Motor drive control device and computer-readable storage medium

Abstract

A motor drive control device according to the present disclosure, on the basis of at least a target value for instructing a position related to a motor, which drives a shaft of an industrial machine to be controlled, and a control amount fed back from the motor, calculates an operation amount to be output to the motor, inputs the control amount, estimates a correction amount obtained by correcting the operation amount using a correction model, which is a highly accurate machine learning model, inputs the control amount, estimates a limit of the correction amount using a limit model, which is a highly-explainable machine learning model, and clamps the correction amount by the limit.

Description

Motor drive control device and computer-readable storage medium

[0001] The present disclosure relates to a motor drive control device and a computer-readable storage medium.

[0002] One of the means for realizing high-precision servo control is repetitive control (learning control) (see, for example, Patent Document 1). Repetitive control learns correction data for reducing errors at each time step in response to repeatedly issued commands.

[0003] Japanese Patent Application Laid-Open No. 2004-234327

[0004] In normal repetitive control, when the amplitude of the position command, command speed, or command acceleration changes, the amount of correction no longer matches, making the correction data learned up to that point unusable. Furthermore, because correction data must be stored in proportion to the repetition period, there is also the issue that a larger memory area is required as the period becomes longer. This issue arises even in non-repetitive motion.

[0005] Currently, there are methods that use machine learning, such as neural networks, to calculate the amount of correction, which can calculate an appropriate amount of correction even if the processing conditions change.

[0006] However, while machine learning such as neural networks demonstrates high accuracy, it suffers from the problem of low explainability (interpretability) due to the complexity of the models. Machine learning with low explainability has unclear criteria and rationale, making it difficult to guarantee that the correct amount of correction is always output, and there is a possibility of overcorrection.

[0007] In motor control, a mechanism that guarantees the amount of machine learning correction is desired.

[0008] The motor drive control device according to the present disclosure comprises: a feedback control unit that calculates an operation amount to be output to a motor based on a target value that commands the position of a motor that drives a shaft of an industrial machine that is at least the object of control, and a control amount that is fed back from the motor; a correction calculation unit that inputs the control amount and estimates a correction amount by correcting the operation amount using a correction model that is a highly accurate machine learning model; a limit calculation unit that inputs the control amount and estimates a limit of the correction amount using a limit model that is a highly explainable machine learning model; and a clamp unit that clamps the correction amount at the limit.

[0009] 1 is a hardware configuration diagram of a control device equipped with a motor drive control device. FIG. 1 is a block diagram of a motor drive control device of a first embodiment. FIG. 2 is a diagram explaining the operation of the motor drive control device of the first embodiment. FIG. 3 is a block diagram of a motor drive control device of a second embodiment. FIG. 4 is a graph showing changes in the amount of correction when a tool is circularly moved about two orthogonal axes at a certain feed rate. FIG. 5 is a graph showing the relationship between learning data (amount of correction) and limits. FIG. 6 is a diagram explaining the operation of the motor drive control device of the second embodiment. FIG. 7 is a graph showing an example of limits for the amount of correction that change with speed and time. FIG. 8 is a block diagram of a motor drive control device of a fourth embodiment.

[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the following description, components having the same or similar functions will be denoted by the same reference numerals. Duplicate descriptions of those components may be omitted.

[0011] In this application, "based on XX" means "based on at least XX," and includes cases where it is based on other elements in addition to XX. Furthermore, "based on XX" is not limited to cases where XX is used directly, but also includes cases where it is based on XX that has been calculated or processed. "XX" is any element (for example, any information).

[0012] 1 is a hardware configuration diagram of a control device according to the present disclosure. The control device according to the present disclosure is a control device that controls industrial machinery such as machine tools.

[0013] The CPU 11 provided in the control device 1 according to this embodiment is a processor that controls the entire control device 1. The CPU 11 reads a system program stored in the ROM 12 via the bus 22 and controls the entire control device 1 in accordance with the system program. The RAM 13 temporarily stores temporary calculation data, display data, various data input from outside, and the like.

