DE102012104359B4 - Method and device for driving a plurality of actuators - Google Patents

Abstract

Method (100) for controlling several actuators (S1, S2, ... Sn) for changing process values (PW) in a process with the following steps: - A manipulated variable (STG1) is used to adjust each actuator (S1) determined, which indicates a deviation of an actual value (ACTUAL) from a target value (TARGET); - An individual adjustment time (VZ1) is calculated for each actuator (S1), which specifies a pulse duration for the actuating variable (STG1) to act on the actuator (S1) (step 120), the adjustment time (VZ1) depending on a slope value (F_M) is calculated, which indicates the speed at which the process value (PW) changed by means of the adjusted actuator (S1) changes; and - the respective actuator (S1) is controlled for the calculated individual adjustment time (VZ1) with the manipulated variable (STG1) (step 130), the individual adjustment time (VZ1) being acted upon by a play time value (F_CLR) (step 123), which specifies the response time with which the respective actuator (S1) is adjusted if the manipulated variable (STG1) has a change with a change of sign, and a learning procedure (101) for calculating a step (101) at least once before the first calculation of the adjustment time (step 120) first slope value (F_M ') is carried out, the learning procedure (101) comprising a learning phase (103) in which the following sub-steps are carried out: the actuator (S1) is at least twice for a predefinable period (A_OTADJ) to a first state (OPEN ) controlled so that a first process value (F_PV1) and a second process value (F_PV2) are achieved in a process Two partial slope values (F_M1, F_M2) are determined, which indicate the speed at which the respective process value is reached when the first state (OPEN) is activated, the first slope value (F_M ') being determined by averaging the partial slope values (F_M1, F_M2 ) is calculated.

Description

The invention relates to a method for driving a plurality of actuators and an apparatus for performing the method. Moreover, the invention relates to the use of such a device in a computerized system or subsystem that drives multiple actuators. For example, the invention also relates to process control systems that perform and monitor processes, in particular glass production processes or processing processes, wherein individual or multiple process values of the respective process can be changed by adjusting the actuators.

Methods and devices for influencing processes, in particular control devices, such. B. classical PID controller (two-point or three-point) are well known. The use of such classical controller is quite material and cost intensive especially in complex processes in which a larger number of actuators is needed. These problems, which the applicant in the field of glass production and processing, for. As in the process monitoring and control of melting units, had experienced, but also generally occur in other areas always where actuators, which have a motor, must be controlled in a larger variety.

The use of conventional PID controllers proves to be costly; besides, the usual regulation is very slow. Since the respective actuator must be controlled continuously, the wear on the actuators is considerable. For example, this wear on large variable transformers has been shown to damage the windings at the taps.

The usual actuators often consist of a driven electric motor with gear or an actuator. Such actuator units inevitably have disturbing up and down times and are generally subject to hysteresis. When changes in direction occur, due to the gear play, disturbing playtime. Because of these typical actuator characteristics, two- or three-position controllers can only be used heavily damped. Many actuators are therefore controlled manually by hand, checked hourly and possibly readjusted. This method is at a number of z. B. about 100 actuators per melting unit very laborious, time-consuming and inaccurate.

From the published patent application DE 43 16 759 A1 is a digital controller structure known, with the help of standard directional valves can be used with push magnets as proportional valves. By suitable dimensioning of the correction factors used in the control loop, the flow in the directional control valve can be kept constant over a wide pressure range, ie the arrangement corresponds to a proportional valve with integrated pressure compensator.

From the patent DD 244 624 A1 a method for controlling the temperature of industrial ovens is known in which several burners or burner groups are operated in an on-off circuit. For this purpose, a pulse controller in response to the control deviation between the actual and desired temperature provides a proportional frequency-modulated pulse train, which is processed in a preprogrammed ring counter. The switch-on commands issued by the ring counter are routed to the switching relays of the burner solenoid valves via a permanently set timer for the firing time. The programming of the ring counter, the selection of the time range for the burning time and the choice of the frequency range at the pulse controller determine the temporally staggered switching on and off of the burners in the context of the circulation control. The burner control is effective both in the period of heating and in the period of cooling, and is preferably used in connection with program timers on gas-fired industrial furnaces.

From the patent DE 10 2004 022 999 B3 A method for determining a control characteristic of a regeneration valve of a fuel vapor retention system for an internal combustion engine is known. In this case, the control characteristic describes the current relationship between a pulse-width-modulated control signal used to control the regeneration valve and an adjusting valve position. The currently valid minimum pulse width of the control signal, which is currently required for opening the regeneration valve, is determined by gradually increasing the pulse width by an amount of change until a deviation of the instantaneous behavior of the internal combustion engine from a stationary behavior of the internal combustion engine is detected. To shorten the calibration time and increase the availability of the calibration process, incrementing the pulse width begins at a given value of the pulse width greater than zero and less than a value corresponding to a minimum pulse width determined earlier.

In the published patent application DE 197 22 431 A1 discloses a method for controlling a lagged process with compensation, in particular for temperature control, in which, after a desired value step, the manipulated variable is set to a first constant value in a controlled operation, an IT1 model is identified on the basis of the step response and a PI Controller is parameterized, which regulates the process in a second stationary state. After that, a more exact process model with compensation is determined and used for controller adjustment in adaptive controllers.

In the published patent application DE 103 50 610 A1 a diagnostic device and a method for quasi-permanent and automatic monitoring of a control loop in terms of its functionality and control performance are described. By a first evaluation stochastic features, determined by another evaluation deterministic features of process variables and compared with associated predetermined reference values. An upstream selection device activates the more suitable of the two evaluation devices for diagnosis depending on the operating state. Preferably, the diagnostic device is implemented by a function block.

