WO2018151215A1 - Dispositif et procédé de commande - Google Patents

Dispositif et procédé de commande Download PDF

Info

Publication number
WO2018151215A1
WO2018151215A1 PCT/JP2018/005267 JP2018005267W WO2018151215A1 WO 2018151215 A1 WO2018151215 A1 WO 2018151215A1 JP 2018005267 W JP2018005267 W JP 2018005267W WO 2018151215 A1 WO2018151215 A1 WO 2018151215A1
Authority
WO
WIPO (PCT)
Prior art keywords
control
learning
control parameter
machine
adjustment unit
Prior art date
Application number
PCT/JP2018/005267
Other languages
English (en)
Japanese (ja)
Inventor
良平 北吉
勝 足立
Original Assignee
株式会社安川電機
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 株式会社安川電機 filed Critical 株式会社安川電機
Priority to JP2018568603A priority Critical patent/JP6774637B2/ja
Publication of WO2018151215A1 publication Critical patent/WO2018151215A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B11/00—Automatic controllers
    • G05B11/01—Automatic controllers electric
    • G05B11/36—Automatic controllers electric with provision for obtaining particular characteristics, e.g. proportional, integral, differential
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric

Definitions

  • the disclosed embodiment relates to a control device and a control method.
  • Patent Document 1 describes a method of automatically setting a control parameter based on an evaluation result of each state quantity detected when a predetermined adjustment sequence operation is executed.
  • the present invention has been made in view of such problems, and an object thereof is to provide a control device and a control method capable of improving the convenience of the control device.
  • a deviation between a reference command and a control amount output by an external control target is input to a controller that controls with a predetermined control parameter, and the control is performed.
  • a control device is applied that includes a feedback control unit that controls the control target with an operation amount output by a device, and an adjustment unit that adjusts the control parameter based on learning content in a machine learning process.
  • an operation that a deviation between a reference command and a control amount output by an external control target is input to a controller that is controlled by a predetermined control parameter, and is output by the controller.
  • a control method is executed that controls the control object by a quantity and adjusts the control parameter based on the learning content in the machine learning process.
  • the convenience of the control device can be improved.
  • FIG. 1 illustrates an example of a schematic system block configuration of a machine control system including a control device according to the present embodiment.
  • This machine control system is a system that controls the linear movement of the slider by controlling the drive of a rotary motor.
  • the machine control system 1 includes a host control device 2, a motor control device 3, a rotary motor 4, and a drive machine 5.
  • the host control device 2 is composed of, for example, a general-purpose personal computer equipped with a CPU, ROM, RAM, operation unit, display unit, and the like (not shown).
  • the host controller 2 generates a position command for positioning a slider of the drive machine 5 described later at a desired position based on various settings and commands input from the operator via the operation unit, and the motor controller 3 is output.
  • the motor control device 3 generates a torque command based on the position command output from the host control device 2 and outputs it to the rotary motor 4. At this time, the motor control device 3 performs position feedback control based on a detection position output from an encoder 41a (described later) included in the rotary motor 4.
  • the motor control device 3 includes a feedback control unit 31 and an adjustment unit 32. The motor control device 3 corresponds to the control device described in each claim.
  • the feedback control unit 31 is a calculation unit that generates a torque command based on the position command and the detected position described above.
  • the control block of the feedback control unit 31 will be described in detail later with reference to FIG.
  • the adjustment unit 32 is a processing unit that adjusts a control parameter (described later) used in the torque command calculation process in the feedback control unit 31 in an appropriate situation based on the position command, the detected position, and the torque command.
  • the processing content of the adjustment unit 32 will be described later in detail.
  • the processing in the feedback control unit 31, the adjustment unit 32, and the like described above is not limited to the example of sharing of these processing, and for example, processing by a smaller number of processing units (for example, one processing unit). Or may be processed by a further subdivided processing unit.
  • the motor control device 3 may be implemented in software by a program executed by a CPU 901 (see FIG. 12) described later, or a part or all of the motor control device 3 may be an ASIC, FPGA, other electric circuit, or the like (neuromorphic). It may be implemented in hardware by an actual device such as a chip.
  • the rotary motor 4 is, for example, a synchronous three-phase AC motor, and integrally includes an encoder 41a that outputs the rotational position of the output shaft of the rotary motor 4 as a detection position.
  • the drive machine 5 is an actuator that converts the rotational position, which is the shaft output of the rotary motor 4 in the illustrated example, into the translational position of the slider 52 via the feed screw 51.
  • the drive machine 5 includes a coupling 53, a feed screw 51, and a slider 52.
  • a feed screw 51 is connected to the output shaft of the rotary motor 4 via a coupling 53.
  • the slider 52 is a pedestal on which an object can be placed above.
  • the feed screw 51 is screwed to the lower part of the slider 52.
  • the slider 52 is fed by the feed screw 51 and is along a direction corresponding to the rotational direction of the rotary motor 4 (left and right direction in the figure). Driven to move straight.
  • the motor control device 3 controls the position feedback of the torque command output to the rotary motor 4 so that the position of the slider 52 follows the position command based on the detection position output from the encoder 41a. Take control.
  • the output position (detected position in the above embodiment) and the output speed, which are control amounts of the servo motor (the rotary motor 4 in the above embodiment), are fed back. Then, a deviation from a reference command (position command or speed command in the case of the above embodiment) that is a target value is input to a controller (described later), and an operation amount (torque in the case of the above embodiment) output from the controller. In many cases, feedback control is performed to control the servo motor by command.
  • semi-closed feedback control that feeds back the output position and output speed of the servo motor itself connected to the drive machine as a controlled variable, or the final output position and output of the drive machine Position control and speed control of the drive machine (for example, the slider 52) can be performed by full-closed feedback control that feeds back speed as a control amount.
  • the controller provided in the motor control device 3 as described above outputs an operation amount by arithmetic processing using a large number of control parameters, and includes the mechanical constants of servo motors and driving machines to be controlled, and disturbances that can be added. It is desirable to appropriately set control parameters corresponding to the above.
  • control target servo motor or drive machine
  • high control accuracy positioning accuracy in the case of the above embodiment
  • short settling time are required. It is necessary to strictly adjust the control parameters in accordance with individual differences (parts manufacturing error, assembly error, etc.), operating conditions, aged deterioration status, and usage environment status of the control target.
  • the motor control device 3 of the present embodiment has an adjustment unit 32 that adjusts the control parameters of the controller based on the learning content in the machine learning process.
