External force model generation device and external force model generation method

Abstract

The present invention generates an external force model which simulates and reproduces external force that acts on a servo mechanism, without the need for a sensor for external force measurement or specialized knowledge for external force analysis.　Provided is an external force model generation device for generating an external force model which simulates and reproduces external force that acts on a servo mechanism, said external force model generation device comprising: an extraction unit which extracts a position and / or a speed from operation history information pertaining to the servo mechanism; a disturbance observer which uses machine configuration information and the operation history information pertaining to the servo mechanism to determine an external force estimation value; and an external force model generation unit which determines a parameter of the external force model and which generates the external force model in such a manner that the position and / or the speed serves as an explanatory variable and the external force estimation value serves as a target variable.

Description

External force model generating device and external force model generating method

[0001] The present disclosure relates to an external force model generating device and an external force model generating method, and more particularly to an external force model generating device and an external force model generating method for generating an external force model that simulates and reproduces an external force acting on a servo mechanism.

[0002] Devices that simulate and reproduce external forces using external force models are described in, for example, Patent Documents 1 to 4. Patent Document 1 describes a robot that can reduce the effects of backdriving. Specifically, Patent Document 1 describes that the robot has joints incorporating motors and reducers, and includes an acquisition means that acquires information about the rotation of the motor, and a control means that detects the application of an external force based on the information acquired by the acquisition means and performs an escape operation. Patent Document 1 also describes that the external force may be estimated using an external force estimation model.

[0003] Patent Document 2 describes an information processing device that appropriately controls the operation of a robot. Specifically, Patent Document 2 describes that the information processing device includes an estimation unit that estimates an external force estimation value that is an estimate of the external force acting on a driven object by executing processing based on a machine learning model that receives as input a force estimation value that is an estimate of the force acting on the driven object and an external force response value that is based on the value of the external force detected by a force sensor that detects the external force acting on the driven object.

[0004] Patent Literature 3 describes a robot control system that accurately controls a robot. Specifically, Patent Literature 3 describes an example of a robot control system that includes an external force estimator that performs an estimation process based on a plurality of disturbance torques and a plurality of axis positions corresponding to a plurality of axes of the robot to calculate an external force acting on an end effector of the robot as an estimated external force, and a robot control unit that controls the robot based on the estimated external force, and describes that, as at least a part of the estimation process, the external force estimator performs calculations using an external force trained model that receives at least one of the plurality of disturbance torques and the plurality of axis positions as input.

[0005] Patent Document 4 describes a force detection method for a force control device that accurately detects external forces in a force control device that performs force control. Specifically, Patent Document 4 describes a learning mechanism configured with a neural network using multilayer perceptrons, and in a learning stage, the learning mechanism inputs motion information as learning input data and uses detected force (disturbance) values ​​as teacher signals, so that the neural network learns to estimate the disturbance. In a further next stage, the learning mechanism determines whether the detected force (disturbance) value and the estimated disturbance value match. If it is determined that the detected force (disturbance) value and the estimated disturbance value match, the learning mechanism outputs an estimated disturbance other than the external force based on this result. Patent Document 4 also describes that in the stage where the force control device performs a force control task, the learning mechanism 4 inputs motion information to calculate an estimated disturbance value Fn other than the external force, and subtracts this estimated disturbance value Fn from a value Fg output from the force detection mechanism using a subtractor to detect a net estimated external force F.

[0006] A machining simulation device for a machine tool is described in Patent Document 5, and an operation model calculation device for a drive unit is described in Patent Document 6. A numerical control device is described in Patent Document 7, a servo control device is described in Patent Document 8, and a disturbance observer is described in Patent Document 9. Furthermore, multiple regression analysis is described in Non-Patent Document 1. Patent Document 5 describes a machining simulation device for a machine tool that is capable of performing a machining simulation with high accuracy while suppressing an increase in time. Specifically, Patent Document 5 describes that the machining simulation device performs a machining simulation of a machine tool that machines a workpiece using a tool based on a machining program, and includes: a machine simulation unit that estimates the position of the tool by simulating the movement of the machine tool when it operates based on the machining program based on a position command and a transfer characteristic of the machine tool; and a machining simulation unit that performs a machining simulation of the workpiece based on tool information and the estimated tool position.

