DE1816068A1 - Method for controlling production or power machine systems by changing an input parameter of the system in order to drive an output parameter of the system to an optimal value - Google Patents

Description

Method for controlling production or power machine systems by changing an input parameter of the system for the purpose of driving an output parameter of the system to a

Optimal value

The invention relates to a method for controlling production or power machinery systems by changing an input parameter of the system for the purpose of driving an output parameter of the system to an optimal value.

According to the invention such a method is characterized in that a ₉ connected to the system digital computer wirdj programmed so as to re in each

JL

BATH

909013/0925

carries out the following activities as a result:

a) generating a fault signal Bl,

b) superposition of this signal on the input parameter 1 for the purpose of generating a corresponding one Fault signal Xl in the output parameter X,

c) reading out the output interference signal Xl,

d) Analyzing the disturbances of the input and output parameters for the purpose of determining the slope Sl des Ratio output / input,

e) Generating a correction signal Zl as a function of this slope value, which indicates the meaning, in which the input parameter must be changed, to drive the output parameter to its optimum value, and

f) superimpose this correction signal on the input parameters.

An example of a method realizing the method according to the invention Tax system will now be referred to with reference to the

909833/0925

accompanying drawings explains in which

Fig. 1 shows a block diagram of this system,

Fig. 2 shows a three-dimensional diagram,

in which various occurring in the system according to the invention Signals are entered,

Fig. 5 shows a block diagram which

various relationships of the block diagram shown in Fig. 1 more explained in detail, and

Fig. 4 shows a timing diagram showing the

signals used in the system.

1 shows, in the form of a block diagram, a gas turbine engine 10 ₉ which, enumerated in the direction of flow, has a compressor 12, a burner 14, a turbine 16 driving the compressor, a jet pipe 18 and an ejector nozzle 20 with a variable cross section held by this jet pipe. The burner is on a fuel source

- 3 909833/0925

connected, which is controlled by means of a valve 22, which is itself actuated by a signal 1. A servomotor 2 ¹ I belongs to the exhaust nozzle 20 and changes the outlet cross section of the exhaust nozzle. The servomotor 2 ¹ I is actuated by a signal 2. The two signals 1 and 2 are also referred to below as parameters.

A change in parameters 1 and 2 results in changes in the power X of the engine, the latter by means of a thrust knife 26 can be measured. The diagram reproduced in FIG. 2 shows this change in the power value X within the surface area 1 · 2 · X. The through Changing the parameters 1 and 2 achievable maximum or optimal thrust is indicated by a point X max. Goes one assumes that the actually delivered power X is at the location of a point Xo, then the power value must be shifted along a line Xr in order to get to the point X max get. The value Xr has components Xl and X2 in the directions of the axes of parameters 1 and 2, respectively.

The engine is controlled by a computer 28, which the parameters 1 and 2 in the sense of an attitude of Value X at point X max changes. For this purpose, the computer according to the invention is programmed as follows:

a) superimpose interference signals Bl, B2 on the

- 4th - - ^ o ■ · ■ * ■; · 909833/0925

Signals 1 and 2. These fault signals are on the summing points 30 and 32 are added to the signals 1 and 2.

b) Recording the signal X, which now contains interference signals that occur due to the signals B1 and B2.

c) analyzing the signal X with reference to the signals Bi and for the purpose of determining the effect of these Signals to the value X.

d) Generation of correction signals Zl and! Z2 25weeks change of parameters 1 and 2 in the sense of wandering point® X to point X max «

So that a distinction between the signals Bi vtnu B2 is possible for the purpose of this analysis, these signals are based on a group of pseudo random binary signals PRBN. PBBH signals are understood to be a signal sequence which has the property of the Hull correlation and all shifted versions of this correlation. This shows aielj in bed on the basis of Fig. 3 * \ which Einselh © iten de © Reöhnes? » 28 vi & ßaw ■ gives * A shift register ^: 30 contains the binary maturity itlOiOO

■ - .. aw '' We ^ t 'η Jfwetit-gleiefe ue »

and also equal to the total number of different series which are possible within a group. The pulse trains Al to An represent the PRBN signals what can be shown by correlating two such sequences.

