JPH0421101A - Saturated processing system for learning controller - Google Patents

Abstract

PURPOSE:To prevent the generation of a positional deviation by integrating an excess value obtained by subtracting a clamp value in band limiting filtering processing from an input value in each sampling and changing positive and negative clamp values so that the integrated value becomes zero. CONSTITUTION:A real speed omega is subtracted from a speed command (r) in each sampling period to find out a speed deviation E and the deviation E and correction data E1 corresponding to the just preceding sampling period are inputted as the input X1 of a saturating processing block 2. When the absolute value of the input X1 is small, the block 2 outputs the input value as an output X0, but when the input X1 exceeds a set positive or negative clamp value, the input value X1 is clamped and outputted as an output X0. Excess values obtained by subtracting respective clamp values from the input value X1 are integrated and a variable clamp value is adjusted in accordance with the integrated value. Thus, the speed loop processing of the output X0 is executed through a band limiting filter processor 3, a delay element 4 and a dynamic characteristic compensating element 5. Thus, the generation of a positional deviation can be prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to the control of servo motors used in machine tools and the like, and particularly to control in which the same pattern is repeatedly commanded at a predetermined period.

Conventional technology In the control of servo motors, a learning control method is used as a method to converge control deviation to zero in response to commands that are repeated in the same pattern at a predetermined period, perform highly accurate motor control, and improve machining accuracy. .

FIG. 5 is a block diagram of essential parts in controlling a servo motor to which the learning control method described above is applied.

In Fig. 5, r is a speed command, E is a speed deviation that is the difference between the speed command r and the actual speed ω, and 6 is a speed loop transfer function, which is a conventional method in which an integrator is added to know the position deviation. As is well known, it performs PI control and the like. and,
A learning controller 10 is added to perform learning control, and the learning controller 10 includes a band-limiting filter 3.
, a delay element 4 that stores data for one period that is repeated at a predetermined period, and a dynamic characteristic compensation element 5 that compensates for phase lag and gain reduction of the controlled object.

The learning controller 10 adds the data El outputted from the delay element 4 at the time of sampling one cycle before to the speed deviation E at every predetermined sampling period T, and performs processing by the band-limiting filter 3 to add the data to the delay element 4. Store data. The delay element 4 has N (=L/T) memories and can store each sampling data for one cycle, and outputs the oldest data at each sampling time. . That is, the input data is shifted by one address for each sampling and stored in the memory at address 0, and
Enter the address data. As a result, the delay element 4 outputs sampling data delayed by one cycle.

Therefore, since the speed command r of the same pattern is given periodically, the speed deviation E and the output of the delay element 4 are added, and the data at the same position on the pattern of the speed command r is stored as correction data. Become.

Further, the output of the delay element 4 is compensated for the phase delay and gain of the controlled object by the dynamic characteristic compensation element 5, and is output as the output y of the learning controller 10, and the output y is the speed deviation E
The speed loop processing is executed using this added data.

As a result, the same pattern of processing is repeated in a predetermined cycle, and if the speed deviation E at the sampling time corresponding to the sampling time in the previous cycle at a certain sampling time is a large value, the learning controller 10
Since a large value of output y is output from and added to the speed deviation, the speed deviation input to the speed loop process changes greatly, and the actual speed ω also changes accordingly, so the speed deviation E is The value is corrected so that it converges to zero, making it possible to control the motor with high precision.

Problems to be Solved by the Invention As mentioned above, in learning control, there are cases where the speed deviation E becomes very large, and depending on the maximum output of the amplifier that drives the servo motor, the maximum output torque of the servo motor, etc. Since the torque that the motor can output is limited, there may be cases where the speed deviation E does not converge. In such cases, the input value of the band-limiting filter 3 in the learning controller 10 gradually increases, and the calculation unit An overflow occurs when the number exceeds the range that can be processed. In such a case, a method has conventionally been adopted in which the input of the band-limiting filter 3 is clamped at a certain fixed value. However, when such a method is used, a problem arises in that the position is shifted.