[0007] Patent Document 6 describes a motion model calculation device that easily creates a motion model of a drive device. Specifically, Patent Document 6 describes that the motion model calculation device is connected to a robot arm that includes a plurality of arms and a joint mechanism that rotatably connects the plurality of arms to connection destinations, outputs a predetermined motion command to the joint mechanism, obtains a drive state of the joint mechanism due to a motion in response to the motion command, and calculates a motion model that represents a relationship between an input value that represents an input to the joint mechanism and an output value of the joint mechanism relative to the input, based on the motion command and the drive state.

[0008] Patent Document 7 describes a numerical control device that reads information about a characteristic shape included in a machining program and generates control commands suitable for machining the characteristic shape. Specifically, Patent Document 7 describes that the numerical control device includes: a characteristic detection unit that detects characteristics of the machining shape from a machining program that commands the movement of a tool or a workpiece; a calculation unit that calculates a relational expression for determining the amount of inner rotation of the machining path relative to the program path based on servo parameters of a servo control device that drives the tool or the workpiece, the characteristics of the machining shape detected from the machining program, and machining requirements that define the machining conditions; and a determination unit that determines an optimized machining speed based on the relational expression.

[0009] Patent Document 8 describes a servo control device equipped with a machine learning device. Specifically, Patent Document 8 describes a servo control device that performs feedback control and includes a machine learning device and at least two or more feedforward calculation units, and that controls a servo motor that drives an axis of a machine tool or industrial machine using feedforward control in which at least two or more feedforward calculation units form multiple loops.

[0010] Patent Document 9 describes a disturbance load estimation method for a servo motor that can accurately estimate a disturbance load applied to the motor from the outside world even when the voltage is saturated. Specifically, Patent Document 9 describes a disturbance load estimation method for estimating a disturbance load that a servo motor receives from the outside world, in which an effective current that actually flows through the servo motor is calculated (DQ conversion term), an acceleration component to the motor is estimated based on this effective current (parameter α term), an acceleration difference between the calculated estimated acceleration and the actual motor acceleration (differential term) is calculated, and the disturbance load is estimated from this acceleration difference (parameter 1 / α term).

[0011] Non-Patent Document 1 describes that multiple regression analysis is a method that uses multiple or single variables (explanatory variables) to explain and predict changes in a certain variable (target variable).

[0012] Japanese Patent Application Laid-Open No. 2021-030344 Japanese Patent Application Laid-Open No. 2021-49597 International Publication No. 2023 / 157137 Japanese Patent Application Laid-Open No. 07-319558 Japanese Patent Application Laid-Open No. 2019-152936 International Publication No. 2020 / 161880 Japanese Patent Application Laid-Open No. 2021-002194 Japanese Patent Application Laid-Open No. 2019-164484 Japanese Patent Application Laid-Open No. 9-146643

[0013] "Introduction to Statistical Data Analysis, Part 3: Introduction to Multivariate Analysis," Intelligence and Information (Journal of the Japanese Society for Fuzzy Theory and Intelligent Informatics) Vol. 23, No. 2 pp. 228-234 (2011)

[0014] Conventionally, in order to perform simulation or control of a servo mechanism taking into account external forces such as friction and gravity, sensors for measuring external forces and specialized knowledge for analyzing external forces are required to analyze the external forces by analyzing the actual operation information of the servo mechanism.

[0015] For this reason, there is a need for an external force model generation device and an external force model generation method that can generate an external force model that simulates and reproduces the external forces acting on a servo mechanism without requiring a sensor for measuring external forces or specialized knowledge for analyzing external forces.

[0016] A first representative aspect of the present disclosure is an external force model generation device for generating an external force model that simulates and reproduces an external force acting on a servomechanism, comprising: an extraction unit that extracts at least one of a position and a velocity from operation record information of the servomechanism; a disturbance observer that obtains an external force estimate using mechanical configuration information of the servomechanism and the operation record information; and an external force model generation unit that determines parameters of the external force model to generate the external force model so that at least one of the position and the velocity is an explanatory variable and the external force estimate is a target variable.

[0017] A second representative aspect of the present disclosure is an external force model generation method in which a computer executes: an extraction step of extracting at least one of position and velocity from operational performance information of a servo mechanism; an external force estimation step of determining an external force estimated value using mechanical configuration information of the servo mechanism and the operational performance information; and an external force model generation step of determining parameters of the external force model to generate the external force model so that at least one of the position and the velocity is an explanatory variable and the external force estimated value is a target variable.