Correlation of two series is understood to mean the formation of the correlation coefficient of these series, which in turn is determined by defines the process of adding up the products of the signs of corresponding digits of two binary numbers is. In this context, the sign is always one Digit assumed to be +1 if it is the binary digit, and assumed to be -1 if the digit has the binary value O. If this rule applies to the two rows Al and A3 3 applied, it can be seen that the products 32 of Pre-mark the sum "make 1" The same sum, i.e. the the same correlation coefficient is determined by correlation any two other rows within the row values Bl until Bn is generated. However, if a correlation is made between any of these series and itself, i. E. a correlation of identical series results in a correlation coefficient lu + 7 * Bar, which results in it mOglieli isfe, each of the rows Al to An each one out all the other ranks you know. This identification process is done by adding the corresponding Digits of the relevant series »This results in a

Row Ax, which is called the collective level. If the series Ax is correlated with any of the series A1 to A7 _S, for example with the series A1, it is found that the correlation products form a series Ax which is identical to the series A1. This means that it is possible to have access to each and every one of the rows which correspond to the row Ax. It is noted that _s for each of two identical rows has the value -1 in the present example, the correlation coefficient. This is because the rows selected each have one "1" more than "0" s. If the number of "O" s and "1" s within the row is made the same by removing a "1" in each case, the correlation coefficient of two identical rows will assume the value -s-β Correlation coefficient for unequal rows would then be zero in the present example. If, for some reason, the PRBN signal is an odd ^{! S} O ^m or "1" and consequently has no correlation to the value zero, then the zero correlation can be ensured that the odd "0" or »" 1 "is neglected. The meaning of this rule will emerge later =

For the purpose of forming correlation signals Zl and Z2

the computer is so programmed that it _s the following calculation

- 7 en «α * - _λ , -« .._ »AD ORIGJNAL

carries out:

n-1

(D Sj =

(n-1) Bjo

fsgn (Bjl) x X "i] - Sgn (Bjp) x Xp

n-1

(2) Cj =

(n + 1) (Bjo)

[s

sgn (Aj

-Sgn (Aj (q + |)) χ Ί (q + |)

i = o

In these equations:

j = the number (1, 2, 3, ... n) of the parameter, Zj = the correction signal for parameter j, Sj s the slope of the parameter j,

vi _{; i} 909833/0926

Cj * the curvature (i.e. the degree of change in the slope) of the parameter j,

Aj a the PRBN series selected in each case to control parameter j,

Bj * the signal Aj in oblique * ltnfctnform, which for Disturbance of the parameter j is used and with a reference to a zero mean value on both sides is changeable

η s dia ttesamtaatil of the digits contained in the PKBK series and also the 6es &"ts" hl of the parameters that can be controlled with an n-liffrigcn borrow j _s

dia »uiffite? (Ο, % g 2, cc n) each of these

the dü * 9h the sign BJ disturbed dor Anlas® «welch®» looks old; Boaug on n & Gh iaciöen Saigon

de?

₈ and

«'Den Wo ^ fe i®F 02 ¹ OSq i» welelies »© iao soga ^ ivo @i? @ Öa

BAD ORlOLNAU

The above formula can easily be used in well-known programs * • written in the language and entered into the computer so that it forms a work program in this. the Values with which this program is to be evaluated are formed from the signals Aj, Bj and Ϊ.

The meaning of the individual steps of the above formula and their connection with the engine will now be described in one piece,

The working cycle of the register 30 is initiated by a start signal 34, which determines the pre-formation period of the 3yitems. Saa Signal 34 triggers atiltrdtt »a side signal 36 C? Ig _e 4} fttut, which determines the Siffemperiod of the digits contained in the PRBM-Reitsen. The rows A and A2 are equivalent to Reeiiteckswelleii and the arrangement is so gft droffest * - that "dl · 2 * lts $ aitJt ·" ine? «Islands i UQV t>» trifferatof! Bite yourself over reel iMpttlsgterioie signals 3 $

Dae ieetmerp ^ QSraasi liami as au® vi «x» sections 38 4a and 44 {fig, 5) dividing aiisaseheon. wtrden, in walchen the WtFte Bj, Sj ₈ Oj ma Ij ö ^ aausfc -be.-

R «ib · Al ■ in

- so ■ * " .

SO0i33 / C? SlS

Converted signal Bl, which represents a leading edge signal, which is based on the fact that the signal level due to the Signals Al does not go back to the zero value. This means, that the individual digits i of the series Al to generate a Sloping edge signal 46 are used and that in each case such successive leading edge signals each have opposite slope, so that the signal is not sharp Signal rise or fall. As already mentioned, the signal Bl is applied to the summing point 30, so that a disturbance is superimposed on signal 1 and it is clear that this disturbance corresponds to the leading edge signal 46, that is to say that the signal 1 then also has a signal rise corresponding to the leading edges of the leading edge signal 46 and Has signal drop. The mean level of the signal B1 is denoted by 0 and corresponds to the level which the Signal 1 would have without the interference superimposed on it. The formation of an arithmetic series in the computer for the purpose of generating the leading edge signal and the definition of the mean value level of this signal are part of programming technology well known.

