DE112021008301T5 - NUMERICAL CONTROL DEVICE, MACHINING SYSTEM, NUMERICAL CONTROL METHOD AND MACHINING METHOD - Google Patents

Abstract

A numerical control device (3) includes: a command generation unit (31) that generates a basic operation command which is an operation command based on a numerical control program (4), and generates a corrected operation command which is the operation command obtained by correcting the basic operation command; a coupled simulation unit (33) that calculates process information reflecting an influence of an operation of a drive system (20) and a dynamics of a structure that generates vibration during operation of a machine tool (2) in a cutting process (M) of a workpiece (W) with a tool (23), the process information indicating results of simulation processing in which the basic operation command and the corrected operation command are respectively given to the machine tool (2); and a process evaluation unit (34) that evaluates, based on a plurality of the process information, a magnitude of a machining error in use of each of a plurality of the operation commands and selects the operation command to be given to the machine tool (2) from the basic operation command and the corrected operation command.

Description

Area

The present disclosure relates to a numerical control apparatus, a machining system, a numerical control method, and a machining method for controlling a machine tool.

background

Machine tools are machining devices that can perform cutting machining, that is, machining in which unnecessary portions are removed from a workpiece by applying force or energy to the workpiece using a tool. A machine tool includes a spindle drive system that rotates a tool or a workpiece, and a feed drive system that changes the relative position between the tool and the workpiece. The numerical control device drives the spindle drive system and the feed drive system based on an operation command generated based on a numerical control program to machine the workpiece. Here, even if the machine tool is controlled according to the command described in the numerical control program, various factors may prevent the machining from being performed according to the command, and a machining error may occur.

Patent Literature 1 proposes a technique for reproducing the property of a machined surface by calculating the displacement of the tool caused by a cutting resistance applied to the tool during cutting. In the method described in Patent Literature 1, a parameter representing a dynamic property of the tool is stored in advance, whereby the displacement of the tool center that occurs when a cutting resistance corresponding to the uncut chip thickness of the tool calculated by simulation is generated is regarded as a machining error.

List of citations

Patent literature

Patent Literature 1: Japanese Laid-Open Publication No. 2013-132733

Brief description

Technical problem

However, the conventional technique described above is problematic in that the machining error cannot be accurately reduced. In the technique described in Patent Literature 1, the deflection amount of the tool is predicted and the displacement of the tool center point is regarded as the machining error. In practice, however, during the operation of the machine tool, the cutting process, the operation of the drive system, and the mechanical dynamics of a structure that generates vibration during the operation of the machine tool interact with each other. Here, the cutting process represents a series of processes in which the cutting edge of the tool penetrates the workpiece to form a machined surface while generating chips, and the mechanical dynamics represents dynamic characteristics of a structure that vibrates when vibrations propagate from vibration sources inside and outside the machine tool. Therefore, the method described in Patent Literature 1 cannot accurately evaluate the machining error and cannot accurately reduce the machining error.

The present disclosure has been made in view of the foregoing, and an object thereof is to achieve a numerical control device capable of accurately reducing the machining error of a machine tool.

the solution of the problem

In order to solve the above problem and achieve an object, the present disclosure is directed to a numerical control device that controls a machine tool by giving an operation command to the machine tool, the machine tool including a drive system that includes a spindle drive system for driving a tool for machining a workpiece or a spindle that rotates the workpiece, and a feed drive system for driving a feed axis that changes a relative position between the tool and the workpiece, the numerical control device including: a command generation unit for generating a basic operation command that is the operation command based on a numerical control program, and generating a corrected operation error that is the operation command obtained by correcting the basic operation error; a coupled simulation unit for generating process information to calculate information reflecting an influence of an operation of the drive system and dynamics of a structure that generates vibration during operation of the machine tool on a cutting process of the workpiece with the tool, the process information indicating results of simulation machining in which the basic operation command and the corrected operation command are respectively given to the machine tool; and a process evaluation unit for evaluating, based on a plurality of the process information, a magnitude of a machining error in use of each of a plurality of the operation commands and selecting the operation command to be given to the machine tool from the basic operation command and the corrected operation command.

Advantageous effects of the invention

The present invention can achieve the effect of accurately reducing the machining error of a machine tool.

Short description of the drawings

• 1 is a diagram illustrating a functional configuration of a machining system according to the first embodiment.
• 2 is a representation showing an exemplary physical configuration of the 1 illustrated machine tool.
• 3 is a diagram illustrating time waveforms of the spindle speed and feed rate in the basic operation command.
• 4 is a representation that illustrates the tool and the workpiece during machining in which the 3 illustrated spindle speed and feed rate can be used.
• 5 is a diagram illustrating time waveforms of the spindle speed and feed rate under the corrected operation command.
• 6 is a representation that illustrates the tool and the workpiece during machining in which the 5 illustrated spindle speed and feed rate can be used.
• 7 is a representation that shows the relationship between the 1 illustrated spindle drive system, mechanical dynamics and the cutting process.
• 8th is a representation which shows the 7 illustrated physical quantities together with the physical configuration of the machine tool.
• 9 is a representation that shows the relationship between the 1 illustrated feed drive system, mechanical dynamics and the cutting process.

• 10 is a representation which shows the 9 illustrated physical quantities together with the physical configuration of the machine tool.
• 11 is a diagram to explain an example of the spindle drive control model in 1 .
• 12 is a diagram to explain an example of the feed drive control model in 1 .
• 13 is a flow chart to explain the operation of the 1 illustrated numerical control device.
• 14 is a diagram illustrating a functional configuration of a machining system according to the second embodiment.
• 15 is a flow chart to explain the operation of the 14 illustrated numerical control device.
• 16 is a diagram showing an exemplary configuration of a learning device with respect to the 14 illustrated numerical control device.
• 17 is a flow chart to explain the learning process of the 16 illustrated learning device.
• 18 is a diagram showing an exemplary configuration of an inference device with respect to the 14 illustrated numerical control device.
• 19 is a flow chart to explain the operation of the 18 illustrated inference device.
• 20 is a diagram illustrating a configuration of a machining system according to the third embodiment.
• 21 is a diagram illustrating dedicated hardware for implementing the functions of the numerical control devices, the learning device, and the inference device according to the first to third embodiments.
• 22 is a diagram illustrating a configuration of the control circuit for implementing the functions of the numerical control devices, the learning device, and the inference device according to the first to third embodiments.

Description of embodiments

Hereinafter, a numerical control apparatus, a machining system, a numerical control method, and a machining method according to embodiments of the present disclosure will be described in detail with reference to the drawings. In the following description, a plurality of components having similar functions can be distinguished from each other by identifying them by a common number followed by a hyphen and a number. In cases where a plurality of components having similar functions do not need to be distinguished from each other, they are only identified by a common number.

First embodiment.

1 is a diagram illustrating a functional configuration of a machining system 1 according to the first embodiment. The machining system 1 includes a machine tool 2 and a numerical control device 3. The numerical control device 3 controls the machine tool 2 by giving the machine tool 2 an operation command generated based on a command described in a numerical control program 4.

The machine tool 2 includes a spindle drive system 21, one or more feed drive systems 22, a tool 23 for machining a workpiece W and a table 24 for holding the workpiece W.

The spindle drive system 21 includes a spindle motor 211 and a spindle drive mechanism 212 driven by the spindle motor 211. The tool 23 is connected to the spindle drive system 21, and the spindle drive system 21 can rotate the tool 23. The spindle motor 211 or the spindle drive mechanism 212 is equipped with a rotary encoder (not illustrated) that represents angle information of the spindle drive system 21.

The feed drive system 22 includes a servo motor 221 and a feed drive mechanism 222 driven by the servo motor 221. The feed drive system 22 can change the relative position between the tool 23 and the workpiece W. The servo motor 221 and the feed drive mechanism 222 are equipped with a rotary encoder (not illustrated) that reflects position information of the feed drive system 22. The table 24 on which the workpiece W is held or the tool 23 is connected to the feed drive system 22 and the feed drive system 22 can change the relative position between the tool 23 and the workpiece W by moving the table 24 or the tool 23. In the embodiment shown in 1 In the example illustrated, the machine tool 2 includes a feed drive system 22-1 that moves the tool 23 and a feed drive system 22-2 that moves the table 24 so that both the tool 23 and the table 24 are moved. However, it is permissible for only the tool 23 to be moved or for only the table 24 to be moved. Any configuration that can change the relative position between the tool 23 and the workpiece W held on the table 24 is available. When the feed drive system 22 changes the relative position between the tool 23 and the workpiece W, the tool 23 cuts the workpiece W along the machining path.

The spindle drive system 21 and the feed drive system 22 are connected to the numerical controller 3, and the spindle motor 211 and the servo motor 221 are controlled by an operation command issued by the numerical controller 3. Hereinafter, the spindle drive system 21 and the feed drive system 22 are collectively referred to as the drive system 20. Note that a series of processes in which the cutting edge of the tool 23 penetrates the workpiece W to form a machined surface while generating chips is referred to as a cutting process M.

