WO2020194751A1 - Numerical control device, electric discharge machining device, and electric discharge machining method - Google Patents

Numerical control device, electric discharge machining device, and electric discharge machining method Download PDF

Info

Publication number
WO2020194751A1
WO2020194751A1 PCT/JP2019/013886 JP2019013886W WO2020194751A1 WO 2020194751 A1 WO2020194751 A1 WO 2020194751A1 JP 2019013886 W JP2019013886 W JP 2019013886W WO 2020194751 A1 WO2020194751 A1 WO 2020194751A1
Authority
WO
WIPO (PCT)
Prior art keywords
workpiece
command value
discharge
electrode
machining
Prior art date
Application number
PCT/JP2019/013886
Other languages
French (fr)
Japanese (ja)
Inventor
中川　孝幸
大介 関本
Original Assignee
三菱電機株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 三菱電機株式会社 filed Critical 三菱電機株式会社
Priority to CN201980094602.9A priority Critical patent/CN113646119B/en
Priority to PCT/JP2019/013886 priority patent/WO2020194751A1/en
Priority to JP2019552298A priority patent/JP6647467B1/en
Publication of WO2020194751A1 publication Critical patent/WO2020194751A1/en

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
    • B23H—WORKING OF METAL BY THE ACTION OF A HIGH CONCENTRATION OF ELECTRIC CURRENT ON A WORKPIECE USING AN ELECTRODE WHICH TAKES THE PLACE OF A TOOL; SUCH WORKING COMBINED WITH OTHER FORMS OF WORKING OF METAL
    • B23H1/00—Electrical discharge machining, i.e. removing metal with a series of rapidly recurring electrical discharges between an electrode and a workpiece in the presence of a fluid dielectric
    • B23H1/02—Electric circuits specially adapted therefor, e.g. power supply, control, preventing short circuits or other abnormal discharges

Definitions

  • the present invention relates to a numerical control device for controlling an electric discharge machine, an electric discharge machine, and an electric discharge machining method.
  • Patent Document 1 discloses an electric discharge machining apparatus that controls a drive shaft for relative movement between a work piece and a machining electrode by using a detection result of a discharge state during machining as a feedback signal.
  • the electric discharge machining apparatus according to Patent Document 1 detects an electric discharge state based on the result of measuring the pole voltage applied between the workpiece and the machining electrode.
  • the machined surface which is the surface of the workpiece facing the machined electrode, moves away from the machined electrode as the machine progresses. Since the pole-to-pole voltage and the pole-to-pole distance correlate with each other, if the actual pole-to-pole distance is longer than the best pole-to-pole distance to generate a high-frequency discharge, the pole-to-pole distance is the best pole-to-pole distance. There is a difference between the command value indicating the voltage and the measurement result of the electrode voltage. When the relative velocity between the workpiece and the processed electrode is controlled only by feedback as in the prior art described in Patent Document 1, the numerical control device detects the difference between the command value and the measurement result of the electrode voltage.
  • the numerical controller does not issue a command to change the relative velocity unless the actual distance between the poles changes from the best distance between the poles.
  • the control output for maintaining the pole-to-pole distance becomes zero, so that the pole-to-pole distance changes such as the retreat of the machined surface from the machined electrode. become.
  • the electric discharge machine does not maintain the best distance between the poles, which makes it difficult to maintain high frequency discharge generation.
  • the numerical control device has a problem that it is difficult for the electric discharge machining device to perform electric discharge machining at a high machining speed by maintaining the generation of electric discharge at a high frequency.
  • the present invention has been made in view of the above, and an object of the present invention is to obtain a numerical control device that enables an electric discharge machine to perform electric discharge machining at a high machining speed by maintaining high frequency discharge generation. And.
  • the numerical control device processes in an electric discharge machining device that processes an workpiece by generating a discharge in a gap between a machining electrode and a workpiece. Controls the relative velocity between the electrode and the workpiece.
  • the numerical control device detects the state of electric discharge and sets the moving speed at which the processed surface, which is the surface of the workpiece facing the processed electrode, moves away from the processed electrode as the processing progresses.
  • a machined surface movement speed estimation unit that estimates based on the state detection result, and a speed command value compensation unit that calculates a speed command value that is a relative speed command value and compensates for the movement speed based on the movement speed estimation value. And.
  • the numerical control device has an effect that the electric discharge machining device can perform electric discharge machining at a high machining speed by maintaining the generation of electric discharge at a high frequency.
  • FIG. 2 shows an example of a hardware configuration of the numerical control device according to the first embodiment.
  • a block diagram showing a functional configuration of a machine learning device included in the machined surface movement speed estimation unit shown in FIG. 7. A flowchart showing an operation flow of a machine learning device using reinforcement learning according to the second embodiment.
  • the numerical control device may be referred to as an NC (Numerical Control) device.
  • FIG. 1 is a diagram showing a schematic configuration of an electric discharge machine having a numerical control device according to a first embodiment of the present invention.
  • the electric discharge machine 100 shown in FIG. 1 is a die-sinking electric discharge machine.
  • the electric discharge machine 100 processes the workpiece 5 by generating an electric discharge in the gap between the machining electrode 4 and the workpiece 5.
  • the NC device 1 controls the electric discharge machine 100.
  • the machining section 2 is a portion of the electric discharge machining device 100 where electric discharge machining is performed, and has a shaft drive section for relatively moving the machining electrode 4 and the workpiece 5.
  • the shaft drive unit moves the machining electrode 4 with respect to the workpiece 5.
  • the NC device 1 controls the servo amplifier 8 by outputting a position command to the servo amplifier 8.
  • the servo amplifier 8 operates the shaft drive unit according to the position command output by the NC device 1.
  • the shaft drive unit moves the processing electrode 4. In FIG. 1, the shaft drive unit is not shown.
  • the machined portion 2 may have a shaft drive portion for moving the workpiece 5.
  • the machining power supply 3 applies a pulse voltage between the machining electrode 4 and the workpiece 5 according to the power supply command output by the NC device 1.
  • the pole voltage measuring device 6 measures the pole voltage applied between the machining electrode 4 and the workpiece 5.
  • the ammeter 7 measures the current supplied from the processing power source 3.
  • the processed surface 5a is a surface of the workpiece 5 facing the tip of the processed electrode 4, and is a surface on which processing is performed.
  • the machined surface 5a moves in a direction away from the machined electrode 4 as the work progresses.
  • the direction in which the machining surface 5a separates from the machining electrode 4 as the machining progresses may be referred to as a machining direction.
  • the machined surface 5a faces the machined electrode 4 in a direction opposite to the machined direction.
  • FIG. 2 is a block diagram showing a functional configuration of the electric discharge machine shown in FIG. FIG. 2 shows a configuration for controlling the relative speed between the machining electrode 4 and the workpiece 5 in the electric discharge machining device 100.
  • the servo amplifier 8 operates the shaft drive unit according to the position command value 23 output from the NC device 1.
  • the NC device 1 is characterized by having a machined surface moving speed estimation unit 20 in addition to a feedback speed command calculation unit 12 for feedback control by an interpole voltage.
  • the machined surface moving speed estimation unit 20 detects the discharge state and estimates the moving speed at which the machined surface 5a moves in the machining direction based on the detection result 24 of the discharge state. The details of the machined surface moving speed estimation unit 20 will be described later.
  • the pole-to-pole voltage measuring device 6 outputs the measured value 26, which is the measurement result of the pole-to-pole voltage, to the NC device 1.
  • the ammeter 7 outputs the measured value 30 of the current to the NC device 1.
  • the user inputs the pole-to-pole voltage command value 21 to the NC device 1 by referring to the machining condition table.
  • the pole voltage command value 21 is a command value of the pole voltage applied between the machining electrode 4 and the workpiece 5.
  • the pole-to-pole voltage and the pole-to-pole distance correlate with each other.
  • the pole voltage command value 21 is the value of the pole voltage when the pole voltage is the best pole distance.
  • the measured value 26 corresponds to the actual pole voltage.