  • the adjustment unit 32 can set appropriate control parameters according to the learning content. That is, autonomous control parameter adjustment by the motor control device 3 itself can be mechanically and automatically performed without depending on human hands. For this reason, it is possible to save labor costs and time costs with respect to adjustment of control parameters in the motor control device 3, and to ensure uniform and stable control accuracy regardless of the skill difference between engineers.
  • FIG. 2 is a control block diagram showing an outline of a control parameter adjustment method in the motor control device 3.
  • a deviation between a reference command r input from the outside and a control amount y output from the control target P is input to the controller K ( ⁇ ).
  • the controller K ( ⁇ ) outputs an operation amount u through arithmetic processing based on the control parameter ⁇ , and the control target P is controlled by the operation amount u.
  • the reference command r, the operation amount u, the control target P, and the control amount y correspond to the position command, the torque command, the rotary motor 4 and the drive machine, and the detection position in the above-described embodiment (controller K). (The correspondence of ( ⁇ ) will be described later).
  • the controller K ( ⁇ ) is a predetermined calculation model using a variable control parameter ⁇ , it can be said that its characteristics and behavior are completely known.
  • the control object P even if a rough model based on the design value is known, as described above, individual differences (part manufacturing error, assembly error, etc.) and operation conditions of the control object P itself, It can be said that the actual and detailed characteristics and behavior are unknown due to changes in aging and usage environment conditions. That is, it is necessary to adjust the control parameter so as to adapt to a specific machine constant or disturbance in the actually applied control object P.
  • the reference command r, the operation amount u, the control amount y, and other operation conditions c at the time of various sequence operations and disturbance addition are associated with the control parameters set at that time.
  • the attached data set is recorded and stored in the database DB.
  • the adjustment unit Cal executes a machine learning process based on these data sets, so that the adjustment is appropriate for the machine constant and disturbance of the controlled object P estimated from the correspondence between the reference command r and the controlled variable y.
  • the control parameter ⁇ new can be set.
  • FIG. 3 shows an example of a transmission / reception relationship of information between the feedback control system including the feedback control unit 31 and the adjustment unit 32.
  • the illustrated feedback control system is represented by a transfer function type control block.
  • a feedback control unit 31, a motor / drive machine 60, an adjustment unit 32, and a database DB are shown.
  • the feedback control unit 31 includes a subtractor 34, a position loop gain Kp, a subtractor 35, an integrator (1 / T ⁇ s), an adder 36, a speed loop gain Kv, and a speed calculator 37.
  • the subtracter 34 subtracts the detected position detected from the motor / drive machine 60 from the position command input from the outside, and outputs a position deviation between them.
  • a speed command is output by multiplying the position deviation by the position loop gain Kp.
  • the position loop gain Kp functions as a position control controller (the controller K ( ⁇ ) in FIG. 2), and is configured to perform so-called position proportional control.
  • the subtracter 35 subtracts the detected speed output from the speed calculator 37 described later from the speed command and outputs a speed deviation between them.
  • the integrator (1 / T ⁇ s) performs an integration operation based on the speed loop integration time constant T with respect to the speed deviation, and the adder 36 adds the output of the integrator (1 / T ⁇ s) and the speed deviation. And output.
  • a torque command is output by multiplying the added output by the speed loop gain Kv.
  • the integrator (1 / T ⁇ s) and the speed loop gain Kv also function as a controller for speed control (the controller K ( ⁇ ) in FIG. 2), so-called speed integral proportionality. It is configured to perform control.
  • the speed calculator 37 is a calculator that outputs a detected speed (output speed of the rotary motor 4) based on the detection position detected from the motor / drive machine 60. Specifically, the speed calculator 37 is configured by a differentiator s. Good.
  • the motor / drive machine 60 corresponds to the rotary motor 4 and the drive machine 5 in FIG. 1, and includes a rotor of the rotary motor 4 and a movable part of the drive machine 5. It is a mathematical model based on the moment of inertia J of the whole movable mechanism connected. Although not particularly shown, when further multiplied by the moment of inertia J 0 of the rotor of the rotary motor 4 to the speed loop gain Kv is a mathematical model in the motor-driving machine 60 defined by the moment of inertia ratio Also good.
  • the feedback control system of the present embodiment configured by the feedback control unit 31 and the motor / drive machine 60 has a double loop configuration of a feedback loop of a position proportional control system and a feedback loop of a speed integral proportional control system.
  • P-IP control so-called P-IP control
  • a current control unit that outputs a drive current by PWM control to the motor / drive machine 60 based on a torque command and a feedback loop of a current control system provided therein are simplified. Omitted for that.
  • the values of the position loop gain Kp, the speed loop integration time constant T, and the speed loop gain Kv corresponding to the control parameter ⁇ are temporarily set.
  • the position command (corresponding to the reference command r), the torque command (corresponding to the operation amount u), and the detected position (corresponding to the control amount y) when a predetermined adjustment sequence operation is executed are recorded.
  • each of these position command, torque command, and detection position (hereinafter collectively referred to as state quantity data) is obtained when the adjustment sequence operation is executed as shown in FIG. It is recorded as time series pattern data in which instantaneous values are sequentially recorded in the same temporal series. In the example shown in FIG.
  • time series pattern data thickness of each state quantity when the sequence operation is executed by the positioning operation in which the position command is increased from 0 to a predetermined position by inching control.
  • d shown in FIG. 4D represents a position deviation, which will be described later.
  • the sequence operation may be executed by speed command control or torque command control.
  • a speed command is input as it is with a trapezoidal time-series pattern having a section where the acceleration at the time of acceleration / deceleration is constant and the steady speed is constant, and the sequence operation is executed. May be.
  • execute a sequence operation by inputting a polynomial that guarantees continuity of acceleration at the time of acceleration / deceleration and constant speed switching, a speed command expressed by a trigonometric function or an exponential function, etc. You may let them.
  • this database DB for each of various adjustment sequence operations, the time series pattern data of each state quantity data and the control parameters Kp, T, Kv temporarily set during the adjustment sequence operation are associated with one adjustment. A part learning data set is created and stored (see the dashed arrow in FIG. 3).
  • this database DB may be comprised by the memory
  • the adjustment unit 32 is subjected to machine learning (data learning) by batch learning (offline learning) using these adjustment unit learning data sets.
  • each operation sequence operation of the machine control system 1 is executed in a state where each control parameter is temporarily set.
  • the time series pattern data of the state quantity data is input to the adjustment unit 32, and the adjustment unit 32 outputs optimal control parameters Kp, T, and Kv to the feedback control unit 31.
  • high control accuracy in the feedback control system at the time of operation can be ensured (see dotted arrows in FIG. 3).
  • FIG. 5 shows an example of a schematic model configuration of the neural network of the adjustment unit 32 when deep learning is applied.