FIGS. 6 and 7 are explanatory diagrams of the occurrence of this positional deviation.

The operation is repeated at a predetermined period, as shown in Figure 6 (
As shown in a), the torque Tdl during deceleration.

Ta2. Ta3... Torque Tal during acceleration,
Ta2. . . . are output alternately, and the position and velocity are repeated in the same pattern at a predetermined period as shown in FIGS. 6(b) and 6(c). "Torque limit" in Fig. 6(a) indicates the limit value of the torque that the servo motor can produce with 8 forces, and the torque command is Ta1, Ta2..., T.
Even if there are dl, Ta2, etc., the torque output from the motor is up to this torque limit value. As a result, the integrated value AI of torque that cannot be output from the motor even if there is a torque command.

In A2..., Bl, B2..., A1=lBI
If A2=I821..., then the torque during acceleration and deceleration is reduced equally, so no positional deviation occurs. FIGS. 6(b) and 6(c) show the state in such a case.

On the other hand, there is a difference in the above integrated values, for example, AI>Bl l
, A2>l 821..., the torque during acceleration becomes smaller, the rising slope of the speed becomes smaller as shown in FIG. 7(b), and the amount of movement decreases accordingly. As a result, as shown in FIG. 7(a), the positional deviation di
, d2... will be generated.

SUMMARY OF THE INVENTION An object of the present invention is to provide a saturation processing method for a learning controller that prevents the positional shift by adjusting the clamp value of the input of the band-limiting filter to an optimal value.

Means for Solving the Problems The present invention provides control of a servo motor in which a speed command is repeated at a predetermined period, and in which correction data corresponding to the sampling period one cycle before the predetermined period is added to the speed deviation at the sampling period. The added value is subjected to band-limiting filter processing and stored as correction data at the sampling time in the cycle, and the speed at the sampling time is calculated based on the correction data corresponding to the sampling time in the previous cycle. In a learning controller using a learning control method that corrects deviations and controls a servo motor, the amount by which the input value exceeds the positive and negative clamp values that clamp the input in the above-mentioned band-limiting filter processing is integrated at each sampling time, and the above-mentioned integrated value is calculated. The above problem was solved by changing the positive and negative clamp values so that the value becomes zero.

For example, in FIG. 6, Al>IBI l, A2>
Suppose that a positional shift occurs in B21... as shown in FIG. In such a case, as shown in Fig. 7(b), the actual speed lags behind the original speed shown by the broken line, and the speed deviation E becomes large with a positive value.As a result, the correction data is added to the speed deviation E. are added, and the integrated value of this added value exceeding the positive and negative clamp values becomes larger than the integrated value detected in the previous cycle. In such a case, the positive (upper limit) and negative (lower limit) clamp values of the band limit filter processing are lowered.

Then, the clamped positive values of the stored correction data become small, and the absolute values of negative values become large. As a result, in the next cycle, the output of the learning controller based on this correction data will be small when positive, and large when negative, and this will be added to the speed deviation, so the speed loop When the torque command output in the process is positive, the torque command becomes small, and when it is negative, the absolute value becomes large. Therefore, the values of Al, A2, . . . in FIG. 6 become small, and the absolute values of Bl, B2, . As a result, the above integrated value A1+B1+A2+
The value of B2... is small (approaching zero), and the integrated value of the amount exceeding the positive and negative clamp values in the added value by adding the correction data to the speed deviation E also approaches zero.

