CCONTROLLER AND METHOD FOR ADAPTIVE TWO-POSITION CONTROL

Abstract

The invention relates to a method for the two-position control of an actuator (1) on the basis of a binary sensor signal (y) of a sensor unit (2), which senses a process variable (P), which can be influenced by the actuator (1), in such a way that the sensor unit outputs a first sensor signal value (y1) when a first switching value (Sw1) is exceeded and a second sensor signal value (y0) when a second switching value (Sw1, Sw2) is fallen below, wherein: the actuator (1) is controlled with a manipulated variable (u), which assumes either a first control value (u1) or a second control value (u2); the first control value (u1) and the second control value (u2) are dynamically adapted during the operation of the actuator (1), in dependence on a fall time (t_fall) corresponding to the duration of the first sensor signal value (y1) and a rise time (t_rise) corresponding to the duration of the second sensor signal value (y0), in such a way that the first and second control values converge. The invention further relates to a two-position controller (10) designed to carry out the method and to an actuator (1) comprising said two-position controller (10).

Claims

1. A method for two-point control of an actuator, the method comprising the steps of: generating a binary sensor signal a sensor that detects a process variable of the actuator in such a way that when a first switching value is exceeded, a first sensor signal value is outputted and when falling below a second switching value a second sensor signal value is outputted, controlling the actuator by a manipulated variable having either a first control value or a second control value, dynamically adjusting the first control value and the second control value in dependence on a fall time corresponding to a duration of the first sensor signal value and a rise time corresponding to a duration of the second sensor signal value such that during operation of the actuator the rise time and fall time converge.

2. The method according to claim 1, wherein the first control value and the second control value are adjusted dynamically so that the fall time and the rise time converge to the same value.

3. The method according to claim 1, wherein the method is repeated and includes the following steps: a first step in which the actuator with the first control value that is the difference between a manipulated variable mean and a corresponds to manipulated variable amplitude, and the fall time is determined when the sensor outputs the first sensor signal value, a second step in which the actuator with the second control value that is the sum of the mean value of the manipulated variable and the manipulated variable amplitude, and the rise time is determined when the sensor outputs the second sensor signal value, and a third step in which a new manipulated variable mean value from the fall time and the rise time is determined.

4. The method according to claim 1, wherein the method is repeated and includes the following steps: a first step in which the actuator is controlled by the first control value that is the difference between a manipulated variable mean value and corresponds to a manipulated variable amplitude, and the rise time is determined when the sensor outputs the second sensor signal value, a second step in which the actuator with the second manipulated variable that is the sum of the manipulated variable mean and the corresponds to the manipulated variable amplitude, and for which the fall time is determined if the sensor outputs the first sensor signal value, and a third step in which a new manipulated variable mean value from the fall time and the rise time is determined.

5. The method according to claim 3 wherein in a fourth step the manipulated variable amplitude is reduced.

6. The method according to claim 5, wherein the fourth method step is carried out repeatedly until a minimum manipulated variable amplitude other than zero is reached.

7. The method according to claim 1, wherein the fall time and the rise time are compared with each other such that, the new mean value of the manipulated variable is increased when the rise time is greater than the fall time, or the new manipulated variable mean value is reduced when the rise time is smaller than the fall time.

8. The method according to claim 7, wherein the new manipulated variable mean is determined according to
u_mean=u_mean+k*delta where k is a correction factor that depends on the fall time and the rise time depends preferably on a ratio of fall time and Rise time.

9. The method according to claim 8, wherein a correction factor is
k=1−t_fall/t_rise when the rise time is greater than the fall time, and the correction factor is
k=t_rise/t_fall−1 when the rise time is less than the fall time.

10. The method according to claim 8, wherein the correction factor is
k=1−sqrt when the rise time is greater than the fall time, and the correction factor is
k=sqrt−1 when the rise time is less than the fall time.

11. The method according to claim 1, further comprising: a start step is carried out in which the actuator is controlled by a manipulated variable, which leads to the first switching value being exceeded or the second switching value is not reached.