[0018] Fig. 1 is a block diagram showing an example configuration of a CNC machine tool including an external force model generating device according to an embodiment of the present disclosure; Fig. 2 is a block diagram showing an example configuration of a servo mechanism; Fig. 3 is a diagram showing an example configuration of a feed axis equipped with a motor, a ball screw, and a table, which serves as an actual machine; Fig. 4 is a perspective view showing an example configuration of a portion of a five-axis machining center, which serves as an actual machine; Fig. 5 is a block diagram showing an example configuration of a disturbance observer; Fig. 6 is a diagram showing a three-layer perceptron-type neural network; Fig. 7 is a characteristic diagram showing position, speed, external force estimated value, and external force of a model; Fig. 8 is a flowchart showing an example of an external force model generating method.

[0019] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an example configuration of a CNC machine tool including an external force model generating device according to an embodiment of the present disclosure. FIG. 2 is a block diagram showing an example configuration of a servomechanism. As shown in FIG. 1 , a CNC (Computer Numerical Control) machine tool 10 includes a CNC unit 100, a servomechanism 200, and an external force model generating device 300. Here, the external force model generating device 300 is included in the CNC machine tool 10, but it may be provided separately from the CNC machine tool including the CNC unit 100 and the servomechanism 200. As shown in FIG. 1 , the external force model generating device 300 includes an extraction unit 310, a disturbance observer 320, and an external force model generating unit 330. The configuration of the external force model generating device 300 will be described in detail below.

[0020] The CNC unit 100 generates control commands such as position commands based on a machining program and outputs them to the servomechanism 200. The configuration of the CNC unit 100 is already known. For example, Patent Document 7 describes a numerical control device including a command analysis unit, an interpolation unit, an acceleration / deceleration control unit, and a command output unit. The command analysis unit, the interpolation unit, the acceleration / deceleration control unit, and the command output unit constitute at least a part of the CNC unit 100.

[0021] The command analysis unit sequentially reads and analyzes blocks containing commands for moving each axis from the machining program, and creates movement command data for commanding the movement of each axis based on the analysis results. The interpolation unit generates interpolation data by interpolating points on the command path at an interpolation period based on the movement commands commanded by the movement command data. The acceleration / deceleration control unit performs acceleration / deceleration processing based on the interpolation data and calculates the machining speed of each axis for each interpolation period. The command output unit generates control commands such as position commands based on the machining speed of each axis.

[0022] The servo mechanism 200 is composed of, for example, multiple feed axes of a machine tool. Each feed axis is controlled by feedback control so that its position, speed, etc. follow a position command, speed command, etc. When there is only one feed axis, the servo mechanism 200 includes a servo controller 210, a mechanical unit 220, and a detector 230, as shown in FIG. 2. The configuration of the servo controller 210 is already known. For example, Patent Document 8 discloses a servo control device that is an example of a servo controller. The servo control device includes a first subtractor, a position control unit, a second subtractor, a speed control unit, a third subtractor, and a current control unit. The first subtractor calculates the difference between an input position command and a detected position (referred to as a position deviation), and the position control unit generates a speed command based on the position deviation. The second subtractor calculates the difference between the speed command and the detected speed (referred to as a speed deviation), and the speed control unit generates a torque command based on the speed deviation. The third subtractor calculates the difference between the torque command and a detected current (referred to as a current deviation), and the current control unit generates a current command based on the current deviation. The motor is controlled based on the current command. The current command or torque command is output to external force model generation device 300 as a manipulated variable. The manipulated variable output to external force model generation device 300 is, for example, a torque command when a disturbance observer shown in Equation 1 described later is used, or a current command when a disturbance observer shown in FIG. 5 described later is used. The manipulated variable is not limited to a current command or a torque command, but is determined by the configuration of disturbance observer 320.

[0023] The mechanical unit 220 is shown as a two-inertia system model in which a driving unit 221 and a driven unit 222 are connected by a spring and a damper. The configuration of the actual machine shown in the two-inertia system model corresponds to, for example, the configuration of a feed axis of the actual machine. FIG. 3 is a diagram showing an example of the configuration of a feed axis of the actual machine, which includes a motor, a ball screw, and a table. In FIG. 3, the mechanical unit 220A, which serves as the feed axis, includes a motor 2201, a coupling 2202, bearings 2203A and 2203B, a table 2204, a screw shaft 2205, and a nut 2206. A workpiece is placed on the table 2204. A detector 230, which will be described later, is attached to the motor 2201.

[0024] The two-inertia system model showing the machine section 220 corresponds to the configuration of the cradle axis used in an actual five-axis machining center. Figure 4 is a perspective view showing an example of the configuration of a portion of an actual five-axis machining center. Figure 4 shows a portion of a table-rotating five-axis machining center equipped with three linear axes, the X-axis, the Y-axis, and the Z-axis, and two rotary axes, the B-axis and the C-axis.