  • the neural network of the adjustment unit 32 corresponds to the position command, torque command, and each time series pattern data of the detected position, which are state data input from each unit. It is designed to output control parameters Kp, T, and Kv that appropriately correspond to the mechanical constants of the motor / driving machine 60 estimated from the relationship, particularly the correspondence between the position command and the detected position.
  • values and signals for example, values and signals representing features
  • control parameters Kp, T, Kv output from each output node of the adjustment unit 32 are output by multi-value output (continuous value) by regression problem processing.
  • the setting process of these control parameters Kp, T, Kv is based on the learning content in the machine learning process in the learning phase of the adjustment unit 32. That is, the neural network of the adjustment unit 32 learns a feature amount that represents a correlation between the state quantity data and an appropriate control parameter Kp, T, Kv.
  • the adjustment unit 32 is made to learn by so-called supervised learning using the part learning data set.
  • the adjustment unit learning data set used here is, for example, as shown in FIG. 6, each state quantity data (time-series pattern data) and evaluation values when various adjustment sequence data are executed, and provisional setting at that time. This is a data set in which the control parameters Kp, T, and Kv are associated with each other.
  • the evaluation value in the illustrated example is an index indicating the evaluation of the responsiveness in the sequence operation of the feedback control system to which the control parameters Kp, T, and Kv of the corresponding data set are applied.
  • This evaluation value is, for example, the positional deviation d shown in FIG. 4 (d), the vibration amplitude at the time of overshoot or undershoot, the settling time until the transition period ends, until the control amount follows the command. May be obtained comprehensively based on the state quantity data of the data set such as the rise time of the torque, the magnitude of the ripple of the torque command value, and the power consumption value that can be estimated from the torque command value and the speed.
  • a torque command combining a plurality of sine waves may be input, and the phase margin, gain margin, and sensitivity function values calculated from the responses may be used as indices.
  • this evaluation value is represented by a two-stage index of “high” and “low”, but it may be represented by an index of three or more stages, or may be represented by a numerical value.
  • the input data and the output layer of the neural network of the adjustment unit 32 are used by using teacher data in a combination of the state quantity data as input data and the control parameter as output data.
  • Learning is performed by a so-called back-propagation process or the like that adjusts the weighting coefficient of each edge connecting the nodes so that the relationship between the nodes is established.
  • this back-propagation process only a data set having a particularly high evaluation value may be extracted from a large number of data sets, and only this may be used as teacher data to adjust the weighting coefficient of each edge.
  • all data sets may be used as teacher data, and adjustment may be performed so that the weighting coefficient of each edge is increased or decreased according to each evaluation value.
  • the processing accuracy is improved by using various known learning methods such as so-called auto encoder, limited Boltzmann machine, dropout, noise addition, and sparse regularization. Also good.
  • the learning phase of the adjustment unit 32 corresponds to the machine learning process described in each claim.
  • the machine learning algorithm of the adjustment unit 32 is not limited to the one based on the illustrated deep learning, and other machine learning algorithms (not shown in particular) using a support vector machine, a Bayesian network, etc. are applied. May be. Even in this case, the basic configuration of outputting a control parameter appropriately corresponding to the input state quantity data is the same.
  • FIG. 7 shows a flowchart of a processing procedure when the CPU 901 (see FIG. 12 described later) of the motor control device 3 executes a machine learning process by data learning in the example of the present embodiment.
  • the data learning process shown in this flow starts when, for example, a command is input from the host control device 2 to execute a machine learning process.
  • step S5 the CPU 901 temporarily sets each control parameter (Kp, T, Kv in this example).
  • the temporary setting of each control parameter is set by a combination of random values within an appropriate settable range.
  • artificial design values that are considered appropriate for the feedback control system to be applied may be set by individually increasing or decreasing the values.
  • step S10 the CPU 901 selects one of various adjustment sequence operations prepared in advance.
  • step S15 the process proceeds to step S15, and the CPU 901 inputs the position command of the adjustment sequence selected in step S10 to the feedback control unit 31, and causes the motor / drive machine 60 to execute the corresponding sequence operation.
  • step S20 the process proceeds to step S20, and the CPU 901 records the state command data of the position command, torque command, and detected position during the execution of the adjustment sequence operation in step S15.
  • step S25 the CPU 901 creates one adjustment unit learning data set with each state quantity data recorded in step S20 and each control parameter temporarily set in step S5.
  • the data set also includes evaluation values for responsiveness obtained based on the state quantity data and other methods.
  • step S30 the CPU 901 determines whether or not a data set has been created for various types of adjustment sequence operations prepared in advance. If an adjustment sequence operation for which a data set has not yet been created remains, the determination is not satisfied, and the procedure returns to step S10 and the same procedure is repeated.
  • step S35 the CPU 901 determines whether or not to create the adjustment unit learning data set by repeating the various adjustment sequence operations while changing the temporarily set control parameter. In other words, it is determined whether or not to end the creation of the data set.
  • the creation of the data set is continued by temporarily setting a new control parameter, the determination is not satisfied, and the process returns to step S5 to temporarily set a new control parameter and repeat the same procedure.
  • step S40 the CPU 901 executes data learning of the adjustment unit 32 using the created adjustment unit learning data set. Then, this flow ends.
  • the machine learning process by the data learning process in the offline learning (batch learning) as described above may be performed before the operation of the machine control system 1 is started, or the aging of the rotary motor 4 and the driving machine 5 is performed. It may be performed when necessary after the start of operation for the purpose of improving the control accuracy in accordance with the deterioration state and the change in the use environment.
  • the machine control system 1 includes the adjustment unit 32 that adjusts the control parameters of the controllers included in the feedback control unit 31 based on the learning content in the machine learning process. .
  • the adjustment unit 32 can set appropriate control parameters according to the learning content. That is, autonomous control parameter adjustment by the motor control device 3 itself can be mechanically and automatically performed without depending on human hands. For this reason, it is possible to reduce the human cost and the time cost with respect to the adjustment of the control parameter in the motor control device 3, and it is possible to ensure a uniform and stable control accuracy regardless of the skill difference among the engineers. As a result, the convenience of the control device can be improved.
  • this embodiment demonstrated the case where a control parameter was adjusted only with respect to a feedback control system, it is not restricted to this.
  • the method of the adjusting unit 32 of the present embodiment may be applied to the control system including the feedforward control system to adjust the control parameters such as the feedforward gain.
  • the same method may be applied to the adjustment of various control parameters used therein.