The cited prior art deals at most with the control of individual actuators. There is therefore a great need to develop a method and a device in order to be able to carry out a control of a plurality of actuators without the disadvantages mentioned above. In particular, a quick and accurate control of multiple actuators to be enabled without having to resort to the known cost and labor intensive solutions.

The object is achieved by a method having the features of claim 1 and by a device operating therewith having the features of claim 16.

Accordingly, a method for driving a plurality of actuators is proposed, comprising the following steps:
It is determined for the adjustment of each actuator a manipulated variable indicating a deviation of an actual value of a target value.
An individual adjustment time is calculated for each actuator, which indicates a pulse duration for the action of the control variable on the actuator, wherein the adjustment time is calculated as a function of a slope value which indicates with which speed the process value can be changed.
The respective actuator is controlled with the manipulated variable for the calculated individual adjustment time.

In the method according to the invention, the individual adjustment time is subjected to a game time value if a compensation of a game (thread) is indicated. This is the case when the manipulated variable has a change with a sign change, so if z. B. a change of direction in the actuator is present. The game time value indicates with which reaction time (delay due to game) the adjustment of the respective actuator takes place.

In addition, in the method, a learning procedure for calculating a first slope value is performed at least once before the first calculation of the adjustment time, wherein the learning procedure comprises a learning phase in which the following partial steps are performed:
the actuator is driven at least twice for a predetermined period of time to a first state, so that in a process, a first process value and a second process value can be achieved, and
at least two partial slope values are determined, which indicate the speed with which the respective process value is reached when the first state is triggered, wherein the first slope value is calculated by averaging the partial slope values.

Likewise, a device for driving a plurality of actuators is proposed, wherein the device is designed to carry out the method according to the invention and has at least one arithmetic unit. The arithmetic unit determines or calculates for the adjustment of each actuator, a manipulated variable indicating a deviation of an actual value of a target value. Furthermore, the arithmetic unit calculates for each actuator an individual adjustment time, which indicates a pulse duration for the action of the control variable on the actuator, wherein the adjustment time is dependent on a slope value indicating the speed with which the process variable (process value) is influenced. The arithmetic unit is connected to at least one driver unit, which controls the respective actuator for the calculated individual adjustment time with the manipulated variable.

The method and the accordingly working thereafter, the z. B. may be a computer system, allow that many actuators individually with a calculated pulse time in the direction of Manipulated variable (eg UP or DOWN) can be adjusted with a high speed and accuracy. The actuator properties required for the calculation (slope and possibly playing time) can be independently determined and stored on request by the at least one arithmetic unit, which is preferably designed as a digital function block.

These and other advantageous embodiments emerge from the subclaims:
In the method according to the invention, a learning procedure for calculating the gradient value and / or the game time value can be carried out at least once before the first calculation of the adjustment time, wherein the learning procedure comprises a learning phase in which the following partial steps are carried out:
the actuator is driven at least once for a predetermined period of time to a first state, so that in a process, a first process value is achieved, and
Thereafter, the actuator is driven at least once for the predetermined period of time to a second state, so that in the process, a further process value is reached; and
at least two partial slope values are determined, which indicate with which speed the respective process value is reached, wherein the slope value is calculated by averaging the partial slope values.

As a result, the respective partial gradients are calculated by at least one adjustment in one direction (toward the first state, eg UP) and by at least one adjustment in the other direction (preferably in the opposite direction), from which then the Total slope is averaged.

Preferably, a first partial slope value can be determined by the actuator being driven at least once in the direction of the first state, so that starting from an initial process value, a first process value is reached, wherein a first duration is determined, which is needed to the first process value to reach, wherein the quotient of the difference of the process values and the first duration indicates the first partial slope value.

In addition, a second partial gradient value can be determined by the actuator being controlled again in the direction of the first state, so that starting from the first process value, a second process value is reached, wherein a second duration is determined, which is required to the second process value reach, wherein the quotient of the difference of the process values and the second duration indicates the second partial slope value.

Furthermore, a third partial slope value can be determined by actuating the actuator in the direction of the second state so that, starting from the second process value, a third process value is reached, whereby a third duration is determined, which is required to reach the third process value , where the quotient of the difference of the process values and the third duration indicates the third partial gradient value.

A fourth partial gradient value can also be determined by the actuator being controlled again in the direction of the second state, so that starting from the third process value, a fourth process value is reached, whereby a fourth duration is determined, which is required to supply the fourth process value reach, wherein the quotient of the difference of the process values and the fourth duration indicates the fourth partial slope value.

The learning phase is aborted, for example, if the difference of the respective process values is smaller than a first permissible error value. This applies to every single step during the learning phase at which the process value should change. If the difference is too small, the process value has not changed appreciably (and will not reach the desired new value). Therefore, there is an assumption that there is an error in the actuator or elsewhere in the system. For example, the engine is defective or the drive jammed, etc.

The slope value is preferably calculated by averaging from the respectively determined partial slope values. In particular, the slope value (F_M) is calculated by averaging from the four determined partial slope values (F_M1, F_M2, F_M3, F_M4) as follows: F_M = (F_M1 + F_M2-F_M3-F_M4) / 4.

In addition, a learning procedure for calculating a second slope value can be performed at least once before the first calculation of the adjustment time, wherein the learning procedure comprises a learning phase in which the following partial steps are performed:
the actuator is driven at least twice for a predetermined period of time to a second state, which is different from the first state, so that in the process a third process value and a fourth process value are achieved, and
at least two partial slope values are determined, which indicate the speed at which the respective process value is reached when the second state is being controlled, the second gradient value being calculated by averaging the partial slope values.