  • the difference device 11 calculates the difference 29 between the pole voltage command value 21 and the measured value 26 input to the NC device 1, and outputs the difference 29 to the feedback speed command calculation unit 12.
  • the feedback speed command calculation unit 12 generates the feedback control amount 22 based on the difference 29.
  • the feedback speed command calculation unit 12 outputs the generated feedback control amount 22 to the adder 13.
  • the adder 13 is a command value of the relative speed between the machined electrode 4 and the workpiece 5 based on the estimated value 27 of the moving speed of the machined surface 5a, and is a speed command value 28 that compensates for the moving speed of the machined surface 5a. Functions as a speed command value compensator for calculating.
  • the volume of the portion removed from the workpiece 5 by electric discharge machining is defined as the removal volume
  • the removal volume per unit time is defined as the removal volume rate.
  • the moving speed of the machined surface 5a corresponds to the result of dividing the removal volume speed by the area of the machined surface 5a.
  • the removal volume velocity can be estimated by multiplying the discharge pulse frequency by the processing energy applied for the discharge.
  • the machined surface moving speed estimation unit 20 detects the state of discharge and multiplies the detection result 24 of the state of discharge by a coefficient 25 to perform processing corresponding to estimation of the removed volume speed and estimation of the moving speed of the machined surface 5a. I do. As a result, the machined surface moving speed estimation unit 20 estimates the moving speed of the machined surface 5a.
  • the machined surface movement speed estimation unit 20 includes a discharge state detection unit 15 that detects the discharge state, a coefficient setting unit 16 that sets a coefficient 25 that is multiplied by the discharge state detection result 24, and a discharge state detection result. It has a multiplier 17 that multiplies 24 by a coefficient 25.
  • the discharge state detection unit 15 counts the number of times a current flows due to discharge based on the transition of the measured value 30 acquired from the ammeter 7.
  • the discharge state detection unit 15 detects the discharge pulse frequency, which is the number of times the discharge has occurred, by counting the number of times the current has flowed.
  • the discharge state detection unit 15 detects the discharge state by detecting the discharge pulse frequency.
  • the discharge state detection unit 15 outputs the detection result 24 of the discharge pulse frequency to the multiplier 17.
  • the machined surface moving speed estimation unit 20 can improve the estimation accuracy by estimating the moving speed based on the discharge pulse frequency. It becomes.
  • the discharge state detection unit 15 may detect the discharge pulse frequency based on the transition of the measured value 26 measured by the interpole voltage measuring device 6.
  • the discharge state detection unit 15 can detect the discharge pulse frequency by counting the number of times the voltage drops during the application of the discharge pulse.
  • the coefficient setting unit 16 outputs a preset coefficient 25.
  • An arbitrary value is set for the coefficient 25.
  • the coefficient 25 takes into account the machining energy per discharge set in the machining conditions and the area of the machining electrode 4 facing the machining surface 5a. Since the processing energy is added to the coefficient 25, the processing surface moving speed estimation unit 20 can perform a process corresponding to the estimation of the removal volume velocity by multiplying the discharge pulse frequency by the coefficient 25. Further, since the area of the surface of the processing electrode 4 facing the processing surface 5a is added to the coefficient 25, the processing surface moving speed estimation unit 20 can move the processing surface 5a from the estimation result of the removal volume velocity. The processing corresponding to the estimation of can be performed.
  • the coefficient setting unit 16 outputs the coefficient 25 to the multiplier 17.
  • the multiplier 17 calculates an estimated value 27 of the moving speed of the machined surface 5a by multiplying the detection result 24 by the coefficient 25.
  • the multiplier 17 outputs the calculated estimated value 27 to the adder 13.
  • the adder 13 calculates the speed command value 28 in which the estimated value 27 is compensated by adding the estimated value 27 input from the multiplier 17 to the feedback control amount 22.
  • the adder 13 outputs the calculated speed command value 28 to the integrator 14.
  • the integrator 14 generates a position command value 23 for changing the relative position between the machining electrode 4 and the workpiece 5 by integrating the speed command value 28.
  • the integrator 14 outputs the generated position command value 23 to the servo amplifier 8.
  • the integrator 14 functions as a position command value calculation unit that calculates the position command value 23 based on the speed command value 28.
  • the coefficient 25 set in the coefficient setting unit 16 is not limited to the one set based on the machining energy per discharge or the area of the machining electrode 4 facing the machining surface 5a.
  • the coefficient 25 is various information regarding electric discharge machining, and can be set based on information that can affect the moving speed of the machined surface 5a.
  • the coefficient 25 may be set based on information such as the polarity of the discharge pulse applied between the processing electrode 4 and the workpiece 5 and the width of the discharge pulse.
  • the coefficient 25 is set by the user of the NC device 1.
  • the coefficient 25 may be set based on a machining model obtained by trial machining by the electric discharge machining device 100.
  • FIG. 3 is a flowchart showing an operation procedure of the numerical control device shown in FIG.
  • the discharge state detection unit 15 detects the discharge pulse frequency.
  • the discharge state detection unit 15 outputs the detection result 24 of the discharge pulse frequency to the multiplier 17.
  • the coefficient setting unit 16 outputs the set coefficient 25 to the multiplier 17.
  • the multiplier 17 calculates the estimated value 27 by multiplying the discharge pulse frequency detection result 24 by the coefficient 25.
  • the multiplier 17 outputs the estimated value 27 to the adder 13.
  • the machined surface moving speed estimation unit 20 estimates the moving speed of the machined surface 5a.
  • step S3 the adder 13 generates a speed command value 28 that compensates for the moving speed of the machined surface 5a by adding the estimated value 27 to the feedback control amount 22.
  • the adder 13 outputs the speed command value 28 to the integrator 14.
  • step S4 the integrator 14 calculates the position command value 23 by integrating the speed command value 28.
  • the integrator 14 outputs the position command value 23 to the servo amplifier 8.
  • the coefficient setting unit 16 may set the coefficient 25 based on the information acquired at the time of processing, in addition to holding the preset coefficient 25.
  • FIG. 4 is an explanatory diagram of a coefficient setting example in the coefficient setting unit of the numerical control device shown in FIG. FIG. 4 shows the waveform of the current flowing between the machining electrode 4 and the workpiece 5 when an electric discharge occurs.
  • the coefficient setting unit 16 can acquire the measured value 30 measured by the ammeter 7.
  • the coefficient setting unit 16 obtains the processing energy per discharge by integrating the measured value 30 over time.
  • the area of the hatched portion in FIG. 4 represents the processing energy per discharge.
  • the total processing energy is the result of multiplying the processing energy per discharge by the discharge pulse frequency.
  • the total processing energy has a high correlation with the removed volume.
  • the removal volume rate can be estimated based on the relationship between the processing energy measured in advance and the removal volume.
  • the moving speed of the machined surface 5a can be estimated based on the removal volume speed and the facing area.
  • the machining surface moving speed estimation unit 20 can estimate the moving speed of the machining surface 5a.
  • the coefficient 25 may be set based on the information acquired at the time of machining other than the total machining energy.
  • the discharge state detection unit 15 is not limited to the one that detects the discharge state by detecting the discharge pulse frequency.
  • the discharge state detection unit 15 may detect the discharge state by detecting the light generated by the discharge or the sound generated by the discharge.
  • the discharge state detection unit 15 can detect the discharge state by detecting a state amount such as a light amount or a volume that changes according to the magnitude of the discharge.
  • the electric discharge machine 100 is not limited to the die-sinking electric discharge machine, and may be a wire electric discharge machine.
  • the function of the NC device 1 is realized by using a processing circuit.
  • the processing circuit is dedicated hardware mounted on the NC device 1.
  • the processing circuit may be a processor that executes a program stored in memory.
  • FIG. 5 is a first diagram showing an example of the hardware configuration of the numerical control device according to the first embodiment.
  • FIG. 5 shows a hardware configuration when the function of the NC device 1 is realized by using dedicated hardware.
  • the NC device 1 includes a processing circuit 51 that executes various processes, and an interface 52 for connecting to an external device of the NC device 1 or inputting / outputting information.
  • the processing circuit 51 and the interface 52 are connected to each other via a bus.