2 is a representation showing an exemplary physical configuration of the 1 illustrated machine tool 2. The table 24 has a table shape having a horizontal plane on which the workpiece W is placed. The spindle drive mechanism 212 is provided such that the tool 23 is positioned above the workpiece W held on the table 24. The spindle motor 211 is provided adjacent to the spindle drive mechanism 212. The spindle of the spindle drive system 21 including the spindle motor 211 and the spindle drive mechanism 212 is perpendicular to the horizontal plane of the table 24, and the spindle drive system 21 rotates the tool 23 about the spindle.

The feed drive mechanism 222-1 of the feed drive system 22-1, which moves the tool 23, is connected to the tool 23 via a structure including the spindle drive mechanism 212 to which the tool 23 is attached. The servo motor 221-1 of the feed drive system 22-1 is provided adjacent to the feed drive mechanism 222-1. The feed axis of the feed drive system 22-1 is parallel to the spindle, and the feed drive system 22-1 moves the tool 23 up and down along the feed axis.

The feed drive mechanism 222-2 of the feed drive system 22-2, which moves the table 24, is connected to the table 24. The servo motor 221-2 of the feed drive system 22-2 is provided adjacent to the feed drive mechanism 222-2. The feed axis of the feed drive system 22-2 is located in a direction in the horizontal plane of the table 24, and the feed drive system 22-2 moves the table 24 in the horizontal direction. Although only a feed drive system 22 for moving the table 24 has been described here, the machine tool 2 may further include the feed drive system 22 having a feed axis in a direction perpendicular to the feed axis of the feed drive system 22-2 and in the horizontal plane of the table 24.

It should be noted that the physical configuration illustrated here is an example shown for ease of explanation, and the physical configuration of the machine tool 2 does not apply to the 2 example illustrated. For example, the number of the feed drive systems 22 of the machine tool 2 may be one, or may be three or more. The directions of the spindle and the feed axis are also examples. The table-shaped table 24 is an example of a mechanism that holds the workpiece W, and may have any configuration that can hold the workpiece W and control the relative position to the tool 23.

Referring again to 1 the numerical control device 3 includes a command generation unit 31, a storage unit 32, a coupled simulation unit 33, a process evaluation unit 34 and a drive control unit 35.

The numerical control program 4 includes a plurality of commands relating to the movement of the spindle and the feed axis of the machine tool 2. The commands included in the numerical control program 4 are, for example, commands specifying a path on which the tool 23 moves as a relative position to the workpiece W. The commands specifying the path of the tool 23 include a plurality of position commands specifying positions on the path. The numerical control program 4 further includes a spindle speed command specifying the rotation speed of the spindle and a feed rate command specifying the movement speed of the feed axis at the position specified by each position command. The numerical control program 4 may be given to the numerical control device 3 from outside the numerical control device 3 or may be held in the numerical control device 3.

The command generation unit 31 analyzes the commands described in the numerical control program 4 and generates from time to time an operation command to be given to the machine tool 2 to control the machine tool 2. The command generation unit 31 generates a basic operation command, which is an operation command for causing the machine tool 2 to execute the command described in the numerical control program 4 as it is without correction, and a corrected operation command, which is an operation command obtained by correcting the basic operation command. The command generation unit 31 may generate one or more corrected operation commands. The corrected operation command may be an operation command in which the relative path along which the tool 23 moves with respect to the workpiece W is the same as that of the basic operation command and at least one of the feed amount of the feed axis and the uncut chip thickness of the workpiece W is changed. The feed amount is the feed amount per process unit, for example, the feed amount per tooth of the tool 23. In this case, in the corrected operation command, at least one of the spindle speed command and the feed rate command at the time when each of the plurality of cutting edges of the tool 23 cuts into the workpiece W is different for each cutting edge of the tool. The spindle speed command and the feed rate command at each time are modulated according to the angle of each of the plurality of cutting edges of the tool 23 and the relative position between the tool 23 and the workpiece W. The command generation unit 31 may generate, as the corrected operation command, an operation command in which the spindle speed command and/or the feed rate command is changed at each time of the basic operation command according to the feed amount or the uncut chip thickness of the workpiece W per tooth for each cutting edge of the tool 23 that inside or outside the command generation unit 31. The command generation unit 31 outputs the generated basic operation command and corrected operation error to both the coupled simulation unit 33 and the drive control unit 35.

3 is a diagram illustrating time waveforms of the spindle speed and the feed rate in the basic operation command. Here, for the sake of simplicity, an example will be described in which two positions P1 and P2 are specified in the numerical control program 4 and the spindle speed and the feed rate are specified at a constant value with respect to the commanded trajectory between the positions P1 and P2. 3 illustrates the spindle speed and the feed rate with respect to the commanded trajectory between the positions P1 and P2. The basic operation command is an operation command for causing the machine tool 2 to execute the commands described in the numerical control program 4 as they are without correction. Therefore, as described in the numerical control program 4, the spindle speed and the feed rate specified by the basic operation command for the commanded trajectory between the positions P1 and P2 are constant.

4 is a diagram illustrating the tool 23 and the workpiece W during machining in which the 3 illustrated spindle speed and feed rate are used. The tool 23 moves from the position P1 toward the position P2 while rotating in the rotation direction R1, and when it comes into contact with the workpiece W, the tooth of the tool 23 cuts the workpiece W. In the basic operation command, both the spindle speed and the feed rate are constant, and thus the feed amount c per tooth is constant. In this case, the cutting area A1 per tooth is also constant.

5 is a diagram illustrating time waveforms of the spindle speed and feed rate in the corrected operation command. Here, while maintaining the commanded trajectory of the control system with reference to the 3 and 4 described basic operation command, the spindle speed and the feed rate vary sinusoidally with respect to the constant values described in the numerical control program 4.

6 is a diagram illustrating the tool 23 and the workpiece W undergoing machining according to the 5 illustrated spindle speed and feed rate. The tool 23 moves from the position P1 toward the position P2 while rotating in the rotation direction R1, and when it comes into contact with the workpiece W, the tooth of the tool 23 cuts the workpiece W. With the corrected operation command, the spindle speed and feed rate constantly change. Therefore, the feed amount c per tooth changes with time, and as a result, the cutting area A1 per tooth also changes.

In the example from 5 the command generation unit 31 generates a corrected operation command by changing both the spindle speed and the feed rate into a sinusoidal variation pattern with respect to the feed amount c or the uncut chip thickness of the workpiece W per tooth, which are based on the basic operation command. However, the variation pattern to be used is not limited to a sinusoidal waveform, and various variation patterns including a triangular waveform and a random waveform may be used. The command generation unit 31 may also generate a corrected operation command by superimposing a predetermined profile variation on at least one of the spindle speed and the feed rate of the basic operation command. The command generation unit 31 may contain in advance information indicating the variation profile to be superimposed on the basic operation command. Although the example in which the command generation unit 31 generates a corrected operation command from a basic operation command has been described here, the command generation unit 31 may generate a plurality of corrected operation commands from a basic operation command.

Referring again to 1 A cutting process model 321, a dynamic model 322, a spindle drive control model 323, a feed drive control model 324, and cutting condition information 325 are stored in the storage unit 32. The storage unit 32 can output the stored information to the coupled simulation unit 33. The cutting condition information 325 includes tool shape information including the number of teeth, the tool diameter, and the helix angle of the tool 23, and the cutting amount when using the tool 23. Details of the cutting process model 321, the dynamic model 322, the spindle drive control model 323, and the feed drive control model 324 will be described later.

Since the cutting performed by the machine tool 2 is a physical phenomenon in which the cutting process and the mechanical dynamics influence each other, it is desirable to carry out an analysis in which the cutting process and the mechanical dynamics are integrated in order to determine the machining process. state. Here, the cutting process represents a series of processes in which the cutting edge of the tool 23 penetrates the workpiece W to form a machined surface while generating chips. The mechanical dynamics represents a dynamic behavior of a structure that generates vibration due to vibration sources inside and outside the machine tool 2. The structure in the present context may include the tool 23 and the workpiece W in addition to the structures that form the machine tool 2.

The drive system 20 is controlled by the numerical control device 3 so that the tool 23 moves to traverse a predetermined path with respect to the workpiece W while rotating. While the tool 23 cuts the workpiece W, the cutting force F _c generated between the tool 23 and the workpiece W is transmitted to the feed drive system 22 as a disturbance force F _d through the structure, and is transmitted to the spindle drive system 21 as a disturbance torque T _d . Since the disturbance force F _d is applied to the feed drive system 22, the position of the feed drive system 22 varies depending on the amplitude and frequency of the disturbance force F _d with respect to the position when the tool 23 is not cutting the workpiece W. Likewise, when the disturbance torque T _d is applied to the spindle drive system 21, the rotation angle of the spindle drive system 21 varies with respect to the rotation angle when the tool 23 is not cutting the workpiece W.