[0025] The table-rotating five-axis machining center 220B shown in FIG. 4 includes a head 2207, which serves as the main axis, and a cradle shaft 2208. The cradle shaft 2208 rotates the table in the direction of the B axis. The five-axis machining center linearly moves the table in the directions of the X, Y, and Z axes, and also rotates it in the direction of the C axis. The head 2207 is fixed in position. A ball end mill, for example, can be used as a tool attached to the head 2207. The two-inertia system model shown in the machine section 220 corresponds to the feed axis configuration of an actual machine using a linear motor (not shown). The linear motor includes a stator and a slider. The slider can be mechanically coupled to the table 2204, for example.

[0026] Detector 230 is a rotary encoder or the like attached to the motor. Detector 230 calculates a detected speed from the detected rotational angle position of the motor and feeds this back to servo controller 210 as the detected speed. Detector 230 also calculates a detected position by integrating the detected speed and feeds this back to servo controller 210. Waveform data of the detected speed and detected position are output as control variables to external force model generating device 300. The control variable may include a detected motor current detected by a current detector. Note that the control variable is determined by the configurations of disturbance observer 320 and external force model generating unit 330. The waveform data of the detected speed and detected position are input to extraction unit 310 of external force model generating unit 330 and used to extract speed and position. The extracted position is used in the disturbance observer shown in Equation 1, which will be described later. The extracted speed is used in the disturbance observer shown in FIG. 5, which will be described later. The control variables and manipulated variables output from servo mechanism 200 to external force model generating device 300 serve as operation performance information and as servo information for servo mechanism 200. As shown in FIG. 1, the control amount and the manipulated variable are output to the extraction unit 310 of the external force model generating device 300 as operation result information.

[0027] The servomechanism 200 or the CNC unit 100 stores mechanical configuration information of the servomechanism 200 and outputs the mechanical configuration information to the disturbance observer 320 of the external force model generation device 300. The mechanical configuration information includes, for example, the inertia of the driver and / or the driven part. The mechanical configuration information includes, for example, the inertia of the moving part. The inertia of the moving part is the sum of the inertia of the driver and the driven part. The mechanical configuration information may also include the rigidity between the driver and the driven part. Note that inertia is mass in a linear system and moment of inertia in a rotary system. In the feed axis of FIG. 3 , the driver is a motor and a ball screw, and the driven part is a nut and a table. There is a spring action between the ball screw and the nut, and the spring constant is the rigidity between the driver and the driven part. In the configuration of an actual feed axis using a linear motor (not shown), the driver is a slider, and the driven part is a table. The slider and table are mechanically connected by bolts or the like, but there is a spring action between them, and the spring constant becomes the stiffness between the driving part and the driven part. The machine configuration information may include other information on inertia and stiffness depending on the configuration of the disturbance observer 320. For example, when using the disturbance observer shown in Equation 1 described below, the machine configuration information includes the sum of the moments of inertia of the motor and the table, as well as the cutoff frequency. Furthermore, when using the disturbance observer shown in FIG. 5 described below, the machine configuration information includes the torque constant of the motor, as well as the inertia (moment of inertia).

[0028] The external force model generating device 300 generates an external force model that simulates and reproduces the external forces acting on the servomechanism 200 based on the operation record information and the machine configuration information. The external forces are forces that act on the motor and / or mechanical parts, etc., and include at least one of friction and gravity. The external forces are also called disturbances.

[0029] The detailed configuration of the external force model generating device 300 will be described below with reference to FIG. 1 . The extraction unit 310 acquires operation history information from the servomechanism 200. The extraction unit 310 extracts at least one of the position and the velocity from the detected position and the detected velocity in the operation history information, and outputs the extracted information to the external force model generating unit 330. FIG. 1 shows an example in which the position and the velocity are output to the external force model generating unit 330. The extraction unit 310 also outputs operation history information according to the configuration of the disturbance observer 320 to the disturbance observer 320. Because the external force of one feed axis may depend on the position or external force of another feed axis, the extraction unit 310 may need to extract multiple pieces of at least one of the positions and the velocities to generate an external force model for one axis.

