SU1325405A1 - Automatic optimization system - Google Patents

Abstract

The invention relates to technical cybernetics and can be used to maintain the optimal state of multidimensional control objects with varying characteristics. The purpose of the invention is to improve the accuracy of the automatic optimization system. The automatic optimization system contains control object 1, sensors of input 2 and output 3 variables, first 4 and second 5 memory blocks, identification block 6, optimization block 7, input variable controller 8, command block 9, experiment planning block 10, block 11 model accuracy analysis. The purpose of the invention is achieved by introducing a model accuracy analysis unit 11. 1 3, p. f-ly, 5 ill. (/) SLE 1C U1 42) About eaten

Description

ten

The invention relates to technical cybernetics and can be used to maintain the optimal state of multidimensional control objects with varying characteristics.

The purpose of the invention is to improve the accuracy of the automatic optimization system.

FIG. 1 shows the functional diagram of the automatic optimization system in FIG. 2 — the functional diagram of the planning block of the experiment in FIG. 3 is a functional on the diagram g of the optimization unit; Fig. 4 is a functional block diagram of the model accuracy analysis; on fig.Z - functional scheme of the command block.

The system contains a control object 1, sensors of input 2 and high 3 variables, first 4 and second 5 memory blocks, identification block 6, optimization block 7, controller 8 of input variables, command block 9., block 10, experiment shan unit 11 analysis of the accuracy of the model. The experiment planning block contains a signal generator, a torus 12, a signal shaping 13 block, a plan memory 14, an expert-1 block. The unit for stitching has a sampling and storage unit 15, the first arithmetic unit 16, a digital-converter converter 17. The ana block, which is for model accuracy, contains an analog-digital converter 18, the second arithmetic unit 19, b. Neural Formation, Comparison Unit 2t, The command unit has a button, the first element OR 23. Reversible

05

Q

40

counter 24, inverter 25, first 26 and second 27 pulse generators, second element of EPI 28, tert ™ 2 fourth 30 and five 31 pulse generators,

The system of automatic optimization works in the following way.

When the Start button 22 is pressed in the command block 9, the positive pulse is fed to the input of the reversing counter 21, which sets the number of pulses generated by the first 26 and second 27 generators of 1 and pulses. While the number of j-pulses is less than the specified number at the output, the reverse the counter 24 is set to a logical unit, and the output of the ignitor 25 is a logical zero. 3 this case is the third 29, fourth 30 and 31

0

g

o

Generators do not form impulses. As soon as the number of pulses generated by the first 26 and second 27 pulse generators reaches the required number, a logical zero will be generated at the output of the reversing counter 24, turning off the first 26 and second 27 pulse generators, and the output of the inverter 25 is the logical one, the third 29 starts. , the fourth 30 and fifth 31 pulse generators, which form the pulses until the second second input of the first element of the FIB 23 FIB receives a positive No. 1 pulse from the unit 11 of the model accuracy analysis. Then the operation of the command block 9 is repeated in accordance with the above description.

 The IlMpuls reading, generated by the first generator 26 of pulses, goes to block 10 n, panning the experiment to the input of block 14 of memory

5 plan of the experiment. In block 10 of the planning of an exer- gant, signals of 1 strobial influences are generated, which arrive at the second input of the regulator 8 of input variables. At the same time, the second gene; 1ator of 27 pulses forms the recording pulses that arrive at the first

4 and second 5 memory blocks, the first 4 and second 5 blocks are stacking memory devices,

B which, upon arrival, singing recording pulses from the command block 9, occurred a shift of previously recorded information and the recording of new information on the vacant space. They can be implemented on the registers of pos, pedovatelnogo records.

So in the first 4 and second

5 blocks of memory (1) accumulation of matrices of input and transient effects, respectively,. P1 pulses read out)) first by KepaTopoiv. 26 pulses, and write pulses ,, (second-wired;; generator 27 i-EK pulses; alternate with each other; 5): recording pulses follow sa read pulses with a delay greater than the duration of transition processes in object 1 right nm and in the controller 8 input variables. After .chfova, n11 the first 26 and the second 27 pulse generators of the required number) 1 pulses. asked

in a progressive f counter 24, the first 26 and Itera 27 generators 1 pulses from

five

0

five

The third 29, fourth 30, and fifth fifth pulse generators that generate the pulses fed into the identification block 6, the optimization block 7, and the first 4 and second 5 memory blocks, respectively, are switched on and on. The third pulse generator 29 generates a pulse that triggers the identification block 6, in which the coefficients S of the model are generated, specified, for example, for the case of a linear object having N inputs and M outputs, in the form N

Zs

hl "- Y

t.

m 1, m.

In this case, the coefficients S of the model are determined from the M systems of linear algebraic equations to

 Yfc, k "ttn,,

btl

the output signal of the model Y and the output signal of the object Y, coming from the sensors 3 output variables through 25 cuts the analog-to-digital converter 18.

The generated square of the mean square error is compared in block 21 of comparison with the threshold pulse:

"

Dm1

YM

in which is the element of the matrix of input signals, Y) is the element of the matrix of signals of output actions.