In addition, in the above integrated values, Al< l BIA2<
In the case of l B2 +..., the deceleration torque is small, the deceleration speed is decreased, and the negative absolute value of the speed deviation E becomes large. As a result, the integrated value of the sum of the speed deviation and the correction data that exceeds the positive and negative clamp values becomes smaller than the integrated value of the previous period. In such a case, the positive (upper limit) and negative (lower limit) clamp values of the band-limiting filter processing are increased. Then, the clamped positive values of the stored correction data become large, and the absolute values of negative values become small. As a result, in the next cycle, the output of the learning controller based on this correction data will be large when positive, and small when negative, and this will be added to the speed deviation, so the speed loop The torque command output in the process increases when it is positive, and its absolute value decreases when it is negative. Therefore, Al in Figure 6
, A2... become large, and the absolute values of Bl, B2... become small. As a result, the above integrated value A
The value of 1+B1+A2+B2... becomes smaller and approaches zero, and the integrated value of the addition of the correction data to the speed deviation E that exceeds the positive and negative clamp values also approaches zero.

Embodiment FIG. 1 is a block diagram of essential parts in controlling a servo motor to which a learning control method according to an embodiment of the present invention is applied.

In FIG. 1, reference numeral 1 denotes a learning controller, which differs from the conventional learning controller 10 shown in FIG. 5 in that a saturation processing block 2 is added. and,
Elements that are the same as those in the conventional system shown in FIG. 5 are designated by the same reference numerals.

If the sampling period (speed loop processing period) of the learning controller 1 is T, and the period of the speed command r repeated in a predetermined processing pattern is L, then the delay element 4 has L/T=N.
It has two memories and can store each sampling data for one cycle, and outputs the oldest data at each sampling time. That is, the input data is shifted by one address for each sampling, and the input data is stored in the memory at address 0, and the data at address N-1 is output.

As a result, the delay element 4 outputs correction data that has been published for one cycle. Therefore, as will be described later, since the same pattern of speed commands r is given periodically, correction data at the same position on the pattern of the speed command r is added to the speed deviation E.

At each sampling period T, the actual speed ω is subtracted from the speed command r to obtain the speed deviation E, and the correction data El corresponding to the speed deviation E at the sampling time one period before (the correction stored at address N-1 in the memory) data) are added, and the added value becomes the input X1 of the saturation processing block 2. The saturation processing block 2 performs processing as shown in the graph in Figure 2, and when the absolute value of input X1 is small, it outputs input X1 as it is as output When it exceeds the value Lh, it is clamped to this value and outputs the output XO. Furthermore, when the input X1 exceeds the negative variable clamp value Ll, the output xO is clamped to this value Ll and output. Note that Lmax and Lmin are the limit values of numerical values that can be processed by the arithmetic unit, Lmax is a positive fixed value, and Lmin is a positive fixed value.
is a fixed negative value. Then, the amount exceeding each clamp value of the input X1 is integrated, and the variable clamp values Lh and Ll are adjusted according to the integrated value.

The data XO output from the saturation processing block 2 in this way
Band limit filter processing 3 is performed, and the output thereof is stored in address 0 of the memory, and the memory contents are shifted one address at a time.

Further, the output of the delay element 4 is compensated for the phase delay and gain of the controlled object by the dynamic characteristic compensation element 5, and is output as the output y of the learning controller 1, and the 8 forces y are added to the speed deviation E. Velocity loop processing is executed using this added data.

The detailed description of the present invention has been described above, and next, one embodiment of the present invention will be described.

FIG. 3 is a block diagram of an embodiment of servo motor control of a machine tool implementing the present invention.

In Figure 3, 20 is a numerical control device that controls the machine tool;
1 is a shared memory for receiving position commands to the servo motor of the machine tool outputted from the numerical control device 20 and passing them to the processor of the digital servo circuit 22; 22 is a digital servo circuit; It controls the position, speed, current, etc. of the servo motor 24, and also performs learning control processing. 23 is a servo amplifier composed of a transistor inverter, etc.; 24 is a servo motor; and 25 is a pulse coder that generates a predetermined number of feedback pulses per servo motor cycle and outputs them to the digital servo circuit 22. Note that 22a is a memory provided in the digital servo circuit 22 and constituted by a ROM and a RAM.

