You are on page 1of 8

Lecture 9 - Review and Assessment

1. Introduction
• Mid course summary

K. J. Åström

– What have we achieved so far
– What remains

1. Introduction
2. Design of Simple Controllers

• Standard models

3. The Basic Feedback Loop

– Block diagrams
– Differential equations
– Transfer functions

4. Specifications
5. Assessment

• Design of simple controllers

6. Summary
Theme: Review of what we learned so far. Assessment of
a control systems is important, particularly if you are a
customer of control.

Block Diagrams

• Insight into feedback and feedforward
• The basic feedback loop
• Fundamental limitations

Standard Model 1 - Differential Equations

The block diagram gives an overview of a system. To draw a
block diagram:
• Understand how the system works.
• Identify the major components and the relevant signals.

A standard model for linear time invariant system
dn y
d n−1 y
d n−1 u
+
a
+
.
.
.
+
a
y
=
b
+ . . . + bnu
1
n
1
dtn
dtn−1
dtn−1

is characterized by two polynomials

• Key questions:

A( s) = sn + a1 sn−1 + a2 sn−2 + . . . + an−1 s + an

Where is the essential dynamics?
What are appropriate abstractions?
Think about physics and storage of mass, energy and
momentum
• Describe the dynamics of the blocks in terms of standard
models.

B ( s) = b1 sn−1 + b2 sn−2 + . . . + bn−1 s + bn

• The roots of A(s) are called poles of the system.
• The roots of B (s) are called zeros of the system.
B ( s)
• The transfer function of the system is G (s) =
A(s)

c K. J. Åström August, 2001
&

1

5 0 −0..5 −1 c K. Experimental determination of step and impulse responses. .25 0.2 0. Graphical representations • Nyquist plot • Bode plot Nyquist’s stability criterion. . Ck (t) polynomials and g is the impulse response. then the output is  y(t) = h G (iω )h sin ω t + arg G (iω ) The number h G (iω )h is called gain ratio or simply gain and the number arg G (iω ) is called phase of the transfer function. stability margins Output y 0. If the input is u(t) = sin ω t.Standard Model 1 . Standard Model 2 . .05 0 −0. + a y = b + . J. which has the form g(t) = n X C¯k−1 (t) eα k t e− st f ( t) dt 0 0 k=1 ∞ where the Laplace transforms are calculated under the assumption that all initial values are zero.. . 2001 & 0 5 2 . A block is simply characterized by Y ( s) = G ( s) U ( s) k=1 Notice appearance of α k again! Signals and and systems have the same representations. + b nu 1 n 1 dtn dtn−1 dtn−1 has the solution Z t n X α kt y(t) = Ck−1(t) e + g(t − τ )u(τ )dτ The transfer function can also be defined as Y ( s) Ly = G ( s) = U ( s) Lu Z L f = F ( s) = where {α k} are roots of the characteristic equation A.05 −0.1 0. Transfer functions and Laplace transforms are ideal to deal with block diagram.1 0 5 10 15 10 15 Input u 1 0.15 0.Transfer Functions dn y d n−1 y d n−1 u + a + . Åström August. Frequency Responses Notice Steady State Responses The complex number G (iω ) tells how a sinusoid propagates through the system in steady state.

For the systems above we have C F = Mu + CM y c K. The Basic Feedback Loop Many Versions of 2DOF r n d Σ F F Σ e u C ki ( R(s) − Y (s)) s Closed loop transfer function from reference r to output y • Behavior of second and third order systems r b s+a Σ x P Σ u C y −1 −1 r Mu My Ingredients: um ym Σ C u • Controller: feedback C. but . parameters ω 0 and ζ can be chosen freely. Åström August. they can be modified or eliminated if the controller has two degrees of freedom. Typical prototypes: s2 + 2ζ ω 0 s + ω 02 (s2 + 2ζ ω 0 s + ω 02)(s + ω 0 ) Controller with 2DOF U (s) = − kY (s) + bki ω 02 Y ( s) = 2 = 2 R( s ) s + (a + bk)s + bki s + 2ζ ω 0 s + ω 02 • The effect of poles and zeros • Some zeros are due to the controller. J. 4. Closed loop system of second order. feedforward F Σ y P −1 • Load disturbance d: Drives the system from desired state • Measurement noise n: Corrupts information about x • Process variable x should follow reference r y P For linear systems all 2DOF configurations have the same properties.. PI controller has two parameters. Design of Simple Controllers PI Control of First Order Systems • Draw block diagram Process transfer function • Obtain simple process model P ( s) = • Select a controller of sufficient flexibility (complexity) • Derive the closed loop characteristic equation • Pick controller parameters to give desired characteristic polynomial (pole placement).3. 2001 & 3 ..