  • the processing circuit 51 which is dedicated hardware, is a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or these. It is a combination.
  • Each function of the difference device 11, the feedback speed command calculation unit 12, the adder 13, the integrator 14, and the machined surface movement speed estimation unit 20 shown in FIG. 2 is realized by using the processing circuit 51.
  • FIG. 6 is a second diagram showing an example of the hardware configuration of the numerical control device according to the first embodiment.
  • FIG. 6 shows a hardware configuration when the function of the NC device 1 is realized by using the hardware for executing the program.
  • the interface 52, the processor 53, and the memory 54 are connected to each other via a bus.
  • the processor 53 is a CPU (Central Processing Unit), a processing device, an arithmetic unit, a microprocessor, a microcomputer, or a DSP (Digital Signal Processor).
  • the functions of the diffifier 11, the feedback speed command calculation unit 12, the adder 13, the integrator 14, and the machining surface movement speed estimation unit 20 shown in FIG. 2 are a combination of the processor 53 and software, firmware, or software and firmware. Realized by.
  • the software or firmware is described as a program and stored in a memory 54 which is a built-in memory.
  • the memory 54 is a non-volatile or volatile semiconductor memory, and is a RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory) or EEPROM (registered trademark) (Electrically Erasable). Programmable Read Only Memory).
  • RAM Random Access Memory
  • ROM Read Only Memory
  • flash memory EPROM (Erasable Programmable Read Only Memory) or EEPROM (registered trademark) (Electrically Erasable). Programmable Read Only Memory).
  • the NC apparatus 1 estimates the moving speed of the machined surface 5a in the machining direction based on the detection result 24 of the discharge state, and compensates the moving speed based on the estimated value 27 of the moving speed.
  • the speed command value 28 is calculated.
  • the NC device 1 can obtain a speed command value 28 that can offset the moving speed of the machining surface 5a in the machining direction.
  • the NC device 1 calculates the speed command value 28 that compensates for the moving speed of the machined surface 5a, so that the electric discharge machine 100 can maintain the best distance between the poles and proceeds with the machining while maintaining a high discharge pulse frequency. be able to.
  • the NC apparatus 1 has an effect that the electric discharge machining apparatus 100 can perform electric discharge machining at a high machining speed by maintaining the generation of electric discharge at a high frequency.
  • FIG. 7 is a block diagram showing a functional configuration of a machined surface moving speed estimation unit included in the NC device according to the second embodiment of the present invention.
  • the NC device 1 according to the second embodiment has a coefficient calculation unit 61 for calculating the coefficient 25 instead of the coefficient setting unit 16 according to the first embodiment.
  • the coefficient calculation unit 61 includes a machine learning device 62 and a decision-making unit 63.
  • the same components as those in the first embodiment are designated by the same reference numerals, and the configurations different from those in the first embodiment will be mainly described.
  • the machine learning device 62 learns a coefficient 25 for obtaining a position command value 23 that can compensate for the moving speed of the machined surface 5a.
  • the decision-making unit 63 determines the coefficient 25 based on the result learned by the machine learning device 62.
  • the discharge state detection unit 15 outputs the discharge pulse frequency detection result 24 to the multiplier 17 and the machine learning device 62.
  • the coefficient calculation unit 61 calculates the coefficient 25 by determining the coefficient 25 in the decision-making unit 63.
  • the coefficient calculation unit 61 outputs the calculated coefficient 25 to the multiplier 17.
  • the difference device 11 shown in FIG. 2 outputs the difference 29 to the feedback speed command calculation unit 12 and the machine learning device 62.
  • the integrator 14 outputs the position command value 23 to the servo amplifier 8 and the machine learning device 62.
  • FIG. 8 is a block diagram showing a functional configuration of a machine learning device included in the machined surface movement speed estimation unit shown in FIG. 7.
  • the machine learning device 62 has a state observation unit 64 and a learning unit 65.
  • the state observation unit 64 observes the discharge pulse frequency detection result 24, the position command value 23, and the difference 29 as state variables.
  • the learning unit 65 learns the coefficient 25 for obtaining the position command value 23 that can compensate the moving speed according to the training data set created based on the state variables.
  • any learning algorithm may be used as the learning algorithm used by the learning unit 65.
  • Reinforcement learning is that an action subject who is an agent in a certain environment observes the current state and decides an action to be taken. Agents get rewarded from the environment by choosing an action and learn how to get the most reward through a series of actions.
  • Q-learning and TD-learning are known as typical methods of reinforcement learning.
  • the behavior value table which is a general update formula of the behavior value function Q (s, a)
  • the action value function Q (s, a) represents the action value Q, which is the value of the action of selecting the action “a” under the environment “s”.
  • the learning unit 65 has a reward calculation unit 66 and a function update unit 67.
  • the reward calculation unit 66 calculates the reward based on the state variable.
  • the function update unit 67 updates the function for determining the coefficient 25 according to the reward calculated by the reward calculation unit 66.
  • the reward calculation unit 66 detects the absolute value of the difference 29.
  • the reward calculation unit 66 calculates the reward "r" based on the change in the absolute value of the difference 29. For example, when the absolute value of the difference 29 becomes smaller as a result of changing the position command value 23, the reward calculation unit 66 increases the reward “r”. The reward calculation unit 66 increases the reward “r” by giving the reward value "1". The reward value is not limited to "1". Further, when the absolute value of the difference 29 becomes large as a result of changing the position command value 23, the reward calculation unit 66 reduces the reward “r”. The reward calculation unit 66 reduces the reward "r” by giving the reward value "-1". The reward value is not limited to "-1".
  • the function update unit 67 updates the function for determining the coefficient 25 according to the reward calculated by the reward calculation unit 66.
  • the function can be updated according to the training data set, for example, by updating the action value table.
  • the action value table is a data set stored in the form of a table in which an arbitrary action is associated with the action value. For example, in the case of Q-learning, it is used as a function for calculating the action value function Q (s t, a t) a coefficient 25 represented by the above formula (1).
  • FIG. 9 is a flowchart showing an operation flow of the machine learning device using the reinforcement learning according to the second embodiment.
  • a reinforcement learning method for updating the action value function Q (s, a) will be described with reference to the flowchart of FIG.
  • step S11 the state observation unit 64 acquires a state variable.
  • the state variables are the discharge pulse frequency which is the detection result 24, the position command value 23, and the difference 29.
  • step S12 the reward calculation unit 66 detects a change in the absolute value of the difference 29 due to the change of the position command value 23.
  • step S13 the reward calculation unit 66 calculates the reward "r" based on the change in the absolute value of the difference 29.
  • step S14 the function update unit 67 updates the action value function Q (s, a) based on the reward “r” calculated in step S13.
  • the function update unit 67 updates the action value function Q (s, a) according to the above equation (1).
  • step S15 the function update unit 67 determines whether or not the action value function Q (s, a) has converged.
  • the function update unit 67 determines that the action value function Q (s, a) has converged because the action value function Q (s, a) in step S14 is not updated.
  • step S15 the machine learning device 62 returns the operation procedure to step S11.
  • step S15 Yes
  • the learning by the learning unit 65 ends.
  • the machine learning device 62 ends the operation according to the procedure shown in FIG.
  • the machine learning device 62 may continue learning by returning the operation procedure from step S14 to step S11 without performing the determination in step S15.
  • the decision-making unit 63 selects the coefficient 25 that gives the most reward based on the result of learning by the learning unit 65, that is, the updated action value function Q (s, a).
  • the multiplier 17 calculates the estimated value 27 by multiplying the detection result 24 by the coefficient 25 determined by the decision-making unit 63.
  • the machined surface moving speed estimation unit 20 outputs the calculated estimated value 27 to the adder 13.
  • the NC device 1 can obtain a position command value 23 capable of compensating for the moving speed of the machined surface 5a by learning the coefficient 25 by the machine learning device 62. As a result, the NC device 1 can obtain the position command value 23 that can maintain the discharge generation at a high frequency.
  • the reward calculation unit 66 may calculate the reward "r” based on the detection result 24 of the discharge pulse frequency.