The above configuration is used to control servo motors of machine tools, etc.
This is a well-known matter as a digital servo circuit, and detailed explanation will be omitted.

Next, the processing of the saturation processing block 2 in the processing performed by the processor of the digital servo circuit 22 is shown in FIG.
) and (b). Note that other processing is the same as the conventional one, so a description thereof will be omitted.

The processor executes the processes shown in FIGS. 4(a) and 4(b) every sampling period (speed processing period) T, and first subtracts the actual speed ω of the motor detected by the pulse coder 25 from the speed command r to obtain the speed deviation E. Calculate the correction data E1 N cycles before the sampling cycle T stored at address N-1 of the delay element in the speed deviation E (since the sampling cycle T is calculated by multiplying by the repeated predetermined cycle and L/T=N, Correction data El corresponding to the sampling period one period before
) is added in step S1. S2), Step S to determine whether the added value X1 is a positive value of zero or more.
3) If it is a positive value of zero or more, it is determined whether the added value X1 is set as the output value XO of the saturation process and passed to the next process, the band-limiting filter process (step S5).

Moreover, if the addition value X1 is larger than the positive variable clamp value Lh, the positive variable clamp value L is used as the output XO of the saturation process.
h is output, and a value obtained by subtracting the positive variable clamp value Lh from the addition value X1 is added to the integrator A (which is initially set to 0) (steps S6 and S7).

Furthermore, if the additional value x1 is negative, the additional value X1 is currently (
Negative variable clamp value Ll (which is initialized at the beginning)
If it is smaller, it is determined whether the added value The value obtained by subtracting the negative variable clamp value Ll from is added (step SI0, 811
) Next, it is determined whether the counter CNT, which is set to 0 in the initial setting and is incremented by 1 each time this saturation process is performed, is equal to the set value n (step 512). If not, 1 is added to the counter CNT. It is determined whether or not the value of the counter CNT is equal to the value N obtained by dividing one period by the sampling period T. If they are not equal, this saturation process for the period is ended, and if they are equal, the counter CNT is set to 0. (step S19.S20.5)
21).

Thereafter, the above process is repeated every sampling period T, and when it is detected in step S12 that the value of the counter CNT has reached the set value n, the value of the counter CNT when the value of the counter CNT was n in the previous period from the integrated value A is Multiply the value obtained by subtracting the value of the register R (A) that stores the integrated value A by the gain K,
The obtained values are set as the current positive and negative variable clamp values Lh, Ll.
, to obtain new positive and negative clamp values Lh and Ll (step 813). Note that the register R(A) is initially set to 0 as an initial setting. Next, register R(A
) is set to the integrated value A (step 514). Then, if the positive variable clamp value Lh exceeds the positive fixed clamp value Lmax that can be processed by the arithmetic unit, this positive variable clamp value Lh is changed to the positive fixed clamp value Lmax,
Similarly, if the negative variable clamp value Ll is smaller than the negative fixed value Lmin, the negative variable clamp value Ll is changed to this negative fixed clamp value. In addition, each variable clamp value Lh, L
If l does not exceed the respective fixed clamp values Lmax and Lmin, the clamp value calculated in step S13 is directly used as the variable clamp value (steps 815 to 5).
18). Then, the processing of steps 319 to S21 is performed to end the processing of the sampling period.

The above processing is performed every sampling period T, and the counter CNT has a count value of N (1 of the repeating pattern).
It is set to 0 every time the number of sampling times within a period (N=L/T) is reached, and each time the value of the counter CNT reaches the set value n, the processing of steps 813 to 318 is performed. The negative variable clamp values Lh and Ll will be adjusted. That is, the variable clamp values Lh and Ll are changed at fixed predetermined positions in a predetermined pattern.

As described above, the variable clamp value Lh is set once at a predetermined position within one cycle of the repeated pattern.