Thereafter, preferably, the displacement time is calculated depending on the first slope value when the actuator is driven toward the first state, or is calculated depending on the second gradient value when the actuator is driven toward the second state; wherein the respective actuator for the calculated individual Verstellzeit is controlled with the manipulated variable.

The method may be aborted if the slope value falls below a predetermined minimum slope value or exceeds a predetermined maximum slope value.

In the method according to the invention, the learning procedure may comprise a preparation phase in which the following sub-steps are carried out in order to set the actuator to an initial state:
The actuator is driven in the direction of a state for a predetermined return or adjustment time, so that a process value, in particular a lowered process value, is reached, which is below the current operating point;
It is waited for a predetermined fixed break time;
After expiration of this fixed pause time, the actuator is driven for a predetermined adjustment time in the direction of another state, so that a process value, in particular the initial process value is reached, which is within the predetermined value range.

In this case, within the learning phase, the game time value can be calculated from a quotient of a process value deviation and the adjustment time, wherein the process value deviation results from the difference between a first process value and a process value deviating therefrom.

The process value deviation can be determined by the actuator being driven from the first process value for the predeterminable period of time to the first state, so that a certain process value is reached, and then from this specific process value for the same period of time to the second state is controlled so that the first process value would have to be reached again, but the process has the deviating process value.

The process may be aborted if the game time value exceeds a predetermined maximum game time value.

The invention and the resulting advantages are described below with reference to an embodiment with reference to the accompanying drawings. The drawings show the following schematic representations:

1 shows in the form of function blocks the structure of a system, here z. B. a process control system, with a device according to the invention for driving a plurality of actuators;

2 shows the time course with the changes caused by a process value (eg current of a heater in a smelting unit);

3 shows a rough flow chart for the inventive method;

4 shows the process in greater detail with its individual process steps; and

5 shows the flow chart for a previously (one-time or on demand) learning procedure to determine the actuator characteristics for the purpose of optimized control.

The 1 schematically shows the structure of a process control system with a device according to the invention BVB, which drives a plurality of actuators S1, S2, ... Sn. The actuators can be various actuators, actuators, potentiometers, valves and the like. Each actuator is individually controlled and adjusted to effect the change to a process value. The actuator S1 is z. Example, a motor-driven and provided with a gearbox potentiometer, which represents the heating current of an electric heating element (electrode) at a melting tank. The process value to be set is then the size of the heating current measured in amperes.

The device BVB represents a computer-aided function block, which calculates the optimum adjustment time (ie here VZ1) individually for each actuator (eg for S1) on the basis of the method which will be described later. The input data used are, in particular, the desired value SET, the actual value IST and data of the parameterization PAR for the actuator to be set or for the process value to be set.

The deviation of the actual value IST from the nominal value results in the manipulated variable (control deviation), here z. B. STG1 based on the first actuator S1. The function block BVB then calculates an adjustment time VZ1, d. H. the pulse duration for which the control value is applied to the actuator S1. It is therefore a pulse-shaped adjustment. In this case, as will be described in detail later, a slope value F_M is taken into account, which is a measure of the speed at which the actuator reaches the desired change of the process value. In addition, if necessary, a game time value F_CLR is taken into account and used to possibly compensate for the play possibly present in the actuator, in particular in the case of a rotation direction or load change in the transmission.

The presented here intelligent control of the actuators is described by the below also with reference to 2 - 5 described function (calculated adjustment) or the underlying method described.

Preferably, the method is implemented by a digital, programmed function that runs on a real-time capable or almost real-time capable computer system. Using analog and digital inputs, the computer system determines directly, or via intelligent subsystems, the associated measured values and status variables of the control loops from the entire process. They are digitized and stored. Digital outputs (s. 1 ) activate the actuators, which then according to the deviation from the setpoint milliseconds (ms) exactly in the direction of the desired state such. B. "ON" or "AB" are controlled. The function block can be instantiated as often as desired, ie the number of control loops per computer system is only limited by the performance of the hardware used. Each actuator has a parameter set of many (approximately 88) data.

