1 of 42

��BASIC CONTROL THEORY�LECTURE 6��

TOPICS

  • Sampled data control systems. Schematic diagram of a sample data control system.
  • Sampling and holding. Physical and mathematical sampling. Shannon sampling theorem.
  • Example of a simple sampled data control system.
  • Description of discrete time signalin the time domain. z-transformation and its basic properties. The z-transformation of elementary time series. Inverse z-transformation.
  • Description of sampled data systems in the time domain (difference equation)

and in the z-operator domain with the pulse transfer function

  • Relation of the pulse transfer function and the difference equation
  • Pulse transfer functions of typical elements. Pole-zero configuration of pulse transfer functions.
  • Discrete state space model

2 of 42

THE STRUCTURE OF A SAMPLED DATA CONTROL SYSTEM

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

2

2019

Nowadays, it is becoming increasingly commonplace for control systems to be implemented by digital devices with adequate real-time services. Depending on the number and complexity of controls within a given application, the range of digital control devices ranges from single-chip microcontrollers to more powerful midi / mini digital computers. Networking of digital equipment also allows for the creation of distributed control systems.

3 of 42

ADVANTAGES OF THE DIGITAL TECHNOLOGY

  • Digital technology is more reliable and cheaper.
  • Digital technology is more flexible concerning the algorithms that can be implemented.
  • Modifications and extensions are easy to implement.
  • Accuracy is guaranteed for a long term.
  • Simple and effective methods are available for setting the reference signal, changing control parameters and monitoring the operation.

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

3

2019

4 of 42

THE STRUCTURE OF A SAMPLED DATA CONTROL SYSTEM

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

4

2019

5 of 42

THE STRUCTURE OF A SAMPLED DATA CONTROL SYSTEM

In computer process control, the basic tasks of continuous process control are performed by the computer in real time. The output signal must be digitized for computer processing. At certain sampling intervals, a continuous signal is sampled and, after quantization and digitization, is assigned a numerical value that can be received by the computer. These operations are performed by the A / D analog/digital converter. Through this the continuous process is connected to the computer.

In the computer, in each sampling interval so-called „tasks” are running, which, when implementing closed-loop control, form a digital reference signal, receive the signal from the process and possibly filter it, compare the reference signal with the measured output signal value, and then generate an actuator signal based on a discrete control algorithm. The discrete control algorithm is implemented by a program.

Then the computer forwards the control signal toward the process. This is a digital signal, which must be converted into an analog physical that can actuate the actuator (e.g., voltage, valve stem position, etc.). This operation is performed by the D / A digital/analog converter, which connects the computer to the continuous process.

It should also be ensured that the input signal of the actuator signal is available between the sampling points. This is provided with a hold element that most often holds the value of the previous signal until a new signal arrives at the next sampling point (zero order hold, ZOH).

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

5

2019

6 of 42

SAMPLING

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

6

2019

Physical sampling

Mathematical sampling

Impulse modulation

Sampling entails information loss.

No information is available between sampling points.

In A / D converter, quantization and coding also mean information loss.

7 of 42

SHANNON THEOREM

The sampling time must be carefully selected: it must be consistent with the dynamics of the process being controlled and with the capabilities of the real-time processing system (operational speed, accuracy of numerical representation).

SHANNON THEOREM:

The sampling frequency should be at least twice the frequency of the component with the highest frequency in the signal. In the time range, this means that at least one sample has to be taken every half period of an oscillating signal.

The smaller the sampling time, the closer the operation of the control system approaches the continuous operation. With larger sampling times, fewer calculations are required. By choosing the sampling time we can influence the dynamics, the speed of the control, the magnitude of the control signal.

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

7

2019

8 of 42

HOLDING

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

8

2019

Function of the holding element: A signal should be provided between two sampling times.

Zero order holding

First order holding

In practice, the use of a zero order holding is common.

9 of 42

ZERO ORDER HOLDING (ZOH)

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

9

2019

Taylor approximation:

Sampling and holding adds extra time delay to the system,� whose value is approx. half sampling time.

10 of 42

ELEMENTS AND SIGNALS IN A SAMPLED DATA CONTROL SYSTEM

Continuous and discrete signals appear in the sampled control system.

The control algorithm is implemented with a real-time program.�New types of sensors have been developed transforming analog inputs to digital outputs.

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

10

2019

11 of 42

REQUIREMENTS SET FOR A CONTROL SYSTEM

The requirements are the same as for continuous systems.