[0014] The nonvolatile memory 14 is configured, for example, by a battery-backed memory or an SSD (Solid State Drive) (not shown), and maintains its stored state even when the power to the control device 1 is turned off. The nonvolatile memory 14 stores programs and data read from an external device 72 via the interface 15, programs and data input via the input device 71, programs and data acquired from the industrial machine 3, and the like. The data stored in the nonvolatile memory 14 may be expanded into the RAM 13 when executed / used. In addition, various system programs such as known analysis programs are written in the ROM 12 in advance.

[0015] The interface 15 is an interface for connecting the CPU 11 of the control device 1 to an external device 72 such as a USB memory, CompactFlash (registered trademark), or SD card. For example, machining programs and various data used to control the industrial machine 3 can be read from the external device 72. Furthermore, machining programs and various data edited within the control device 1 can be stored in the external device 72. A programmable logic controller (PLC) 16 outputs signals to the industrial machine 3 and its peripheral devices (e.g., tool changers, actuators such as robots, sensors attached to the industrial machine 3, etc.) via an I / O unit 17 to control the industrial machine 3 and its peripheral devices (e.g., tool changers, actuators such as robots, sensors attached to the industrial machine 3, etc.) according to a sequence program built into the control device 1. The PLC 16 also receives signals from various switches on an operation panel installed on the main body of the industrial machine 3 and from peripheral devices, performs necessary signal processing, and then passes the signals to the CPU 11.

[0016] The display device 70 displays various data loaded into the memory, data obtained as a result of executing programs, etc., output via the interface 18. An input device 71, which is comprised of a keyboard, pointing device, etc., passes instructions, data, etc., based on operations by an operator to the CPU 11 via the interface 19.

[0017] The interface 20 is an interface for connecting the CPU 11 of the control device 1 to a wired or wireless network 5. The network 5 may communicate using technologies such as serial communication such as RS-485, Ethernet (registered trademark), optical communication, wireless LAN, Wi-Fi (registered trademark), Bluetooth (registered trademark), etc. Computers such as a fog computer 6 and a cloud server 7 are connected to the network 5, and data is exchanged between the network 5 and the control device 1.

[0018] The axis control circuit 30 for controlling the control axes of the industrial machine 3 receives position commands for the control axes from the CPU 11 and outputs commands for the control axes to the servo amplifier 40. The servo amplifier 40 receives these commands and drives the servo motors 50 for the control axes, moving each component of the industrial machine 3 along the respective control axes. Each servo motor 50 has a built-in position detector, and feeds back a position feedback signal from this position detector or an external position detector to the axis control circuit 30. The axis control circuit 30 performs feedback control of the servo motor 50 based on this position feedback signal. Note that while the hardware configuration diagram in FIG. 1 shows only one axis control circuit 30, one servo amplifier 40, and one servo motor 50, in reality, there are as many axis control circuits 30, as there are control axes of the industrial machine 3 to be controlled. For example, to control a typical machine tool with three linear axes, three sets of axis control circuits 30, servo amplifiers 40, and servo motors 50 are provided to move a spindle to which a tool is attached and a workpiece relatively in the three linear axes (X-axis, Y-axis, and Z-axis).

[0019] The spindle control circuit 60 receives a spindle rotation command and outputs a spindle speed signal to a spindle amplifier 61. The spindle amplifier 61 receives this spindle speed signal and rotates a spindle motor 62 of the industrial machine 3 at the commanded rotation speed, thereby driving the spindle. A position coder 63 is connected to the spindle motor 62. The position coder 63 outputs a feedback pulse in synchronization with the rotation of the spindle, and this feedback pulse is read by the CPU 11. Note that while the hardware configuration diagram in FIG. 1 shows only one spindle control circuit 60, one spindle amplifier 61, and one spindle motor 62, in reality, there are as many spindle control circuits 60, spindle amplifiers 61, and spindle motors 62 as there are control axes provided in the industrial machine 3 to be controlled. Furthermore, industrial machines 3 that do not have a spindle may not have these components.

[0020] [First embodiment] Figure 2 is a block diagram of a motor drive control device 2 according to the first embodiment. Each function of the motor drive control device 2 according to this embodiment is realized by the axis control circuit 30 shown in Figure 1. Note that some of the functions of the motor drive control device 2 may be executed by the CPU 11, or a machine learning device (not shown) may be provided.