• – Alle Datenfelder des digitalen Funktionsblocks sind so angelegt, dass sie in Verbindung mit einem Prozess-Leitsystem angezeigt und somit auch bedient oder ggf. verändert werden können.
• – Ein automatisch ablaufender Lernalgorithmus (wird manuell gestartet) ermittelt und speichert die Steigung und die Spielzeit, (die bei Drehrichtungswechseln anfällt) individuell für jedes Stellglied. Die Steigung gibt an, um wieviel sich die Regelgröße bei einem Stellimpuls von 1000 ms ändert. Die Stellgliedeigenschaften wie Hoch- und Nachlaufzeit und die Hysterese werden automatisch berücksichtigt.
• – Der Regelalgorithmus benötigt keine Rückmeldesignale von den Stellgliedern.
• – Aus der Differenz zwischen Soll- und Istwert wird mit der Steigung und unter Berücksichtigung der letzten Verstellrichtung ggf. mit der Spielzeit eine Verstellzeit berechnet. Entsprechend dem Vorzeichen wird der Digitalausgang des Stellgliedes für AUF oder AB ms genau angesteuert.
• – Ein Gut-Toleranzband verhindert weitere unnötige Verstellungen.
• – Auch nichtlineare Steigungskennlinien bearbeitet das System erfolgreich, da im Arbeitspunkt ein annähernd linearer Zusammenhang besteht.
• – Sehr kurze Einstellzeit (Verstellzeit plus Einstell-Wartezeit) nach einer Sollwertänderung.
• – Geringer Verschleiß an den Stellgliedern und den ggf. nachgeschalteten Einrichtungen.
• – Umfangreiche Plausibilitätsprüfungen verhindern fehlerhafte Verstellungen, ggf. werden Alarmflags gesetzt.
• – Nach einer Verstellung wartet die Funktion eine definierbare Zeit (Totzeit der Regelstrecke und zur Glättung des Messwertes), damit der Istwert sich auf den neuen Endwert einschwingen kann. Diese Wartezeit ist sehr vorteilhaft wenn mehrere Regelkreise eines Systems sich gegenseitig beeinflussen.
• – Jeder Verstellschritt wird gezählt und erst auf null gesetzt wenn der Messwert innerhalb des Gut-Toleranzbandes liegt.
• – Wird nach einer parametrierbaren Anzahl von Verstellschritten das Gut-Toleranzband nicht erreicht werden die Verstellzeiten der nachfolgenden Schritte um x Prozent reduziert. Dieser X%-Wert kann je Stellglied bestimmt werden.
• – Kommt ein Istwert nach n Verstellschritten nicht in das Gut-Toleranzband wird ein Alarmflag gesetzt.
• – Die BV-Funktion erreicht in der Regel (in ca. einer Minute) nach n Verstellschritten das Gut-Toleranzband. Danach finden über einen längeren Zeitraum keine Aktionen mehr statt, es sei denn der Sollwert ändert sich.
• – Die realisierte Stellfunktion ist sehr kostengünstig, pro Stellglied sind nur 2 Relais erforderlich.
• – Die Inbetriebnahme einer Regelstrecke ist mittels der Lernfunktion innerhalb weniger Minuten getan.
• – Durch das sehr schnelle und präzise Erreichen der Sollwerte haben die Volumenströme für Brenner (Öl, Gas, Luft, Sauerstoff) stets das optimale stöchiometrische Verhältnis. Die Verbrennung ist somit Energieoptimal und die Abgase enthalten die niedrigsten Schadstoffmengen.

The device BVB operates according to a special digital control algorithm with numerous advantages, in particular with the following advantages:

• - All data fields of the digital function block are designed so that they can be displayed in conjunction with a process control system and thus also operated or changed if necessary.
• - An automatic learning algorithm (started manually) determines and stores the slope and the playing time (which occurs when the direction of rotation changes) individually for each actuator. The slope indicates how much the controlled variable changes with an actuating pulse of 1000 ms. The actuator properties such as acceleration and retardation time and hysteresis are automatically taken into account.
• - The control algorithm requires no feedback signals from the actuators.
• - From the difference between the setpoint and actual value is calculated with the slope and taking into account the last adjustment, possibly with the playing time an adjustment. According to the sign, the digital output of the actuator for UP or DOWN ms is precisely controlled.
• - A good tolerance band prevents further unnecessary adjustments.
• - Nonlinear slope characteristics are also processed successfully by the system, since there is an approximately linear relationship at the operating point.
• - Very short setting time (adjusting time plus setting waiting time) after a setpoint change.
• - Low wear on the actuators and any downstream equipment.
• - Extensive plausibility checks prevent faulty adjustments, if necessary, alarm flags are set.
• - After an adjustment, the function waits for a definable time (dead time of the controlled system and for smoothing the measured value) so that the actual value can settle to the new final value. This waiting time is very advantageous if several control loops of a system influence each other.
• - Each adjustment step is counted and only set to zero if the measured value is within the good tolerance band.
• - If the material tolerance band is not reached after a configurable number of adjustment steps, the adjustment times of the following steps are reduced by x percent. This X% value can be determined per actuator.
• - If an actual value does not enter the good tolerance band after n adjustment steps, an alarm flag is set.
• - The BV function usually reaches (within approx. One minute) the material tolerance band after n adjustment steps. Thereafter, no actions take place over a longer period, unless the target value changes.
• - The realized control function is very cost-effective, only 2 relays per actuator are required.
• - The commissioning of a controlled system is done within a few minutes using the learning function.
• - Due to the very fast and precise reaching of the setpoints, the volume flows for burners (oil, gas, air, oxygen) always have the optimal stoichiometric ratio. The combustion is thus energy optimal and the exhaust gases contain the lowest amounts of pollutants.

• – Die Funktion der Vorrichtung BVB ermittelt mit dem internen Lernalgorithmus innerhalb weniger Minuten die exakten Stellgliedeigenschaften der Regelgröße. Danach kann sofort die Prozessgröße geregelt werden.
• – Bei der Berechnung des Stellimpulses werden die Zeiten für Hoch-, u. Nachlauf, Spiel und Hysterese berücksichtigt. Durch den ms genauen Stellimpuls wird meistens mit dem 1. Verstellschritt das Gut-Toleranzband erreicht.
• – Stellglieder die bereits Verschleiß aufweisen (eine deutlich erhöhte Spielzeit) können mit der BV-Funktion problemlos weiter genutzt werden, da die Spielzeit bekannt ist und berücksichtigt wird.
• – Die erheblich geringere Anzahl von Verstellschritten reduziert drastisch den Verschleiß in der Anlage.
• – Ein Dreipunktregler kann das mechanische Spiel eines Stellgliedantriebes nicht selbständig erfassen und berücksichtigt bei der Berechnung des Stellimpulses keine Spielzeit.
• – Ein Dreipunktregler ermittelt nicht die Steigung, die erzeugten Stellimpulse des Dreipunktreglers verringern nur stufenweise (gemächlich) die Regelabweichung.