Stability

Prescribed static accuracy for reference signal tracking and

disturbance rejection

Prescriptions for dynamic behaviour (overshoot, settling time)

Keeping the control signal within the prescribed limits

Features different from the continuous case:

Impact of sampling (loss of Information)

The value of the control signal is constant between two sampling

points

Investigation of behaviour between sampling points

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

11

2019

12 of 42

EXAMPLE FOR A SAMPLED DATA CONTROL SYSTEM

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

12

2019

The process is an integrator, the controller is a constant element. The sampling time is Ts, zero order hold is applied at the input of the process.

Give the difference equation of the closed loop control system, calculate the values of the output signal in the first 5 sampling points for unit step reference signal, if K=1 and Ts=0.5, 1, 1.5 and 3 sec.

Rearranging:

If Ts tends to 0,

The continuous system is structurally stable.

13 of 42

EXAMPLE: THE VALUES OF THE OUTPUT SIGNAL

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

13

2019

For Ts=0.5: 0; 0.5; 0.75; 0.875; 0.9375

For Ts=1: 0; 1; 1; 1; 1

For Ts=1.5: 0; 1.5; 0.75; 1.125; 0.9375

For Ts=3: 0; 3; -3; 9; -15

The sampled system did not preserve the structurally stable nature

of the continuous system.

At Ts=1 the settling is reached in finite time within one sampling step.

Such a phenomenon does not occur for continuous systems.

14 of 42

DESCRIPTION OF A SAMPLED SIGNAL IN THE TIME DOMAIN. � Z-TRANSFORMATION

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

14

2019

The Laplace transform of the discrete signal:

Introduce the following notation:

This is the z – transformation.

This is the shift operator.

or

This is the z –transform of the signal.

15 of 42

Z-TRANSFORMATION MEANS COMFORM MAPPING

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

15

2019

The z-transformation maps the complex plane s to another complex plane z. The imaginary axis of plane s is formed into a unit radius circle. The left half plane is inside the circle, the right half plane is outside the circle.

Because the entire left half plane "shrinks" to the inside of the unit radius during mapping, greater precision is required in the z-domain when specifying each point than in the s-domain.�(Coefficients, poles, zeros in the s-domain are sufficient to give with 2 decimal accuracy, z-domain requires at least 4 decimals.)

16 of 42

SOME PROPERTIES OF Z-TRANSFORMATION

  • Linearity

  • Shift theorem

  • Finite value theorem

Initial value:

Final value:

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

16

2019

17 of 42

Z – TRANSFORMS OF SOME BASIC SIGNALS

  • Unit pulse

  • Unit step

  • Exponential function

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

17

2019

18 of 42

Z – TRANSFORMS OF SOME BASIC SIGNALS

  • Unit ramp signal

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

18

2019

But this is not a geometric series, it cannot be converted into a closed form.�The signal can be specified as the sum of shifted jumps.

19 of 42

Z – TRANSFORMS OF SOME BASIC SIGNALS

  • Sine and cosine signal

can be derived from the z-transform of the exponential signal.

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

19

2019

20 of 42

A METHOD TO DETERMINE THE Z-TRANSFORMS

The form of a continuous signal:

Its Laplace transform:

Its z-transform:

and then bring it to a common denominator.

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

20

2019

There are z-transformation tables.

(See the Lecture notes.)

21 of 42

INVERSE Z-TRANSFORMATION

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

21

2019

Inversion integral

is located inside the circle of radius C sugarú around the origo.

Each pole of

Practical methods:

With polynomial division

With partial fractional expansion

Example

22 of 42

PULSE TRANSFER FUNCTION

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

22

2019

Calculation of step equivalent pulse transfer function:

23 of 42

PULSE TRANSFER FUNCTIONS OF BASIC ELEMENTS

Integrator

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

23

2019

Or with step equivalent calculation: for unit step the output of the integrator is a ramp.

Remark: In continuous case the pole is in s1=0, in discrete case it is in z1=1.

24 of 42

PULSE TRANSFER FUNCTIONS OF BASIC ELEMENTS

First order lag element

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

24

2019

Remark: In continuous case the pole is at s1=-1/T1, in discrete case the pole is at z1=exp(-Ts/T1)

25 of 42

PULSE TRANSFER FUNCTIONS OF BASIC ELEMENTS

Second order lag element

If there is a sampler and a holder between two first order lag elements connected in series, the pulse transfer functions are simply multiplied.

If there is no sampler in between:

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

25

2019

Remark:

There are two poles and one zero,

Because of sampling and holding an additional zero is introduced!

