JP5886717B2 - Iterative learning position controller - Google Patents

Description

  The present invention relates to an iterative learning position control device, particularly a numerical control machine, which is repeatedly mounted on a control axis (high-speed synchronous axis) that causes a cutting tool to perform high-speed synchronous operation in the radial direction of the workpiece in accordance with the rotation angle of the workpiece. The present invention relates to a learning position control device.

  FIG. 5 is a diagram schematically showing an example of a conventional structure of a numerical control machine equipped with a high-speed synchronous axis. One end of the workpiece is held by a chuck, and the other end is supported by a tailstock. For this reason, the workpiece rotates in accordance with the main shaft rotation. The tool is fixed to the tool post, and the tool post is driven in the X-axis direction (the workpiece radial direction) and the Z-axis direction (spindle rotation axis direction) via a plurality of ball screws (not shown). In FIG. 5, the operation of the cutting tool in the X-axis direction is a high-speed synchronization axis. In many cases, the cutting tool moves back and forth in the X-axis direction and is fed at a constant speed in one direction in the Z-axis direction in synchronization with the angle of rotation of the main shaft (main shaft rotation angle). By such a combined operation, the workpiece can be processed into a rope screw shape or a cam shape.

  The workpiece is defined by a position command value in the X-axis direction set for each main shaft rotation angle at a constant angular pitch interval in each of a plurality of cross-sections (multi-sections) arranged at equal intervals in the Z-axis direction. This is created by a host controller (not shown) such as a CAD device as an X-axis position command value data table (hereinafter simply referred to as a command value table) for each constant angle pitch, It is set in the position control device. In general, in the high-speed synchronous axis control, the main axis is controlled at a constant rotation, and, for example, the X-axis (high-speed synchronous axis) direction and the Z-axis direction are simultaneously started at a main axis rotation angle: 0 deg passage timing. This establishes position synchronization between the main axis, the X axis (high-speed synchronization axis), and the Z axis.

  Here, the X-axis (hereinafter also referred to as a high-speed synchronous axis) has a steep change in the command value with respect to the main axis and the Z-axis operating at a constant speed. Even if a simple tracking control system is configured, high machining accuracy cannot be obtained. Therefore, a position control Dc that is (position command value−position detection value) is collected for all cross sections corresponding to the spindle rotation angle, and a learning control system is added to create a correction value Cc by applying a learning calculation operation. Many position control devices adopt iterative learning control that achieves reduction of the position deviation Dc while repeating this correction value generation cycle.

  FIG. 6 is a block diagram illustrating the configuration of a high-speed synchronous axis iterative learning position control apparatus 100 equipped with an iterative learning control unit 101. The command value generation unit 50 is set in advance with a command value table created from a host controller (not shown) at a constant angular pitch corresponding to the multi-spindle spindle rotation angle. In the position control of the high-speed synchronous axis, the position command value Xc is read from the command value generation unit 50 corresponding to the spindle rotation angle.

  The feedback configuration of the iterative learning position control device 100 is as follows. The tool post driving unit 200 is provided with a position detector (not shown) for detecting the position of the cutting tool in the high-speed synchronous axis direction, and its output is a position detection value xf. The subtractor 51 subtracts the position detection value xf, which is position feedback, from the position command value Xc, and outputs a position deviation Dc.

  The position deviation Dc is added by the adder 56 with the correction value Cc read out from the correction value table set in the correction value generating unit 54 in accordance with the spindle rotation angle for multiple sections, and the position deviation amplifier 57 It is amplified to a position loop gain Kp times to become a speed command value Vc. The subtractor 58 subtracts a motor speed vm of a servo motor (not shown) that drives the tool rest driving unit 200 from the speed command value Vc, and outputs a speed deviation Ve.