[0021] The control device 1 of this embodiment includes a control unit 100. The motor drive control device 2 also includes a feedback control unit 200, a correction calculation unit 210, a limit calculation unit 220, and a clamp unit 230. A machining program for controlling the industrial machine 3 is stored in advance on the RAM 13 to the nonvolatile memory 14 of the control device 1.

[0022] The control unit 100 analyzes the machining program and creates command data for controlling the industrial machine 3 equipped with the servo motor 50 and spindle motor 62, and the peripheral devices of the industrial machine 3. The control unit 100 then controls each part of the industrial machine 3 and the peripheral devices based on the created command data. For example, the control unit 100 generates target values ​​for position control based on commands to the servo motor 50 that move the drive unit along each axis of the industrial machine 3, and outputs the target values ​​to the feedback control unit 200 of the motor drive control device 2. Furthermore, for example, the control unit 100 generates data for controlling the rotation of the spindle based on commands to rotate the spindle of the industrial machine 3, and outputs the data to the spindle motor 62. Furthermore, the control unit 100 generates predetermined signals for operating the peripheral devices of the industrial machine 3 based on commands to operate the peripheral devices, and outputs the signals to the PLC 16.

[0023] The feedback control unit 200 performs feedback control by calculating and outputting a manipulated variable to be output to the servo motor 50. The feedback control unit 200 calculates the manipulated variable based on a target value for position control of the servo motor 50 that drives the axis of the industrial machine 3 that is the control target, a control variable related to the position or speed fed back from the servo motor 50, and a correction variable. The controlled variable and / or manipulated variable includes position, speed, acceleration, current, etc. Outputting the manipulated variable from the controlled variable is an existing technology. The feedback control unit 200 of this embodiment further corrects the manipulated variable using the correction variable.

[0024] The correction calculation unit 210 includes a correction model. The correction model is a highly accurate learning model such as a neural network. The correction model receives as input the control amount fed back from the industrial machine 3 and estimates the correction amount used to correct the manipulated variable.

[0025] Calculating the amount of correction can improve the accuracy of motor control. However, while models such as neural networks have high accuracy, they have low interpretability and cannot predict the amount of correction for the manipulated variable.

[0026] The limit calculation unit 220 calculates the limit of the correction amount. The limit calculation unit 220 holds a trained model with high interpretability, such as linear regression or polynomial regression. The trained model inputs control variables such as position, velocity, and acceleration, and outputs the limit of the correction amount.

[0027] The clamping unit 230 monitors the correction amount of the feedback control unit 200, and clamps the correction amount at the limit when the correction amount exceeds the limit.

[0028] The operation of the motor drive control device 2 will be described with reference to Fig. 3. The control unit 100 calculates a target value from a machining program (step S1). Typically, the target value is position information.

[0029] The correction calculation unit 210 receives the control amount and estimates the correction amount (step S2). A highly accurate correction model is used to calculate the correction amount. The limit calculation unit 220 receives the control amount and estimates the limit of the correction amount (step S3). The clamp unit 230 compares the correction amount with the limit (step S4), and if the correction amount exceeds the limit, clamps the correction amount at the limit (step S5).

[0030] The feedback control unit 200 receives the target value, the controlled variable, and the correction variable, and corrects the manipulated variable (step S6). The manipulated variable is a quantity to be added to the controlled object (industrial machine 3). The controlled variable is a value to be fed back from the industrial machine 3. The feedback control unit 200 controls the manipulated variable so that the controlled variable approaches the target value.

[0031] As described above, the motor drive control device 2 of the first embodiment prepares two models for correcting the manipulated variable output by the feedback control unit 200. The first model (correction model) is a highly accurate model such as a neural network, and estimates the correction amount used to correct the manipulated variable. The second model (limit model) is a highly interpretable model, and estimates the limit of the correction amount. The motor drive control device 2 clamps the correction amount at a limit. A highly accurate model has low interpretability, and may not be able to predict the output. Therefore, excessive correction is avoided by estimating the limit using a highly interpretable limit model and clamping the correction amount.