  • the discharge pulse frequency increases as the actual pole-to-pole distance approaches the best pole-to-pole distance. For this reason, the reward calculation unit 66 may increase the reward "r” when the discharge pulse frequency increases, and decrease the reward "r” when the discharge pulse frequency decreases.
  • the learning unit 65 may perform machine learning according to other known methods such as neural networks, genetic programming, functional logic programming, support vector machines, and the like.
  • the NC device 1 has a machine learning device 62 that learns a coefficient 25 for obtaining a position command value 23 capable of compensating for the moving speed of the machined surface 5a, thereby generating a high frequency discharge.
  • the position command value 23 that makes it possible to maintain the above can be obtained.
  • the NC device 1 has an effect that the discharge processing device 100 can perform the discharge processing at a high processing speed by maintaining the generation of the discharge at a high frequency.
  • the configuration shown in the above-described embodiment shows an example of the content of the present invention, can be combined with another known technique, and is one of the configurations without departing from the gist of the present invention. It is also possible to omit or change the part.

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Electrical Discharge Machining, Electrochemical Machining, And Combined Machining (AREA)

Abstract

A numerical control device (1) controls the relative speed between a workpiece and a machining electrode in an electric discharge machining device (100) that machines the workpiece by generating electric discharge in a gap between the machining electrode and the workpiece. The numerical control device (1) comprises: a machined surface movement speed estimation unit (20) that detects the state of discharge and estimates, on the basis of the discharge state detection result, the movement speed at which a machined surface of the workpiece, which is the surface thereof facing the machining electrode, moves in a direction away from the machining electrode as machining progresses; and an adder (13) that is a speed command value compensation unit for calculating, on the basis of the estimated value (27) of movement speed, a speed command value (28) which is a relative speed command value and which compensates for the movement speed.

Description

数値制御装置、放電加工装置および放電加工方法Numerical control device, electric discharge machine and electric discharge machining method
　本発明は、放電加工装置を制御する数値制御装置、放電加工装置および放電加工方法に関する。 The present invention relates to a numerical control device for controlling an electric discharge machine, an electric discharge machine, and an electric discharge machining method.
　放電加工装置は、被加工物と加工電極との間隙において放電を発生させることによって被加工物を加工する。放電加工装置を制御する数値制御装置は、安定した加工を行うために、被加工物と加工電極との距離である極間距離を一定に保つように、被加工物と加工電極との相対速度を制御する。特許文献１には、加工中の放電状態の検出結果をフィードバック信号とすることによって被加工物と加工電極との相対移動のための駆動軸を制御する放電加工装置が開示されている。特許文献１にかかる放電加工装置は、被加工物と加工電極との間に印加される極間電圧を測定した結果に基づいて放電状態を検出する。 The electric discharge machine processes the workpiece by generating an electric discharge in the gap between the workpiece and the machining electrode. The numerical control device that controls the electric discharge machining device has a relative speed between the workpiece and the machining electrode so as to keep the distance between the poles, which is the distance between the workpiece and the machining electrode, constant in order to perform stable machining. To control. Patent Document 1 discloses an electric discharge machining apparatus that controls a drive shaft for relative movement between a work piece and a machining electrode by using a detection result of a discharge state during machining as a feedback signal. The electric discharge machining apparatus according to Patent Document 1 detects an electric discharge state based on the result of measuring the pole voltage applied between the workpiece and the machining electrode.
特開平９－１１０４３号公報Japanese Unexamined Patent Publication No. 9-11043
　放電加工において、被加工物のうち加工電極と対向する面である加工面は、加工の進行によって加工電極から離れる方向へ移動していく。極間電圧と極間距離とは互いに相関するため、高い周波数の放電を発生させるための最良の極間距離よりも実際の極間距離が長くなると、最良の極間距離であるときの極間電圧を表す指令値と、極間電圧の測定結果とに差が生じる。上記特許文献１に記載される従来技術のようにフィードバックのみによって被加工物と加工電極との相対速度を制御する場合、数値制御装置は、指令値と極間電圧の測定結果との差分が検出されることによって、被加工物と加工電極との相対速度を変化させる指令を発生させる。この場合、数値制御装置は、実際の極間距離が最良の極間距離から変化しない限り、相対速度を変化させる指令を発生させない。実際の極間距離が最良の極間距離と同じであるときには、極間距離の維持のための制御出力がゼロになるため、加工電極からの加工面の後退といった極間距離の変化が生じることになる。放電加工装置は、最良の極間距離が維持されないことにより、高い周波数の放電発生を維持することが困難となる。このように、従来技術によると、数値制御装置は、高い周波数の放電発生の維持によって速い加工速度での放電加工を放電加工装置に行わせることが困難であるという課題があった。 In electric discharge machining, the machined surface, which is the surface of the workpiece facing the machined electrode, moves away from the machined electrode as the machine progresses. Since the pole-to-pole voltage and the pole-to-pole distance correlate with each other, if the actual pole-to-pole distance is longer than the best pole-to-pole distance to generate a high-frequency discharge, the pole-to-pole distance is the best pole-to-pole distance. There is a difference between the command value indicating the voltage and the measurement result of the electrode voltage. When the relative velocity between the workpiece and the processed electrode is controlled only by feedback as in the prior art described in Patent Document 1, the numerical control device detects the difference between the command value and the measurement result of the electrode voltage. By doing so, a command to change the relative velocity between the workpiece and the machining electrode is generated. In this case, the numerical controller does not issue a command to change the relative velocity unless the actual distance between the poles changes from the best distance between the poles. When the actual pole-to-pole distance is the same as the best pole-to-pole distance, the control output for maintaining the pole-to-pole distance becomes zero, so that the pole-to-pole distance changes such as the retreat of the machined surface from the machined electrode. become. The electric discharge machine does not maintain the best distance between the poles, which makes it difficult to maintain high frequency discharge generation. As described above, according to the prior art, the numerical control device has a problem that it is difficult for the electric discharge machining device to perform electric discharge machining at a high machining speed by maintaining the generation of electric discharge at a high frequency.
　本発明は、上記に鑑みてなされたものであって、高い周波数の放電発生の維持によって速い加工速度での放電加工を放電加工装置に行わせることを可能とする数値制御装置を得ることを目的とする。 The present invention has been made in view of the above, and an object of the present invention is to obtain a numerical control device that enables an electric discharge machine to perform electric discharge machining at a high machining speed by maintaining high frequency discharge generation. And.
　上述した課題を解決し、目的を達成するために、本発明にかかる数値制御装置は、加工電極と被加工物との間隙において放電を発生させることにより被加工物を加工する放電加工装置における加工電極と被加工物との相対速度を制御する。本発明にかかる数値制御装置は、放電の状態を検出して、被加工物のうち加工電極と対向する面である加工面が加工の進行によって加工電極から離れる方向へ移動する移動速度を放電の状態の検出結果に基づいて推定する加工面移動速度推定部と、移動速度の推定値に基づいて、相対速度の指令値であって移動速度を補償する速度指令値を算出する速度指令値補償部と、を備える。 In order to solve the above-mentioned problems and achieve the object, the numerical control device according to the present invention processes in an electric discharge machining device that processes an workpiece by generating a discharge in a gap between a machining electrode and a workpiece. Controls the relative velocity between the electrode and the workpiece. The numerical control device according to the present invention detects the state of electric discharge and sets the moving speed at which the processed surface, which is the surface of the workpiece facing the processed electrode, moves away from the processed electrode as the processing progresses. A machined surface movement speed estimation unit that estimates based on the state detection result, and a speed command value compensation unit that calculates a speed command value that is a relative speed command value and compensates for the movement speed based on the movement speed estimation value. And.
　本発明にかかる数値制御装置は、高い周波数の放電発生の維持によって速い加工速度での放電加工を放電加工装置に行わせることができるという効果を奏する。 The numerical control device according to the present invention has an effect that the electric discharge machining device can perform electric discharge machining at a high machining speed by maintaining the generation of electric discharge at a high frequency.