The speed deviation Ve is normally proportionally integrated and amplified by the speed deviation amplifier 59. Equation (1) shows this speed loop gain Gv (Gp is a proportional gain, Gi is an integral gain, and s is an operator of Laplace transform).
Gv = Gp + (Gi / s) (1)
The output of the speed deviation amplifier 59 becomes the control input u of the tool post driving unit 200. The control input u is torque controlled (not shown) so as to be equal to the torque generated by the servo motor.

  Next, the iterative learning control unit 101 of FIG. 6 will be described with reference to the operation flowchart of FIG. When the repeated learning control is started, first, the command value table position control is turned ON in S100. This indicates that the current learning cycle is started, the position command value Xc is read from the command value table of the command value generating unit 50, and the position control operation is started.

  S101 Deviation value collection: ON indicates that the above-mentioned position deviation Dc is collected by the deviation value collection unit 52 of FIG. 6 and the creation of a deviation value table corresponding to the spindle rotation angle for multiple sections is indicated. When all reading of the position command values corresponding to all the spindle rotation angle points for the multi-section set in the command value table is completed and the deviation value collection corresponding to all points is completed, the creation of the deviation value table is completed in S102. Once established, the command value table position control is turned OFF in S103. The created deviation value table is stored and set in the deviation value collection unit 52.

  Next, the learning control unit 53 obtains the absolute value average of the deviation values Def of all points (average value of the absolute values of all the obtained position deviations Dc) from the created deviation value table, and pre-control device ( If the absolute value average of the deviation value Def is smaller than the deviation value convergence reference value set from (not shown), the deviation value convergence is determined in S104, and the iterative learning control is terminated. On the contrary, when the absolute value average of the deviation value Def is larger than the deviation value convergence reference value, it is determined that the deviation value has not converged, and in S105, the learning calculation process is executed on the created deviation value table.

  The learning calculation process of S105 is usually a filtering for removing a noise component included in the deviation value table, a phase advance compensation process for creating a correction value with the phase advanced in consideration of a response delay of the position control system, or the like. The learning control unit 53 is configured to output a correction value Cmp0 of the current learning cycle that has been subjected to the learning calculation process.

  The correction value table for the next learning cycle in S106 is created according to the following procedure. First, the correction value Cmpb of the current correction value table set in the correction value generation unit 54 is read out and added to the correction value Cmp0 of the current learning cycle by the adder 55 in FIG. 6 to obtain the correction value Cmp for the next learning cycle. . The correction value Cmp is calculated for all the main axis rotation angle points for the multi-section, and a correction value table for the next learning cycle is set in the correction value generator 54.

  When the creation of the correction value table is finished, the current learning cycle is finished, the process returns to S100, and the next learning cycle is started. The correction value table created in S106 is a table of correction values Cc read in accordance with the spindle rotation angle in the next learning cycle (that is, the correction value Cc (i) at the i-th learning cycle is (I-1) The correction value Cmp (i-1) calculated during the first learning cycle is obtained).

  The shape of the workpiece processed with the high-speed synchronous axis is developed in two dimensions in the X-axis and Z-axis directions with a CAD device, etc., and the Z-axis position is multi-sectioned based on the amount of Z-axis movement per main spindle rotation Is converted into a spindle rotation angle and set as a command value table. In this case, when the workpiece shape is a rope screw or the like, generally, the shape of the screw cut and the raised portion often include a shape sudden change portion.

FIG. 8 is a three-dimensional graph showing position command values for multiple cross sections. In FIG. 8, the “command value” on the vertical axis indicates the X-axis position command value, the “spindle rotation angle” indicates the main shaft rotation angle within one cross section, and the “cross section No” indicates cross sections arranged at equal intervals in the Z axis direction The numbers are shown respectively. The cross section to be displayed is thinned out at a three cross section pitch. Further, the command value, for each cross No, since the display by connecting the spindle rotation angle of direction to express the passing track of the actual command value. As shown in FIG. 8, in the case of a rope screw, a shape suddenly changing portion is included in a cut-in portion or a raised portion. In this example, the area from the 15th section to the 25th section of the section No is a cut-in part, and the area from the 310th section of the section No to the 360th section is a rounded-up section.