The above relationship will be described with reference to drawings. 7 is a representation that shows the relationship between the 1 illustrated spindle drive system 21, the mechanical dynamics and the cutting process M. 8th is a representation which shows the 7 illustrated together with the physical configuration of the machine tool 2. When the numerical controller 3 issues an operation command to the spindle drive system 21, the spindle motor 211 drives the spindle drive mechanism 212, the structure of the machine tool 2 including the tool 23 rotates, and the workpiece W is machined. Here, the spindle drive system 21 is controlled according to the spindle drive system angle θ1 based on the operation command, and then the actual angle of the tool 23 becomes the tool angle θ2 under the influence of the tool-side mechanical dynamics MD1. The series of cutting processes M in which the tool 23 penetrates the workpiece W is carried out to form a machined surface while generating chips. The cutting torque T _c generated thereby is influenced by the tool-side mechanical dynamics MD1 through the structure and returns to the spindle drive system 21 as a disturbance torque T _d . The machine tool 2 outputs a feedback signal to the numerical controller 3. If the state of the spindle drive system 21 which has received the disturbance torque T _d is different from the operation command, the numerical controller 3 changes the operation command based on the feedback signal transmitted from the spindle drive system 21.

9 is a representation that shows the relationship between the 1 illustrated feed drive system 22-2, the mechanical dynamics and the cutting process M. 10 is a representation which shows the 9 illustrated together with the physical configuration of the machine tool 2. The numerical controller 3 issues an operation command to the feed drive system 22-2, and then the workpiece W is machined by the relative movement between the tool 23 and the workpiece W. At this time, the servo motor 221-2 of the feed drive system 22-2 drives the feed drive mechanism 222-2 based on the operation command, and thus the drive system displacement r1 occurs in the table 24. The actual displacement generated in the workpiece W becomes the structure displacement r2 reflecting the influence of the workpiece-side mechanical dynamics MD2 when the drive system displacement r1 occurs. The cutting force F _c generated thereby returns to the feed drive system 22-2 as a disturbance force F _d through the structure. If the state of the feed drive system 22-2 which has received the disturbance force F _d is different from the operation command, the numerical controller 3 changes the operation command based on the feedback signal transmitted from the feed drive system 22-2.

Although the spindle drive system 21 and the feed drive system 22 have been described separately above for explanation with reference to 7 to 10 As described, displacement and force propagation during machining occur simultaneously in the spindle drive system 21 and the feed drive system 22.

As described above, in cutting, the cutting process M, the mechanical dynamics and the drive system 20 form a coupled system, and the numerical control device 3 participates in the cutting process M through the drive system 20 and the mechanical dynamics. In addition, in the cutting process, between the tool 23 and the workpiece W, the machining point at which the cutting force F _c is generated along with the generation of chips and thus it is not possible to directly detect the cutting force F _c by installing a sensor. Therefore, in order to accurately evaluate cutting including the movement of the tool 23 and the workpiece W, it is necessary to perform a simulation that includes the operation of the spindle drive system 21 and the feed drive system 22 in addition to the cutting process M and the mechanical dynamics.

Next, concrete examples of the cutting process model 321, the dynamic model 322, the spindle drive control model 323, and the feed drive control model 324 stored in the storage unit 32 will be described. These models are used when the coupled simulation unit 33 described later performs simulation.

The cutting process model 321 represents a cutting characteristic between the tool 23 and the workpiece W. More specifically, the cutting process model 321 is a mathematical model that expresses the cutting force F _c generated according to the positional relationship between the tool 23 and the workpiece W. The following formula (1) is an example of a formula that expresses the cutting force F _c generated while the cutting edge of the tool 23 is in contact with the workpiece W. The formula (1) represents the exact cutting force ΔF _c per cross section of the tool 23 using the specific cutting resistance K _c , the edge force coefficient K _e , the exact thickness Δa of the cross section of the tool 23, the uncut chip thickness h of the workpiece W, the rotation angle φ of the tool 23, and the time t. The total cutting force F _c generated by cutting with the tool 23 can be calculated by adding the exact cutting force ΔF _c given by the formula (1) in the axial direction of the tool 23. The uncut chip thickness h of the workpiece W is the distance between the previous machined surface and the current machining target surface in the radial direction of the tool 23. The formula (1) indicates that the cutting force F _c can be calculated by the sum of a force proportional to the uncut chip thickness h and a certain amount of force, which is called the edge force.
Formula 1: $$\Delta F_c=\Delta a\left(K_{c}H\left(\phi \left(t\right)\right)+K_e\right)$$

The uncut chip thickness h can be expressed by the formula (2) below. The uncut chip thickness h is represented by the sum of a component representing the nominal uncut chip thickness determined by the feed amount c of the tool 23 per tooth, a component representing the relative vibration of the tool 23 and the workpiece W, and a component representing the increase or decrease of the uncut chip thickness due to a difference in the turning radius of cutting edges of the tool 23 when the tool 23 includes a plurality of cutting edges. The component representing the relative vibration of the tool 23 and the workpiece W is represented by the difference between the tool 23 radial component u _r of the relative displacement between the tool 23 and the workpiece W at the time of current cutting of the machining target surface and the tool radial component w _r of the relative displacement between the tool 23 and the workpiece W transferred to the previous machined surface. The component representing the increase or decrease of the uncut chip thickness due to the difference in the turning radius of the cutting edges is represented by the turning radius correction amount Δe of the cutting edge of the tool 23.
Formula 2: $$H\left(\phi \left(t\right)\right)=csin\phi \left(t\right)+\left\{u_r\left(\phi ,t\right)-w_r\left(\phi ,t\right)\right\}+\Delta e\left(\phi ,t,\right)$$

Formula (1) is an example of the cutting process model 321, and the cutting process model 321 is not limited to the above. An example is a model that calculates the cutting force F _c using voxels expressing the shape of the tool 23 and the shape of the workpiece W.

The dynamic model 322 represents a dynamic property of a structure that generates vibration during operation of the machine tool 2. Specifically, the dynamic model 322 is a mathematical model that shows that a structure is dynamically displaced when a dynamic force is applied to the structure. For example, the behavior of the workpiece W when the cutting force F _c is applied to the workpiece W connected to the drive system 20 can be represented by the formula (3) below.
Formula 3: $$\{\begin{array}{c}{F}_c=m\ddot{u}+C\left(\dot{u}-\dot{v}\right)+K\left(u-v\right)\\ F_d=+C\left(\dot{u}-\dot{v}\right)+K\left(u-v\right)\end{array}$$

Formula (3) is an example of a formula expressing the vibration of the workpiece W. The cutting force F _c generated between the tool 23 and the workpiece W is calculated using the relative displacement u between the tool 23 and the workpiece W, the relative displacement v of the drive system 20, the equivalent mass m of the workpiece W, the equivalent viscosity coefficient C of the workpiece W and the equivalent spring constant K of the workpiece W. Formula (3) represents the mechanical dynamics in which the cutting force F _c is transmitted as a disturbance force F _d through the workpiece W to the drive system 20.

Note that the dynamic model 322 is not limited to the formula (3). An example is a model in which the shape of the workpiece W is expressed by voxels and the displacement when a structure vibrates is calculated using a finite element method (FEM) analysis. Although the dynamic model 322 described here only expresses the vibration of the workpiece W, the dynamic model 322 may express the vibration of the tool 23 or another structure instead of the workpiece W. Alternatively, the dynamic model 322 may express the vibrations of both the tool 23 and the workpiece W.

The spindle drive control model 323 is a mathematical model representing the spindle drive system 21 included in the machine tool 2 and a spindle drive controller included in the drive control unit 35 of the numerical control device 3 and controlling the spindle drive system 21. 11 is a diagram for explaining an example of the spindle drive control model 323 in 1 . The spindle drive control model 323 is a mathematical model for the case where the position and speed of the spindle drive system 21 are controlled by the position controller and the speed controller included in the spindle drive controller in a situation where the disturbance torque T _d caused by the cutting torque T _c is transmitted to the spindle drive system 21 when the spindle rotation angle command is issued. This mathematical model outputs the actual spindle rotation angle θ in response to the spindle rotation angle command θ _r input to the controller _. Here, K _pp1 , K _vp1 , and K _vi1 are control gains, which are the proportional gain K _pp1 for position control, the proportional gain K _vp1 for speed control, and the integral gain K _vi1 for speed control, respectively. P ₁ (s) is a torque-to-position transfer function for the entire spindle drive system 21, where s is a complex number. P ₁ (s) can be identified from the actual response of the spindle drive system 21 by using a known system identification method. Although the spindle drive system 21 is modeled here as a single inertia system, the spindle drive system 21 can be modeled as a multiple inertia system. In addition, feedforward control can be added to the spindle drive control.