  • the method is not limited to the so-called motion control as in the present embodiment, and the same method may be applied to the adjustment of the control parameters for so-called process control.
  • the adjustment unit 32 sets a control parameter corresponding to a mechanical constant (or disturbance described later) of the motor / drive machine 60 estimated from the correspondence between the position command and the detected position.
  • the adjustment part 32 is appropriate for the machine constant (or disturbance to be described later) of the motor / drive machine 60 (control target P) specifically estimated for each controller of the feedback control part 31. Control parameters can be set.
  • the adjustment unit 32 is machine-learned. Accordingly, it is possible to implement a specific machine learning process by supervised learning using a data set for the adjustment unit 32. Further, by performing online learning of data, it is possible to set appropriate control parameters at all times in response to changes in aging deterioration conditions and usage environment conditions during operation of the machine control system 1 in particular. In addition, environmental data (such as the ambient temperature and attitude of the drive machine 5) that may affect the control accuracy in the sequence operation of the motor / drive machine 60 is detected by a separate sensor and included in the adjustment unit learning data set. 32 may learn data.
  • the position command, torque command, and detection position of the adjustment unit learning data set are time-series pattern data during the same sequence operation.
  • the adjustment unit 32 can perform machine learning corresponding to the state of change of each state quantity data, and can set a highly versatile control parameter appropriately corresponding to various sequence operations.
  • the same feedback control unit 31 can be changed to various motors / drive machines 60 to create an adjustment unit learning data set and cause the same adjustment unit 32 to learn, so that the adjustment unit 32 can have various motors. -Adjustment of the control parameter corresponding to the drive machine 60 flexibly becomes possible, and the versatility of the feedback control unit 31 is improved.
  • the adjustment unit learning data set includes an evaluation value, and the machine learning process performs machine learning based on the evaluation value.
  • the machine learning process of the adjustment unit 32 the data learning of each adjustment unit learning data set can be machine-learned according to the evaluation value, and the control accuracy of the feedback control unit 31 can be further improved.
  • control parameter includes at least one of the position loop gain Kp, the speed loop gain Kv, and the speed loop integration time constant T (and the inertia moment ratio of the motor / drive machine 60).
  • control parameters such as model following control gain, torque command filter time constant, notch filter frequency, notch filter Q value, or notch filter depth may be included.
  • ⁇ Modification 1 When the machine learning process is performed by reinforcement learning>
  • the machine learning process of an appropriate control parameter corresponding to the state quantity data is performed by data learning.
  • the present invention is not limited to this.
  • the neural network of the adjustment unit 32 may be learned by a machine learning process of reinforcement learning processing as illustrated in FIG. Note that the reinforcement learning process of the present modification shown in the drawing is performed based on a so-called Q learning method.
  • step S105 the CPU 901 temporarily sets each control parameter.
  • the initial temporary setting of each control parameter is set with an artificial design value that is considered appropriate for the feedback control system to be applied.
  • step S110 the CPU 901 searches for a new control parameter.
  • this new control parameter from the control parameter set as appropriate until that time (the so-called Q value described later is maximum), a control parameter that is corrected in a random direction with a small probability, or For most other probabilities, control parameters that are considered appropriate are left as they are.
  • step S115 the CPU 901 executes a predetermined adjustment sequence operation with the control parameter searched and set in step S110.
  • step S120 the process proceeds to step S120, and the CPU 901 records the state command data of the position command, torque command, and detected position during the execution of the adjustment sequence operation in step S115.
  • step S125 the CPU 901 evaluates the current adjustment sequence operation based on each state quantity data recorded in step S120, and calculates a reward according to this evaluation.
  • the CPU 901 performs reinforcement learning on the neural network of the adjustment unit 32 by so-called Q learning based on the control parameter searched and set in step S110 and the reward calculated in step S125.
  • the control parameter is learned so that the Q value indicating the effectiveness of the set control parameter is maximized.
  • Q-learning a known method may be used, and detailed description thereof is omitted here.
  • step S135 the CPU 901 determines whether or not machine learning has been performed with a sufficient number of trials for the neural network of the adjustment unit 32. If the machine learning is insufficient and the reinforcement learning process is not terminated, the determination is not satisfied, and the same procedure is repeated by returning to step S110.
  • the machine learning process performs machine learning of control parameters by reinforcement learning (Q learning or the like) based on at least one of a position command, a torque command, and a detected position. .
  • Q learning reinforcement learning
  • reinforcement learning is reinforcement learning based on a reward for setting a control parameter during a predetermined adjustment sequence operation.
  • the adjustment unit 32 can set control parameters with high responsiveness that appropriately correspond to various sequence operations.
  • the adjustment unit 32 sets the control parameter from the viewpoint of improving the responsiveness in the sequence operation of the feedback control system.
  • the present invention is not limited to this.
  • the adjustment unit 32 may set the control parameter from the viewpoint of improving robustness against disturbance added to the control target (particularly dynamically added disturbance).
  • the adjustment unit learning data including the time series pattern data of each state quantity data recorded at the time of the same disturbance addition that can be assumed when the machine control system 1 is operated. Create sets and learn data.
  • the adjustment unit learning data set in the illustrated example illustrates each state quantity data when an object having a predetermined weight is placed on the moving slider 52 and an evaluation value for robustness.
  • the difference between the addition of disturbance and the sequence operation in the above embodiment may be distinguished from the adjustment unit learning data set as the operation condition c shown in FIG.
  • the position command, torque command, and detection position state quantity data of the adjustment unit learning data set are time-series pattern data when the same disturbance is applied. .
  • the adjustment unit 32 can set control parameters that can improve robustness against various disturbances.
  • the adjustment unit 32 may learn the control parameter setting for such disturbance addition by the above-described reinforcement learning.
  • reinforcement learning is reinforcement learning based on a reward for setting a control parameter when a predetermined disturbance is added.
  • the adjustment unit 32 can set control parameters that can improve robustness against various disturbances.
  • ⁇ Modification 3 When adjusting unit is divided and machine constant is output>
  • the neural network constituting the adjustment unit 32 may be divided into two in the layer direction, and a mechanical constant (or disturbance) may be output between them.
  • the neural network on the input layer side (left side in the figure) that inputs state quantity data is, for example, the first neural network 71
  • the neural network on the output layer side (right side in the figure) that outputs control parameters is, for example, the second neural network.
  • a neural network 72 is assumed. In the drawing, in order to avoid the complexity of illustration, nodes and edges are not shown and each neural network is shown as a rectangular block.