本発明の実施の形態１にかかる数値制御装置を有する放電加工装置の概略構成を示す図The figure which shows the schematic structure of the electric discharge machining apparatus which has the numerical control apparatus which concerns on Embodiment 1 of this invention. 図１に示す放電加工装置の機能構成を示すブロック図A block diagram showing a functional configuration of the electric discharge machine shown in FIG. 図２に示す数値制御装置の動作手順を示すフローチャートA flowchart showing the operation procedure of the numerical control device shown in FIG. 図２に示す数値制御装置が有する係数設定部における係数の設定例についての説明図Explanatory drawing about the setting example of the coefficient in the coefficient setting part of the numerical control device shown in FIG. 実施の形態１にかかる数値制御装置のハードウェア構成の例を示す第１の図The first figure which shows the example of the hardware configuration of the numerical control apparatus which concerns on Embodiment 1. 実施の形態１にかかる数値制御装置のハードウェア構成の例を示す第２の図FIG. 2 shows an example of a hardware configuration of the numerical control device according to the first embodiment. 本発明の実施の形態２にかかるＮＣ装置が有する加工面移動速度推定部の機能構成を示すブロック図A block diagram showing a functional configuration of a machined surface moving speed estimation unit included in the NC apparatus according to the second embodiment of the present invention. 図７に示す加工面移動速度推定部が有する機械学習装置の機能構成を示すブロック図A block diagram showing a functional configuration of a machine learning device included in the machined surface movement speed estimation unit shown in FIG. 7. 実施の形態２にかかる強化学習を用いた機械学習装置の動作フローを示すフローチャートA flowchart showing an operation flow of a machine learning device using reinforcement learning according to the second embodiment.
　以下に、本発明の実施の形態にかかる数値制御装置、放電加工装置および放電加工方法を図面に基づいて詳細に説明する。なお、この実施の形態によりこの発明が限定されるものではない。以下の説明では、数値制御装置をＮＣ（Numerical　Control）装置と称することがある。 Hereinafter, the numerical control device, the electric discharge machining device, and the electric discharge machining method according to the embodiment of the present invention will be described in detail with reference to the drawings. The present invention is not limited to this embodiment. In the following description, the numerical control device may be referred to as an NC (Numerical Control) device.
実施の形態１．
　図１は、本発明の実施の形態１にかかる数値制御装置を有する放電加工装置の概略構成を示す図である。図１に示す放電加工装置１００は、形彫放電加工装置である。放電加工装置１００は、加工電極４と被加工物５との間隙において放電を発生させることによって被加工物５を加工する。ＮＣ装置１は、放電加工装置１００を制御する。 Embodiment 1.
FIG. 1 is a diagram showing a schematic configuration of an electric discharge machine having a numerical control device according to a first embodiment of the present invention. The electric discharge machine 100 shown in FIG. 1 is a die-sinking electric discharge machine. The electric discharge machine 100 processes the workpiece 5 by generating an electric discharge in the gap between the machining electrode 4 and the workpiece 5. The NC device 1 controls the electric discharge machine 100.
　加工部２は、放電加工装置１００のうち放電加工が行われている部分であって、加工電極４と被加工物５とを相対移動させるための軸駆動部とを有する。実施の形態１では、軸駆動部は、被加工物５に対して加工電極４を移動させる。ＮＣ装置１は、サーボアンプ８へ位置指令を出力することによって、サーボアンプ８を制御する。サーボアンプ８は、ＮＣ装置１が出力する位置指令にしたがって軸駆動部を動作させる。軸駆動部は、加工電極４を移動させる。図１では、軸駆動部の図示を省略する。なお、加工部２は、被加工物５を移動させる軸駆動部を有しても良い。 The machining section 2 is a portion of the electric discharge machining device 100 where electric discharge machining is performed, and has a shaft drive section for relatively moving the machining electrode 4 and the workpiece 5. In the first embodiment, the shaft drive unit moves the machining electrode 4 with respect to the workpiece 5. The NC device 1 controls the servo amplifier 8 by outputting a position command to the servo amplifier 8. The servo amplifier 8 operates the shaft drive unit according to the position command output by the NC device 1. The shaft drive unit moves the processing electrode 4. In FIG. 1, the shaft drive unit is not shown. The machined portion 2 may have a shaft drive portion for moving the workpiece 5.
　加工電源３は、ＮＣ装置１が出力する電源指令にしたがって、加工電極４と被加工物５との間にパルス電圧を印加する。極間電圧測定器６は、加工電極４と被加工物５との間に印加される極間電圧を測定する。電流計７は、加工電源３から供給される電流を測定する。 The machining power supply 3 applies a pulse voltage between the machining electrode 4 and the workpiece 5 according to the power supply command output by the NC device 1. The pole voltage measuring device 6 measures the pole voltage applied between the machining electrode 4 and the workpiece 5. The ammeter 7 measures the current supplied from the processing power source 3.
　加工面５ａは、被加工物５のうち加工電極４の先端と対向する面であって、加工が行われている面である。加工面５ａは、加工の進行によって、加工電極４から離れる方向へ移動する。以下の説明では、加工の進行によって加工面５ａが加工電極４から離れる方向を、加工方向と称することがある。加工面５ａは、加工方向とは逆の方向において加工電極４と対向する。 The processed surface 5a is a surface of the workpiece 5 facing the tip of the processed electrode 4, and is a surface on which processing is performed. The machined surface 5a moves in a direction away from the machined electrode 4 as the work progresses. In the following description, the direction in which the machining surface 5a separates from the machining electrode 4 as the machining progresses may be referred to as a machining direction. The machined surface 5a faces the machined electrode 4 in a direction opposite to the machined direction.
　図２は、図１に示す放電加工装置の機能構成を示すブロック図である。図２には、放電加工装置１００のうち加工電極４と被加工物５との相対速度を制御するための構成を示している。サーボアンプ８は、ＮＣ装置１から出力される位置指令値２３にしたがって上記軸駆動部を動作させる。 FIG. 2 is a block diagram showing a functional configuration of the electric discharge machine shown in FIG. FIG. 2 shows a configuration for controlling the relative speed between the machining electrode 4 and the workpiece 5 in the electric discharge machining device 100. The servo amplifier 8 operates the shaft drive unit according to the position command value 23 output from the NC device 1.
　実施の形態１にかかるＮＣ装置１は、極間電圧によるフィードバック制御のためのフィードバック速度指令演算部１２に加えて、加工面移動速度推定部２０を有することを特徴とする。加工面移動速度推定部２０は、放電の状態を検出して、加工方向へ加工面５ａが移動する移動速度を放電の状態の検出結果２４に基づいて推定する。加工面移動速度推定部２０の詳細については後述する。 The NC device 1 according to the first embodiment is characterized by having a machined surface moving speed estimation unit 20 in addition to a feedback speed command calculation unit 12 for feedback control by an interpole voltage. The machined surface moving speed estimation unit 20 detects the discharge state and estimates the moving speed at which the machined surface 5a moves in the machining direction based on the detection result 24 of the discharge state. The details of the machined surface moving speed estimation unit 20 will be described later.