[0032] 4 is a block diagram of a motor drive control device 2 according to a second embodiment. The motor drive control device 2 according to the second embodiment includes a feedback control unit 200, a correction calculation unit 210, a limit calculation unit 220, a clamp unit 230, a learning data storage unit 240, a correction model creation unit 250, and a limit model creation unit 260.

[0033] A learning data storage unit 240 is provided on the RAM 13 to the non-volatile memory 14 of the control device 1. The learning data storage unit 240 stores learning data of a correction model. The learning data of the correction model is, for example, a pair of a target value and a correction amount. The target value and the correction amount include position, velocity, acceleration, etc. Note that the above-mentioned learning data is an example and is not limited to this. Furthermore, it is also possible to calculate velocity by differentiating position, or to calculate acceleration by differentiating velocity.

[0034] The correction model creation unit 250 creates a correction model that corrects the operation amount of the motor of the industrial machine 3. The correction model creation unit 250 learns a correction amount that reduces the deviation, for example, based on the deviation, which is the difference between the target value during repeated operation and the control amount fed back from the industrial machine 3. The method of creating a correction model is an existing technology, so a detailed description will be omitted.

[0035] The manipulated variable correction model is a highly accurate machine learning model such as a neural network. Complex machine learning models such as neural networks input multiple variables and numerically derive the relationships behind the variables. While such models are highly accurate, they suffer from the problem of low interpretability. With low interpretability, the output cannot be predicted, which can lead to over-correction.

[0036] The limit model creation unit 260 creates a limit model that outputs a limit of the correction amount. The limit model uses a model that has relatively low accuracy but high interpretability, such as linear regression or polynomial regression. Note that the learning data may be read from the learning data storage unit 240, as with the correction model creation unit 250. Learning data may also be collected by specially operating the industrial machine 3.

[0037] The method for creating a limit model will be described. First, the method for collecting learning data will be described. To collect learning data for a limit model, the same operation is repeated multiple times while changing the feed rate. For example, the same operation is repeated 20 times at a feed rate of 2000 mm / s, then 20 times at a feed rate of 2100 mm / s, and so on, to collect learning data. Figure 5 shows the change in the amount of compensation when the tool is circularly moved about two orthogonal axes at a certain feed rate. The amount of compensation drops sharply just before time 200 ms and rises sharply around time 1100 ms. For example, in a circular movement about the X and Y axes, when the direction is reversed, a sudden increase or decrease in the amount of compensation (quadrant protrusion) occurs, as shown in Figure 5. The learning data storage unit 240 stores the amount of compensation when the same operation is performed multiple times at the same feed rate.

[0038] The black circles in Fig. 6 represent the learning data. The curve in Fig. 6 is a fourth-order polynomial fitted to the learning data, resulting in the graph in Fig. 6. The white circles in Fig. 6 are a graph of the estimated maximum value of the correction amount at each feed rate, calculated from the fourth-order polynomial, plus a margin.

[0039] Although not shown, in this example, two limit models with different polarities (positive or negative) are learned. For the negative limit, a graph is created by subtracting a margin from the estimated minimum correction amount. The graph in FIG. 5 plots the positive limit (upper limit) and negative limit (lower limit) of the correction amount at a certain speed. The limits change according to changes in speed.

[0040] The clamp unit 230 receives the limit from the limit calculation unit 220, and clamps the correction amount at the limit if the correction amount calculated by the correction calculation unit exceeds the limit. Clamping the correction amount prevents excessive correction. The feedback control unit 200 corrects the manipulated variable with the correction amount. The industrial machine 3 drives the motor with the manipulated variable received from the feedback control unit 200.

[0041] The operation of the motor drive control device 2 will be described with reference to Figure 7. The correction model creation unit 250 creates a correction model that estimates a correction amount from a control amount (step S10). The learning data storage unit 240 stores learning data for creating the correction model. The learning data is, for example, a set of a target value and a correction amount.

[0042] The limit model creation unit 260 creates a limit model, which is a limit of the correction amount, from the control amount (step S11). The learning data stored in the learning data storage unit 240 can also be used to create the limit model. The motor drive control device 2 controls the motor using the created correction model and limit model (step S12). The control method is the same as in the first embodiment.