12. The method according to claim 1, further comprising: a test step, in which the fall time and/or the rise time is compared with an average value from both times and, if there is a difference, a reset is carried out, in particular to reset the manipulated variable mean and the manipulated variable amplitude to their initial values.

13. The method according to claim 1, wherein the first switching value and the second switching value have the same value.

14. The method according to claim 1, wherein the actuator is a pump, valve, heater, or cooler.

15. A two-point controller for controlling an actuator based on a binary sensor signal having an input for the binary sensor signal and an output for a manipulated variable to control an actuator, the two-position controller being set up to carry out the method according to claim 1.

16. An actuator having a controller according to claim 15.

17. The method according to claim 4, wherein in a fourth step the manipulated variable amplitude is reduced.

Description

[0051] Embodiments of the invention are described below with reference to the drawing in which:

[0052] FIG. 1a shows an application in which the process variable depends positively from the manipulated variable, namely a container or room that is heated by a heater,

[0053] FIG. 1b shows an application in which the process variable is negatively related to the manipulated variable, namely a container that is emptied by a pump,

[0054] FIG. 2a is a time graph of a process variable that is determined that influences the underlying process,

[0055] FIG. 2b is a time graph of a sensor signal that a single sensor outputs,

[0056] FIG. 2c is a time graph of the manipulated variable that controls the actuator, as well as the corresponding manipulated variable mean value and the corresponding manipulated variable amplitude,

[0057] FIG. 3 is a schematic diagram showing the features of the underlying method that lead to the time courses from FIGS. 2a, 2b and 2c, and

[0058] FIG. 4 schematically shows a two-point controller according to the invention formed by analog components.

[0059] An application of the underlying method is shown in FIG. 1a where the process variable P is positively influenced by a manipulated variable u of an actuator 1. This is a container 3 that is heated by a heater 1, the level P, in the example the temperature, being measured within the container by a sensor 2. The sensor delivers a binary sensor signal y that depends on the level P. In particular, the sensor 2 outputs a first sensor signal value y1 when the process variable P exceeds a first switching value Sw1. If the process variable P drops below a second switching value Sw2, the sensor 2 outputs a second sensor signal value y0. In the case shown, the heater 1 is the actuator and the heat outputted by the heater 1 is the manipulated variable u that is controlled by the two-point controller 10. The straight arrows represent the direction of flow of the medium to be heated. The flow rate of the medium varies as required.

[0060] FIG. 1b shows an embodiment in which the process variable P depends negatively on the manipulated variable u. This is a container 3 that is emptied by a pump 1, the level P inside the container 3 being measured by a sensor 2 that produces a binary sensor signal y. In this application, the level P drops as the manipulated variable u or the speed of the pump 1 increases. The manipulated variable u is controlled by the two-point controller 10 that is set up to carry out the method according to the invention.

[0061] FIGS. 2a, 2b, and 2c show an example of the method with respect to time, the process variable P depends positively on the manipulated variable u, i.e. when the manipulated variable u increases. The time t is given in seconds. With the maximum manipulated variable (u=10), as it is at the beginning of the time line, the process variable P increases. FIG. 2b shows this sensor signal y, the first sensor signal value y1 being applied when the process variable P exceeds the first switching value Sw1. The first signal value y1 is applied until the process variable P drops below the second switching value Sw2, whereupon the second sensor outputs signal value y0. As shown in FIG. 2c, the controlled variable u of the actuator 1 is controlled based on the binary sensor signal y, in which case, manipulated variable u assumes a first manipulated variable u1 when the first sensor signal value y1 is present. When the second sensor signal value y0 is present, the manipulated variable rises to the second manipulated variable u2. This can be seen particularly well in the time range between t=800 sec and t=1000 sec, since the switching times between the first control value u1 and the second control value u2 are relatively large. As long as the actuator 1 with the first manipulated variable u1 is controlled, the process variable P drops, so that the duration of the first sensor signal value y1 corresponds to a fall time tfall. The process variable P drops if that actuator 1 assumes the second manipulated variable u2, so that the duration of the second sensor signal value y0 corresponds to a rise time t_rise.