The invention may, for. B. in smelting and metallurgical plants, power and heating plants and in manufacturing facilities with a larger number of electrically driven actuators are used. It is to be preferred to the well-known PID controllers, in particular three-point controllers, because it offers, among other things, the following advantages:

• - The function of the device BVB determined with the internal learning algorithm within a few minutes, the exact actuator properties of the controlled variable. Thereafter, the process size can be regulated immediately.
• - When calculating the actuating pulse, the times for high, u. Caster, play and hysteresis are considered. Due to the ms precise setting pulse, the good tolerance band is usually reached with the 1st adjustment step.
• - Actuators that already have wear (a significantly increased playing time) can be easily used with the BV function, as the playing time is known and taken into account.
• - The significantly smaller number of adjustment steps drastically reduces wear in the system.
• - A three-position controller can not detect the mechanical play of an actuator drive independently and takes into account in the calculation of the control pulse no play time.
• - A three-position controller does not determine the slope, the generated control pulses of the three-position controller reduce the control deviation only gradually (slowly).

The basic structure of the process according to the invention is in 3 shown schematically. The procedure 100 contains three main areas (steps or blocks 110 . 120 and 130 ), where in a first block 110 the manipulated variable STGi is determined, for. B. the manipulated variable or control deviation STG1 for the first actuator S1. For this purpose, the actual value IST is compared with the setpoint nominal (see also 1 ).

In a block 120 the individual adjustment time VZi for the respective actuator (here STG1 for S1) is calculated. After that, then in step 130 actuates the actuator accordingly.

Based on 4 becomes the procedure 100 now described in more detail:
After the start of the device BVB, in which a loop counter is reset to zero (i = 0) is followed by a first incrementing of the loop counter i = i + 1 in step 111 , This step 111 is also approached from the return point A, which is part of an iterative loop, so that the loop counter may already have a value greater than zero and then in step 111 increased by one. In step 112 it is then checked whether the loop counter i has reached a final value (maximum number of all actuators). Are z. B. 100 actuators available and would be i = 101 then the procedure would abort (YES branch based on query 111 ). If not, then it goes to the step 113 , which checks if with a learning procedure 101 (see also 3 ) should be started or not. This condition is z. B. determined by a manual operation request. If YES, then the learning procedure starts 101 , later still based on the 5 and the values F_M and F_CLR for the calculation of the individual adjustment time VZi (see step 120 ).

If the learning function fails (step 105 ), the return point A is approached, from which then (i = i + 1) the next actuator is processed. Otherwise (NO) it goes to the step 118 which is above the main strand 112 . 113 . 114 and 115 is approached. This means that starting from the step 113 , the step 114 follows if no learning procedure is to be performed; and then in the step 115 It is checked whether the control element of the actuator (eg the control relay) is already switched on, ie whether an adjustment is active. If NO, then the step follows 118 , which checks whether the predetermined waiting time has expired, ie whether so much time has expired that the respective process value (measured value) is stable or stabilized. If YES, then follow the steps 119a and 119b , If NO (in step 118 ), then it goes back to the return point A.

Starting from the step 115 leads a branch to the step 116 when the relay is switched on (YES). In this case follows in the step 116 the verification of the adjustment time; This means that it is checked for each cycle (here: every 50 ms) whether the calculated adjustment time (eg 970 ms) has been reached. If YES, then the relay is opened and a transient waiting time is set (also the last direction of rotation, in this case OPEN, is stored). Only then it goes back to the return point A. If NO, then it goes directly back to the return point A.

In step 119a the manipulated variable STGi (or control deviation) is calculated from the difference between the actual value and the setpoint. Then in step 119b Checked whether the control deviation or manipulated variable STGi (see also 3 ) is large enough so that a recalculation of the adjustment time VZi is indicated. If STGi is not smaller than a tolerance value or band TB, then the sequence of steps is 120 approached to calculate VZi. Otherwise (YES) the control deviation is very small and the procedure goes back to the return point A.

The sequence of steps for calculating VZi (see also 3 ) will now be based on the 4 explained in more detail:
In the block 120 will be in a first step 121 the adjustment time VZ1 is calculated by forming the quotient of the gradient value F_M and the manipulated variable STG1 (the gradient value F_M is determined on the basis of the learning procedure described later, see 5 ). Then in the sub-step 122 checked whether there is a change of direction and therefore a play of the actuator must be considered. If this is not the case (NO), then the calculated adjustment time is the current adjustment time, ie z. For example, VZ1 = VZi, for i = 1. However, if a game is to be considered (YES), then it will be in a sub-step 123 , the adjustment time is applied to the time value F_CLR: VZ1 = VZi + F_CLR, for i = 1 (The time value F_CLR is also determined by the learning procedure described later, see 5 ).

With each in step 120 calculated adjustment time (here VZ1 for the actuator S1) can then in step 130 the actuator can be controlled individually.

The 5 illustrates the learning procedure 101 (see also 4 ), which serves at the beginning and when needed to check the actuator properties and to calculate corresponding values, namely the slope value F_M and possibly the time value F_CLR. The learning procedure can be carried out during the ongoing process and is based on the 2 described in more detail below:
As part of a preparatory phase I (see also 2 as well as step 102 in 5 ), the process value PW is significantly different from the current operating point (outside the working area) changed, z. For example, the process value is first significantly lowered (within a predetermined period of time) starting from the current value (between L_PV3 and L_PV1). Then a certain pause time is waited to allow the actuator (here electric motor or actuator) to come to rest before the opposite direction of rotation (OPEN) is activated (time 3 in the 2 ). Subsequently, the process value is raised to an initial process value F_PV0, again awaiting a time for stabilization (time 5). By the previous lowering so an actuator or geared actuator is approached without play "from below"; the subsequent adjustments "up" (from time 5 to time 8) so done so that the transmission can not come into the state of play. The preparation phase serves to stay as symmetrical as possible around the working point during the subsequent learning phase.