[0043] As described above, the motor drive control device 2 of the second embodiment includes a correction model creation unit 250 that acquires learning data while operating the industrial machine 3 and creates a correction model for the feedback manipulated variable. Since the industrial machine 3 often repeats the same operation, during repetitive operations, the industrial machine 3 is actually operated while correcting the manipulated variable, and a correction model can be created that sufficiently converges the deviation between the target value and the controlled variable. The correction model uses a model that is highly accurate but has relatively low interpretability, such as a neural network. A model with low interpretability may result in excessive correction, so a limit model is used to limit the correction amount and avoid excessive correction.

[0044] Third Embodiment As a third embodiment, an application example of a limit model will be described. The limit model can be applied to correcting feed irregularities due to motor torque ripple, reducing vibrations when a robot is stopped, and the like. For correcting feed irregularities due to motor torque ripple and vibrations when a robot is stopped, learning of the limit model involves collecting learning data. In collecting the learning data, for example, as in the first embodiment, the same operation is repeated multiple times while changing the feed speed, and correction amounts at various feed speeds are collected.

[0045] The amount of compensation changes over time. The maximum, minimum, and extreme values ​​of the amount of compensation at a certain feed rate are obtained as samples. The obtained samples are then trained using a highly explanatory model to create a limit model.

[0046] Multiple limits may be combined. For example, limit models for the motor torque ripple and quadrant projection may be created, and different limits may be output from the two limit models. A new limit may then be created from the different limits, or one of multiple limits may be selected.

[0047] There is also a method of using data other than minimum and maximum values ​​as learning data for creating a limit model. In FIG. 8, the correction amount limit is calculated over the entire graph. In this case, for example, data on the correction amount at a certain feed rate and a certain time is acquired as a sample. Then, a limit model is created for each time of the acquired sample. This makes it possible to calculate the motor feed rate and the limit for each time. Note that in this embodiment, speed is used as learning data, but position or acceleration may also be used as learning data. Also, acceleration obtained by differentiating speed may be used as learning data.

[0048] [Fourth embodiment] Figure 9 is a block diagram of a motor drive control device 2 of the fourth embodiment. The motor drive control device 2 of the fourth embodiment includes an operation amount clamping unit 270. The operation amount clamping unit 270 stores information related to the specifications of the motor and industrial machine 3 to be controlled. The information related to the specifications includes catalog specifications published by the product manufacturer and actual measured values ​​obtained by actually operating the motor or industrial machine 3. The operation amount clamping unit 270 inputs an operation amount, and clamps the operation amount if the operation amount exceeds the specifications of the motor or industrial machine 3.

[0049] In the motor drive control device 2 of the fourth embodiment, the operation amount is limited based on the specifications of the motor or industrial machine 3 to be controlled, which provides even greater safety than clamping using a limit model output from a machine learning model.

[0050] Fifth Embodiment The motor drive control device 2 can also be realized by simulation. The control device 1 and industrial machine 3 can be reproduced in a virtual space, and a limit can be placed on the amount of correction. It is also possible to acquire real-time data and control the motor in the virtual space.

[0051] Although the embodiments of the present disclosure have been described in detail above, the present disclosure is not limited to the individual embodiments described above. Various additions, substitutions, modifications, partial deletions, etc. are possible in these embodiments without departing from the gist of the invention or the idea and intent of the present disclosure derived from the content described in the claims and their equivalents. For example, in the above-described embodiments, the order of each operation and the order of each process are shown as examples and are not limited to these. The same applies when numerical values ​​or mathematical expressions are used in the description of the above-described embodiments.