  • the first neural network 71 with respect to each time series pattern data of the position command, torque command, and detection position, which are the state quantity data input from each unit, from the correspondence between these state quantity data It is designed to output estimated machine constants (or disturbances).
  • the moment of inertia, the elastic coefficient, the first resonance frequency, and the second resonance frequency are illustrated as mechanical constants (in the case of a disturbance, for example, disturbance mass or disturbance elasticity to be added; illustration is omitted).
  • each state quantity data is recorded during a predetermined adjustment sequence operation, and the machine constant of the control target P is accurately detected by a separate detection machine, a detection test, or the like. Then, a learning data set for the first neural network 71 is created. Then, by executing data learning using a large number of learning data sets, the first neural network 71 learns feature quantities representing correlations between state quantity data and machine constants.
  • the temporarily set control parameters (Kp ′, Kv ′, T ′ shown in the figure) are temporarily set as the control parameters temporarily set during the adjustment sequence operation, and the first neural network 71 together with the state quantity data as shown in FIG. You may improve the estimation precision of the mechanical constant (or disturbance) output by inputting.
  • the temporary setting control parameter is also recorded together with the learning data set and used for data learning of the first neural network 71.
  • control parameters to be output to the second neural network 72 are referred to as adaptive control parameters as shown in the figure.
  • the adaptive control parameters Kp, Kv that are appropriate for the machine constants (or disturbances) input from the first neural network 71 from the correspondence between these machine constants. , T are designed to output.
  • adaptive control parameters are artificially set separately in order to perform a predetermined adjustment sequence operation on a large number of control objects each having a known machine constant (or disturbance).
  • a learning data set for the second neural network 72 is created that is appropriately set and includes these machine constants and adaptive control parameters. Then, by executing data learning using a large number of learning data sets, the second neural network 72 learns a feature amount representing the correlation between each machine constant (or disturbance) and the adaptive control parameter. .
  • the machine control system 1 of the present modification can output the machine constant (or disturbance) as a specific value, and by referring to this, the adaptive control parameter set by the adjustment unit 32 can be obtained. Validity can be easily examined.
  • ⁇ Modification 4 When letting the adjustment unit learn in advance>
  • the method of causing the adjustment unit 32 to perform data learning using only the adjustment unit learning data set from the beginning has been described.
  • a control parameter that is roughly appropriate by a known calculation method may be calculated, and the adjustment unit 32 may be pre-learned using the calculated control parameter.
  • an estimation unit 33 that calculates a temporary setting control parameter based on each state quantity data is provided in the motor control device 3 ⁇ / b> A, and the calculated temporary setting control parameter is input to the adjustment unit 32.
  • the estimation unit 33 obtains a machine constant of the control target P from each state quantity data detected by a predetermined adjustment sequence operation by a known calculation method, and calculates the machine constant and a state equation corresponding to the feedback control system. Based on this, the temporary setting control parameter is estimated.
  • the temporarily set control parameters estimated by the estimation unit 33 do not strictly correspond to individual differences (part manufacturing error, assembly error, etc.), operating conditions, aging deterioration conditions, and usage environment conditions of the control target P.
  • the design mathematical model is roughly appropriate, and there is only a slight difference from the adaptive control parameter to be finally obtained. Therefore, by temporarily setting with the temporary setting control parameter estimated by the estimation unit 33, the machine learning process of the adjustment unit 32 performs only fine adjustment (fine tuning), that is, prior learning can be performed.
  • the machine control system 1 includes the estimation unit 33 that estimates the control parameter by a predetermined sequence operation, and the machine learning process performs pre-learning using the control parameter estimated by the estimation unit 33. .
  • the learning time required in the machine learning process of the adjustment unit 32 can be significantly shortened.
  • the estimation unit 33 may estimate only the machine constant of the control target P from each state quantity data based on a known calculation method and input the estimated value to the adjustment unit 32. In this case, pre-learning of the portion corresponding to the first neural network shown in FIG. 10 is possible, and the learning time can be greatly shortened.
  • the motor control device 3 includes, for example, a CPU 901, a ROM 903, a RAM 905, a dedicated integrated circuit 907 constructed for a specific application such as an ASIC or FPGA, an input device 913, and an output device. 915, a recording device 917, a drive 919, a connection port 921, and a communication device 923. These components are connected to each other via a bus 909 and an input / output interface 911 so that signals can be transmitted to each other.
  • the program can be recorded in, for example, the ROM 903, the RAM 905, the recording device 917, or the like.
  • the program can also be recorded temporarily or permanently on, for example, a magnetic disk such as a flexible disk, various optical disks such as CD / MO disks / DVDs, and a removable recording medium 925 such as a semiconductor memory. .
  • a recording medium 925 can also be provided as so-called package software.
  • the program recorded on these recording media 925 may be read by the drive 919 and recorded on the recording device 917 via the input / output interface 911, the bus 909, or the like.
  • the program can be recorded on, for example, a download site, another computer, another recording device (not shown), or the like.
  • the program is transferred via a network NW such as a LAN or the Internet, and the communication device 923 receives this program.
  • the program received by the communication device 923 may be recorded in the recording device 917 via the input / output interface 911, the bus 909, or the like.
  • the program can be recorded in, for example, an appropriate external connection device 927.
  • the program may be transferred via an appropriate connection port 921 and recorded in the recording device 917 via the input / output interface 911, the bus 909, or the like.
  • the CPU 901 executes various processes according to the program recorded in the recording device 917, whereby the processes by the feedback control unit 31, the adjustment unit 32, and the like are realized.
  • the CPU 901 may directly read and execute the program from the recording device 917 or may be executed after it is once loaded into the RAM 905. Further, for example, when the program is received via the communication device 923, the drive 919, and the connection port 921, the CPU 901 may directly execute the received program without recording it in the recording device 917.
  • the CPU 901 may perform various processes based on signals and information input from the input device 913 such as a mouse, a keyboard, and a microphone (not shown) as necessary.
  • the input device 913 such as a mouse, a keyboard, and a microphone (not shown) as necessary.
  • the CPU 901 may output the result of executing the above processing from an output device 915 such as a display device or an audio output device, and the CPU 901 may send the processing result to the communication device 923 or the connection device as necessary. It may be transmitted via the port 921 or recorded on the recording device 917 or the recording medium 925.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Feedback Control In General (AREA)
  • Control Of Electric Motors In General (AREA)

