1 of 22

�BASIC CONTROL THEORY�LECTURE 7�

TOPICS

  • Frequency function of a sampled data control system and its relation to the frequency function of the continuous system
  • Design of discrete PID controllers using pole cancellation technique
  • Determining the start and end values of the signals in the control system for a unit step reference signal.
  • State feedback in sampled control systems. State feedback with state estimation.

 

2 of 22

2

3 of 22

DEMONSTRATING EXAMPLE: INTEGRATOR

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

3

2019

correction

The approximation is valid, if ω<1/Ts

Thus, in the low frequency range, the amplitude-frequency diagram of the continuous and sampled systems are

roughly the same, the discrete phase-frequency diagram differs from that of the continuous one with the angle

of the additional dead time of value Ts / 2.

4 of 22

SAMPLING AND HOLDING INTRODUCES ADDITIONAL DEAD TIME

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

4

2019

Taylor approximation:

Sampling and holding introduces additional dead time whose value

is approximately half of the sampling time.

5 of 22

DEMONSTRATING EXAMPLE: FIRST ORDER LAG ELEMENT

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

5

2019

If Ts/T1<1 and ωTs<1

Example: T1=0.1, Ts=0.1

Bode diagram

6 of 22

THE SAMPLED CONTROL SYSTEM

Continuous and discrete signals appear in the sampled control system.

The control algorithm is implemented with a real-time program.�When designing the controller, the additional dead time resulting from sampling should be taken into account.

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

6

2019

7 of 22

REQUIREMENTS SET FOR THE CONTROL SYSTEM

The requirements are the same as for continuous systems.

Stability

Prescribed static accuracy for reference signal tracking and

disturbance rejection

Dynamic prescriptions (overshoot, settling time)

Keeping the the control signal within the prescribed limits

Features different from continuous case:

The effect of sampling and signal conversion (loss of information)

The value of the control signal is constant between two sampling

points

Investigation of behavior between sampling points

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

7

2019

8 of 22

THE CONTROLLER IS DESIGNED� FOR THE MODEL OF THE PROCESS

The controller should be designed so that the controller meets the quality specifications.

When designing the controller the model of the process is considered.

The process model can be the pulse transfer function or the state equation.

The control loop can be examined and designed in the time,

the z-operator, and the frequency domain.

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

8

2019

9 of 22

DISCRETE PID CONTROLLERS

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

9

2019

CONTINUOUS

CONTROLLER

DISCRETE CONTROLLER

DIFFERENCE

EQUATION

STEP

RESPONSE

Ideal PD: -----

10 of 22

THE IDEAL DISCRETE PD CONTROLLER IS REALIZABLE

as its overexcitation (the ratio of its initial and final value for step input) is not infinity. (In continuous case the overexcitation is infinity, therefore the continuous ideal PD controller is not realizable.)

Calculate the value of the overexcitation for unit step input.

Initial value:

Final value:

The overexcitation:

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

10

2019

11 of 22

DESIGN OF DISCRETE PID CONTROLLER BASED ON CONSIDERATIONS IN THE CONTINUOUS FREQUENCY DOMAIN

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

11

2019

Pole cancellation technique can be used.

For example

With PI controller the biggest time constant is „cancelled” and instead an integration effect is introduced.

With PD controller the second biggest time constant is cancelled and instead a smaller one is introduced.

The gain factor of the regulator kc is chosen from the Bode diagram of the open loop ensuring

phase margin of ~60°. For this the cut-off frequency should be located at the straight line

of slope -20dB/decade and . Td is the continuous dead time, Ts is the sampling time.

The discrete Bode amplitude diagram approximately coincides with the continuous one until ω ~1/Ts, it is drawn until this point. From control point of view, this range is of interest.

12 of 22

FIRST EXAMPLE FOR CONTROLLER DESIGN

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

12

2019

The transfer function of the continuous process:

The sampling time: Ts=0.5 sec

The pulse transfer function:

Requirements: follow the unit step reference signal without static error;

The control should be as quick as possible; the phase margin should be approx. 60 °.

The task can be solved with PI controller using pole cancellation.

The dead time together with the additional dead time: Th=2+0.25, the cut off frequency should be located

approximately at 1/2Th=0.218.