　極間電圧測定器６は、極間電圧の測定結果である測定値２６をＮＣ装置１へ出力する。電流計７は、電流の測定値３０をＮＣ装置１へ出力する。ユーザは、加工条件表を参照することにより、ＮＣ装置１へ極間電圧指令値２１を入力する。極間電圧指令値２１は、加工電極４と被加工物５との間に印加される極間電圧の指令値である。極間電圧と極間距離とは互いに相関する。極間電圧指令値２１は、極間電圧が最良の極間距離であるときの極間電圧の値である。測定値２６は、実際の極間電圧に対応する。差分器１１は、ＮＣ装置１へ入力される極間電圧指令値２１と測定値２６との差分２９を算出し、差分２９をフィードバック速度指令演算部１２へ出力する。フィードバック速度指令演算部１２は、差分２９に基づいてフィードバック制御量２２を生成する。フィードバック速度指令演算部１２は、生成されたフィードバック制御量２２を加算器１３へ出力する。加算器１３は、加工面５ａの移動速度の推定値２７に基づいて、加工電極４と被加工物５との相対速度の指令値であって加工面５ａの移動速度を補償する速度指令値２８を算出する速度指令値補償部として機能する。 The pole-to-pole voltage measuring device 6 outputs the measured value 26, which is the measurement result of the pole-to-pole voltage, to the NC device 1. The ammeter 7 outputs the measured value 30 of the current to the NC device 1. The user inputs the pole-to-pole voltage command value 21 to the NC device 1 by referring to the machining condition table. The pole voltage command value 21 is a command value of the pole voltage applied between the machining electrode 4 and the workpiece 5. The pole-to-pole voltage and the pole-to-pole distance correlate with each other. The pole voltage command value 21 is the value of the pole voltage when the pole voltage is the best pole distance. The measured value 26 corresponds to the actual pole voltage. The difference device 11 calculates the difference 29 between the pole voltage command value 21 and the measured value 26 input to the NC device 1, and outputs the difference 29 to the feedback speed command calculation unit 12. The feedback speed command calculation unit 12 generates the feedback control amount 22 based on the difference 29. The feedback speed command calculation unit 12 outputs the generated feedback control amount 22 to the adder 13. The adder 13 is a command value of the relative speed between the machined electrode 4 and the workpiece 5 based on the estimated value 27 of the moving speed of the machined surface 5a, and is a speed command value 28 that compensates for the moving speed of the machined surface 5a. Functions as a speed command value compensator for calculating.
　ここで、放電加工によって被加工物５から除去される部分の体積を除去体積とし、単位時間当たりの除去体積を除去体積速度とする。加工面５ａの移動速度は、除去体積速度を加工面５ａの面積で除算した結果に相当する。除去体積速度は、放電のために投入される加工エネルギーを放電パルス周波数に乗算することによって推定できる。加工面移動速度推定部２０は、放電の状態を検出し、放電の状態の検出結果２４へ係数２５を乗算することによって、除去体積速度の推定および加工面５ａの移動速度の推定に相当する処理を行う。これにより、加工面移動速度推定部２０は、加工面５ａの移動速度を推定する。 Here, the volume of the portion removed from the workpiece 5 by electric discharge machining is defined as the removal volume, and the removal volume per unit time is defined as the removal volume rate. The moving speed of the machined surface 5a corresponds to the result of dividing the removal volume speed by the area of the machined surface 5a. The removal volume velocity can be estimated by multiplying the discharge pulse frequency by the processing energy applied for the discharge. The machined surface moving speed estimation unit 20 detects the state of discharge and multiplies the detection result 24 of the state of discharge by a coefficient 25 to perform processing corresponding to estimation of the removed volume speed and estimation of the moving speed of the machined surface 5a. I do. As a result, the machined surface moving speed estimation unit 20 estimates the moving speed of the machined surface 5a.
　加工面移動速度推定部２０は、放電の状態を検出する放電状態検出部１５と、放電の状態の検出結果２４に乗算される係数２５を設定する係数設定部１６と、放電の状態の検出結果２４に係数２５を乗算する乗算器１７とを有する。 The machined surface movement speed estimation unit 20 includes a discharge state detection unit 15 that detects the discharge state, a coefficient setting unit 16 that sets a coefficient 25 that is multiplied by the discharge state detection result 24, and a discharge state detection result. It has a multiplier 17 that multiplies 24 by a coefficient 25.
　放電状態検出部１５は、電流計７から取得された測定値３０の推移を基に、放電によって電流が流れた回数をカウントする。放電状態検出部１５は、電流が流れた回数をカウントすることによって、放電が発生した回数である放電パルス周波数を検出する。放電状態検出部１５は、放電パルス周波数を検出することによって、放電の状態を検出する。放電状態検出部１５は、放電パルス周波数の検出結果２４を乗算器１７へ出力する。 The discharge state detection unit 15 counts the number of times a current flows due to discharge based on the transition of the measured value 30 acquired from the ammeter 7. The discharge state detection unit 15 detects the discharge pulse frequency, which is the number of times the discharge has occurred, by counting the number of times the current has flowed. The discharge state detection unit 15 detects the discharge state by detecting the discharge pulse frequency. The discharge state detection unit 15 outputs the detection result 24 of the discharge pulse frequency to the multiplier 17.
　放電パルス周波数は加工面５ａの移動速度との相関を有する情報であることから、加工面移動速度推定部２０は、放電パルス周波数を基に移動速度を推定することによって、推定精度の向上が可能となる。なお、放電状態検出部１５は、極間電圧測定器６によって測定された測定値２６の推移を基に、放電パルス周波数を検出しても良い。放電状態検出部１５は、放電パルスの印加中に電圧が降下した回数をカウントすることによって放電パルス周波数を検出することができる。 Since the discharge pulse frequency is information having a correlation with the moving speed of the machined surface 5a, the machined surface moving speed estimation unit 20 can improve the estimation accuracy by estimating the moving speed based on the discharge pulse frequency. It becomes. The discharge state detection unit 15 may detect the discharge pulse frequency based on the transition of the measured value 26 measured by the interpole voltage measuring device 6. The discharge state detection unit 15 can detect the discharge pulse frequency by counting the number of times the voltage drops during the application of the discharge pulse.
　係数設定部１６は、あらかじめ設定された係数２５を出力する。係数２５には、任意の値が設定される。例を挙げると、係数２５には、加工条件にて設定される放電１つ当たりの加工エネルギーと加工電極４のうち加工面５ａと対向する面の面積とが加味される。加工エネルギーが係数２５に加味されていることで、加工面移動速度推定部２０は、放電パルス周波数への係数２５の乗算によって、除去体積速度の推定に相当する処理を行うことができる。さらに、加工電極４のうち加工面５ａと対向する面の面積が係数２５に加味されていることで、加工面移動速度推定部２０は、除去体積速度の推定結果からの加工面５ａの移動速度の推定に相当する処理を行うことができる。係数設定部１６は、乗算器１７へ係数２５を出力する。 The coefficient setting unit 16 outputs a preset coefficient 25. An arbitrary value is set for the coefficient 25. For example, the coefficient 25 takes into account the machining energy per discharge set in the machining conditions and the area of the machining electrode 4 facing the machining surface 5a. Since the processing energy is added to the coefficient 25, the processing surface moving speed estimation unit 20 can perform a process corresponding to the estimation of the removal volume velocity by multiplying the discharge pulse frequency by the coefficient 25. Further, since the area of the surface of the processing electrode 4 facing the processing surface 5a is added to the coefficient 25, the processing surface moving speed estimation unit 20 can move the processing surface 5a from the estimation result of the removal volume velocity. The processing corresponding to the estimation of can be performed. The coefficient setting unit 16 outputs the coefficient 25 to the multiplier 17.
　乗算器１７は、検出結果２４へ係数２５を乗算することによって、加工面５ａの移動速度の推定値２７を算出する。乗算器１７は、算出された推定値２７を加算器１３へ出力する。加算器１３は、乗算器１７から入力された推定値２７をフィードバック制御量２２に加算することによって、推定値２７が補償された速度指令値２８を算出する。加算器１３は、算出された速度指令値２８を積分器１４へ出力する。積分器１４は、速度指令値２８を積分することによって、加工電極４と被加工物５との相対位置を変化させるための位置指令値２３を生成する。積分器１４は、生成された位置指令値２３をサーボアンプ８へ出力する。積分器１４は、速度指令値２８に基づいて位置指令値２３を算出する位置指令値算出部として機能する。 The multiplier 17 calculates an estimated value 27 of the moving speed of the machined surface 5a by multiplying the detection result 24 by the coefficient 25. The multiplier 17 outputs the calculated estimated value 27 to the adder 13. The adder 13 calculates the speed command value 28 in which the estimated value 27 is compensated by adding the estimated value 27 input from the multiplier 17 to the feedback control amount 22. The adder 13 outputs the calculated speed command value 28 to the integrator 14. The integrator 14 generates a position command value 23 for changing the relative position between the machining electrode 4 and the workpiece 5 by integrating the speed command value 28. The integrator 14 outputs the generated position command value 23 to the servo amplifier 8. The integrator 14 functions as a position command value calculation unit that calculates the position command value 23 based on the speed command value 28.