As part of the subsequent learning phase II (see also step 103 in 5 ), the actuator is then actuated and moved at least once in the direction of UP and once in the direction AB in order to determine several partial gradients after passing through this learning phase, from which an averaged slope value F_M and a time value F_CLR can be calculated, which will be described in more detail below ,

The learning procedure adjusts the control function individually for each actuator to the current operating point and the conditions of the controlled system. The learning phase is carried out on the basis of successive adjustments and waiting times. The following parameters must be specified for each actuator:
A_OTADJ [0 ... 30 s] = adjustment time
A_OTCLR [0 ... 30 s] = playing time
A_OTPAU [0 ... 100 s] = pause time
A_OCMIN [0 ... 1000] = error criteria - minimum change. Value by which the process variable must at least change if an impulse of length A_OTADJ was output during the learning phase. If the process value does not change or changes only slightly, the learning phase is aborted.
A_OMMIN [0 ... 200] = error criteria - minimum slope, which may be determined during the learning phase. If the determined slope is lower, the actuator is considered not optimized and the learning phase is stopped.
A_OMMAX [0 ... 200] = error criteria - maximum slope, which may be determined during the learning phase. If the determined slope is greater, the actuator is considered not optimized and the learning phase is stopped.
A_OCMAX [0 ... 30s] = Error Criteria - Maximum playing time allowed during the learning phase. If the determined game time is greater, the actuator is considered not optimized and the learning phase is stopped.

Based on 2 the individual processes are described in more detail at the times 1 to 14 given there:
At time 1: The actuator or actuator is closed for the period A_OTADJ + A_OTCLR.
At time 2: It follows a pause, the fixed z. B. is programmed to 1 second.
At time 3: The actuator is raised for the time A_OTCLR. This impulse should eliminate any existing drive clearance due to the change of direction.
At time 4: waiting time of length A_OTPAU. At the end of the waiting period, the current process value is stamped in F_PVO.
At time 5: The actuator is raised for the time A_OTADJ. The actual duration of this output pulse is recorded in F_T1.
At time 6: waiting time of length A_OTPAU. At the end of the waiting time, the current process value is stamped in F_PV1. If the difference F_PV1-F_PVO is less than the learning phase error criteria or the minimum change (A_OCMIN), the learning phase is aborted prematurely and the error flag F_OECHG is set.
At time 7: The actuator or actuator is ramped up for the time A_OTADJ. The actual duration of this output pulse is recorded in F_T2.
At time 8: waiting time of length A_OTPAU. At the end of the waiting period, the current process value is stamped in F_PV2. If the difference F_PV2-F_PV1 is less than the learning phase error criteria or the minimum change (A_OCMIN), the learning phase is aborted prematurely and the error flag F_OECHG is set.
At time 9: The actuator is closed for the time A_OTADJ. The actual duration of this output pulse is recorded in F_T3.
At time 10: waiting time of length A_OTPAU. At the end of the waiting time, the current process value is stamped in F_PV3. If the difference F_PV3 - F_PV2 is less than the learning phase error criteria or the - minimum change (A_OCMIN), the learning phase is aborted prematurely and the error flag F_OECHG is set.
At time 11: The actuator is closed for the time A_OTADJ. The actual duration of this output pulse is recorded in F_T4.
At time 12: waiting time of length A_OTPAU. At the end of the waiting time, the current process value is stamped in F_PV4. If the difference F_PV4 ~ F_PV3 is less than the learning phase error criteria or the minimum change (A_OCMIN), the learning phase is aborted prematurely and the error flag F_OECHG is set.
At time 13: In this time, the slope and the game time are determined from the previously determined data. This only takes one cycle and is therefore practically invisible.
At time 14: If no errors occurred in the previous learning phase steps (times), the counter is set to 14. This number is the indicator for a valid or completed learning phase.

The partial slopes F_M1 ... F_M4 are calculated as follows:
F_M1 = (F_PV1 - F_PVO) / F_T1
F_M2 = (F_PV2 - F_PV1) / F_T2
F_M3 = (F_PV3 - F_PV2) / F_T3
F_M4 = (F_PV4 - F_PV3) / F_T4

From the part gradients, the total slope, d. H. the slope value for the calculation of the adjustment time, averaged: F_M = (F_M1 + F_M2 - F_M3 - F_M4) / 4

From the difference of the process value at the end of step 6 and step 10, the play time of the drive is determined: F_CLR = Δ / F_M = (F_PV3 - F_PV1) / F_M

The slope value for the calculation of the adjustment time can also be calculated separately for the direction UP or DOWN as follows (see also 2 ):
For this purpose, the actuator is actuated at least twice in the direction OPEN, so that the partial slopes F_M1 and F_M2 can be calculated. The averaging results in the slope value for the direction OPEN: F_M '= (F_M1 + F_M2) / 2

In addition, the actuator is controlled at least twice in the direction AB, so that the partial slopes F_M3 and F_M4 can be calculated. From the averaging, the slope value for the direction AB results: F_M '' = - (F_M3 + F_M4) / 2

If the actuator is to be adjusted in the direction of OPEN, the first gradient value F_M 'is used. If the actuator is to be adjusted in the direction of AB, the second slope value F_M '' is used. Thus, differentiation is made between the driving directions.

• – der Wichtungsfaktor > Null ist
• – die letzte und die neue Verstellrichtung gleich sind (kein Drehrichtungswechsel)
• – die Verstellzeit > n% von der GesamtVerstellzeit (von AB bis AUF) ist

The device BVB may optionally include a module providing adaptive functionality. The slope value F_M is thus adapted independently to current process conditions. This module will only run through if an actuator is considered to have:

• - the weighting factor> zero
• - the last and the new adjustment direction are the same (no change of direction)
• - the adjustment time is> n% of the total operating time (from DOWN to OPEN)

First we calculate: m = ABS / VZ
where ABS is the absolute value of the difference "measured value before adjustment - measured value after adjustment", ie an unsigned value.