26 of 42

EXAMPLE

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

26

2019

The sampling time is Ts=5 sec

If there is a sampler and holder between the two first order lags:

If there is no sampler and holder between the two first order lags:

27 of 42

PULSE TRANSFER FUNCTIONS OF BASIC ELEMENTS

Pulse transfer functions of elements containing more lags, pulse transfer functions of elements given by rational can be calculated similarly, by partial fractional expansion.

In the pulse transfer function of element containing more lags, the number of zeros is 1 less than the number of poles. There will be zeros also outside the unit circle.

There are tables providing the pulse transfer functions of the most important elements.

Dead time element:

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

27

2019

28 of 42

EXAMPLE

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

28

2019

The pulse transfer function of the closed loop system:

Notation:

For unit step reference signal:

K=1 and for Ts=0.5 :

The result is the same as obtained with the difference equation.

29 of 42

ABOUT STABILITY

The roots of the characteristic equation (the poles of the pulse transfer function of the closed loop) in the z-domain should be inside the unit circle.

For our example:

The range of stability:

(The same result was obtained analysing the difference equation.)

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

29

2019

30 of 42

TRANSITION FROM THE PULSE TRANSFER FUNCTION TO THE DIFFERENCE EQUATION, DISCRETIZATION OF THE INTEGRATOR

Example: Integrator

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

30

2019

Transition to the difference equation:

This is a recursive relationship,

which is easily programmable.

Zero order hold corresponds to the left hand rectangle rule.

According to the right hand rectangle rule:

The corresponding pulse transfer function:

According to the trapezoid rule:

The corresponding pulse transfer function:

31 of 42

DISCRETE STATE SPACE MODEL

The discrete state equation can be determined by sampling the continuous state variables.

The continuous state equation and its solution:

Zero order hold is applied. Let us execute the integration between two consecutive sampling points.. Let be the beginning of the sampling interval and the end of the interval.

In this range is constant.

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

31

2019

32 of 42

DISCRETE STATE SPACE MODEL

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

32

2019

where

If matrix A is invertible,

The pulse transfer function on the basis of the discrete state equation:

33 of 42

RECURSIVE SOLUTION OF THE STATE EQUATION

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

33

2019

Recursive solution:

34 of 42

PREDICTIVE STATE SPACE MODEL

With a so called predictive model, we intend to determine and predict the values of the future output signal or state variables of the system based on the available information and data on the input signal and the state variables (or the output signal) up to the given k sampling moment. Assume d = 0.

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

34

2019

35 of 42

PREDICTIVE STATE SPACE MODEL

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

35

2019

The vertical lines mean that the predicted value of the signal is determined based on the information available up to the current time k. In vector / matrix form up to 4 predicted points:

The first part of the right side of the equation is the so-called free response. The second part gives the forced response. The free response is the effect of past interventions on the future time horizon, and we cannot change this, we must bear that effect. The generated forced response gives the effect of the current and next input interventions on the output signal. These interventions can be chosen so that the output signal follows a given reference signal as closely as possible over a specified future time horizon. We can only influence the future behavior of the system with input effects at current and future times.

36 of 42

EXAMPLE: DISCRETE STATE SPACE MODEL OF A DOUBLE INTEGRATOR

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

36

2019

The continuous state equation:

The state matrices of the discrete state equation:

The pulse transfer function:

For Ts=1:

37 of 42

DETERMINING THE STATE EQUATION FROM THE PULSE TRANSFER FUNCTION

For a given pulse transfer function, an infinite number of state space descriptions can be generated which give the same output to a given input.

Consider a third-order pulse transfer function.

Observable canonical form

(Remark: d is not zero only if the degree of the numerator and denominator of G are equal.)

Restructuring:

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

37

2019

or

38 of 42

OBSERVABLE CANONICAL FORM

Further conversion supposing d=0:

Rearranging:

Based on this the block scheme

and the state equation:

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

38

2019

The state variables can be recorded at the output of the shift elements.

39 of 42

CONTROLLABLE CANONICAL FORM

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

39

2019

Conversion of the pulse transfer function by introducing the auxiliary variable V :

Based on this expression:

The block scheme:

The controllable form of the state equation:

40 of 42

CANONICAL FORM

With partial fractional expansion:

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

40

2019

Introduce the state variables as follows:

The block scheme:

The canonical state equation:

41 of 42

THE CONTROLLER IS DESIGNED� FOR THE MODEL OF THE PROCESS

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

When designing the controller, we consider the process model.

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 domains.

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

41

2019

42 of 42

THANK YOU FOR YOUR ATTENTION���