[0030] The disturbance observer 320 acquires machine configuration information from the servomechanism 200 or the CNC unit 100, and acquires operation record information from the extraction unit 310. The machine configuration information may be stored in advance in the disturbance observer 320. The disturbance observer 320 uses the machine configuration information and operation record information to determine an external force estimate and outputs the external force estimate to the external force model generation unit 330. Because the external force of one feed axis may depend on the position or external force of another feed axis, the disturbance observer 320 may need to determine multiple external force estimates to generate an external force model for one axis.

[0031] The disturbance observer 320 may be, for example, either a first or second disturbance observer, which will be described below. The first disturbance observer 320 calculates an estimated external force T ｄ The first disturbance observer 320 acquires the position and torque command as operation result information from the extraction unit 310. In Equation 1, θ represents the position, Tm represents the torque command, J represents the sum of the inertia moments of the motor and the table, and ω ｃ indicates the cutoff frequency. The total moment of inertia J and the cutoff frequency ω ｃ is machine configuration information acquired from the servo mechanism 200 or the CNC unit 100.

[0032] The second disturbance observer 320 is, for example, the disturbance observer described in Patent Document 9. An example in which the configuration of the disturbance observer described in Patent Document 9 is applied to the second disturbance observer 320 will be briefly described with reference to FIG. 5. Details of the configuration and operation of the second disturbance observer are described in Patent Document 9. FIG. 5 is a block diagram showing an example of the configuration of the disturbance observer. When the disturbance observer 320 shown in FIG. 5 is used, the extraction unit 310 extracts a current command and a speed from the operation result information and outputs them to the disturbance observer 320.

[0033] The second disturbance observer 320 includes a three-phase AC-two-phase DC converter 3201 , an acceleration estimator 3202 , an actual acceleration calculator 3203 , a subtractor 3204 , and an external force estimator 3205 .

[0034] The three-phase AC-two-phase DC conversion unit 3201 acquires three-phase current commands Ir, Is, and It, which constitute a current command as one piece of operation result information, from the extraction unit 310, and converts them into two-phase DC currents Iq and Id. The acceleration estimation unit 3202 estimates acceleration using the DC current Iq, which is the q-phase current of the two-phase DC currents Iq and Id. The estimated acceleration value can be obtained by multiplying the DC current Iq by a parameter α. The parameter α is a value obtained by dividing the torque constant Kt of the motor by the inertia Jm. The torque constant Kt and inertia Jm of the motor are machine configuration information acquired from the servomechanism 200 or the CNC unit 100.

[0035] On the other hand, the actual acceleration calculation unit 3203 acquires the speed v as other operation result information from the extraction unit 310, and calculates the actual acceleration of the motor using the speed v. This actual acceleration of the motor can be calculated by differentiating the speed v. A low-pass filter may be added downstream of the actual acceleration calculation unit 3203.

[0036] The cutoff frequency ω in the first disturbance observer ｃIn addition, it is preferable that the cutoff frequency of the low-pass filter added to the subsequent stage of the actual acceleration calculation unit 3203 in the second disturbance observer is set to a value lower than the anti-resonance frequency of the two-inertia system model. The anti-resonance frequency is the natural frequency of the driven part when the driving part is fixed, and is determined by the inertia of the driven part and the rigidity between the driving part and the driven part.

[0037] In the first disturbance observer, the velocity v may be used instead of the position θ. In this case, Equation 1 becomes Equation 2 (Equation 2 below). In addition, in the second disturbance observer, the position θ may be used instead of the velocity v. In this case, the actual acceleration calculation unit differentiates the position θ to obtain the velocity v, and further differentiates the velocity v to obtain the actual acceleration of the motor.

[0038] Next, subtractor 3204 calculates the difference between the acceleration estimate α·Iq and the actual acceleration dv / dt. External force estimator 3205 multiplies this difference by the reciprocal of parameter α, 1 / α=Jm / Kt, to calculate an external force estimate and outputs the calculated value to external force model generator 330. In FIG. 5 , disturbance observer 320 calculates the external force estimate based on the current command and the detected speed. However, the external force estimate may also be calculated based on the torque command and the detected current. In this case, a torque command is input to subtractor 3204 in FIG. 5 instead of acceleration estimate α·Iq. Furthermore, instead of actual acceleration calculator 3203, an actual torque calculator that calculates an actual torque value based on the current acquired from extractor 310 may be provided, and subtractor 3204 receives the actual torque value calculated by the actual torque calculator. The actual torque calculator can calculate the actual torque value by multiplying the current by a torque constant. Then, the difference between the torque command and the actual torque value is output to the external force model generation unit 330 as an external force estimated value, without going through the external force estimation unit 3205 .