Abstract

Le problème décrit par la présente invention est d'améliorer la commodité d'un dispositif de commande. La solution selon la présente invention concerne : un système de commande de rétroaction qui entre une déviation entre une instruction de référence (r) et une valeur de commande (y) provenant d'un objet externe (P) devant être commandé par un dispositif de commande (K(ρ)) qui exécute une commande sur la base d'un paramètre de commande prescrit (ρ), et commande l'objet (P) au moyen d'une valeur de fonctionnement (u) provenant du dispositif de commande (K(ρ)) ; et une unité de réglage (32) qui définit un paramètre de commande ρnew correspondant à une perturbation externe ou à la constante machine de l'objet (P) estimée à partir d'une correspondance entre l'instruction de référence (r) et la valeur de commande (y), sur la base d'un contenu appris pendant un processus d'apprentissage automatique. Pendant le processus d'apprentissage automatique, une unité de réglage (Cal) effectue un apprentissage automatique par le biais d'un apprentissage de données qui utilise un ensemble de données associant le paramètre de commande ρnew et/ou l'instruction de référence (r) et/ou la valeur de fonctionnement (u) et/ou la valeur de commande (y).
PCT/JP2018/005267 2017-02-20 2018-02-15 Dispositif et procédé de commande WO2018151215A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2018568603A JP6774637B2 (ja) 2017-02-20 2018-02-15 制御装置及び制御方法

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2017029383 2017-02-20
JP2017-029383 2017-12-27

Publications (1)

Publication Number Publication Date
WO2018151215A1 true WO2018151215A1 (fr) 2018-08-23

Family

ID=63169375

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2018/005267 WO2018151215A1 (fr) 2017-02-20 2018-02-15 Dispositif et procédé de commande

Country Status (2)

Country Link
JP (1) JP6774637B2 (fr)
WO (1) WO2018151215A1 (fr)