Here, since the command value table is created at a constant angle pitch, the command value speed of the high-speed synchronous axis is proportional to the spindle rotational speed, and the command value acceleration is proportional to the square of the spindle rotational speed. become. 9 and 10 are diagrams showing the command value and command value acceleration of the round-up portion when the spindle rotation speed: 800 min ⁻¹ is selected, and the horizontal axis indicates the cross section No. In particular, FIG. FIG. 38 is an enlarged view of the 337th section from the 335th section. From this, it can be seen that an excessive command value acceleration of 50G exceeding the physical acceleration limit of the high-speed synchronous axis occurs in each of positive and negative directions at almost constant points (main shaft rotation angle) at each rotation.

  Next, FIG. 11 is an enlarged view of the command value and the command value acceleration for the 150th to 152nd cross sections in FIG. 8 at the same spindle rotational speed. As described above, even under the same spindle rotation speed, the shape change is gentle in the complete screw portion, and therefore the command value acceleration is less than 1G.

  When it is determined that the deviation value has converged in the operation flowchart of FIG. 7 and the operation of the iterative learning control unit 101 of FIG. 6 is completed, a final correction value table is completed. 12 and 13 are diagrams showing the command value Xc, the correction value Cc, and the position deviation Dc when the machining operation is executed at the same spindle rotational speed as described above, using the final correction value table. is there. Since FIG. 12 shows a complete screw portion, the positional deviation Dc can be minimized by a sine wave-like gentle correction value Cc.

  On the other hand, FIG. 13 shows a rounded-up portion having a shape suddenly changing portion and includes an excessive command value acceleration, so that a steep and large amplitude correction value Cc is created, but eventually has a vibration amplitude of several tens of μm. The position deviation Dc remains.

  As described above, in the conventional iterative learning position control device mounted on the control shaft that performs high-speed synchronous operation, an excessive command acceleration exceeding the physical acceleration limit is generated in the shape sudden change section, so that iterative learning control As a result, there is a problem in that sufficient deviation value convergence due to the above cannot be obtained, and the machining surface becomes rough due to vibration generated by the sharply created correction value. The problem to be solved by the present invention is to provide an iterative learning position control device that reduces the occurrence of vibration by suppressing the learning control effect at the shape sudden change portion.

  The present invention constantly detects the command value acceleration during the learning cycle operation, adds a reduction calculation to the deviation value table according to the command value acceleration at the time of collecting the deviation value, and outputs it to the learning control unit. Thereby, in the shape sudden change part, the correction value created by iterative learning control can be reduced, the learning control effect is suppressed, and the above-mentioned problem can be solved.

  In the iterative learning position control device according to the present invention, by suppressing the learning control effect at the shape sudden change portion, vibration can be reduced and the quality of the machined surface can be improved.

It is a block diagram which shows the structural example of the iterative learning position control apparatus which is embodiment of this invention. It is a figure which shows the structural example of the learning removal control part in embodiment of FIG. It is a figure explaining operation | movement of the learning removal control part of FIG. It is a figure which shows an example of the control result in the shape sudden change part by an iterative learning position control apparatus. It is the schematic of the structural example of the numerical control machine carrying a high-speed synchronous axis. It is a block diagram which shows the structural example of the conventional iterative learning position control apparatus. It is a flowchart explaining the operation | movement of iterative learning control. It is the figure which expressed an example of the command value table of a rope screw in three dimensions. It is a figure which shows an example of the command value and command value acceleration in a rope screw round-up part. It is the figure which expanded and expressed a part of section No. of FIG. It is a figure which shows an example of the command value and command value acceleration in a rope screw complete screw part. It is a figure which shows an example of the control result in a complete screw part. It is a figure which shows an example of the control result in the shape sudden change part by the conventional repetition learning position control apparatus.