[0039] The external force model generating unit 330 generates an external force model that simulates and reproduces the external force acting on the servo mechanism. The external force model generating unit 330 determines the parameters of the external force model so that at least one of the position and the velocity is the explanatory variable and the external force estimated value is the target variable, thereby generating the external force model.

[0040] One method by which the external force model generating unit 330 generates an external force model is a method using multiple regression analysis, which will be described below. Details of multiple regression analysis are described in, for example, Non-Patent Document 1.

[0041] The external force model is shown in Equation 3 (hereinafter, Equation 3). In Equation 3, x １ｔ is the actual speed at time t, x ２ｔ is the actual speed one sample before time t, x ｊｔ is the actual speed result j-1 samples before time t, x ｊ＋１ｔ is the position result at time t, x ｊ＋２ｔ is the position result one sample before time t, and x ｍｔ is the position result m-j-1 samples before time t, y ｔ ' indicates external forces. Actual speed and actual position are explanatory variables. The external force estimated value y at time t is given as the objective variable. ｔ The residual between ' and the estimated external force y is the residual e ｔ The parameter b (b ０ , b １ , ..., b ｊ , b ｊ＋１ , b ｊ＋２ , ...b ｍ ) is calculated so as to minimize the residual sum of squares. In this way, the external force model generating unit 330 generates the external force model shown in Equation 3.

[0042] Another method by which the external force model generating unit 330 generates an external force model is a method using supervised machine learning, which will be described below. Supervised machine learning can be performed using a neural network, which will be described below. FIG. 6 is a diagram showing a three-layer perceptron type neural network. As shown in FIG. 6, the three-layer perceptron type neural network is composed of three layers: an input layer, a hidden layer (also called an intermediate layer), and an output layer. In FIG. 6, w １１ , w ２１ , w ｊ１ , w ｊ２ , w １ , w ｊ is the weight, b １ , b ｊ, b indicates bias. The weight is a value for adjusting the importance of the input, and the bias is a value for adjusting the values ​​aggregated in the intermediate layer and the output layer.

[0043] When a neural network is configured for each axis, there is one output layer, but when one neural network is used to estimate external forces for multiple axes, there are multiple output layers. Figure 6 shows the case where there is one output layer.

[0044] The three-layer perceptron type neural network shown in FIG. １ and velocity x ２ When input as explanatory variables, a three-layer perceptron-type neural network outputs an external force. The weights and biases shown in Figure 6 are model parameters specific to each axis. The external force estimates obtained from the disturbance observer shown in Equation 1 are used as training data, and the model parameters are learned by backpropagation. The velocity used for learning can also be the time difference of position. Backpropagation is a method in which, when estimating model parameters in a neural network, the gradient of the loss function is calculated from the output layer to the input layer, and the model parameters are updated so that the difference between the external force estimates, which serve as label data, and the output data (which serves as training data) is reduced.

[0045] The design parameters of the external force model include the cutoff frequency of the disturbance observer, the number of hidden layers of the neural network, and learning settings of the backpropagation method, which are given by the user. The external force model generated by the external force model generation unit 330 is output to a machining simulation device for predicting machining results or a control device for compensating for the external force of a servo mechanism.

[0046] A specific example is given below. In the case where a five-axis machining center is used as the CNC machine tool 10 and the servo mechanism 200 is equipped with a cradle axis, the model output obtained using the external force model generated by supervised machine learning using the neural network shown in FIG. 6 is described. The cradle axis shown in FIG. 4 exhibits a load torque due to gravity that varies depending on the position (rotation angle). An external force model was constructed using such a cradle axis as an example, and the results of plotting the external force estimates of the actual machine and the output of the external force model are shown in FIG. 7. FIG. 7 is a characteristic diagram showing the position, velocity, external force estimates, and model external force. The external force on the vertical axis in FIG. 7 represents the external force estimates or the model external force values. As shown in FIG. 7, it can be seen that the model external force over a period of constant velocity accurately reproduces the external force estimates.

[0047] 8 is a flowchart showing an example of an external force model generation method. In the following explanation, an example will be described in which the external force model generation method of the present disclosure is executed by external force model generation device 300, but the external force model generation method of the present disclosure can also be executed by configurations other than external force model generation device 300. Note that in the following explanation, an example will be described in which the disturbance observer shown in FIG. 5 is used as the disturbance observer, but even when another disturbance observer is used, an external force model can be generated in a similar manner depending on the information extracted from the operation result information and the configuration of the disturbance observer.