The feed drive control model 324 is a mathematical model representing the feed drive system 22 of the machine tool 2 and a feed drive controller included in the drive control unit 35 of the numerical controller 3. 12 is a diagram for explaining an example of the feed drive control model 324 in 1 . The feed drive control model 324 is a mathematical model for the case where the position and speed of the feed drive system 22 are controlled by the position control and the speed control included in the feed drive control in a situation where the disturbance force F _d caused by the cutting force F _c is transmitted to the feed drive system 21 when the position command for the feed drive system is issued. In response to the input of the position command x _r for the feed drive system, the mathematical model outputs the actual position x of the feed drive system. Here, K _pp2 , K _vp2 and K _vi2 are control gains, which are the proportional gain K _pp2 for position control, the proportional gain K _vp2 for speed control and the integral gain K _vi2 for speed control, respectively. P ₂ (s) is a force to position transfer function for the entire feed drive system 22, where s is a complex number. P ₂ (s) can be identified from the actual response of the feed drive system 22 by using a known system identification method. Although the feed drive system 22 is modeled here as a single inertia system, the feed drive system 22 can be modeled as a multiple inertia system. In addition, feedforward control can be added to the feed drive control.

The coupled simulation unit 33 simulates machining in which each of a plurality of operation commands output from the command generation unit 31 is given to the machine tool 2, and calculates process information indicating the simulation result. The process information includes parameters that enable comparison of machining errors, and includes, for example, the uncut chip thickness of the workpiece W, the cutting force F _c , the disturbance force F _d and the like. Here, the uncut chip thickness of the workpiece W is For example, the uncut chip thickness of the workpiece W per tooth of the tool 23. The coupled simulation unit 33 can simulate the machining performed by the machine tool 2, reflecting the influence of the operation of the drive system 20 including the spindle drive system 21 and the feed drive system 22, and the influence of the dynamics of the structure that generates vibration in the cutting process M during the operation of the machine tool 2. The coupled simulation unit 33 performs the number of simulations equal to the number of operation commands generated by the command generation unit 31, and generates the number of process information indicating simulation results equal to the number of operation errors. The coupled simulation unit 33 outputs the plurality of generated process information to the process evaluation unit 34.

The coupled simulation unit 33 issues the operation command under the predetermined cutting condition output from the command generation unit 31 to the cutting process model 321, the dynamic model 322, the spindle drive control model 323, and the feed drive control model 324, whereby the coupled simulation unit 33 simulates machining performed by the machine tool 2 and calculates process information indicating the simulation result. At this time, the coupled simulation unit 33 may use the cutting process model 321, the dynamic model 322, the spindle drive control model 323, the feed drive control model 324, and the cutting condition information 325 stored in the storage unit 32. When the cutting condition information 325 stored in the storage unit 32 is used, the predetermined cutting condition is the cutting condition indicated by the cutting condition information 325.

The coupled simulation unit 33 carries out a simulation in which the cutting process M between the tool 23 and the workpiece W, the mechanical dynamics of the structure of the machine tool 2, the operation of the spindle drive system 21 and the operation of the feed drive system 22 are coupled. Based on the 7 to 10 In the relationships illustrated, the coupled simulation unit 33 performs a coupled simulation in the situation that the basic operation command and the corrected operation command are each given to a coupled model under the cutting condition described in the cutting condition information 325, and the coupled simulation unit 33 calculates the time series information and the frequency component information thereof. The coupled model is a model obtained by combining the cutting process model 321, the dynamic model 322, the spindle drive control model 323, and the feed drive control model 324. The coupled simulation is a simulation of the drive signal, the spindle drive system angle θ1, the drive system displacement r1, the tool angle θ2, the structural displacement r2 of the feed system, the uncut chip thickness h of the workpiece, the cutting torque T _c , the cutting force F _c , the disturbance torque T _d , the disturbance force F _d and the feedback signal.

The process evaluation unit 34 evaluates the amount of machining error by using each of the plurality of operation commands based on the plurality of process information output from the coupled simulation unit 33, and selects an operation command to be given to the machine tool 2 from the basic operation command and the corrected operation command generated by the command generation unit 31. The process evaluation unit 34 outputs a command selection signal indicating the selected operation command to the drive control unit 35.

An example of an evaluation method in the process evaluation unit 34 will be described below. The process evaluation unit 34 may evaluate the magnitude of the machining error based on the temporal change of the uncut chip thickness h of the workpiece W. The process evaluation unit 34 evaluates that the smaller the increase in the uncut chip thickness h of the workpiece W, the smaller the magnitude of the machining error. The process evaluation unit 34 may select an operation command that minimizes the increase in the uncut chip thickness h as the operation command to be given to the machine tool 2. The uncut chip thickness h represents the vibration between the tool 23 and the workpiece W. When vibration called chatter vibration occurs between the tool 23 and the workpiece W, the amplitude increases with the passage of time, resulting in deterioration of the machining error. Therefore, the process evaluation unit 34 can select an operation command that minimizes the vibration between the tool 23 and the workpiece W by evaluating the temporal change of the uncut chip thickness h. The operation command that minimizes the vibration between the tool 23 and the workpiece W can minimize the machining error caused by the vibration between the tool 23 and the workpiece W.

The process evaluation unit 34 may also evaluate the magnitude of the machining error based on the maximum amplitude of the disturbance force F _d or the disturbance torque T _d obtained when each operation command is executed. The process evaluation unit 34 evaluates that the smaller the maximum amplitude of the disturbance force F _d or the disturbance torque T _d , the smaller the magnitude of the machining error. The process evaluation unit 34 may select an operation command that minimizes the maximum amplitude as the operation command to be given to the machine tool 2. The smaller the maximum amplitude of the disturbance force F _d or the disturbance torque T _d , the smaller the vibration of the drive system 20 caused by the disturbance force F _d or the disturbance torque T _d . Therefore, by selecting the operation command that minimizes the maximum amplitude of the disturbance force F _d or the disturbance torque T _d , it is possible to minimize the machining error caused by the vibration of the drive system 20.

In addition, the process evaluation unit 34 may compare the time waveform of the process information calculated by the coupled simulation unit 33 with a preset target profile and evaluate the magnitude of the machining error based on the deviation from the target profile. The target profile is a profile in which the machining error becomes less than or equal to an allowable value, and is set in advance in the process evaluation unit 34, for example. The process evaluation unit 34 evaluates that the smaller the deviation from the target profile, the smaller the magnitude of the machining error. The process evaluation unit 34 may evaluate the deviation from the target profile based on a loss function such as residual sum of squares, or may evaluate the deviation from the target profile based on a machine learning method such as pattern-based search. The process evaluation unit 34 may minimize the machining error by selecting an operation command that minimizes the deviation from the target profile.

The process evaluation unit 34 may evaluate the amount of machining error using any of the above-described variety of evaluation methods, or may use a combination of the above-described variety of evaluation methods.

The drive control unit 35 controls the drive system 20 of the machine tool 2 based on the operation command indicated by the command selection signal output from the process evaluation unit 34 from the plurality of operation commands generated by the command generation unit 31. The drive control unit 35 includes therein the spindle drive controller for controlling the spindle drive system 21 and the feed drive controller for controlling the feed drive system 22. While monitoring signals from the encoder provided in the spindle drive system 21, the spindle drive controller outputs a command to the spindle motor 211 such that the position and speed of the spindle drive system 21 have amounts specified by the operation command. While monitoring signals from the encoder provided in the feed drive system 22, the feed drive controller outputs a command to the servo motor 221 such that the position and speed of the feed drive system 22 have amounts specified by the operation command.

13 is a flow chart to explain the operation of the 1 illustrated numerical control device 3. Once the machining system 1 starts operation, the command generation unit 31 of the numerical control device 3 reads out the numerical control program 4 and analyzes the read out numerical control program 4. Then, the command generation unit 31 generates: a basic operation command for causing the machine tool 2 to execute the command described in the numerical control program 4, and a corrected operation command obtained by correcting the basic operation command (step S101). After generating a basic operation command and one or more patterns of corrected operation commands, the command generation unit 31 outputs the generated operation commands to the coupled simulation unit 33.

The coupled simulation unit 33 performs a coupled simulation for each of the operation commands output from the command generation unit 31 to calculate a variety of process information (step S102). The coupled simulation unit 33 outputs the calculated process information to the process evaluation unit 34.

The process evaluation unit 34 compares and evaluates the plurality of process information, evaluates the amount of machining error when using each operation command, and selects an operation command to be given to the machine tool 2 from the basic operation command and the corrected operation command (step S103). The process evaluation unit 34 outputs a command selection signal indicating the selected operation command to the drive control unit 35.

The drive control unit 35 controls the operation of the machine tool 2 using of the selected operation command based on the selection signal output from the process evaluation unit 34 (step S104). The command generation unit 31 determines whether the reading of all the commands described in the numerical control program 4 is completed (step S105). In response to determining that the reading is not completed (step S105: No), the command generation unit 31 repeats the processing of step S101. In response to determining that the reading is completed (step S105: Yes), the machining system 1 terminates the operation.