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200010982A (ko) * 2018-06-25 2020-01-31 군산대학교산학협력단 심층 강화 학습을 이용한 자율 이동체의 충돌 회피 및 자율 탐사 기법 및 장치
JP2020035213A (ja) * 2018-08-30 2020-03-05 ファナック株式会社 機械学習装置、制御システム及び機械学習方法
CN110888398A (zh) * 2018-09-11 2020-03-17 发那科株式会社 控制装置、cnc装置以及控制方法
JP2020046964A (ja) * 2018-09-19 2020-03-26 ファナック株式会社 特性判定装置、特性判定方法及び特性判定プログラム
JP2020047228A (ja) * 2018-09-21 2020-03-26 ファナック株式会社 モータ制御装置
JP2020057211A (ja) * 2018-10-02 2020-04-09 ファナック株式会社 機械学習装置、制御装置及び機械学習方法
CN111103849A (zh) * 2018-10-25 2020-05-05 发那科株式会社 输出装置、控制装置、以及学习参数的输出方法
DE102019217130A1 (de) 2018-11-09 2020-05-14 Fanuc Corporation Ausgabevorrichtung, steuervorrichtung und verfahren zum ausgeben von bewertungsfunktionen und maschinellen lernergebnissen
KR20200082909A (ko) * 2018-12-31 2020-07-08 한국기술교육대학교 산학협력단 디바이스 원격 제어를 위한 모방 강화학습 시스템 및 그 방법
JP2020119295A (ja) * 2019-01-24 2020-08-06 ファナック株式会社 フィルタの係数を最適化する機械学習システム、制御装置及び機械学習方法
JP2020134375A (ja) * 2019-02-22 2020-08-31 株式会社日立インダストリアルプロダクツ 油圧駆動式試験装置
JP2020154989A (ja) * 2019-03-22 2020-09-24 ファナック株式会社 機械学習装置、制御システム及び機械学習方法
JP2020177257A (ja) * 2019-04-15 2020-10-29 ファナック株式会社 機械学習装置、制御装置及び機械学習方法
WO2022124281A1 (fr) * 2020-12-11 2022-06-16 キヤノン株式会社 Dispositif de commande, procédé d'ajustement, dispositif de lithographie et procédé de fabrication d'article
WO2023167074A1 (fr) * 2022-03-02 2023-09-07 株式会社Ｓｃｒｅｅｎホールディングス Dispositif d'aide à la commande et procédé d'aide à la commande
WO2023203933A1 (fr) * 2022-04-21 2023-10-26 パナソニックＩｐマネジメント株式会社 Procédé de traitement d'informations, procédé d'optimisation d'informations, dispositif de traitement d'informations et programme
JP7403270B2 (ja) 2019-10-09 2023-12-22 株式会社日立産機システム 巻上機、巻上機システム、吊荷質量推定装置及び吊荷質量推定方法
WO2024005136A1 (fr) * 2022-07-01 2024-01-04 パナソニックＩｐマネジメント株式会社 Procédé de génération de paramètre de commande, programme, support d'enregistrement et dispositif de génération de paramètre de commande
WO2024090126A1 (fr) * 2022-10-25 2024-05-02 キヤノン株式会社 Procédé de commande, dispositif de commande, dispositif de lithographie et procédé de fabrication d'article

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2017034844A (ja) * 2015-07-31 2017-02-09 ファナック株式会社 ゲインの最適化を学習する機械学習装置及び機械学習装置を備えた電動機制御装置並びに機械学習方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04302304A (ja) * 1991-03-29 1992-10-26 Toshiba Corp 非線形プロセス制御装置

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2017034844A (ja) * 2015-07-31 2017-02-09 ファナック株式会社 ゲインの最適化を学習する機械学習装置及び機械学習装置を備えた電動機制御装置並びに機械学習方法