[0048] In step S1, the extraction unit 310 collects waveform data of servo information from the servo mechanism 200 while the actual machine is in operation, as operational performance information.

[0049] In step S2, the disturbance observer 320 acquires machine configuration information from the servo mechanism 200 or the CNC unit 100.

[0050] In step S3 , the extraction unit 310 extracts the position, speed, and current command from the operation result information, outputs the position and speed to the external force model generation unit 330 , and outputs the speed and current command to the disturbance observer 320 .

[0051] In step S4, the disturbance observer 320 uses the machine configuration information to obtain an estimated external force value based on the speed and current command. The disturbance observer 320 outputs the estimated external force value to the external force model generator 330.

[0052] In step S5, the external force model generation unit 330 performs supervised machine learning using the position and velocity as input data and the external force estimated value as label data to generate an external force model. The external force model may be generated using multiple regression analysis.

[0053] According to the external force model generation device and external force model generation method of the present embodiment described above, it is possible to generate an external force model that simulates and reproduces the external forces acting on a servomechanism, without requiring a sensor for measuring external forces or specialized knowledge for analyzing external forces.

[0054] To realize the functional blocks included in the external force model generation device in each embodiment described above, the external force model generation device can be realized by hardware, software, or a combination of these. The external force model generation method can also be realized by hardware, software, or a combination of these. Here, "realized by software" means that the method is realized by a computer reading and executing a program.

[0055] In order to realize the components included in the external force model generation device by software or a combination thereof, the external force model generation device includes a processing unit such as a CPU (Central Processing Unit). The processing unit functions as an execution unit. The external force model generation device also includes an auxiliary storage device such as an HDD (Hard Disk Drive) that stores various control programs such as application software or an OS (Operating System), and a main storage device such as a RAM (Random Access Memory) that stores data temporarily required for the processing unit to execute the programs.

[0056] In the external force model generating device, the arithmetic processing unit reads application software or an OS from the auxiliary storage device, and while expanding the read application software or OS into the main storage device, performs arithmetic processing based on the application software or OS. Furthermore, based on the results of this calculation, various pieces of hardware included in the external force model generating device are controlled. In this way, the functional blocks of this embodiment are realized. Furthermore, the external force model generating method can also be realized with a configuration similar to that of the external force model generating device.

[0057] The components included in the external force model generation device can be realized by hardware including electronic circuits, etc. When the external force model generation device is configured by hardware, some or all of the functions of the components included in the external force model generation device can be configured by an integrated circuit (IC), such as an ASIC (Application Specific Integrated Circuit), a gate array, an FPGA (Field Programmable Gate Array), or a CPLD (Complex Programmable Logic Device).

[0058] The program can be stored and supplied to a computer using various types of non-transitory computer-readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (e.g., hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)). The program may also be supplied to the computer by various types of transient computer readable media.

[0059] According to the external force model generation device and external force model generation method disclosed herein, including the embodiments described above, it is possible to generate an external force model that simulates and reproduces the external forces acting on a servomechanism, without requiring a sensor for measuring external forces or specialized knowledge for analyzing external forces.

[0060] Although the above-described embodiments are preferred embodiments of the present invention, the scope of the present invention is not limited to the above-described embodiments, and the present invention can be implemented in various modified forms within the scope that does not deviate from the gist of the present invention.

[0061] The following supplementary note is further disclosed regarding the above embodiment: (Supplementary Note 1) An external force model generation device (300) for generating an external force model that simulates and reproduces an external force acting on a servomechanism (200), comprising: an extraction unit (310) that extracts at least one of a position and a velocity from operation record information of the servomechanism, a disturbance observer (320) that determines an external force estimated value using mechanical configuration information of the servomechanism and the operation record information, and an external force model generation unit (330) that determines parameters of the external force model to generate the external force model so that either the position or the velocity serves as an explanatory variable and the external force estimated value serves as a target variable.

[0062] (Supplementary Note 2) The external force model generation device according to Supplementary Note 1, wherein the external force model generation unit (330) generates the external force model by supervised machine learning using the explanatory variables as input data and the objective variables as label data.

[0063] (Supplementary Note 3) The external force model generation device according to Supplementary Note 1 or 2, wherein the operation history information includes a control amount and an operation amount of the servo mechanism (200) of the axis for which the external force model is to be created, and at least one of the position and the velocity is extracted from the control amount, and the mechanical configuration information includes an inertia of a movable part of the axis for which the external force model is to be created, or an inertia of a driving part and an inertia of a driven part.