As described above, in the machining system 1 according to the first embodiment, the numerical controller 3 calculates process information reflecting the influence of the operation of the drive system 20 and the dynamics of the structure that generates vibration during the operation of the machine tool 2 in the cutting process M of the workpiece W with the tool 23, the process information indicating a result of simulation of machining in which the basic operation command generated based on the numerical control program and the corrected operation error obtained by correcting the basic operation command are given to the machine tool 2. Then, the numerical controller 3 selects an operation command to be given to the machine tool 2 based on the evaluation result of the process information. Therefore, the numerical controller 3 can reduce the machining error even when the machining error occurs due to the mutual influence of the cutting process, the operation of the drive system 20, and the mechanical dynamics of the structure that generates vibration during the operation of the machine tool 2.

The coupled simulation unit 33 calculates the process information when the operation command under a predetermined cutting condition is given to the cutting process model 321 representing a cutting property between the tool 23 and the workpiece W, the dynamics model 322 representing a dynamic property of the structure that generates vibration during the operation of the machine tool 2, the spindle drive control model 323 representing the spindle drive system 21 and the spindle drive controller controlling the spindle drive system 21, and the feed drive control model 324 representing the feed drive system 22 and the feed drive controller controlling the feed drive system 22. By performing the coupled simulation using the mathematical models, it is possible to accurately evaluate the influence of the operation command on the cutting process M by the drive system 20 and the mechanical dynamics.

In the first embodiment, the storage unit 32 in which the cutting process model 321, the dynamic model 322, the spindle drive control model 323, the feed drive control model 324, and the cutting condition information 325 indicating the cutting conditions are stored is provided in the numerical control device 3; however, the storage unit 32 may be provided outside the numerical control device 3.

The command generation unit 31 may generate, as the corrected operation command, a command in which the relative path along which the tool 23 moves with respect to the workpiece W is the same as that of the basic operation command and the feed amount or the uncut chip thickness h of the workpiece W per tooth is changed. For example, the command generation unit 31 may set, as the corrected operation command, a command in which the spindle speed and/or the feed rate of the basic operation command is changed according to the feed amount or the uncut chip thickness h of the workpiece W per tooth. Concretely, the command generation unit 31 may generate a corrected operation command by superimposing a predetermined profile variation on the spindle speed and/or the feed rate of the basic operation command. By generating the corrected operation command in this way, it is possible to generate the operation command that reduces the machining error without changing the shape of the workpiece W.

The storage unit 32 can store different models and cutting conditions depending on the cutting process described in the numerical control program 4. The coupled simulation unit 33 can perform simulation using different models and cutting conditions depending on the cutting process. In the first embodiment, the machine tool 2 includes a spindle drive system 21 and one or more feed drive systems 22, but the machine tool 2 may include a plurality of spindle drive systems 21. Even if the machine tool 2 includes a plurality of spindle drive systems 21, the simulation unit 33 described in 13 The operation illustrated can be carried out in a similar way.

In the first embodiment, the machine tool 2 in which the tool 23 is connected to the spindle drive system 21 and the tool 23 rotates, such as a machining center, has been described. The machine tool 2 however, may have a configuration in which the workpiece W is connected to the spindle drive system 21 and the workpiece W rotates, such as a numerically controlled (NC) lathe. In this case, by replacing the feed amount per tooth with the feed amount per rotation of the spindle, the command generation unit 31 can select an operation command that reduces the machining error among a plurality of operation commands without changing the path determined by the numerical control program 4.

Second embodiment.

14 is a diagram illustrating a functional configuration of a machining system 1a according to the second embodiment. Functional components having the same functions as those in the first embodiment are denoted by the same reference numerals as in the first embodiment, and redundant explanations are omitted. Differences from the first embodiment will be mainly described below. The machining system 1a differs from the machining system 1 in that an operation command is generated based on a simulation result.

The machining system 1a includes the machine tool 2 and a numerical control device 3a. Like the numerical control device 3, the numerical control device 3a controls the machine tool 2 based on the command described in the numerical control program 4. The numerical control device 3a includes a command generation unit 31a, a storage unit 32a, a coupled simulation unit 33a, the process evaluation unit 34, and the drive control unit 35.

When generating a corrected operation command, the command generation unit 31a may use the process information output from the coupled simulation unit 33a. The command generation unit 31a may set an operation command obtained by correcting the basic operation command based on the process information as a corrected operation command. Here, the command generation unit 31a may use the cutting process model 321, the dynamic model 322, the spindle drive control model 323, the feed drive control model 324, and the cutting condition information 325 stored in the storage unit 32a. More concretely, the command generation unit 31a generates a corrected operation command by adding, to the basic operation command, a variation that compensates for the amplitude or phase of the dynamic vibration component imposed on the uncut chip thickness of the workpiece W included in the process information. The dynamic vibration component corresponds to the second term on the right side of the above formula (2). The command generation unit 31a may apply to the basic operation command a variation that attenuates the amplitude or phase of the vibration component using a band-stop filter that attenuates the amplitude of the dynamic vibration component imposed on the uncut chip thickness of the workpiece W, or using a phase compensation filter that compensates the phase delay of the vibration component with respect to the timing at which the cutting edge of the tool 23 cuts.

Like the storage unit 32, the storage unit 32a stores the cutting process model 321, the dynamic model 322, the spindle drive control model 323, the feed drive control model 324, and the cutting condition information 325, and outputs the stored information to the coupled simulation unit 33a. The storage unit 32a can further output the stored information to the command generation unit 31a.

Like the coupled simulation unit 33, the coupled simulation unit 33a calculates process information indicating results of a machining simulation in which the basic operation command and the corrected operation command are respectively issued to the machine tool 2. The coupled simulation unit 33a outputs the calculated process information to the process evaluation unit 34 and also to the command generation unit 31a.

15 is a flow chart to explain the operation of the 14 illustrated numerical control device 3a. Once the machining system 1a starts operation, the command generation unit 31 of the numerical control device 3a reads out the numerical control program 4, analyzes the read out numerical control program 4, and generates a basic operation command for causing the machine tool 2 to execute the command described in the numerical control program 4 (step S201). The command generation unit 31a outputs the generated basic operation command to the coupled simulation unit 33a.

The coupled simulation unit 33a executes a coupled simulation in which the basic operation command output from the command generation unit 31a is executed by the machine tool 2 to generate process information (step S202). The coupled simulation unit unit 33a outputs the generated process information to both the process evaluation unit 34 and the command generation unit 31a.

The command generation unit 31a corrects the basic operation command based on the process information output as a result of executing step S202 and generates a corrected operation command (step S203). The command generation unit 31a outputs the generated corrected operation command to the coupled simulation unit 33a.

The coupled simulation unit 33a performs a coupled simulation in which the corrected operation command output from the command generation unit 31a is executed by the machine tool 2 to generate process information (step S204). The coupled simulation unit 33a outputs the generated process information to both the process evaluation unit 34 and the command generation unit 31a.

The process evaluation unit 34 compares and evaluates the plurality of process information, evaluates the amount of machining error when using each operation command, and selects an operation command to be given to the machine tool 2 from the basic operation command and the corrected operation command (step S205). The process evaluation unit 34 outputs a command selection signal indicating the selected operation command to the drive control unit 35.

The drive control unit 35 controls the operation of the machine tool 2 using the selected operation command based on the selection signal output from the process evaluation unit 34 (step S206). The command generation unit 31a determines whether the reading of all the commands described in the numerical control program 4 is completed (step S207). In response to determining that the reading is not completed (step S207: No), the command generation unit 31a repeats the processing of step S201. In response to determining that the reading is completed (step S207: Yes), the machining system 1a terminates the operation.

In the above example, the command generation unit 31a generates a corrected operation command based on the process information indicating a simulation result when the basic operation command is issued to the machine tool 2. However, the command generation unit 31a may further generate a corrected operation command based on the process information indicating a simulation result when the corrected operation command is issued to the machine tool 2. In this case, it is possible to adopt a method of searching for a corrected operation command capable of reducing the vibration of the uncut chip thickness of the workpiece W using a machine learning method while setting the amplitude or phase of the vibration component of the uncut chip thickness of the workpiece W as an evaluation value.

16 is a diagram showing an exemplary configuration of a learning device 50 with respect to the 14 numerical control device 3a illustrated. For example, the learning device 50 in the numerical control device 3a illustrated in 14 or may be an information processing device different from the numerical control device 3a. The learning device 50 includes a learning data acquisition unit 51 and a model generation unit 52.

The learning data acquisition unit 51 acquires, as learning data, the operation command generated by the command generation unit 31a and the process information corresponding to the operation command, that is, the process information indicating a simulation result when the operation command is given to the machine tool 2. The learning data acquisition unit 51 may output the acquired learning data to the model generation unit 52. Note that the learning data acquisition unit 51 may acquire all of the process information or may acquire a part of the process information. For example, the learning data acquisition unit 51 may acquire, as learning data, a parameter indicating the size of the machining error in the process information. For example, the learning data acquisition unit 51 may acquire, as learning data, the uncut chip thickness of the workpiece W or the amplitude or phase of the vibration component of the uncut chip thickness of the workpiece W.