It is reminded that the signal is a "1" more as "0" s. This was therefore determined in such a way that it is ensured that the signal B1 comes on without a vertical step the respective corresponding signal B1 of the next

- ti -

.:., · ■ «ϊ 3 ,,,, 909833/0925 * ^AD ° ^mmL

Trial period follows »This becomes odd" 1 " but eliminated later in the program,

An advantage of using a leading edge signal, which does not go back to its zero value, your disturbance of the signal i lies in the fact that the engine is on such a thing Signal responds with some accuracy. If the signal had the The form of a square wave instead of the form of a leading edge signal would make it difficult for the engine to follow the sharp signal rises and falls of such a square wave exactly and the engine would also inadmissibly rough stressed by such a signal *

The signal B2 is similar to the signal Bi as well a leading edge signal that does not return to its zero value and is based on signal A2 and is used to control signal 2 in a corresponding manner.

The combined effect of the signals B1 and B2 on the engine causes the power X of the same in the in 4 is changed by an amount of interference X * will. Fig. 4 also shows the component signal Xl, which symbolizes the effect that the signal B1 would have, if it were fed alone. A corresponding component signal X2 shows the individual effect of the signal B2. the

- 12 -9 09833/09 2 5

Signals Xl and X2 do not exist as such if parameters 1 and Z are disturbed at the same time * They are only shown «because the purpose of the correlation is to enable the slope and curvature of these signals Xl and X2 to be identified»

In section * t0 of the program, the slope S1 of the signal Xl is made by correlating the signals B1 and X according to the formula (1). The slope 32 of the signal X2 is determined in the same way. If, for example, the value B1 is included in the formula, the result is that the expression Sgn CBIi) χ Xi results in the product of the sign of the signal B1 for each of the seven digits i from 0 to 6. The expression Sgn (BIp) χ Xp is ^{subtracted for the purpose of eliminating the odd 11} I ". The summation of these products gives the correlation coefficient. A division of this summation by the expression η - 1 gives the mean change value of the quantity X, which is equal to the mean Change value of Q size X and a further division by the module of B1 for the relationship I = O gives the slope S1. As far as the signal X is concerned, it is easy to average the interference values of the signal X and the signal X in each case as a positive or negative deviation from this mean value This is done in section 39 (FIG. 3) of the program by adding the highest and

- 13 -90 983 3/09 2 5

lowest values of the deviations of the size X respectively during the trial formation period and by dividing the result by the number 2 for the purpose of forming the mean value. This mean value is then for each digit i in the series of is subtracted from the value of Xi, producing row X. It is clear that the shape of the series X * is actually of those corresponds to the quantity X, but its mean value is zero is equivalent to"

It can also be seen that the signal Z is equivalent to the signal Ax of the general correlation example given above and that the correlation between the quantities B1 and X * can also be applied to the component Xl of the signal X, just as in the general example given the Al series is also valid for the AIx series. However, while in the general example the individual rows must be added to form the row Ax, each of these rows remaining unchanged, ie each of these rows being increased by the value 1, when the rows A1 and A2 are fed to the engine these latter rows each of 1 different growth rates, as can be seen from Fig. 4, the signal X door. However, the correlation properties of the signal are not significantly affected by these incremental amounts.

- 14 - ■

, OfflQiNAL INSPECTED

909833/0925

In formula (2), in the PpogrÄBßiabechnitfc kZ, the value Cj is determined by correlating the quantities Aj and £. Of the

Expression Β & ά CMIi ♦ §> 5 χ IT (i ^ | j results in the correlation «· products and their suramation results in the correlation cosients · The expression Sgn (Ai (C * · |)) χ X (q + 5) becomes for the purpose of eliminating the odd "0." Division by the expression (n + 1) / 2 gives the mean deviation of the signal X at the midpoints of the digits i, ie at the points i + | and this deviation represents one A multiplication by the expression ~ 2 / (Bjo) ² results in an approximation of the first derivative of the slope, ie aft the curvature.

In the formula (3), the slope is in the program portion 44 divided by the curvature ₉ are created whereby the correction signals Zi and ZZ which the summing points are eye leads 30 and 32, the parameters 1 and the point X max jeu drive. The fact that the slope is divided by the curvature means that where the slope is zero, the correction signal will also be zero, whether or not there are any curvature errors. It can be shown that when the area 1 * 2 * X is based on a quadratic equation, the transfer from point Xo to point X max by means of a single correlation step

ORIGINAL INSPECTED 909833/0925

can be executed. Otherwise there are two or more of them Correlation steps necessary to reach the point X max and it is clear that the computer continues to work to follow these steps. In this way, the engine is in an operating state corresponding to point X max held.