[0064] (Supplementary Note 4) The external force model generation device according to Supplementary Note 1 or 2, wherein the disturbance observer (320) uses mechanical configuration information of the servo mechanism and determines the external force estimation value based on the position or velocity and a torque command extracted from the operation record information, or based on the position or velocity and a current command extracted from the operation record information.

[0065] (Supplementary Note 5) An external force model generation method in which a computer executes the following steps: an extraction step of extracting at least one of position and velocity from operation history information of a servo mechanism (200); an external force estimation step of determining an external force estimated value using mechanical configuration information of the servo mechanism and the operation history information; and an external force model generation step of determining parameters of the external force model to generate the external force model so that at least one of the position and the velocity is an explanatory variable and the external force estimated value is a target variable.

[0066] (Supplementary Note 6) The external force model generating method according to Supplementary Note 5, wherein the external force model generating step generates the external force model by supervised machine learning using the explanatory variables as input data and the objective variables as label data.

[0067] (Supplementary Note 7) The external force model generation method described in Supplementary Note 5 or 6, wherein the operation history information includes a control amount and an operation amount of the servo mechanism (200) of the axis for which the external force model is to be created, and at least one of the position and the velocity is extracted from the control amount, and the mechanical configuration information includes the inertia of the movable part of the axis for which the external force model is to be created, or the inertia of the driving part and the inertia of the driven part.

[0068] (Supplementary Note 8) The external force model generation method according to Supplementary Note 5 or 6, wherein the external force estimation step uses mechanical configuration information of the servo mechanism and determines the external force estimation value based on the position or velocity and a torque command extracted from the operation performance information, or based on the position or velocity and a current command extracted from the operation performance information.

[0069] REFERENCE SIGNS LIST 10 CNC machine tool 100 CNC unit 200 Servo mechanism 210 Servo controller 220 Machine unit 230 Detector 300 External force model generating device 310 Extraction unit 320 Disturbance observer 330 External force model generating unit

Claims

1. An external force model generation device for generating an external force model that simulates and reproduces external forces acting on a servo mechanism, comprising: an extraction unit that extracts at least one of position and velocity from operation record information of the servo mechanism; a disturbance observer that determines an external force estimate using mechanical configuration information of the servo mechanism and the operation record information; and an external force model generation unit that determines parameters of the external force model and generates the external force model so that at least one of the position and the velocity is an explanatory variable and the external force estimate is a target variable.

2. The external force model generating device according to claim 1, wherein the external force model generating unit generates the external force model by supervised machine learning using the explanatory variables as input data and the objective variables as label data.

3. An external force model generating device as described in claim 1 or 2, wherein the operational performance information includes the control amount and operation amount of the servo mechanism of the axis for which the external force model is to be created, and at least one of the position and the velocity is extracted from the control amount, and the mechanical configuration information includes the inertia of the movable part of the axis for which the external force model is to be created, or the inertia of the driving part and the driven part, and the rigidity between the driving part and the driven part.

4. An external force model generation device as described in claim 1 or 2, wherein the disturbance observer uses mechanical configuration information of the servo mechanism to determine the external force estimation value based on the position or velocity and a torque command extracted from the operation performance information, or based on the position or velocity and a current command extracted from the operation performance information.

5. An external force model generation method in which a computer executes the following steps: an extraction step of extracting at least one of position and velocity from operational performance information of a servo mechanism; an external force estimation step of determining an external force estimation value using mechanical configuration information of the servo mechanism and the operational performance information; and an external force model generation step of determining parameters of an external force model that simulates and reproduces the external force acting on the servo mechanism, so that at least one of the position and the velocity is an explanatory variable and the external force estimation value is a target variable, to generate the external force model.

6. The external force model generating method according to claim 5, wherein the external force model generating step generates the external force model by supervised machine learning using the explanatory variables as input data and the objective variables as label data.

7. An external force model generation method as described in claim 5 or 6, wherein the operation history information includes the control amount and operation amount of the servo mechanism of the axis for which the external force model is to be created, and at least one of the position and the velocity is extracted from the control amount, and the mechanical configuration information includes the inertia of the movable part of the axis for which the external force model is to be created, or the inertia of the driving part and the driven part, and the rigidity between the driving part and the driven part.

8. An external force model generation method as described in claim 5 or 6, wherein the external force estimation step uses mechanical configuration information of the servo mechanism to determine the external force estimation value based on the position or velocity and a torque command extracted from the operation performance information, or based on the position or velocity and a current command extracted from the operation performance information.

Images