　係数設定部１６に設定される係数２５は、放電１回当たりの加工エネルギーまたは加工電極４のうち加工面５ａと対向する面の面積に基づいて設定されるものに限られない。係数２５は、放電加工に関する種々の情報であって、加工面５ａの移動速度へ影響を及ぼし得る情報に基づいて設定されることができる。係数２５は、加工電極４および被加工物５の間に印加される放電パルスの極性、放電パルスの幅といった情報に基づいて設定されても良い。係数２５は、ＮＣ装置１のユーザによって設定される。係数２５は、放電加工装置１００による試し加工によって得られた加工モデルに基づいて設定されても良い。 The coefficient 25 set in the coefficient setting unit 16 is not limited to the one set based on the machining energy per discharge or the area of the machining electrode 4 facing the machining surface 5a. The coefficient 25 is various information regarding electric discharge machining, and can be set based on information that can affect the moving speed of the machined surface 5a. The coefficient 25 may be set based on information such as the polarity of the discharge pulse applied between the processing electrode 4 and the workpiece 5 and the width of the discharge pulse. The coefficient 25 is set by the user of the NC device 1. The coefficient 25 may be set based on a machining model obtained by trial machining by the electric discharge machining device 100.
　図３は、図２に示す数値制御装置の動作手順を示すフローチャートである。ステップＳ１において、放電状態検出部１５は、放電パルス周波数を検出する。放電状態検出部１５は、放電パルス周波数の検出結果２４を乗算器１７へ出力する。係数設定部１６は、設定された係数２５を乗算器１７へ出力する。ステップＳ２において、乗算器１７は、放電パルス周波数の検出結果２４に係数２５を乗算することによって推定値２７を算出する。乗算器１７は、推定値２７を加算器１３へ出力する。ステップＳ１およびステップＳ２により、加工面移動速度推定部２０は、加工面５ａの移動速度を推定する。 FIG. 3 is a flowchart showing an operation procedure of the numerical control device shown in FIG. In step S1, the discharge state detection unit 15 detects the discharge pulse frequency. The discharge state detection unit 15 outputs the detection result 24 of the discharge pulse frequency to the multiplier 17. The coefficient setting unit 16 outputs the set coefficient 25 to the multiplier 17. In step S2, the multiplier 17 calculates the estimated value 27 by multiplying the discharge pulse frequency detection result 24 by the coefficient 25. The multiplier 17 outputs the estimated value 27 to the adder 13. In step S1 and step S2, the machined surface moving speed estimation unit 20 estimates the moving speed of the machined surface 5a.
　ステップＳ３において、加算器１３は、フィードバック制御量２２に推定値２７を加算することによって、加工面５ａの移動速度を補償する速度指令値２８を生成する。加算器１３は、速度指令値２８を積分器１４へ出力する。ステップＳ４において、積分器１４は、速度指令値２８を積分することにより、位置指令値２３を算出する。積分器１４は、サーボアンプ８へ位置指令値２３を出力する。これにより、ＮＣ装置１は、図３に示す手順による動作を終了する。 In step S3, the adder 13 generates a speed command value 28 that compensates for the moving speed of the machined surface 5a by adding the estimated value 27 to the feedback control amount 22. The adder 13 outputs the speed command value 28 to the integrator 14. In step S4, the integrator 14 calculates the position command value 23 by integrating the speed command value 28. The integrator 14 outputs the position command value 23 to the servo amplifier 8. As a result, the NC device 1 ends the operation according to the procedure shown in FIG.
　係数設定部１６は、あらかじめ設定された係数２５を保持する以外に、加工時に取得された情報に基づいて係数２５を設定しても良い。図４は、図２に示す数値制御装置が有する係数設定部における係数の設定例についての説明図である。図４には、放電が発生したときに加工電極４と被加工物５との間に流れる電流の波形を示している。係数設定部１６は、電流計７によって測定された測定値３０を取得することができる。 The coefficient setting unit 16 may set the coefficient 25 based on the information acquired at the time of processing, in addition to holding the preset coefficient 25. FIG. 4 is an explanatory diagram of a coefficient setting example in the coefficient setting unit of the numerical control device shown in FIG. FIG. 4 shows the waveform of the current flowing between the machining electrode 4 and the workpiece 5 when an electric discharge occurs. The coefficient setting unit 16 can acquire the measured value 30 measured by the ammeter 7.
　係数設定部１６は、測定値３０を時間で積分することによって放電１回当たりの加工エネルギーを求める。図４においてハッチングを付した部分の面積は、放電１回当たりの加工エネルギーを表している。放電１回当たりの加工エネルギーに放電パルス周波数を乗算した結果が全加工エネルギーとなる。全加工エネルギーは、除去体積と高い相関を有する。単位時間当たりの加工エネルギーが算出されると、事前に測定された加工エネルギーと除去体積との関係とに基づいて、除去体積速度が推定できる。また、除去体積速度と対向面積とを基に、加工面５ａの移動速度が推定できる。放電の状態の検出結果２４に乗算される係数２５が全加工エネルギーに基づいて設定されることによって、加工面移動速度推定部２０は、加工面５ａの移動速度を推定することができる。係数２５は、全加工エネルギー以外に加工時において取得された情報に基づいて設定されても良い。 The coefficient setting unit 16 obtains the processing energy per discharge by integrating the measured value 30 over time. The area of the hatched portion in FIG. 4 represents the processing energy per discharge. The total processing energy is the result of multiplying the processing energy per discharge by the discharge pulse frequency. The total processing energy has a high correlation with the removed volume. Once the processing energy per unit time is calculated, the removal volume rate can be estimated based on the relationship between the processing energy measured in advance and the removal volume. Further, the moving speed of the machined surface 5a can be estimated based on the removal volume speed and the facing area. By setting a coefficient 25 to be multiplied by the discharge state detection result 24 based on the total machining energy, the machining surface moving speed estimation unit 20 can estimate the moving speed of the machining surface 5a. The coefficient 25 may be set based on the information acquired at the time of machining other than the total machining energy.
　放電状態検出部１５は、放電パルス周波数を検出することによって放電の状態の検出するものに限られない。放電状態検出部１５は、放電によって発生する光、あるいは放電によって発生する音を検出することによって放電の状態を検出しても良い。放電状態検出部１５は、放電の大きさにしたがって変化する光量あるいは音量といった状態量を検出することによって、放電の状態を検出することができる。放電加工装置１００は、形彫放電加工装置に限られず、ワイヤ放電加工装置であっても良い。 The discharge state detection unit 15 is not limited to the one that detects the discharge state by detecting the discharge pulse frequency. The discharge state detection unit 15 may detect the discharge state by detecting the light generated by the discharge or the sound generated by the discharge. The discharge state detection unit 15 can detect the discharge state by detecting a state amount such as a light amount or a volume that changes according to the magnitude of the discharge. The electric discharge machine 100 is not limited to the die-sinking electric discharge machine, and may be a wire electric discharge machine.
　次に、ＮＣ装置１が有するハードウェア構成について説明する。ＮＣ装置１が有する機能は、処理回路を使用して実現される。処理回路は、ＮＣ装置１に搭載される専用のハードウェアである。処理回路は、メモリに格納されるプログラムを実行するプロセッサであっても良い。 Next, the hardware configuration of the NC device 1 will be described. The function of the NC device 1 is realized by using a processing circuit. The processing circuit is dedicated hardware mounted on the NC device 1. The processing circuit may be a processor that executes a program stored in memory.
　図５は、実施の形態１にかかる数値制御装置のハードウェア構成の例を示す第１の図である。図５には、ＮＣ装置１の機能が専用のハードウェアを使用して実現される場合におけるハードウェア構成を示している。ＮＣ装置１は、各種処理を実行する処理回路５１と、ＮＣ装置１の外部の機器との接続あるいは情報の入出力のためのインタフェース５２とを備える。処理回路５１とインタフェース５２とは、バスを介して相互に接続されている。 FIG. 5 is a first diagram showing an example of the hardware configuration of the numerical control device according to the first embodiment. FIG. 5 shows a hardware configuration when the function of the NC device 1 is realized by using dedicated hardware. The NC device 1 includes a processing circuit 51 that executes various processes, and an interface 52 for connecting to an external device of the NC device 1 or inputting / outputting information. The processing circuit 51 and the interface 52 are connected to each other via a bus.