Cited By (38)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20200010982A (ko) * 2018-06-25 2020-01-31 군산대학교산학협력단 심층 강화 학습을 이용한 자율 이동체의 충돌 회피 및 자율 탐사 기법 및 장치
KR102124553B1 (ko) 2018-06-25 2020-06-18 군산대학교 산학협력단 심층 강화 학습을 이용한 자율 이동체의 충돌 회피 및 자율 탐사 기법 및 장치
JP2020035213A (ja) * 2018-08-30 2020-03-05 ファナック株式会社 機械学習装置、制御システム及び機械学習方法
US11029650B2 (en) 2018-08-30 2021-06-08 Fanuc Corporation Machine learning device, control system, and machine learning method
CN110888398B (zh) * 2018-09-11 2024-04-30 发那科株式会社 控制装置、cnc装置以及控制方法
CN110888398A (zh) * 2018-09-11 2020-03-17 发那科株式会社 控制装置、cnc装置以及控制方法
JP2020046964A (ja) * 2018-09-19 2020-03-26 ファナック株式会社 特性判定装置、特性判定方法及び特性判定プログラム
CN110941242B (zh) * 2018-09-21 2024-04-02 发那科株式会社 电动机控制装置
CN110941242A (zh) * 2018-09-21 2020-03-31 发那科株式会社 电动机控制装置
JP2020047228A (ja) * 2018-09-21 2020-03-26 ファナック株式会社 モータ制御装置
JP2020057211A (ja) * 2018-10-02 2020-04-09 ファナック株式会社 機械学習装置、制御装置及び機械学習方法
US11521119B2 (en) 2018-10-02 2022-12-06 Fanuc Corporation Machine learning device, control device, and machine learning method
CN111103849A (zh) * 2018-10-25 2020-05-05 发那科株式会社 输出装置、控制装置、以及学习参数的输出方法
CN111103849B (zh) * 2018-10-25 2024-03-15 发那科株式会社 输出装置、控制装置、以及学习参数的输出方法
CN111176114B (zh) * 2018-11-09 2023-02-17 发那科株式会社 输出装置、控制装置、以及评价函数和机器学习结果的输出方法
DE102019217130A1 (de) 2018-11-09 2020-05-14 Fanuc Corporation Ausgabevorrichtung, steuervorrichtung und verfahren zum ausgeben von bewertungsfunktionen und maschinellen lernergebnissen
CN111176114A (zh) * 2018-11-09 2020-05-19 发那科株式会社 输出装置、控制装置、以及评价函数和机器学习结果的输出方法
JP2020077310A (ja) * 2018-11-09 2020-05-21 ファナック株式会社 出力装置、制御装置、及び評価関数と機械学習結果の出力方法
US11592789B2 (en) 2018-11-09 2023-02-28 Fanuc Corporation Output device, control device, and method for outputting evaluation functions and machine learning results
KR102198450B1 (ko) * 2018-12-31 2021-01-05 한국기술교육대학교 산학협력단 디바이스 원격 제어를 위한 모방 강화학습 시스템 및 그 방법
KR20200082909A (ko) * 2018-12-31 2020-07-08 한국기술교육대학교 산학협력단 디바이스 원격 제어를 위한 모방 강화학습 시스템 및 그 방법
JP2020119295A (ja) * 2019-01-24 2020-08-06 ファナック株式会社 フィルタの係数を最適化する機械学習システム、制御装置及び機械学習方法
JP2020134375A (ja) * 2019-02-22 2020-08-31 株式会社日立インダストリアルプロダクツ 油圧駆動式試験装置
JP7189801B2 (ja) 2019-02-22 2022-12-14 株式会社日立インダストリアルプロダクツ 油圧駆動式試験装置
JP7000371B2 (ja) 2019-03-22 2022-01-19 ファナック株式会社 機械学習装置、制御システム及び機械学習方法
JP2020154989A (ja) * 2019-03-22 2020-09-24 ファナック株式会社 機械学習装置、制御システム及び機械学習方法
CN111722530A (zh) * 2019-03-22 2020-09-29 发那科株式会社 机器学习装置、控制系统以及机器学习方法
US11243501B2 (en) 2019-03-22 2022-02-08 Fanuc Corporation Machine learning device, control system, and machine learning
CN111722530B (zh) * 2019-03-22 2023-12-15 发那科株式会社 机器学习装置、控制系统以及机器学习方法
JP7000373B2 (ja) 2019-04-15 2022-01-19 ファナック株式会社 機械学習装置、制御装置及び機械学習方法
JP2020177257A (ja) * 2019-04-15 2020-10-29 ファナック株式会社 機械学習装置、制御装置及び機械学習方法
US11256220B2 (en) 2019-04-15 2022-02-22 Fanuc Corporation Machine learning device, control device and machine learning method
JP7403270B2 (ja) 2019-10-09 2023-12-22 株式会社日立産機システム 巻上機、巻上機システム、吊荷質量推定装置及び吊荷質量推定方法
WO2022124281A1 (fr) * 2020-12-11 2022-06-16 キヤノン株式会社 Dispositif de commande, procédé d'ajustement, dispositif de lithographie et procédé de fabrication d'article
WO2023167074A1 (fr) * 2022-03-02 2023-09-07 株式会社Ｓｃｒｅｅｎホールディングス Dispositif d'aide à la commande et procédé d'aide à la commande
WO2023203933A1 (fr) * 2022-04-21 2023-10-26 パナソニックＩｐマネジメント株式会社 Procédé de traitement d'informations, procédé d'optimisation d'informations, dispositif de traitement d'informations et programme
WO2024005136A1 (fr) * 2022-07-01 2024-01-04 パナソニックＩｐマネジメント株式会社 Procédé de génération de paramètre de commande, programme, support d'enregistrement et dispositif de génération de paramètre de commande
WO2024090126A1 (fr) * 2022-10-25 2024-05-02 キヤノン株式会社 Procédé de commande, dispositif de commande, dispositif de lithographie et procédé de fabrication d'article

Also Published As

Publication number Publication date
JPWO2018151215A1 (ja) 2019-11-07
JP6774637B2 (ja) 2020-10-28

Similar Documents

Publication Publication Date Title
WO2018151215A1 (fr) Dispositif et procédé de commande
CN109274314B (zh) 机器学习装置、伺服电动机控制装置、伺服电动机控制系统以及机器学习方法
CN108628355B (zh) 伺服控制装置及系统、机器学习装置及方法
CN106020124B (zh) 伺服马达控制装置以及碰撞检测方法
CN109116811B (zh) 机器学习装置和方法、伺服控制装置、伺服控制系统
US10824121B2 (en) Machine learning device, servo motor controller, servo motor control system, and machine learning method
JP6832498B2 (ja) システムおよび評価装置ならびに評価方法
JP6748135B2 (ja) 機械学習装置、サーボ制御装置、サーボ制御システム、及び機械学習方法
CN111103849B (zh) 输出装置、控制装置、以及学习参数的输出方法
US11009837B2 (en) Machine learning device that adjusts controller gain in a servo control apparatus
US11592789B2 (en) Output device, control device, and method for outputting evaluation functions and machine learning results
JP2018180764A (ja) 調整装置及び調整方法
CN111552237B (zh) 机器学习装置、控制装置、以及机器学习的搜索范围的设定方法
CN110941242A (zh) 电动机控制装置
JP2019008707A (ja) サーボ制御装置、サーボ制御方法及びシステム
JP6740263B2 (ja) 機械学習装置、サーボモータ制御装置、サーボモータ制御システム、及び機械学習方法
CN110888398A (zh) 控制装置、cnc装置以及控制方法
JP7469476B2 (ja) 制御支援装置、制御システム及び制御支援方法
JP7381751B2 (ja) 制御支援装置、制御装置及び制御支援方法
CN110727242B (zh) 机器学习装置、控制装置、以及机器学习方法
WO2021176619A1 (fr) Dispositif de commande de positionnement et procédé de positionnement
CN116301079A (zh) 一种轮式机器人速度精确控制方法以及装置
CN117032211A (zh) 一种双轴扰动控制方法和改进型扩张状态观测器
JP2005085111A (ja) 機械システムの制御方法及び装置
Fangfang et al. Adaptive pole-assignment speed control of ultrasonic motor

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18753659

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2018568603

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18753659

Country of ref document: EP

Kind code of ref document: A1