Ll has been changed, and this change is due to the fact that the integrated value detected in the previous cycle (the integrated value of the sum of the speed deviation E and correction data El exceeding the variable clamp value Lh, Ll) is changed in the current cycle. If the integrated value is smaller than the above-mentioned integrated value, the positive and negative variable clamp values Lh and Ll are changed to a smaller value than the previous cycle, and if larger, they are changed to a larger value. The fact that the integrated value detected in the previous cycle is smaller than the integrated value detected in the current cycle means that the speed deviation E is increasing in the positive direction, and the positive torque command value is large and the motor The amount exceeding the torque limit is large (A1 in Figure 6)
．． A2... is large). So, Tadashi,
If the negative variable clamp values Lh and Ll are made small (i.e., the clamped values stored as correction data are made small), the output y of the learning controller that is added to the speed deviation E will become positive in the next cycle. When the value is a negative value, the absolute value is small, and as a result, the positive integrated value of the torque command exceeding the motor torque limit value (AI, A2, etc. in Figure 6) becomes smaller, the absolute value of the negative integrated values (Bl, B2, etc. in Figure 6) becomes larger, and the sum of these integrated values approaches zero, resulting in a variable clamp of the value obtained by adding the correction data to the speed deviation. The integrated value for the amount exceeding the torque limit also approaches zero, and the positive and negative variable clamp values converge to an optimal constant value.Then, the positive and negative integrated values of the torque command for the amount exceeding the torque limit become equal. Therefore, positional deviation is eliminated.Also, in the opposite case, the integrated value detected in the previous cycle (the integrated value of the sum of the speed deviation E and correction data El that exceeds the variable clamp values Lh and Ll) is larger than the above-mentioned integrated value for the corresponding cycle, the positive and negative variable clamp values Lh and Ll are changed to values larger than those of the previous cycle, thereby determining the optimal positive and negative variable clamp values to eliminate positional deviation. It will be automatically set to Lh and Ll.

Effects of the Invention The present invention changes the clamp value that clamps the value obtained by adding correction data to the speed deviation input as saturation processing so that the integrated value exceeding the clamp value becomes zero. Since the value corrected to the speed deviation by the learning controller is modified so that the integrated value of the torque command exceeding the motor torque limit becomes zero, it is possible to prevent positional deviation from occurring.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of the main parts in controlling a servo motor to which the learning controller method of an embodiment of the present invention is applied, Fig. 2 is a diagram showing the input/output relationship of the saturation processing block in the embodiment, and Fig. 3 is a diagram showing the input/output relationship of the saturation processing block in the embodiment. 4(a) and 4(b) are flowcharts of the process executed as saturation processing by the processor of the digital servo circuit in the same embodiment, and FIG. FIGS. 6 and 7 are block diagrams of main parts in controlling a servo motor that performs learning control using a conventional learning controller, and are explanatory diagrams for explaining positional deviation. 1.10 Learning controller, 2 Saturation processing block, 3 Band limit filter, 4 Delay element, Dynamic characteristic compensation element, 6 Velocity loop transfer function, 20 ... Numerical control device, 21 ... Shared memory, 2
2... Digital servo circuit, 23... Servo amplifier, 24... Servo motor, 25... Pulse coder. Figure fa)

Claims (1)

[Claims] Control of a servo motor in which a speed command is repeated at a predetermined period, in which correction data corresponding to the sampling period one cycle before the predetermined period is added to the speed deviation at the sampling period, and the added value is applied to a band-limiting filter. Learning to perform processing and store it as correction data at the time of the sampling in the relevant cycle, and to control the servo motor by correcting the speed deviation at the time of the sampling in question based on the correction data corresponding to the time of the sampling in the previous cycle. In the learning controller in the control method, the amount by which the input value exceeds the positive and negative clamp values that clamp the input in the band-limiting filter processing is integrated at each sampling time, and the positive and negative clamp values are integrated so that the integrated value becomes zero. A saturation processing method for a learning controller characterized by changing a clamp value.