[0062] The durations of the sensor signal values y1 and y2 are used for the underlying method so that the first control value u1 and the second control value u2 adjust when operating the actuator 1 dynamically so that they converge to one another. As can be seen in FIG. 2c, the first control value u1 and the second control value u2 converge when the actuator 1 is in operation. This makes the switching steps between the first manipulated variable u1 and the second manipulated variable u2 smaller, which leads to a reduction in switching losses when operating the actuator 1. In this way, more energy-efficient and less expensive operation of the actuator 1, for example a pump or a heater, can be attained. The convergence of the first control value u1 and the second control value u2 ensures that the differences between the control values are smaller, so that the change in process variable P slowly follows. A slower change in the process variable P has the consequence that the fall time t_fall and the rise time t_rise increase from cycle to cycle, so that the manipulated variable u is switched less and less. This also leads to the fact that the actuator 1 consumes less energy and the wear on actuator 1 is reduced. The convergence of the control values u1 and u2 can be set particularly well in the time ranges around t=200 sec, t=700 sec and t=1200 sec.

[0063] The first control value u1 and the second control value u2 are adapted so that the fall time t_fall and the rise time t_rise converge toward the same value. Alignment of fall time t_fall to rise time t_rise can be seen in FIG. 2c in the time ranges before t=500 sec, before t=1000 sec and before t=1500 sec.

[0064] The first setting value u1 and the second setting value u2 are preferably determined using a manipulated variable mean u_mean and a manipulated variable amplitude delta are calculated. in the illustrated case, the first manipulated variable u1 corresponds to the difference between the manipulated variable mean u_mean and the manipulated variable amplitude delta. It is applied when the sensor 2 outputs the first sensor signal value y1, i.e. when the process variable P exceeds the first switching value Sw1. As a result of the control with a reduced manipulated variable u, the process variable P, until it drops below the second switching value Sw2 and the sensor 2 outputs the second sensor signal value y2. The duration of the first sensor signal value y1 corresponds to the fall time t_fall. As soon as the second sensor signal value y2 is present, the actuator 1 is controlled by the second manipulated variable u2 that then corresponds to sum of the manipulated variable mean u_mean and the manipulated variable amplitude delta. The increased manipulated variable u causes the process variable P to rise, whereupon the second sensor signal value y2 is applied until the process variable P has exceeded the first switching value Sw1. Then the actuator 1 is controlled by the first manipulated variable u1, with a newly determined manipulated variable mean u_mean and a newly determined manipulated variable amplitude delta that are determined on the basis of the fall time t_fall and the rise time t_rise. As can be seen from FIG. 2c, the manipulated variable amplitude is successively reduced, so that the first control value u1 and the second control value u2 converge to one another. In addition, the manipulated variable mean value is successively adjusted to u_mean, which after a plurality of cycles converges to a value at which the fall time t_fall and rise time t_rise are the same.

[0065] The time curves from FIGS. 2a, 2b and 2c are based on a method that is shown in the process diagram of FIG. 3 that is a structured chart. The manipulated variable u is dependent on the state of the control procedure controlled, with a distinction between states 1, 2 and 3.

[0066] State 1 corresponds to a start step 101 in which the actuator 1 is controlled by a manipulated variable u which leads to the first switching value Sw1 being exceeded, or the first sensor signal value y1, i.e. y=1, is present. For this, a manipulated variable u is applied, which corresponds to the sum of the manipulated variable mean u_mean and the manipulated variable amplitude corresponds to delta, where the output values u_mean=5 and delta=5. As soon as the first sensor signal value y1 is outputted by the sensor 2, so y=1, is switched to state 2 that is represented by state=2 and a dashed arrow.