  Hereinafter, the best mode for carrying out the present invention will be described with reference to examples (hereinafter referred to as examples). FIG. 1 is a block diagram illustrating the configuration of an iterative learning position control device 1 equipped with an iterative learning control unit 2 according to the present invention. In the following, the parts different from the conventional examples described so far will be mainly described. First, the command value generation unit 50 is set with a command value table corresponding to the spindle rotation angle for a multi-step surface having a constant angular pitch.

  The learning removal control unit 3 reads the position command value Xc corresponding to the spindle rotation angle from the command value generation unit 50, and calculates the command value acceleration A from the amount of change in the position command value Xc for each angle pitch. Further, the learning removal control unit 3 calculates a deviation value Def from the command value acceleration A and the deviation value Def0 of the deviation value table created by the deviation value collection unit 52 based on learning removal control, which will be described later. To the unit 53.

Here, the command value speed V and the command value acceleration A are defined by equations (2) and (3).
In the following equation, n is a table point number, and Xc (n) represents a command value set in the table of the nth point, but the notation of (n) is omitted as before except in the following equation.
V (n) = {Xc (n) -Xc (n-1)} (6N / H) (2)
A (n) = {V (n) −V (n−1)} (6N / H) = {Xc (n) −2Xc (n−1) + Xc (n−2)} (6N / H) ^2. (3)
N is the spindle rotation speed [min ⁻¹ ], and H is the table angle pitch [deg / pitch] described above.

  FIG. 2 is a block diagram illustrating the configuration of the learning removal control unit 3 of the present invention. In the necessity determination unit 4, an acceleration reference value Arej and a learning removal end waiting time Wrej are set in advance. The necessity determination unit 4 first reads the position command value Xc corresponding to the spindle rotation angle from the command value generation unit 50. Next, when the command value acceleration A calculated by the equation (3) is larger than the acceleration reference value Arej, it is determined that learning removal processing is necessary. On the contrary, when the command value acceleration A is smaller than the acceleration reference value Arej, first, after confirming that this state continues for one rotation of the main spindle, the continuation confirmation is further performed during the learning removal end waiting time Wrej. Then, it is determined that the learning removal process is unnecessary.

  A learning removal fade-in / fade-out time Irej is set in advance in the learning effective coefficient calculation unit 5, and a learning removal processing necessity determination is input from the necessity determination unit 4. The learning effective coefficient Crej is an initial value 1, and if it is determined that the learning removal process is unnecessary, the learning effective coefficient Crej is not changed. On the other hand, when it is determined that the learning removal process is necessary, the learning effective coefficient Crej is linearly shifted from 1 to 0 at the fade-in / fade-out time Irej. Even after the learning effective coefficient Crej = 0 is reached, Crej = 0 is maintained while the necessary state of the learning removal process continues.

The learning effective coefficient calculation unit 5 outputs the learning effective coefficient Crej described above to the multiplier 6. The multiplier 6 multiplies the learning effective coefficient Crej and the deviation value Def0 of the deviation value table, calculates the deviation value Def, and outputs it to the learning control unit 53, as shown in Expression (4).
Def = Crej · Def0 (4)

As described above, an excessive command value acceleration is generated at two points per one rotation of the main shaft in the rope screw round-up portion. When | A | ≧ Arej is established at timing t1 in FIG. 3 , learning removal processing is required, and thereafter, the learning effective coefficient Crej is changed from 1 to 0 at the fade-in / fade-out time Irej.

  When the command value acceleration decreases after passing through the rounding-up unit, the learning removal process becomes unnecessary at the timing t2 when the state continuation of | A | <Aej is confirmed during the learning removal end waiting time Wej. Thereafter, Crej returns from 0 to 1 in Irej time. Note that learning removal ON / OFF in the figure indicates the presence or absence of learning removal processing. ON corresponds to the learning effective coefficient Crej = 1, and OFF corresponds to the learning effective coefficient Crej ≠ 1.