The model generation unit 52 learns a newly corrected operation command based on the learning data including the operation command and the process information indicating a simulation result obtained when the operation command is given to the machine tool 2. That is, the model generation unit 52 generates a learned model for deriving a newly corrected operation command from the process information of the numerical control device 3a. The model generation unit 52 outputs the generated learned model to a learned model storage unit 53.

The learning algorithm used by the model generation unit 52 may be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, a case where reinforcement learning is applied will be described. In reinforcement learning, an agent (subject of an action) in an environment observes an environmental parameter indicating the current state and determines the action to be taken. The environment changes dynamically due to the agent's behavior, and the agent is given a reward according to the change in the environment. The agent repeats this to learn an action policy that maximizes the reward through a series of actions. Q-learning and TD-learning are known as representative methods of reinforcement learning. In the case of Q-learning, for example, a general update expression for the action value function Q(s, a) is represented by the formula (4) below.
Formula 4: $$Q\left(s_t,a_t\right)\leftarrow Q\left(s_t,a_t\right)+\alpha \left(r_{t+1}+\gamma \underset{a}{\text{Max}}Q\left(s_{t+1},a\right)-Q\left(s_t,a_t\right)\right)$$

In the formula (4), s _t represents the state of the environment at a time t, and at represents the action at the time t. The action at changes the state to s _t+1 . In addition, r _t+1 represents the reward that can be obtained by changing the state, γ represents a discount rate, and α represents a learning coefficient. Note that γ has a value in the range of 0<γ≤1, and α has a value in the range of 0<α≤1. The corrected operation command serves as the action at, the process information serves as the state s _t, and the learning device 50 learns the best action at at the state at the time t.

The update expression represented by formula (4) increases the action value Q if the action value Q of the action "a" having the highest Q value at time t+1 is larger than the action value Q of the action "a" executed at time t, and decreases the action value Q otherwise. In other words, the action value function Q(S,a) is updated such that the action value Q of the action "a" at time t is brought closer to the best action value at time t+1. As a result, the best action value Q in a certain environment propagates to the action values Q in the previous environments one after another.

As described above, in the case where a learned model is generated by reinforcement learning, the model generating unit 52 includes a reward calculating unit 54 and a function updating unit 55.

The reward calculation unit 54 calculates a reward based on the operation command and the process information. The reward calculation unit 54 calculates a reward r based on reward criteria D including a reward increase criterion D1 and a reward decrease criterion D2. For example, the reward criteria D is determined based on the magnitude of the machining error indicated by the process information. As the parameter indicating the magnitude of the machining error, for example, the amplitude of the vibration component of the uncut chip thickness of the workpiece W is used. For example, the reward increase criterion D1 may be defined that the amplitude of the vibration component of the uncut chip thickness of the workpiece W is less than a threshold value, and the reward decrease criterion D2 may be defined that the amplitude of the vibration component of the uncut chip thickness of the workpiece W is greater than or equal to a threshold value. For example, the reward calculation unit 54 increases the reward r by giving a reward of “+1” when the reward increase criterion D1 is satisfied, and decreases the reward r by giving a reward of “-1” when the reward decrease criterion D2 is satisfied. The reward calculation unit 54 outputs the calculated reward r to the function updating unit 55. As another example, as the parameter indicating the magnitude of the machining error, the phase of the vibration component of the uncut chip thickness of the workpiece W may be used in addition to the amplitude of the vibration component of the chip thickness. Here, the phase of the vibration component of the uncut chip thickness is the phase of the vibration superimposed on the chip shape at the moment when the cutting edge of the tool 23 starts cutting the workpiece W. In this case, the reward increase criterion D1 may be defined such that the phase of the vibration component of the uncut chip thickness of the workpiece W is a value within a predetermined range, and the reward decrease criterion D2 may be defined such that the phase of the vibration component of the uncut chip thickness of the workpiece W is a value outside the above range.

The function updating unit 55 updates the function for determining a corrected operation command according to the reward r calculated by the reward calculating unit 54 and outputs the updated function to the learned model storage unit 53. For example, in the case of Q-learning, the action value function Q (s _t , a _t ) represented by formula (4) is used as a function for calculating a corrected operation command.

The above learning is repeatedly performed. The learned model storage unit 53 stores the action value function Q (s _t , a _t ) updated by the function updating unit 55, that is, the learned model.

Next, the learning process by the learning device 50 is described with reference to 17 described. 17 is a flow chart to explain the learning process of the 16 illustrated learning device 50.

The learning data acquisition unit 51 acquires, as learning data, the operation command generated by the command generation unit 31a and the process information indicating a simulation result when the operation command is issued to the machine tool 2 (step S301).

The model generation unit 52 calculates the reward r based on the operation command and the process information included in the learning data acquired by the learning data acquisition unit 51 (step S302). Specifically, the reward calculation unit 54 acquires the operation command and the process information and determines whether to increase the reward r or to decrease the reward r based on the predetermined reward criteria D (step S303).

In response to determining that the reward r is to be increased (step S303: increase), the reward calculation unit 54 increases the reward r (step S304). In response to determining that the reward r is to be decreased (step S303: decrease), the reward calculation unit 54 decreases the reward r (step S305).

The function updating unit 55 updates the action value function Q (s _t , a _t ) stored in the learned model storage unit 53 based on the reward r calculated by the reward calculation unit 54 (step S306).

The learning device 50 repeatedly executes the above processing from step S301 to step S306 and stores the generated action value function Q (s _t , a _t ) as a learned model.

In 16 the learned model storage unit 53 is provided outside the learning device 50, but the learning device 50 may include the learned model storage unit 53 therein. When the learning device 50 is provided in the numerical control device 3a, the learned model storage unit 53 may be provided in the same storage device as the storage unit 32a, or may be provided in a different storage device.

18 is a diagram showing an exemplary configuration of an inference device 60 with respect to the 14 illustrated numerical control device 3a. The inference device 60 includes a data acquisition unit 61 and an inference unit 62. The inference device 60 may be provided in the numerical control device 3a, or may be an information processing device different from the numerical control device 3a. The inference device 60 is provided, for example, in the command generation unit 31a of the numerical control device 3a.

The data acquisition unit 61 acquires the process information output from the coupled simulation unit 33a. The data acquisition unit 61 outputs the acquired data to the inference unit 62.

The inference unit 62 uses the learned model stored in the learned model storage unit 53 to derive a newly corrected operation command from the process information acquired by the data acquisition unit 61. That is, the inference unit 62 can derive a corrected operation command suitable for the process information by inputting the process information output by the data acquisition unit 61 into the learned model.

In the above description, the inference device 60 outputs the corrected operation command using the learned model as a result of performing machine learning using the data acquired by the numerical control device 3a. However, the learned model may be acquired from another numerical control device 3a and the corrected operation command may be output based on the learned model.

19 is a flow chart to explain the operation of the 18 illustrated inference device 60. The data acquisition unit 61 of the inference device 60 acquires the process information as inference data (step S401) and outputs the acquired process information to the inference unit 62.

The inference unit 62 inputs the process information, which is the inference data acquired in step S401, into the learned model stored in the learned model storage unit 53 (step S402). The inference unit 62 outputs a corrected operation command as a result of inputting the process information into the learned model (step S403). Note that the command generation unit 31 of the numerical control device 3a acquires the corrected operation command output by the inference unit 62 and outputs the acquired corrected operation command to the coupled simulation unit 33a.

Although the inference unit 62 uses reinforcement learning as the learning algorithm in the above description, the learning algorithm used by the inference unit 62 is not limited to reinforcement learning. In addition to reinforcement learning, the inference unit 62 may also use supervised learning, unsupervised learning, semi-supervised learning, or the like as the learning algorithm.

The learning algorithm used by the model generation unit 52 may also be deep learning that learns feature extraction directly. Alternatively, other known methods such as neural networks, genetic programming, functional programming with inference, and support vector machines may be used to perform machine learning.

Note that the machine learning device 50 and the inference device 60 may each be a device separate from the numerical control device 3a and connected to the numerical control device 3a via a network, for example. Moreover, the learning device 50 and the inference device 60 may each be built into the numerical control device 3a. Furthermore, the learning device 50 and the inference device 60 may each be provided on a cloud server.

The model generation unit 52 may learn a corrected operation command using the learning data acquired from a plurality of numerical control devices 3a. Note that the model generation unit 52 may acquire learning data from a plurality of numerical control devices 3a used in the same area, or may learn a corrected operation command using learning data collected from a plurality of numerical control devices 3a independently operating in different areas. Moreover, in the middle of learning, it is possible to start collecting learning data from a new numerical control device 3a or stop collecting learning data from a numerical control device 3a. Moreover, the learning device 50 that has learned the corrected operation command for one numerical control device 3a can be applied to another numerical control device 3a, and the corrected operation command for the other numerical control device 3a can be relearned and updated.