(Draw the Bode diagramot until ω=2.)

13 of 22

FIRST EXAMPLE FOR CONTROLLER DESIGN (continued)

The pulse transfer function of the open loop:

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

13

2019

Its frequency function:

Its absolute value at the cut off frequency is 1.

whence kc=0.1724

The phase margin: 61.89°. u(0)=kc; u()=1.

The difference equation of the controller:

The behavior of the control system over time can be simulated with matlab and with simulink.

14 of 22

SECOND EXAMPLE FOR CONTROLLER DESIGN

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

14

2019

Design a discrete serial PID controller, fulfilling the

following quality specifications:

- Phase margin ≅ 60°

- The control should be as quick as possible;

- The static error for step reference signal should be 0.

The transfer function of the continuous process:

The sampling time: Ts=1sec

The pulse transfer function:

The pulse transfer function of the PID controller

with pole cancellation:

The pulse transfer function of the open loop:

15 of 22

SECOND EXAMPLE FOR CONTROLLER DESIGN (continued)

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

15

2019

The frequency range is considered till 1/Ts=1. Draw the Bode diagram.

The cut off frequency should be located approximately at 1/(2(1+1)=0.25 to ensure the required phase margin.

This can be calculated by Matlab. Kc is obtained appr. as 15.

Behavior between sampling points

can be checked with the simulink program.

16 of 22

CHECKING THE INITIAL AND FINAL VALUES� OF THE OUTPUT AND THE CONTROL SIGNALS

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

16

2019

The control signal should be kept within limits for practical reasons. (Avoid saturation of the actuator.)

The values can be calculated from the limits of the resulting transfer functions, or based on simple considerations.

For step reference signal:

With simple considerations: if the process contains lags, the output value will initially be 0. If the controller

has an integrator, the output signal is set exactly to 1 in steady state. The final value of the control signal

is the reciprocal of the gain of the process. In t = 0, the feedback signal value is 0, u(0) can be calculated

as if the reference signal is directly exciting the controller.

17 of 22

OTHER CONTROLLER DESIGN METHODS

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

17

2019

whose parameters can be derived from the parallel connected discrete P, I and D effects.

a./ Intermediate continuous controller design and its discretization.

The model of the continuous process is supplemented with the holding element, and for this we design

a continuous controller. Then the discrete equivalent of the controller is determined.

b./ Design of discrete controller based on the discretised process model.

The pulse transfer function of the continuous process is determined. A discrete controller is designed

considering this pulse transfer function which ensures that the pulse transfer function

of the closed loop system fulfills given design specifications. (See Lecture 8.)

c./ Design of discrete controller based on the continuous process model.

(based on frequency domain considerations, this method was dealt with previously.)

18 of 22

STATE FEEDBACK IN SAMPLED DATA SYSTEMS

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

18

2019

The discrete state equation:

The poles of the system are obtained by the roots of the characteristic equation:

The aim of control is to accelerate the dynamic behavior, one of the ways of which is to feed back the state variables. By specifying the poles of the closed loop system, we determine the degree of acceleration.

The control signal is obtained by feeding back of discrete state variables:

The characteristic equation of the closed loop system:

F and g are known, the prescribed poles are given. The state feedback vector can be calculated (Ackermann formula).

As there is no integrating effect in the control system, static compensation is required.

19 of 22

CALCULATION OF THE STATIC COMPENSATION

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

19

2019

whence

and the correction factor:

20 of 22

STATE FEEDBACK IN SAMPLED DATA SYSTEMS

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

20

2019

The block scheme of the state equation

of the discrete process

Block scheme of state feedback in discrete system

It can be enhanced with an additional integrating state variable.

21 of 22

STATE FEEDBACK WITH STATE ESTIMATION

BARS RUTH*, KEVICZKY LÁSZLÓ**, HETTHÉSSY JENŐ*, MAX GYULA*, VÁMOS TIBOR**, *BME AAIT, **SZTAKI

21

2019

The state feedback loop should be faster

than the process.

The state estimation loop should be faster

than the state feedback loop.

Example in matlab exercises material.

22 of 22

THANK YOU FOR YOUR ATTENTION����