J.4.2 for Ms.Part 1 1 L . Mt < 1.5 ω gc p 0. Åström August. 2001 & 0.Designing System with Two Degrees of Freedom The Gang of Six n d Design procedure: • Design the feedback C to achieve r F – Small sensitivity to load disturbances d – Low injection of measurement noise n – High robustness to process variations e Σ u C Σ y P PC PCF D− N+ R 1 + PC 1 + PC 1 + PC P 1 PCF D+ N+ R Y= 1 + PC 1 + PC 1 + PC PC C CF U =− D− N+ R 1 + PC 1 + PC 1 + PC For many problems in process control the load disturbance response is much more important than the set point response. Z ∞ X π log h S(iω )hdω = π Re pk − lim sL(s) 2 s→∞ 0 for Ms. S= S= x P −1 • Then design the feedforward F to achieve desired response to command signals r. 4 .  ω gcT ≤ • A RHP pole p c K.37 for Ms . Mt = max hT (iω )h 1+ L 1+ L The value 1/ Ms is the shortest distance from the Nyquist curve of the loop transfer function L(iω ) to the critical point −1. Mt < 2 0.7 for Ms . T= . Mt < 1. Few textbooks and papers show more than set point responses. Sensitivity Functions Σ Ycl(s) V log T = V log P Yol (s) • A RHP zero z ω gc z  ≤ • A time delay T How much can the process be changed without making the system unstable? h∆ Ph 1 < h Ph hT h Bode’s integral the water bed effect. X = Rules of Thumb for Limitations . System inversion. The set point response is more important in motion control. Mt < 2 0. Ms = max h S(iω )h. Mt < 2 5 for Ms .4.  ≥ 2 for Ms .4. Mt < 1.

2001 & 5 . Mt < 2 pT ≤ 0.5 for Ms .4 14.4 p • If you are designing control system it is important to assess their properties • A RHP pole-zero pair with z < p  p 6. Assessment of a Control System • A RHP pole-zero pair with z > p  z 6.4 z • If you are using control systems it is important to assess their performance • If you are a customer to control system designers is is useful to assess their properties • You must consider all properties not just the response to reference signals • A RHP pole p and a time delay T  0.4. Mt < 2 ≥ for Ms . Åström August. Mt < 2 ≥ for Ms. Mt < 1. We use our new insight into the properties of the basic feedback loop and specifications to make an assessment of the design. J.5 for Ms. Examples PI Control of a First Order System Previously we have seen that it was straight forward to use pole placement to design a controller for a given process. Mt < 1.16 for Ms.4. Mt < 1.05 for Ms. Two simple examples are used as illustration. 14.Part 2 5. • PI Control of a first order system • PI Control of a second order system Some interesting questions Process model: P = b s+a ki (R − Y) s b( ks + ki) Loop transfer function: L = s ( s + a) Controller with 2 DOF: U = − kY + Closed loop characteristic equation: s2 + s(a + bk) + bki = 0 Choose k = (2ζ ω 0 − 1)/b and ki = ω 02/b to give • How should the closed loop poles be chosen? s2 + 2ζ ω 0 s + ω 02 = 0 • Are there any restrictions? Are there any restrictions on ζ and ω 0 ? c K.Rules of Thumb for Limitations .