Any number of parameters up to the number η - 1 can be used provided that between these parameters there are any possible functional relationships within the engine.

The main advantages of the invention result from the fact that PRBN series in connection with a two-level, non-square interference signal are used. Under a A two-level signal is understood to be a signal that continues to rise between a maximum value and a minimum value and falls and does not remain at any intermediate level for appreciable periods of time. The leading edge signal represents an example of such a bi-level signal and such signals are of course easy to generate and to process. A non-rectangular signal is understood to be a signal that does not have any sharp signal increases or Has signal drops, such as those for square wave signals are characteristic, but which exemplary

- 16 909833/0925

wise has the shape of a leading edge signal, or the shape a sign wave or any other waveform that enables the individual digits of the PRBN series without any sharp steps. Such waves are called non-rectangular waves or waves with finite Called slope. The importance of this requirement is that square waves make it difficult to trace the slope of the Signal X to be determined when this slope is either equal to or approximately equal to zero. To make this clearer a signal XIa is entered in Fig. 4, indicating a change of the signal Xl when the parameter 1 is close to the maximum point Xl max on the area 1 * 2 * X. Like this is also the case with the signals Xl, X2 and X, then the signal XIa is also determined with reference to a zero mean value, to enable the formation of positive and negative values of the signal for the purpose of the slope correlation. If the signal B1 were a square wave and not a non-square wave, then the signal XIa would be in the form of a sharp peak pulse 48 and this would be very make it difficult or even impossible to make an accurate "measurement of the signal size using the peak of the peak pulse, however, as already mentioned, such a measurement is necessary in order to form the mean value for the zero line. The invention makes it possible, especially at the

Cf

Place at which the signal X reaches the value X max, a

- 17 90983 3/09 2 5

To achieve good working accuracy and it is actually desirable that the system the engine at this maximum performance point holds. It should be noted that this great accuracy can be achieved by means of a single two-level signal is achieved, for example by means of a leading edge signal and that it is not necessary, any more complicated Use signals, such as three-level signals, to achieve the required accuracy at the signal peaks.

The invention also makes it possible to obtain an accurate measurement of the curvature by using a simple two-level Signal is used, for example a leading edge signal. It is clear that the signals Xl, X2, X essentially vertical signal edges 50 would have if the signal Bj had the shape of a square wave. Then it would be but also impossible to form the value of the quantity X for the position i + ^ in the formula (2), unless an essential more complicated signal would be used, for example a three-level signal.

The computer has switches, not shown in the drawings, by means of which it is sent to the summing points 30 and 32 and can be connected to the thrust knife 26 or disconnected from these locations. If the computer is not in Activity, the engine can, for example, by hand

- 18 -

S0S833 / 0925

be controlled, in which case the summing points 30 and 32 signals 50 and 52 are supplied.

- 19 909833/0925

Claims (6)

Claims 1. Method for controlling production or power machine systems by changing an input parameter of the system for the purpose of driving an output parameter of the system to an optimum value, characterized in that a digital computer connected to the system is programmed in such a way that it follows in a repetitive sequence Performs activities: a) generating a fault signal (Bl), b) superimposition of this signal on the input parameter (1) for the purpose of generating a corresponding one Fault signal (Xl) in the output parameter (X), c) reading out the output interference signal (Xl), d) Analyzing the disturbances of the input and output parameters for the purpose of determining the slope (Sl) of the Ratio output / input, e) Generating a correction signal (Zl) as a function of this slope value, which indicates the sense in which the input parameter must be changed, - 20 -909833/0925 to drive the output parameter to its optimum value, and f) Superimposing this correction signal on the input parameter. 2. The method according to claim 1, characterized in that the input interference signal (Bl) is in the form of a pseudo-random binary series (Al) has. 3 * Method according to claim 1 or 2, characterized in that that the input disturbance signal (Bl) has a non-rectangular shape Has. 4. The method according to claim 3 * characterized in that that the input interference signal (Bl) has the form of a leading edge signal that does not return to its zero value. 5. The method according to any one of claims 1 to 4, characterized in that the respectively repeating sequence of activities of the digital computer also the determination of the Curvature value (Cl) of the output disturbance values (Xl) and the formation of the correction signal (Zl) as a quotient (Sl / Cl) includes slope and curvature signal. - 21 9 0 9 8 3 3/0925 6. The method according to any one of claims 1 to 5, characterized in that the respectively repeating sequence of Activities of the digital computer also include the generation of a signal representing a pseudo-random binary series (A1), a Conversion of this signal into a non-rectangular signal that does not go back to its zero value and a determination of the called curvature value (Cl) by correlation between this pseudo-random binary series (Al) and the output interference signal (Xl). - 22 909833/0925 -I- Blank page