As described above, the numerical control device 3a according to the second embodiment generates a basic operation command generated from the numerical control program 4 and a corrected operation command obtained by correcting the basic operation command using the process information, and selects an operation command to be given to the machine tool 2 based on the evaluation result of each operation command. Since the command generation unit 31a newly generates a corrected operation command based on the simulation result executed by the coupled simulation unit 33a, it is possible to generate a corrected operation command based on the characteristics of the drive system 20, the mechanical dynamics, and the cutting process M. Therefore, the numerical control device 3a can efficiently reduce the machining error.

From a corrected operation command, the numerical control device 3a may generate another corrected operation command. In this case, the corrected operation command is learned based on the learning data including the corrected operation command and the process information by machine learning using the learning device 50. The numerical control device 3a may use the corrected operation command output by the inference device 60 that derives the corrected operation command using the learned model that is the learning result of the learning device 50. By using machine learning, the numerical control device 3a can generate a corrected operation command in an exploratory manner, so that the machining system 1a can generate a corrected operation command capable of reducing the machining error without creating a rule for correcting the operation command in advance.

Third embodiment.

20 is a diagram illustrating a configuration of a machining system 1b according to the third embodiment. E is to Note that components having the same functions as those in the first embodiment are designated by the same reference numerals as those in the first embodiment, and redundant explanations are omitted. The following mainly describes differences from the first and second embodiments.

The machining system 1b includes a machine tool 2b and a numerical control device 3b. The machine tool 2b includes the spindle drive system 21, the feed drive system 22, the tool 23, the table 24 and a sensor 25.

The sensor 25 detects vibration of a structure that generates vibration during operation of the machine tool 2b. The sensor 25 is, for example, an acceleration sensor or a force sensor. Alternatively, the sensor 25 may be an encoder provided in advance within the drive system 20 for feedback control of the drive system 20. The sensor 25 is connected to the numerical control device 3b, and a signal detected by the sensor 25 is output as a sensor signal to the numerical control device 3b.

The numerical control device 3b includes a command generation unit 31, the storage unit 32, the coupled simulation unit 33, the process evaluation unit 34, and the drive control unit 35. The numerical control device 3b differs from the first and second embodiments in generating an operation command based on the sensor signal output from the sensor 25.

The command generation unit 31b generates a basic operation command similarly to the command generation unit 31 in the first embodiment. Furthermore, the command generation unit 31b can generate a corrected operation command by correcting the basic operation command based on the sensor signal output from the sensor 25. Concretely, the command generation unit 31b determines the feed amount or the uncut chip thickness of the workpiece W per tooth based on the sensor signal to generate a corrected operation command. For example, learning for the feed amount or the uncut chip thickness of the workpiece W per tooth is performed in advance according to the time waveform or the frequency spectrum of the sensor signal, and when the sensor signal is input, the command generation unit 31b determines the feed amount or the uncut chip thickness of the workpiece W per tooth using a machine learning method such as pattern matching. Alternatively, a correspondence table including the amplitude of the sensor signal and the feed amount or the uncut chip thickness of the workpiece W per tooth may be recorded in advance, and the command generation unit 31b may determine the feed amount or the uncut chip thickness of the workpiece W per tooth based on the correspondence table.

Since the operation of the numerical control device 3b is similar to the operation of the 13 numerical control device 3a illustrated except that the sensor signal is used to generate a corrected operation command, a detailed description thereof is omitted here.

In the above description, the command generation unit 31b generates a corrected operation command by correcting the basic operation command using the sensor signal, but the operation command may be corrected sequentially. That is, the command generation unit 31b may further generate a corrected operation command using a sensor signal detected when the corrected operation command generated using the sensor signal is given to the machine tool 2b. In this case, it is possible to use a method of searching for a corrected operation command that reduces the vibration of the sensor signal using a machine learning method such as reinforcement learning using the amplitude or phase of the vibration component of the sensor signal as an evaluation value.

In the case of using machine learning, for example, the learned model can be optimized using the 16 illustrated learning device 50 and a corrected operating command can be derived from the learned model using the 18 illustrated inference device 60. In this case, in the description provided in the second embodiment, the description of the method for generating a corrected operation command used by the numerical control device 3b according to the third embodiment is omitted by replacing the "process information" acquired by the learning data acquisition unit 51 and the data acquisition unit 61 with the "sensor signal". In this case, the operation command acquired by the learning data acquisition unit 51 is the operation command corresponding to the sensor signal, specifically, the operation command given to the machine tool 2b when the sensor signal is acquired.

As described above, in the numerical control device 3b according to the third embodiment, the machine tool 2b includes the sensor 25, and the command generation unit 31b of the numerical control device 3b can generate a corrected operation command based on the sensor signal. Therefore, the command generation unit 31b can correct the operation command according to the vibration state actually generated in the machine tool 2b, and can generate the operation command that effectively reduces the machining error.

Furthermore, by generating a corrected operation command in an exploratory manner using machine learning, the machining system 1b can generate a corrected operation command that reduces the machining error without creating a correction rule for the operation command in advance.

Next, a hardware configuration of the numerical control devices 3, 3a, and 3b, the learning device 50, and the inference device 60 according to the first to third embodiments will be described. The command generation units 31, 31a, and 31b, the coupled simulation units 33 and 33a, the process evaluation unit 34, and the drive control unit 35 of the numerical control devices 3, 3a, and 3b, the learning data acquisition unit 51 and the model generation unit 52 of the learning device 50, and the data acquisition unit 61 and the inference unit 62 of the inference device 60 are implemented by a processing circuit. The processing circuit may be implemented by dedicated hardware or may be a control circuit using a central processing unit (CPU).

If the above processing circuit is implemented by dedicated hardware, the processing circuit is controlled by the 21 illustrated processing circuit 90 is implemented. 21 is a diagram illustrating dedicated hardware for implementing the functions of the numerical control devices 3, 3a and 3b, the learning device 50 and the inference device 60 according to the first to third embodiments. The processing circuit 90 is a single circuit, a compound circuit, a programmed processor, a parallel programmed processor, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or a combination thereof.

When the above processing circuit is implemented by a control circuit using a CPU, this control circuit is, for example, a control circuit 91 having the 22 illustrated configuration. 22 is a diagram illustrating a configuration of the control circuit 91 for implementing the functions of the numerical control devices 3, 3a and 3b, the learning device 50 and the inference device 60 according to the first to third embodiments. As shown in 22 As illustrated, the control circuit 91 includes a processor 92 and a memory 93. The processor 92 is a CPU and is also referred to as an arithmetic device, microprocessor, microcomputer, digital signal processor (DSP), or the like. Examples of the memory 93 include a nonvolatile or volatile semiconductor memory, a magnetic disk, a flexible disk, an optical disk, a compact disc, a mini disc, a digital versatile disc (DVD), and the like. Examples of nonvolatile or volatile semiconductor memories include a random access memory (RAM), a read only memory (ROM), a flash memory, an erasable programmable ROM (EPROM), an electrically operated EPROM (EEPROM, registered trademark), and the like.

When the above processing circuit is implemented by the control circuit 91, the processor 92 reads and executes the program corresponding to the process of each component stored in the memory 93, thereby implementing the processing circuit. The memory 93 is also used as a temporary storage for each process executed by the processor 92.

It should be noted that the program executed by the processor 92 may be provided by being stored on a storage medium or provided via a communication path. Moreover, the functions of the numerical control devices 3, 3a and 3b, the learning device 50 and the inference device 60 according to the first to third embodiments may be implemented using any of the methods described in 21 illustrated processing circuit 90 and the in 22 illustrated control circuit 91 or using a combination of the processing circuit 90 and the control circuit 91.

The configurations described in the above-mentioned embodiments are examples. The embodiments may be combined with other well-known technology and with each other, and some of the configurations may be omitted or changed in a range not deviating from the gist.

List of reference symbols

1, 1a, 1b machining system; 2, 2b machine tool; 3, 3a, 3b numerical control device; 4 numerical control program; 20 drive system; 21 spindle drive system; 22, 22-1, 22-2 feed drive system; 23 tool; 24 table; 25 sensor; 31, 31a, 31b command generation unit; 32, 32a storage unit; 33, 33a coupled simulation unit; 34 process evaluation unit; 35 drive control unit; 50 learning device; 51 learning data acquisition unit; 52 model generation unit; 53 learned model storage unit; 54 reward calculation unit; 55 function updating unit; 60 inference device; 61 data acquisition unit; 62 inference unit; 90 processing circuit; 91 control circuit; 92 processor; 93 memory; 211 spindle motor; 212 spindle drive mechanism; 221, 221-1, 221-2 servo motor; 222, 222-1, 222-2 feed drive mechanism; 321 cutting process model; 322 dynamics model; 323 spindle drive control model; 324 feed drive control model; 325 cutting condition information; A1 cutting area per tooth; c feed amount; F _c cutting force; F _d disturbance force; M cutting process; MD1 tool-side mechanical dynamics; MD2 workpiece-side mechanical dynamics; P1, P2 position; R1 rotation direction; r1 drive system displacement; r2 structure displacement; T _c cutting torque; T _d disturbance torque; W workpiece; θ1 spindle drive system angle; θ2 tool angle.