  FIG. 4 shows an example of the above-described rope screw, a rounded-up portion having a shape suddenly changing portion (cross section 335) when the learning removal control of the present invention is applied to the cutout portion having the shape suddenly changing portion shown in FIG. FIG. 6 is a diagram showing a command value Xc, a correction value Cc, and a position deviation Dc of (cross section 337). In this example, since the acceleration reference value Arej = 6 [G] and the learning removal fade-in / fade-out time Irej = 3 [sec] are set, the learning effective coefficient Crej≈0. About 6 are selected.

  Here, since this round-up part is a region that requires responsiveness exceeding the physical acceleration limit, vibration suppression is required rather than tracking accuracy. Compared with FIG. 13 which is the control result of the same section by the conventional iterative learning position control device, FIG. 4 shows that the correction value Cc having a smooth and small amplitude is created by the learning removal control, and the amplitude of the position deviation Dc is Although increasing, the vibration tendency can be reduced.

  As described above, in the iterative learning position control device according to the present invention, during the learning operation, the shape suddenly changing portion is always detected from the command value acceleration, and the deviation value at the shape suddenly changing portion is reduced. The obtained correction value can be smoothed and reduced to improve the quality of the machined surface by reducing vibration. In the above-described embodiment, in addition to setting the acceleration reference value Arej to be proportional to the commanded acceleration for the necessity of removal of learning control, the learning removal end waiting time Wrej and the fade-in / fade-out time Irej are set. Although both forms are shown, the present invention is not limited to this, and only the acceleration reference value Arej and the learning removal end waiting time Wrej are set, or only the acceleration reference value Arej is set and waiting for the learning removal end. It is also possible to execute without setting both the time Wrej and the fade-in / fade-out time Irej.

  1,100 Iterative learning position control device 2,101 Iterative learning control unit 3 Learning removal control unit 4 Necessity determination unit 5 Learning effective coefficient calculation unit 6 Multiplier 50 Command value generation unit 51 Subtractor 52 deviation value collection unit, 53 learning control unit, 54 correction value generation unit, 55, 56 adder, 57 position deviation amplifier, 58 subtractor, 59 speed deviation amplifier, 200 turret drive unit.

Claims (3)

A repetitive learning position control device mounted on a control shaft that causes the cutting tool to perform a high-speed synchronous operation in the radial direction according to the rotation angle of the workpiece gripped by the main shaft,
The iterative learning position control device includes:
A deviation value collecting unit that collects a position deviation that is a deviation between a position command value of the cutting tool and a position detection value;
A learning control unit that adds a learning calculation process to the collected deviation value and outputs a correction value for the current learning cycle;
A correction value generator that reads the correction value adopted in the current learning cycle, adds the correction value of the current learning cycle, and determines the correction value to be adopted in the next learning cycle;
The command value acceleration calculated from the position command value corresponding to the spindle rotation angle is compared with a preset acceleration reference value. When the command value acceleration increases, it is determined that learning removal is necessary, and the collected A learning removal control unit that replaces the deviation value with zero and outputs to the learning control unit;
An iterative learning position control device comprising:   In the learning removal control unit, when the command value acceleration decreases and transitions from a state larger than the acceleration reference value to a smaller state, the collected value is collected while a preset learning removal end waiting time elapses. 2. The iterative learning position control device according to claim 1, further comprising a function of continuing zero replacement of the deviation value, determining that learning removal is unnecessary after elapse, and stopping zero replacement of the collected deviation value. .   When it is determined that learning removal is necessary for the learning removal control unit, the collected deviation value is gradually decreased at a preset fade-in / fade-out time, and when learning removal is determined to be unnecessary, 3. The iterative learning position control according to claim 1, further comprising a function of gradually increasing the collected deviation value from zero during the fade-in / fade-out time and outputting the same to a learning control unit. apparatus.