[0067] The state 2 corresponds to a first step 102 in which the actuator with a manipulated variable u that is the difference between the manipulated variable mean u_mean and that corresponds to the manipulated variable amplitude delta. For the first cycle this means that u=0 because u_mean=5 and delta=5. The control with such a manipulated variable u leads to increases of the process variable P. As soon as the process variable P reaches the first switching value Sw1, the sensor 2 outputs the second sensor signal value y0, that is y=0, so that the right column of state 2 is executed. The fall time t_fall is determined and change in the state 3 changed, which is represented by state=3 and a dashed arrow.

[0068] State 3 corresponds to a second step 103 in which a manipulated variable u is set such that of the sum of the manipulated variable mean u_mean corresponds to the manipulated variable amplitude delta. In the first cycle the manipulated variable is u=10, since the output values u_mean=5 and delta=5. Because of the control with the maximum manipulated variable, the process variable P increases, whereby the sensor 2 continues to output the second sensor signal value y0, so that say y=0. As soon as the process variable P exceeds the first switching value Sw1, the sensor 2 outputs the first sensor signal value y1, so y=1. As a result, the left column of state 3 is executed, and the rise time t_rise is determined.

[0069] Subsequently, in a third step 104 based on the rise time t_rise and the fall time t_fall a new manipulated variable mean value u_mean is determined. Here, when determining the new manipulated variable mean value u_mean, it is first asked whether the rise time t_rise is greater than the fall time t_fall. If this is the case, a correction factor is determined according to

k=1−sqrt(t_fall/t_rise)

The new manipulated variable mean u_mean is calculated based on the correction factor k where

u_mean=u_mean+k*delta

In this case, the ratio of fall time tfall and rise time t_rise is greater than 1, so that the correction factor k is positive and thus the new manipulated variable mean value is greater than that previous manipulated variable mean. In the event that the fall time is greater than the rise time, the correction factor k is calculated according to

k=sqrt(t_rise/t_fall)−1

In this case the correction factor k is negative because the ratio of both times is less than 1. The new manipulated variable mean value u_mean is therefore smaller than the old manipulated variable mean u_mean. Such a correction ensures that the new manipulated variable mean value u_mean approaches a value at which the rise time t_rise is equal to the fall time t_fall. This adaptation of the manipulated variable mean value u_mean is shown in FIG. 2c by the light gray line.

[0070] After calculation of the new manipulated variable mean value u_mean, in this embodiment, a fourth step 105 is carried out, in which the manipulated variable amplitude delta is reduced. In the present case, the new manipulated variable amplitude corresponds to 0.9 times the old manipulated variable amplitude delta. If the reduction is repeated, the control amplitude delta thus decreases exponentially. After reducing the manipulated variable amplitude, state 2 is executed, with the newly determined by the calculation of the manipulated variable u being from manipulated variable mean u_mean and the newly determined manipulated variable amplitude delta. In the present case, the manipulated variable u corresponds to the difference between the new control value mean u_mean and the new control value amplitude delta. Due to the process, variable P decreases again when controlled by a reduced manipulated value. As soon as the process variable P drops below the second switching value Sw2, the sensor 2 outputs the second sensor signal value y0, i.e. y=0, so that the controller is switched to state 3, i.e. the second step 103 executes. Accordingly, the first step 102 becomes the second step 103, the third step 104 and the fourth step 105 are executed successively and repeatedly, and the manipulated variable mean value u_mean is adapted and the manipulated variable amplitude delta is gradually reduced. After several cycles, the manipulated variable mean value u_mean approaches a value at which the fall time t_fall is equal to the rise time t_rise.

[0071] The manipulated variable amplitude delta is reduced until a minimum manipulated variable amplitude value delta min is reached. In the diagram of FIG. 2c this is in the ranges before t=500 sec, before t=1,000 sec and before t=1500 sec the case. A minimal manipulated variable amplitude delta min is required to determine fall times t_fall and rise times t_rise then seek to be able to determine when the manipulated variable mean value u_mean reaches a value in which the rise time t_rise is equal to the fall time t_fall. The continuing fluctuation of the process variable P and the associated repeated determination of the times t_fall and t_rise is necessary in order to be able to recognize any disturbances of process role P from outside.