Then the new slope value is calculated: F_M = [F_M old * (WF-1) + m] / WF where WF is the weighting factor.

The invention proposed here can also achieve significant improvements with respect to the accuracy of an actuator control. This is illustrated by the following accuracy considerations for a Windows operating system with 50 ms time slices:
Most actuators set with this procedure have a total run time of 0% (AB) to 100% (UP) of 90 seconds.

For a cycle time of 50 ms, 25 ms rounding sets each adjustment time with a maximum error of 25 ms, so the mean error is 12.5 ms.

The maximum setting error = (25 · 100) / 90000 - = 0.027777% of the setting or measuring range.

The mean adjustment error = (12.5 · 100) / 90000 - = 0.013888% of the setting or measuring range.

The good tolerance band for an actuator is set to +/- 0.05 or 0.1% of the measuring range.

The maximum remaining control deviation in this constellation is therefore 0.1% of the measuring range.

The time required to eliminate a control deviation is essentially determined by the adjustment time (F_VZ) of the actuator and by the waiting time for damping and smoothing the measured value acquisition. The calculation step for further adjustment requires only 50 ms under Windows.

The 1st adjustment step usually eliminates 97% of the control deviation.

The analog-to-digital converters for the measured value acquisition should work with a resolution of 12 to 16 bits under these conditions.

With real time operating systems that provide time slices or time slots of 1 or 5 ms, better results can be achieved by a factor of 10 to 50.

In the following, the calculation processes are illustrated by means of tabular flow charts using the example of an actuator which influences the current [A] of a process variable (eg heating current for an electrode) via a motor-driven potentiometer. In this example, the schedule ( 4 ) in several loops: Parameters of the operating system Duration of a time slice (assigned resource timeslot): 50 [ms] Parameters and data of this actuator: Gradient (F_M): 0.002002 [A / ms] Playing time (F_CLR): 270 [ms] Actuator time: 90 [s] Transient waiting time (eg for PAR in FIG. 1): 2000 [ms] Tolerance band (eg for PAR in FIG. 1): ± 0.12 [A] Last direction of rotation (eg for PAR in Fig. 1): FROM Addr. Digital output AB: 47: 1 Addr. Digital output OPEN: 47: 2 old target value: 100 [A] new setpoint: 120 [A] Measuring range actual value: 0-180 [A] Current actual value: 100.02 [A] Block No. (see Fig. 4) Result Note / calculation 111 I = 1 i = i + 1 112 N ? The End 113 N ? Learn 114 N ? any errors exist 115 N ? Is still an adjustment 118 J ? Transient waiting time expired 119A 19.98 Control value [A] = new setpoint - actual value (result is positive -> relay OPEN must be activated (change of PW direction OPEN) 119B N ? Control value <tolerance band 121 9980.01998 VZ [ms] = control value / slope 122 J ? Change of direction From (old) OPEN (new) 123 10,250.01998 VZ = VZ + play time (F_CLR) 130 47: 2 = 1 Switch on the relay for OPEN (address 47: 2) (= 1) → A The relay of this actuator must remain closed 205 cycles a '50 ms = 10250 ms to reach the positioning time VZ. Process in the 204th cycle (here: penultimate cycle) with i = 1 111 I = 1 i = i + 1 112 N ? The End 113 N ? Learn 114 N ? any errors exist 115 J ? Is still an adjustment 116 N ? (Resetting time + 25 ms) <50 ms (time slice) (50,02 + 25) <50 ms → A Process in the 205th cycle with i = 1 111 I = 1 i = i + 1 112 N ? The End 113 N ? Learn 114 N ? any errors exist 115 J ? Is still an adjustment 116 J ? (Resetting time + 25 ms) <50 ms (time slice) (0.02 + 25) <50 ms 117 47: 2 = 0 Switch off the relay for OPEN and set the last direction of rotation (value = OPEN); Set residual settling waiting time to 2000 ms = 40 cycles a '50 ms → A After another 39 cycles (remaining settling waiting time) with i = 1 111 I = 1 i = i + 1 112 N ? The End 113 N ? Learn 114 N ? any errors exist 115 N ? Is still an adjustment 118 N ? (Residual settling delay + 25 ms) <50 ms (75.0 + 25) <50 ms → A Process in the 40th cycle with i = 1 111 I = 1 i = i + 1 112 N ? The End 113 N ? Learn 114 N ? any errors exist 115 N ? Is still an adjustment 118 J ? (Residual settling delay + 25 ms) <50 ms (0.0 + 25) <50 ms 119A 0.06 Control value [A] = setpoint - actual value (120,0 - 119,94) 119B J ? Control value <tolerance band (0.06 <0.12) → A

Claims (10)