[0052] The following are supplementary notes related to embodiments of the present disclosure. (Supplementary Note 1) A motor drive control device (2) includes: a feedback control unit (200) that calculates an operation amount to be output to a motor based on a target value that commands a position of a motor that drives a shaft of an industrial machine that is a control target, and a control amount fed back from the motor; a correction calculation unit (210) that receives the control amount and estimates a correction amount by correcting a deviation of the operation amount using a correction model that is a highly accurate machine learning model; a limit calculation unit (220) that receives the control amount and estimates a limit of the correction amount using a limit model that is a highly interpretable machine learning model; and a clamping unit (230) that clamps the correction amount at the limit. (Supplementary Note 2) The motor drive control device (2) includes: a learning data storage unit (240) that actually drives the industrial machine, collects and stores learning data for learning the correction model, and a correction model creation unit (250) that creates a correction model that receives the control amount and estimates a correction amount by correcting a deviation of the operation amount based on the learning data. (Supplementary Note 3) The motor drive control device (2) comprises a learning data storage unit (240) that actually drives the industrial machine and collects and stores learning data for learning the correction model, and a limit model creation unit (260) that creates a limit model that estimates a limit for the correction amount based on the learning data stored in the learning data storage unit. (Supplementary Note 4) The motor drive control device (2) comprises an operation amount clamp unit (270) that corrects the operation amount based on information related to the motor specifications. (Supplementary Note 5) The limit calculation unit (220) estimates a limit from limit models with different polarities. (Supplementary Note 6) The limit calculation unit (220) estimates multiple limits from multiple different limit models. (Supplementary Note 7) The correction model is a neural network. (Supplementary Note 8) The correction model is at least one of linear regression and polynomial regression.(Supplementary Note 9) The computer-readable storage medium (12, 13, 14) stores instructions for causing one or more processors (11) to execute the following processes: calculate an operation amount to be output to a motor based on a target value that commands a position related to at least a motor that drives a shaft of industrial machinery that is a control target and a control amount fed back from the motor; input the control amount; estimate a correction amount obtained by correcting the operation amount using a correction model that is a highly accurate machine learning model; input the control amount; estimate a limit for the correction amount using a limit model that is a highly interpretable machine learning model; and clamp the correction amount at the limit.

[0053] REFERENCE SIGNS LIST 1 Control device 2 Motor drive control device 3 Industrial machine 11 CPU 12 ROM 13 RAM 14 Non-volatile memory 100 Control unit 200 Feedback control unit 210 Correction calculation unit 220 Limit calculation unit 230 Clamp unit 240 Learning data storage unit 250 Correction model creation unit 260 Limit model creation unit 270 Operation amount clamp unit

Claims

1. A motor drive control device comprising: a feedback control unit that calculates an operation amount to be output to a motor based on a target value that commands the position of a motor that drives a shaft of at least an industrial machine that is the object of control, and a control amount that is fed back from the motor; a correction calculation unit that inputs the control amount and estimates a correction amount that corrects for a deviation of the operation amount using a correction model that is a highly accurate machine learning model; a limit calculation unit that inputs the control amount and estimates a limit of the correction amount using a limit model that is a highly explainable machine learning model; and a clamp unit that clamps the correction amount at the limit.

2. A motor drive control device as claimed in claim 1, comprising: a learning data storage unit that actually drives the industrial machine and collects and stores learning data for learning the correction model; and a correction model creation unit that creates a correction model that inputs a control variable and estimates a correction amount that corrects for a deviation in the manipulated variable based on the learning data.

3. A motor drive control device as claimed in claim 1, comprising: a learning data storage unit that actually drives the industrial machine and collects and stores learning data for learning the correction model; and a limit model creation unit that creates a limit model that estimates a limit for the correction amount based on the learning data stored in the learning data storage unit.

4. The motor drive control device according to claim 1, further comprising an operation amount clamping unit that corrects the operation amount based on information relating to the motor specifications.

5. The motor drive control device according to claim 1, wherein the limit calculation unit estimates the limits from limit models of different polarities.

6. The motor drive control device according to claim 1, wherein the limit calculation unit estimates a plurality of limits from a plurality of different limit models.

7. The motor drive control device according to claim 1, wherein the correction model is a neural network.

8. The motor drive control device according to claim 1, wherein the correction model is at least one of linear regression and polynomial regression.

9. A computer-readable storage medium storing instructions to cause one or more processors to execute the following processes: calculate an operation amount to be output to a motor based on a target value that commands the position of at least a motor that drives an axis of industrial machinery that is the object of control and a control amount fed back from the motor; input the control amount and estimate a correction amount by correcting the operation amount using a correction model that is a highly accurate machine learning model; input the control amount and estimate a limit for the correction amount using a limit model that is a highly interpretable machine learning model; and clamp the correction amount at the limit.

Images