[0072] In the present embodiment, in state 2, when the first sensor signal value y1 (y=1) is applied, a test step 106 is carried out in which the fall time is compared with a mean value from the two times t_span (see FIG. 3 left column State 2). If the fall time t_fall is more than twice as great as twice the mean value from both times t_span, a reset takes place by adding the manipulated variable u_mean and the manipulated variable amplitude delta are reset to their initial values. In the present case u_mean=5 and delta=5. In addition, the control value suppression no_adj=1 is switched on, so that in the cycle that follows an error, no new mean value of the manipulated variable u_mean and the manipulated variable amplitude delta are calculated. In FIG. 3, the query of the disturbance is represented at 107. Suppression of the recalculation ensures that the control quickly reacts to malfunctions and can adjust so that the process variable P does not deviate too much from the desired value. In state 3, test step 106 is carried out when the first sensor signal value y0 (y=0) is applied.

[0073] In FIGS. 2a, 2b and 2c, a first disturbance is induced at time t=500 sec and a second disturbance induced at time t=1000 sec. The first problem is to correct the manipulated variable mean value u_mean upward, so that after several switching cycles a manipulated variable mean value u_mean is set that is larger than the manipulated variable mean value u_mean at t=500 sec. In the example of heating a container with a heater, this disturbance can be synonymous with the fact that the temperature has dropped due to an opening of the container so that the output of the heater must be increased to compensate for the temperature drop. The second fault is corrected when the mean value of the manipulated variable u_mean goes down, with a value after approximately t=1350 sec that sets the mean value of the manipulated variable u_mean that corresponds to that at t=500 sec. The first failure will be thus reduced by the second disturbance, or the temperature drop due to the opening compensated for the first failure.

[0074] In this embodiment, the manipulated variable u can be varied between 0 and 10 and represents an abstract value. It can be, for example, a voltage with which an actuator is operated. Based on the application example from FIG. 1a, the heating power might correspond to a value that is proportional to the manipulated variable u from FIG. 2c.

[0075] In an unillustrated variant the method derived from the diagram in FIG. 3 can be used with an application example of FIG. 1b.

[0076] An embodiment of a controller 10 according to the invention is shown in FIG. 4. The circuit diagram shown corresponds to an upper structure level in which the individual blocks are each formed from analog circuits. The binary output signal y of the sensor 2 is applied in this case to an input 11 and by analog components, for example operational amplifiers and storage elements, outputted as manipulated variable u at output 12. The binary sensor signal y is applied to an input of a status machine 13 that outputs five states s1 to s5. State s1 is used to initialize the controller. State s2 is active when the first sensor signal value y1, i.e. y=1, is present. State s3 is active when the second sensor signal value y2 is present, i.e. y=0. In state s4 the new manipulated variable u_mean is calculated. In state s5, the manipulated variable amplitude is reduced to delta and the timing is reset. The states are changed to the sequence 1, 2, 3, 4, 5, 2, 3, 4, 5, 2 . . . by the status machine 13. The status machine 13 outputs a signal rise, with which a first switch 21 is controlled, so that the second control value u2 is present at the output 12, which leads to an increase in the process variable P. If the status machine 13 gives a drop signal, a switch 22 is activated so that the first control value u1 is present at output 12 and the process variable P drops.