Procedure ( 100 ) for controlling a plurality of actuators (S1, S2, ... Sn) for changing process values (PW) in a process comprising the following steps: - a manipulated variable (STG1) is determined for the adjustment of each actuator (S1) indicates a deviation of an actual value (IST) from a nominal value (SOLL); - For each actuator (S1) an individual adjustment time (VZ1) is calculated, which indicates a pulse duration for the action of the manipulated variable (STG1) on the actuator (S1) (step 120 ), wherein the adjustment time (VZ1) is calculated as a function of a slope value (F_M) which indicates at what speed the process value (PW) changed by means of the displaced actuator (S1) changes; and - the respective actuator (S1) is controlled for the calculated individual displacement time (VZ1) with the manipulated variable (STG1) (step 130 ), wherein the individual adjustment time (VZ1) is applied with a game time value (F_CLR) (step 123 ), which indicates with which reaction time the adjustment of the respective actuator (S1) takes place when the manipulated variable (STG1) has a change with a sign change, and wherein at least once before the first calculation of the adjustment time (step 120 ) a learning procedure ( 101 ) for calculating a first slope value (F_M '), wherein the learning procedure ( 101 ) a learning phase ( 103 ), in which the following sub-steps are carried out: the actuator (S1) is actuated at least twice for a predeterminable period of time (A_OTADJ) to a first state (AUF), so that in a process a first process value (F_PV1) and a second process value (F_PV2) are achieved, and at least two partial slope values (F_M1, F_M2) are determined which indicate the speed with which the respective process value is reached when the first state (AUF) is activated, the first slope value (F_M ' ) is calculated by averaging the partial slope values (F_M1, F_M2). Procedure ( 100 ) according to claim 1, wherein at least once before the first calculation of the adjustment time a learning procedure ( 101 ) for calculating a second slope value (F_M ''), the learning procedure ( 101 ) a learning phase ( 103 ), in which the following sub-steps are performed: the actuator (S1) is driven at least twice for a predeterminable period of time (A_OTADJ) to a second state (AB), which is different from the first state, so that in the process third process value (F_PV3) and a fourth process value (F_PV4) are achieved, and at least two partial slope values (F_M3, F_M4) are determined, which indicate the speed with which the respective process value is reached, if the second state (AB) is activated where the second slope value (F_M '') is calculated by averaging the partial slope values (F_M3, F_M4). Procedure ( 100 ) according to claim 1 and 2, wherein the adjustment time (VZ1) is calculated as a function of the first slope value (F_M ') when the actuator is driven towards the first state (AUF) or depending on the second slope value (F_M''). is calculated when the actuator is driven to the second state (AB) out; and wherein the respective actuator (S1) for the calculated individual adjustment time (VZ1) with the manipulated variable (STG1) is controlled. Procedure ( 100 ) according to one of the preceding claims, wherein the learning phase ( 103 ) is aborted if the slope value (F_M) falls below a predetermined minimum slope value (A_OMMIN) or exceeds a predetermined maximum slope value (A_OMMAX). Procedure ( 100 ) according to one of claims 1 to 4, wherein the learning procedure ( 101 ) a preparatory phase ( 102 ), in which the following sub-steps are performed to set the actuator (S1) in an initial state: - the actuator is for a predetermined period t (A_OTADJ + A_OTCLR) in the direction of a state (AB) is driven, so that a Process value, in particular a lowered process value is reached, which is outside a given work area; - Waiting for a given fixed break time; - After expiration of this fixed pause time, the actuator for a predetermined adjustment time (A_OTADJ) is driven in the direction of another state (AUF), so that a process value, in particular the initial process value (F_PV0), is reached, which is within the specified working range. Procedure ( 100 ) according to one of claims 1 to 5, wherein within the learning phase ( 103 ) the game time value (F_CLR) is calculated from a quotient of a process value deviation (Δ) and the adjustment time (F_M), the process value deviation (Δ) being the difference between a first process value (F_PV1) and a different process value ( F_PV3). Procedure ( 100 ) according to claim 6, wherein the process value deviation (Δ) is determined by the actuator (S1) starting from the first process value (F_PV1) for the predeterminable time period (A_OTADJ) is driven to the first state (AUF), so that a certain process value (F_PV2) is reached, and then based on this specific process value (F_PV2) for the same period (A_OTADJ) is driven to the second state (AB), so that the first process value (F_PV1) would have to be reached again Process but the deviating process value (F_PV3) has. Procedure ( 100 ) according to claim 6 or 7, wherein the learning phase ( 103 ) is aborted if the game time value (F_CLR) exceeds a predetermined maximum game time value (A_OCMAX). Device (BVB) for driving a plurality of actuators (S1, S2, ... Sn), wherein the device is designed to carry out the method according to claim 1 and has at least one arithmetic unit, which for the adjustment of each actuator (S1) is a manipulated variable ( STG1) which indicates a deviation of an actual value (IST) from a nominal value (SOLL), wherein the at least one arithmetic unit calculates an individual adjustment time (VZ1) for each actuator (S1) which has a pulse duration for the action of the manipulated variable (STG1). indicates the actuator (S1), wherein the arithmetic unit calculates the adjustment time (VZ1) as a function of a slope value (F_M) which indicates the speed at which the adjustment of the actuator (S1) takes place; and wherein the arithmetic unit is connected to at least one driver unit, which controls the respective actuator (S1) for the calculated individual adjustment time (VZ1) with the manipulated variable (STG1), wherein the arithmetic unit acts on the individual adjustment time (VZ1) with a play time value (F_CLR) indicating with which reaction time the adjustment of the respective actuator (S1) occurs when the manipulated variable (STG1) has a change with a sign change, and wherein the arithmetic unit at least once before the first calculation of the Verstellzeit a learning procedure ( 101 ) for calculating a first slope value (F_M '), wherein the learning procedure ( 101 ) a learning phase ( 103 ) in which the device carries out the following substeps: the driver unit actuates the actuator (S1) at least twice for a predefinable time duration (A_OTADJ) to a first state (AUF), so that in a process a first process value (F_PV1) and a second process value (F_PV2) are achieved, and the arithmetic unit determines at least two partial slope values (F_M1, F_M2) which indicate at which speed the respective process value is reached when the first state (AUF) is activated, the arithmetic unit the first slope value (F_M ') is calculated by averaging the partial slope values (F_M1, F_M2). Use of the device (BVB) according to claim 9 for controlling a plurality of actuators in a process control system, which performs and monitors a process, in particular a glass production process or processing, wherein by adjusting the actuators one or more process values of the process can be changed.

Images