steady-state gain K matrix. λij = Kij H ij (18 − 37 ) where Kij is the (i.

135 3.286 1.080 −0.919 ⎦ Recommended controller pairing? 26 .931 0.150 0.900 1.429 0.154 ⎥ ⎢ ⎥ y3 ⎢ −0.314 −0.164 ⎤ y2 ⎢ −0.270 −1.910 ⎥ ⎢ ⎥ y4 ⎣ 0.011 −0.215 −2.Example: Hydrocracker The RGA for a hydrocracker has been reported as. Chapter 18 Λ = u1 u2 u3 u4 y1 ⎡ 0.030 0.

…. Thus. • The columns of matrices W and V are orthonormal. • Condition number (CN) is defined to be the ratio of the largest to the smallest singular value. 27 . T T WW = I and VV = I • Can calculate Σ.Singular Value Analysis • Any real m x n matrix can be factored as. Chapter 18 CN σ1 σr • A large value of CN indicates that K is ill-conditioned. σ2. σr) • The singular values are the positive square roots of the T T eigenvalues of K K ( r = the rank of K K). svd. T K=WΣV • Matrix Σ is a diagonal matrix of singular values: Σ = diag (σ1. and V using MATLAB command. W.

> 10). • Consider the RGA for a 2x2 process.. 28 . ⎡ 1 0⎤ K =⎢ ⎥ 10 1 ⎣ ⎦ ⇒ Λ= I • If K12 changes from 0 to 0.Condition Number Chapter 18 • CN is a measure of sensitivity of the matrix properties to changes in individual elements. Thus small changes in the model for this process can make it very difficult to control. ⎡10. then K becomes a singular matrix.1 0 ⎤ ∑ (K ) = ⎢ ⎥ ⎣ 0 0. • RGA and SVA used together can indicate whether a process is easy (or difficult) to control.1⎦ CN = 101 • K is poorly conditioned when CN is a large number (e.1.g. which corresponds to a process that is difficult to control.

Example: ⎡ 0 .0 0 6 ⎤ 0 .Selection of Inputs and Outputs • Chapter 18 • • Arrange the singular values in order of largest to smallest and look for any σi/σi-1 > 10.0 0 8 ⎥ ⎥ 0 .5 2 ⎢ ⎣ 0 . Delete one row and one column of K at a time and evaluate the properties of the reduced gain matrix.0 2 0 ⎥ ⎦ 29 .9 5 − 0 .9 5 − 0 .4 8 K = ⎢ ⎢ 0 . then one or more inputs (or outputs) can be deleted.9 0 0 .9 0 0 .

5 (σ1/σ3) The RGA is: ⎡ −2. 30 .5407 − ⎢ ⎥ ⎢ ⎣ 2.2211 0. y2-u3.3766 0.9999 ⎥ ⎦ • CN = 166.⎡ 0.5714 0.0241 0.0068 ⎢ ⎥ ⎢ ⎣ −0.7292 ⎤ ⎥ W =⎢ ⎢ 0. y3-u1.0458 ⎥ ⎦ Preliminary pairing: y1-u2.0151⎤ ⎡ 0.4093 −0.618 ⎥ 0 1.2623 0. CN suggests only two output variables can be controlled.6035 0.4376 3. Eliminate one input and one output (3x3→2x2).6843⎥ ⎢ ⎣ −0.0097 ⎥ ⎣ 0 ⎦ 0.0154 −0.5561 0.0541 0.0060 0.9984 ⎥ − − V =⎢ 0.143 0 ∑=⎢ ⎢ ⎥ ⎢ 0 0.9985 0.0540 0.4135 ⎤ ⎥ Λ = ⎢ 1.8311 0.2165 −1.7617 0.0066 ⎥ ⎦ Chapter Chapter18 18 0 0 ⎤ ⎡1.

Chapter 18 Question: How sensitive are these results to the scaling of inputs and outputs? 31 .

"Detune" one or more FB controllers. Use a decoupling control scheme. 2. decoupling control allows setpoint changes to affect only the desired controlled variables. Use some other type of multivariable control scheme. e. 4.g.Alternative Strategies for Dealing with Undesirable Control Loop Interactions 1... Select different manipulated or controlled variables. Decoupling Control Systems • Basic Idea: Use additional controllers to compensate for process interactions and thus reduce control loop interactions • Ideally. a steady-state model or transfer function model) 32 Chapter 18 . decoupling controllers are designed using a simple process model (e. nonlinear functions of original variables 3. • Typically.g.

Chapter 18 33 .

Thus. T12 GP11U 22 + GP12U 22 = 0 (18 − 79 ) (18 − 80 ) (18 − 76 ) (18 − 78) Chapter 18 Because U22 ≠ 0 in general. then GP12 T12 = − GP11 T21GP 22U11 + GP 21U11 = 0 GP 21 ∴ T21 = − GP 22 Similarly. Compare with the design equations for feedforward control based on block diagram analysis 34 . to cancel the effect of U2 on Y1. Thus. we would like. we require that. T12.Decoupler Design Equations We want cross-controller. we want T12 to cancel the effect of U1 on Y2.

Static Decoupling: Design to eliminate SS interactions Ideal decouplers are merely gains: K P12 T12 = − K P11 K P 21 T21 = − K P 22 3.” Chapter 18 2. 35 .Variations on a Theme 1. Nonlinear Decoupling (18 − 85) (18 − 86 ) Appropriate for nonlinear processes. Partial Decoupling: Use only one “cross-controller.

composition (wt.Wood-Berry Distillation Column Model (methanol-water separation) Chapter 18 CT Feed F Reflux R Distillate D. %) XD Steam S CT Bottoms B. composition (wt. %) XB 36 36 .

6e ⎢ ⎣10.7 s + 1 ⎡ y1 ( s ) ⎤ ⎢ ⎢ y ( s) ⎥ = ⎢ ⎣ 2 ⎦ ⎢ −7 s ⎢ 6.4e −3s ⎥ 14. %MeOH u1 = R = reflux flow rate. %MeOH y2 = xB = bottoms composition.4 s + 1 ⎥ ⎦ Chapter 18 ⎡ u1 ( s ) ⎤ ⎢u ( s ) ⎥ ⎣ 2 ⎦ (18 − 12) 37 37 .9e −3s ⎤ ⎥ 21s + 1 ⎥ ⎥ ⎥ −19. lb/min −18.8e− s ⎢ 16. lb/min u1 = S = reflux flow rate.Wood-Berry Distillation Column Model ⎡ 12.9 s + 1 where: y1 = xD = distillate composition.

Chapter 18 .