[0077] In the state s1 that corresponds to the starting step 101, a first memory block 14 for the manipulated variable mean value u_mean set a voltage Vin1 that corresponds to the voltage from a voltage source 24, in this case 5 volts. This supply voltage is set a second memory block 15 for the manipulated variable amplitude delta. Both memory blocks 14 and 15 are sample-and-hold circuits with two specifiable input values. The first memory block 14 and the second memory block supply output voltages of 5 volts that are applied to the inputs of an adder 18 and a differentiator 19 are applied. The adder 18 supplies the sum of the two output voltages, so that when the first switch 21 is activated at the output u=u_mean+delta is applied. The differentiator 19 delivers as output voltage the difference between the voltages, so that a manipulated variable u=u_mean−delta is applied to the output 12 when the second switch 22 is activated.

[0078] In state s2, that is, when the first sensor signal value y1 is present, a signal is sent to an input of a first integrator 16 whose duration corresponds is outputted as voltage Vout by the integrator 16. This voltage is applied to a mean value calculation block 20 as t_fall.

[0079] If the state s3 is active, that is when the second sensor signal value y0 is present, a signal is applied to a second integrator 17 whose output voltage Vout corresponds to the rise time t_rise. The output voltage of the second integrator 17 is also used as the input of the average value calculation block 20. The mean value calculation block 20 determines from the fall time t_fall, the rise time t_rise, the manipulated variable mean u_mean and the manipulated variable amplitude delta a new manipulated variable mean value newUmean that is sent to an input Vin2 of the first memory block 14.

[0080] As soon as the state s4 is activated, the new manipulated variable mean value newUmean is calculated and set as output voltage by application to the input setVin2 of the first memory block 14. This voltage is applied to the adder 18 and the differentiator 19.

[0081] When the state s5 is subsequently activated, the first integrator 16 and the second and the integrator 17 are reset. In addition, the value for the manipulated variable amplitude delta is reduced in the memory block 15. The manipulated variable u at output 12 depends on the newly set manipulated variable mean value of the first memory block 14 and the reduced manipulated variable amplitude of the second memory block 15.

[0082] The manipulated variable amplitude will not be reduced if the mean value calculation block 20 is reset by a flip-flop 30 and the second AND gate 29 together. In the same way a new manipulated variable mean value u_mean is not set if the reset is present, since the output s4 of the status machine and the output Q of the flip-flop 30 are linked via a first AND gate 28 to each other. When the reset is activated, a first OR gate 25 at the first memory block 14 sets the supply voltage of the voltage source 24 as the manipulated variable mean value. In addition, a second OR gate 26 is activated when the reset is activated in the second memory block 15, and the supply voltage of the voltage source 24 is set as the manipulated variable amplitude. The reset is set by the mean value calculation block 20 if the fall time t_fall or the rise time t_rise deviates from the mean value of both times t_span.

[0083] With the circuit of FIG. 4, the underlying method can be handled by an analog circuit, with only simple components such as operational amplifiers and logic gates required. Such an analog circuit can be designed as an integrated module that is integrated into an actuator 1 that is for example, a pump, a valve, a heater or a cooling system. Advantageously, one can do this without significant effort to implement an energy-efficient and low-wear control of an actuator 1, if a sensor 2 is used that makes a binary sensor signal y available.

TABLE-US-00001 List of reference symbols  1 actuator  2 sensor  10 controllers  11 input  12 output  13 status machine  14 first memory block (u_mean)  15 second memory block (delta)  16 first integrator (t_fall)  17 second integrator (t_rise)  18 adder  19 differentiators  20 Average value calculation block  21 first switch (u1)  22 second switch (u2)  23 mass  24 voltage source  25 first OR gate  26 second OR gate  27 third OR gate  28 first AND gate  29 second AND gate  30 flip flops 101 start step 102 first step 103 second step 104 third step 105 fourth step 106 test step 107 Suppression query Y sensor signal (binary) yl first sensor signal value y0 second sensor signal value P process variable, level Swl first switching value Sw2 second switching value u manipulated variable ul first control value u2 second control value t_fall fall time t_rise rise time t_span mean value from fall and rise time u_mean manipulated variable mean delta manipulated variable amplitude delta_min minimum manipulated variable amplitude k correction factor State of the controller no_adj Suppression of the calculation of the mean value of the manipulated variable