QUOTES INCLUDED IN THE DESCRIPTION

This list of documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions.

Cited patent literature

• JP2013132733 [0004]

Claims (20)

A numerical control device that controls a machine tool by giving an operation command to the machine tool, the machine tool including a drive system that includes a spindle drive system for driving a tool for machining a workpiece or a spindle that rotates the workpiece, and a feed drive system for driving a feed axis that changes a relative position between the tool and the workpiece, the numerical control device comprising: a command generation unit for generating a basic operation command that is the operation command based on a numerical control program, and generating a corrected operation command that is the operation command obtained by correcting the basic operation command; a coupled simulation unit for calculating process information reflecting an influence of an operation of the drive system and dynamics of a structure that generates vibration during operation of the machine tool in machining the workpiece with the tool, the process information indicating results of simulation machining in which the basic operation command and the corrected operation command are respectively given to the machine tool; and a process evaluation unit for evaluating, based on a plurality of the process information, a magnitude of a machining error when using each of a plurality of the operation commands and selecting the operation command to be given to the machine tool from the basic operation command and the corrected operation command. Numerical control device according to Claim 1 , wherein the coupled simulation unit calculates the process information when using the operation command under a predetermined cutting condition for a cutting process model, the cutting process model being a mathematical model representing a cutting property between the tool and the workpiece, a dynamic model being a mathematical model representing a dynamic property of the structure, a spindle drive control model being a mathematical model representing the spindle drive system and a spindle drive controller controlling the spindle drive system, and a feed drive control model being a mathematical model representing the feed drive system and a feed drive controller controlling the feed drive system. Numerical control device according to Claim 2 , further comprising a storage unit for storing the cutting process model, the dynamic model, the spindle drive control model, the feed drive control model, and cutting condition information representing the cutting condition, wherein the coupled simulation unit calculates the process information using the cutting process model, the dynamic model, the spindle drive control model, the feed drive control model, and the cutting condition information stored in the storage unit. Numerical control device according to one of the Claims 1 until 3 wherein the command generation unit generates, as the corrected operation command, a command in which a relative path along which the tool moves with respect to the workpiece is the same as that of the basic operation command and at least one of a feed amount of the feed axis and an uncut chip thickness of the workpiece is changed. Numerical control device according to Claim 4 wherein the command generation unit sets, as the corrected operation command, an operation command in which at least one of a spindle speed and a feed rate of the basic operation command is changed according to the feed amount or the uncut chip thickness of the workpiece. Numerical control device according to one of the Claims 1 until 5 wherein the command generating unit generates the corrected operation command by superimposing a predetermined profile variation on at least one of a spindle speed and a feed rate of the basic operation command. Numerical control device according to one of the Claims 1 until 6 , wherein the command generation unit generates the corrected operation command based on the process information. Numerical control device according to Claim 7 wherein the command generation unit further generates a corrected operation command based on the process information when the corrected operation command is issued to the machine tool. Numerical control device according to Claim 7 or 8th , further comprising: a learning data acquisition unit for acquiring learning data including the process information and the operation command corresponding to the process information; and a model generation unit for generating a learned model for deriving a newly corrected operating command from the process information using the learning data. Numerical control device according to one of the Claims 7 until 9 , further comprising: a data acquisition unit for acquiring the process information; and an inference unit for outputting a newly corrected operation command from the process information acquired by the data acquisition unit using a learned model for deriving a newly corrected operation command from the process information. Numerical control device according to one of the Claims 1 until 6 wherein the machine tool further includes a sensor for detecting vibration of the structure during operation and outputting a sensor signal, and the command generating unit generates the corrected operation command based on the sensor signal. Numerical control device according to Claim 11 , further comprising: a learning data acquisition unit for acquiring learning data including the sensor signal and the operation command corresponding to the sensor signal; and a model generation unit for generating a learned model for deriving a newly corrected operation command from the sensor signal using the learning data. Numerical control device according to Claim 11 or 12 , further comprising: a data acquisition unit for acquiring the sensor signal; and an inference unit for outputting a newly corrected operation command from the sensor signal acquired by the data acquisition unit using a learned model for deriving a newly corrected operation command from the sensor signal. A machining system comprising: a machine tool including a drive system including a spindle drive system for driving a tool for machining a workpiece or a spindle rotating the workpiece, and a feed drive system for driving a feed axis that changes a relative position between the tool and the workpiece, the machine tool configured to machine the workpiece based on an operation command generated based on a numerical control program; and a numerical control device for controlling the machine tool by issuing the operation command to the machine tool, the numerical control device including: a command generation unit for generating a basic operation command that is the operation command based on the numerical control program and generating a corrected operation command that is the operation command obtained by correcting the basic operation command; a coupled simulation unit for calculating process information reflecting an influence of an operation of the drive system and dynamics of a structure that generates vibration during operation of the machine tool in a cutting process of the workpiece with the tool, the process information indicating results of simulation machining in which the basic operation command and the corrected operation command are respectively given to the machine tool; and a process evaluation unit for evaluating, based on a plurality of the process information, a magnitude of a machining error in use of each of a plurality of the operation commands and selecting the operation command to be given to the machine tool from the basic operation command and the corrected operation command. Processing system according to Claim 14 , further comprising a learning device including: a learning data acquisition unit for acquiring learning data including the operation command generated by the numerical control device and the process information indicating a simulation result when the operation command is given to the machine tool; and a model generation unit for generating a learned model for deriving a newly corrected operation command from the process information of the numerical control device using the learning data. Processing system according to Claim 14 or 15 , further comprising an inference device including: a data acquisition unit for acquiring the process information generated by the numerical control device; and an inference unit for outputting a newly corrected operation command from the process information acquired by the data acquisition unit using a learned model for deriving a newly corrected operation command from the process information. Processing system according to Claim 14 , wherein the machine tool further includes a sensor for detecting vibration of the structure during operation and outputting a sensor signal, and the machining system further comprises a learning device including: a learning data acquisition unit for acquiring learning data including the sensor signal obtained during control of the machine tool by the numerical control device and the operation command corresponding to the sensor signal; and a model generation unit for generating a learned model for deriving a newly corrected operation command from the sensor signal using the learning data. Processing system according to Claim 14 or 17 , wherein the machine tool further includes a sensor for detecting vibration of the structure during operation and outputting a sensor signal, and the machining system further comprises an inference device including: a data acquisition unit for acquiring the sensor signal when the numerical controller issues the operation command to the machine tool; and an inference unit for outputting a newly corrected operation command from the sensor signal acquired by the data acquisition unit by using a learned model for deriving a newly corrected operation command from the sensor signal. A numerical control method executed by a numerical control device that controls a machine tool by giving an operation command to the machine tool, the machine tool including a drive system including a spindle drive system for driving a tool for machining a workpiece or a spindle that rotates the workpiece, and a feed drive system for driving a feed axis that changes a relative position between the tool and the workpiece, the numerical control method comprising: a step of generating a basic operation command that is the operation command based on a numerical control program; a step of generating a corrected operation command that is the operation command obtained by correcting the basic operation command; a step of calculating process information reflecting an influence of an operation of the drive system and a dynamic of a structure that generates vibration during operation of the machine tool in a cutting process of the workpiece with the tool, the process information indicating results of simulation machining in which the basic operation command and the corrected operation command are respectively given to the machine tool; and a step of evaluating, based on a plurality of the process information, a magnitude of a machining error when using each of a plurality of the operation commands and selecting the operation command to be given to the machine tool from the basic operation command and the corrected operation command. A machining method for machining a workpiece by issuing an operation command to a machine tool, the machine tool including a drive system that includes a spindle drive system for driving a tool for machining a workpiece or a spindle that rotates the workpiece, and a feed drive system for driving a feed axis that changes a relative position between the tool and the workpiece, the machining method comprising: a step of generating a basic operation command that is the operation command based on a numerical control program; a step of generating a corrected operation command that is the operation command obtained by correcting the basic operation command; a step of calculating process information that reflects an influence of an operation of the drive system and dynamics of a structure that generates vibration during operation of the machine tool in a cutting process of the workpiece with the tool, the process information indicating results of simulation machining in which the basic operation command and the corrected operation command are respectively issued to the machine tool; a step of evaluating, based on a plurality of the process information, a magnitude of a machining error when using each of a plurality of the operation commands and selecting the operation command to be given to the machine tool from the basic operation command and the corrected operation command; a step of giving the selected operation command to the machine tool; and a step of machining the workpiece using the tool while the Drive system is operated according to the operating command.

Images