7. T ( s ) = s + 2ζ ω 0 s + ω 02 s2 + 2ζ ω 0 s + ω 02 The Sensitivity Function The Complementary Sensitivity Function Gain curve of Bode plot for ζ = 0. Notice this occurs even if relative damping ζ is well chosen! c K. Notice this occurs even if relative damping ζ is well chosen! 10 −2 10 −1 10 0 ω 10 1 10 2 10 Choosing ω 0 too small in comparison with a gives a system with poor robustness. a = 1 and ω 0/a = 0. ((2ζ ω 0 − a)s + ω 02 ) s(s + a) S ( s) = 2 .The Gang of Six The Sensitivity Function Response to reference signal Y ( s) = ω 02 s2 + 2ζ ω 0 s + ω 02 R(s). 2001 & 6 . Åström August. J. ζ = 0. 1 (full) and 10 (dotted) 1 10 10 hT (iω )h h S (iω )h 0 10 0 10 −1 10 −2 10 −2 10 −4 10 −3 −2 10 −1 10 0 ω 10 1 10 2 10 Choosing ω 0 too small compared to a gives a system with large Ms large disturbance amplification. a = 1 and ω 0 /a = 0.1 (dashed). 1 (solid) and 10 (dotted) 2 Gain curve of Bode plot for ζ = 0.7.. U (s) = ω 02 (s + a) R( s b(s2 + 2ζ ω 0 s + ω 02 ) Response to load disturbance Y ( s) = bs s2 + 2ζ ω 0 s + ω 02 Sketch the gain curve of the Bode plot for the sensitivity function s ( s + a) S ( s) = 2 s + 2ζ ω 0 s + ω 02 for a = 1.1 D ( s) Response to measurement noise U ( s) = − (s + a)((2ζ ω 0 − a)s + ω 02 ) (s + a)( ks + ki ) N ( s ) = − N ( s) s2 + 2ζ ω 0 s + ω 02 s2 + 2ζ ω 0 s + ω 02 Sensitivity functions The Audience is Thinking ..1 (dashed).5 and ω 0 = 0.

(ω 0 ≥ 0. 2001 & 7 . Has he given you a good controller? Response to reference signal Y ( s) = 1 (s + 1)(s + 0. J.02) 0 20 40 60 80 100 120 140 Very slow recover of load disturbance! Why does the control signal not eliminate it? c K. U (s) = R( s s2 + s + 1 s2 + s + 1 1.02)(s + 1) U ( s) = − N ( s) s2 + s + 1 −0.02 D ( s) U ( s) = − 2 D ( s) 2 (s + 0.Conclusion Control of a Thermal Process Process model: P(s) = s+b a Design a PI controller to give a closed loop system with the characteristic polynomial You are in charge of a project concerning temperature control for a wafer production. Åström August. He also showed that the maximal values of the sensitivity functions are reasonable. T ( s) = 2 2 s +s+1 s +s+1 0 −0.2 −1.5 s s + 0.6 (s + 0. He showed you ice step responses to changes in the reference signal.2 Response to measurement noise S ( s) = Process output 2 Response to load disturbance Y ( s) = 1 (R − Y) 50s He said that the response to a step in the reference signal originally had large overshoot which was eliminated by not applying proportional action to the reference. b ki = P ( s) = ω 20 b To obtain reasonable robustness the parameter ω 0 cannot be too small.4 −0. Too high values of ω 0 give systems that are too sensitive to measurement noise k = 2ζ ω 0 ! Model uncertainty also restricts how large ω 0 can be! Even if we can place the poles arbitrarily the process imposes limitations! The Gang of Six An engineer that you supervised has proposed U = −Y + Response to Step in Load Disturbance 1 (s + 1)(s + 0.02)(s + s + 1) s +s+1 0 20 40 60 80 100 120 140 160 180 200 160 180 200 Control signal 0 −0.5 1 0.02) R(s).8 −1 −1.4 Sensitivity functions s( s + 1) 1 . A simple process model is s2 + 2ζ ω 0 s + ω 02 Controller parameters: k = 2ζ ω 0 − a .5844a gives sensitivities less than 2.

Summary Changing Ti from 50 to 4 gives a drastic improvement • Insight into feedback and feedforward Process output 2 • A good understanding of the basic feedback loop 1. 2001 & 8 .4 0 20 40 60 80 100 120 140 160 180 How to obtain the models? How to deal with many inputs and outputs? Polynomials are bad numerically for high order More efficient computations 200 c K.Drastically Improved Control 6.6 −0. Åström August. J.5 • Summarizing what we learned so far 0 20 40 60 80 100 120 140 160 180 200 Control signal 0 • Assessment gives insight into the design problem and guidance for choice of parameters • What remains to do −0.5 0 −0.2 −1.2 – – – – −0.4 −0.8 −1 −1.5 • Important to look at all six transfer functions (“Gang of Six”) 1 0.