The fourth pulse generator 30 generates a pulse triggering optimization unit 7 and following the pulse .JQ before the specified amplitude value generated by the third pulse generator 29, with a delay equal to the time for generating the coefficients S of the model in the identification block 6. The optimization unit 7, using the coefficients S of the model and the signals from the sensors of the 3 output variables, generates the optimal control signals acting through the controller 8 of the input variables on the object 1 of the control. For example, if the square of the standard deviation of the signals is used as the optimality criterion

Z

35

40

A fifth pulse generator 31 generates a write pulse arriving in the first 4 and second 5 memory blocks, and following the pulse, we form a fourth pulse generator 30 with a delay equal to the formation time in block 7 of the optimization of optimal signals and the duration of transition signals. processes in the controller 8 input variables and control object 1. If the recorded conditions are fulfilled, then at the output of the model accuracy analysis block 11 a positive impulse is not formed, and the system continues to work in optimization mode when, in accordance with the pulses generated by the third 29, fourth 30 and fifth 31 pulse generators, in block 6 identification, the new model coefficients S are formed, in the optimization unit 7, new optic signals are received, coming through the controller 8 input variables to the input of the control object 1, the first and second 5 memory units are stored

M and

-III

Wsl h: i

M

,,

OUTPUT variables Y of the control object 1 from some given YQ signals, then the optimal control signals can be determined by solving a system of linear equations

N M-

IlY, Smi, J 1, N.

Ix (I.s, s.)

n: 1 # r

Perv arithmetic unit 16, using the coefficients S model, formed in block 6 identification, and the signals from the sensors 3 output trans

The variables generated by the sampling and storage unit 15, synchronized by the steering wheels themselves, coming from the fourth generator 30 of the pulses, form optimal control signals.

The unit 11 for analyzing the accuracy of the model from incoming signals from sensors 2 in the common variables X. and coefficients S of the model, coming from identification block 6, generates output variables f in the second arithmetic unit 19

 V

 m -

and forms the B block 20 forming the square of the mean square error m

11.- Y

(Mil

(Ml

the output signal of the model Y and the output signal of the object Y, coming from the sensors 3 output variables through the analog-to-digital converter 18.

The generated square of the mean square error is compared in block 21 of comparison with a predetermined amplitude value.

threshold pulse:

"

Dm1

YM

unit specified value

Z

before a given amplitude value

The fifth pulse generator 31 generates a write pulse that enters the first 4 and second 5 memory blocks, and follows the pulse generated by the fourth pulse generator 30 with a delay equal to the formation time in block 7 of optimizing optimal signals and the duration of transients in the controller 8 input variables and control object 1. If the recorded conditions are complete, then the output of the model accuracy analysis block 11 does not generate a positive pulse, and the system continues to operate in optimization mode when, in accordance with the pulses generated by the third 29, fourth 30 and fifth 31 pulse generators, in the block 6 identifications form new coefficients of the S model, in block 7 of optimization new optimal signals are generated, coming through the controller 8 input variables to the input of the control object 1, in the first 4 and second 5 memory blocks are stored

513

optimum signals, acting from: sensors of 2 input variables, and filing from sensors of 3 output variables, respectively. And so it goes until the condition is not violated. Then, at the output of the model accuracy analysis unit 11, a positive impulse is generated, which arrives at the command unit 9 at the second input of the first element of WIT 23. Thus, the logical unit is formed at the output of the reversing counter 24, the starting first 26 and second 27 pulse generators and disabling the third 29, fourth 30 and fifth 31 pulse generators. Thus, the system switches. The v mode of information accumulation and identification is described, and its operation begins again in accordance with the above description.

Thus, during the operation of the proposed automatic optimization system in the optimization mode, along with the formation of optimal signals, the model of the object is refined. control, while instead of the test signals are used optimal signals generated in the optimization block. Refining the object model in optimization mode leads to an increase in the accuracy of the automatic optimization system. In addition, the model's accuracy analysis unit stops the transition from the optimization mode to the data accumulation and identification mode only when it is sharp; a change in object characteristics or a control model does not meet the accuracy criterion.

Claims (2)

1. An automatic optimization system, containing sensors of input variables, connected by inputs to the inputs of the control object, and outputs to the first inputs of the first memory block, the output of which is connected to the first input of the identification block connected to the output from the first input optimization unit, the output of which is connected to the first input of the input variable controller, connected by the output with the inputs of the control object, 056 sensors of output variables connected by inputs to the outputs of the control object, and outputs to the first inputs of the second memory block, output to: which is connected to the second input of the identification block, the command block connected to the first output through the planning block of the experiment with the second input the input variable controller, and the second output - with the second inputs of the first and second memory blocks, characterized in that, in order to improve the accuracy of the automatic optimization system, a model accuracy analysis unit was introduced, the first input of which go is connected to the output of the i.centification block, and the second and third inputs are connected to the sensor outputs of the input variables, the second input of the optimization block is connected to the third output of the command block, and the third input to the first inputs of the second memory block and the fourth input of the model accuracy analysis block , connected by Bjodom with the input of the command block, the third input of the identification block is connected to the fourth output of the command block. 2. The system is populated with the fact that the command block contains the first and second elements ILR, the reversible counter, the inverter, the first, second, third, fourth and fifth generators of impulses and the Start button connected to the first input of the first element I.PI whose second input is the input of the command block, the 3 output is connected to the input of a reversible counter, connected via an output of the first pulse generator to the counting input of a reversible counter and the first output of the command block, through a second pulse generator to the first input the second element OR, through the inverter, with the inputs of the third, fourth, fifth pulse generators, the output of the fifth pulse generator is connected to the second input of the second SHTI element, the output of which is the second output, and the outputs of the third and fourth pulse generators, respectively, are fourth to third command block outputs, FIG. 2 Editor E. Papp 5 Compiled by V. Bashkirov Tehred L.Serdyukova Proofreader L.Pnlipekko Order 3106 / 41Tyr already 863Subscription VNIIGO State Committee of the USSR for inventions and discoveries 113035, Moscow, Zh-35, Raushsk nab., 4/5 Production and printing company, Uzhgorod, st. Project, 4 FIG. 3 Figm