　専用のハードウェアである処理回路５１は、単一回路、複合回路、プログラム化されたプロセッサ、並列プログラム化したプロセッサ、ＡＳＩＣ（Application　Specific　Integrated　Circuit）、ＦＰＧＡ（Field-Programmable　Gate　Array）、又はこれらの組み合わせである。図２に示す差分器１１、フィードバック速度指令演算部１２、加算器１３、積分器１４および加工面移動速度推定部２０の各機能は、処理回路５１を用いて実現される。 The processing circuit 51, which is dedicated hardware, is a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or these. It is a combination. Each function of the difference device 11, the feedback speed command calculation unit 12, the adder 13, the integrator 14, and the machined surface movement speed estimation unit 20 shown in FIG. 2 is realized by using the processing circuit 51.
　図６は、実施の形態１にかかる数値制御装置のハードウェア構成の例を示す第２の図である。図６には、ＮＣ装置１の機能がプログラムを実行するハードウェアを用いて実現される場合におけるハードウェア構成を示している。インタフェース５２とプロセッサ５３とメモリ５４は、バスを介して相互に接続されている。 FIG. 6 is a second diagram showing an example of the hardware configuration of the numerical control device according to the first embodiment. FIG. 6 shows a hardware configuration when the function of the NC device 1 is realized by using the hardware for executing the program. The interface 52, the processor 53, and the memory 54 are connected to each other via a bus.
　プロセッサ５３は、ＣＰＵ（Central　Processing　Unit）、処理装置、演算装置、マイクロプロセッサ、マイクロコンピュータ、又はＤＳＰ（Digital　Signal　Processor）である。図２に示す差分器１１、フィードバック速度指令演算部１２、加算器１３、積分器１４および加工面移動速度推定部２０の各機能は、プロセッサ５３と、ソフトウェア、ファームウェア、またはソフトウェアとファームウェアとの組み合わせによって実現される。ソフトウェアまたはファームウェアは、プログラムとして記述され、内蔵メモリであるメモリ５４に格納される。メモリ５４は、不揮発性もしくは揮発性の半導体メモリであって、ＲＡＭ（Random　Access　Memory）、ＲＯＭ（Read　Only　Memory）、フラッシュメモリ、ＥＰＲＯＭ（Erasable　Programmable　Read　Only　Memory）またはＥＥＰＲＯＭ（登録商標）（Electrically　Erasable　Programmable　Read　Only　Memory）である。 The processor 53 is a CPU (Central Processing Unit), a processing device, an arithmetic unit, a microprocessor, a microcomputer, or a DSP (Digital Signal Processor). The functions of the diffifier 11, the feedback speed command calculation unit 12, the adder 13, the integrator 14, and the machining surface movement speed estimation unit 20 shown in FIG. 2 are a combination of the processor 53 and software, firmware, or software and firmware. Realized by. The software or firmware is described as a program and stored in a memory 54 which is a built-in memory. The memory 54 is a non-volatile or volatile semiconductor memory, and is a RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory) or EEPROM (registered trademark) (Electrically Erasable). Programmable Read Only Memory).
　実施の形態１によると、ＮＣ装置１は、加工方向における加工面５ａの移動速度を放電の状態の検出結果２４に基づいて推定して、移動速度の推定値２７に基づいて、移動速度を補償する速度指令値２８を算出する。これにより、ＮＣ装置１は、加工方向への加工面５ａの移動速度を相殺可能な速度指令値２８を得ることができる。加工面５ａの移動速度を補償する速度指令値２８をＮＣ装置１が算出することによって、放電加工装置１００は、最良の極間距離を維持可能とし、高い放電パルス周波数を維持しながら加工を進めることができる。これにより、ＮＣ装置１は、高い周波数の放電発生の維持によって速い加工速度での放電加工を放電加工装置１００に行わせることができるという効果を奏する。 According to the first embodiment, the NC apparatus 1 estimates the moving speed of the machined surface 5a in the machining direction based on the detection result 24 of the discharge state, and compensates the moving speed based on the estimated value 27 of the moving speed. The speed command value 28 is calculated. As a result, the NC device 1 can obtain a speed command value 28 that can offset the moving speed of the machining surface 5a in the machining direction. The NC device 1 calculates the speed command value 28 that compensates for the moving speed of the machined surface 5a, so that the electric discharge machine 100 can maintain the best distance between the poles and proceeds with the machining while maintaining a high discharge pulse frequency. be able to. As a result, the NC apparatus 1 has an effect that the electric discharge machining apparatus 100 can perform electric discharge machining at a high machining speed by maintaining the generation of electric discharge at a high frequency.
実施の形態２．
　図７は、本発明の実施の形態２にかかるＮＣ装置が有する加工面移動速度推定部の機能構成を示すブロック図である。実施の形態２にかかるＮＣ装置１は、実施の形態１にかかる係数設定部１６に代えて、係数２５を算出する係数算出部６１を有する。係数算出部６１は、機械学習装置６２と意思決定部６３とを有する。実施の形態２では、上記の実施の形態１と同一の構成要素には同一の符号を付し、実施の形態１とは異なる構成について主に説明する。 Embodiment 2.
FIG. 7 is a block diagram showing a functional configuration of a machined surface moving speed estimation unit included in the NC device according to the second embodiment of the present invention. The NC device 1 according to the second embodiment has a coefficient calculation unit 61 for calculating the coefficient 25 instead of the coefficient setting unit 16 according to the first embodiment. The coefficient calculation unit 61 includes a machine learning device 62 and a decision-making unit 63. In the second embodiment, the same components as those in the first embodiment are designated by the same reference numerals, and the configurations different from those in the first embodiment will be mainly described.
　機械学習装置６２は、加工面５ａの移動速度を補償可能とする位置指令値２３を得るための係数２５を学習する。意思決定部６３は、機械学習装置６２が学習した結果に基づいて係数２５を決定する。放電状態検出部１５は、放電パルス周波数の検出結果２４を乗算器１７と機械学習装置６２とへ出力する。係数算出部６１は、意思決定部６３において係数２５を決定することによって、係数２５を算出する。係数算出部６１は、算出された係数２５を乗算器１７へ出力する。また、図２に示す差分器１１は、フィードバック速度指令演算部１２と機械学習装置６２とへ差分２９を出力する。積分器１４は、サーボアンプ８と機械学習装置６２とへ位置指令値２３を出力する。 The machine learning device 62 learns a coefficient 25 for obtaining a position command value 23 that can compensate for the moving speed of the machined surface 5a. The decision-making unit 63 determines the coefficient 25 based on the result learned by the machine learning device 62. The discharge state detection unit 15 outputs the discharge pulse frequency detection result 24 to the multiplier 17 and the machine learning device 62. The coefficient calculation unit 61 calculates the coefficient 25 by determining the coefficient 25 in the decision-making unit 63. The coefficient calculation unit 61 outputs the calculated coefficient 25 to the multiplier 17. Further, the difference device 11 shown in FIG. 2 outputs the difference 29 to the feedback speed command calculation unit 12 and the machine learning device 62. The integrator 14 outputs the position command value 23 to the servo amplifier 8 and the machine learning device 62.
　図８は、図７に示す加工面移動速度推定部が有する機械学習装置の機能構成を示すブロック図である。機械学習装置６２は、状態観測部６４と学習部６５とを有する。状態観測部６４は、放電パルス周波数の検出結果２４と位置指令値２３と差分２９とを、状態変数として観測する。学習部６５は、状態変数に基づいて作成される訓練データセットに従って、移動速度を補償可能とする位置指令値２３を得るための係数２５を学習する。 FIG. 8 is a block diagram showing a functional configuration of a machine learning device included in the machined surface movement speed estimation unit shown in FIG. 7. The machine learning device 62 has a state observation unit 64 and a learning unit 65. The state observation unit 64 observes the discharge pulse frequency detection result 24, the position command value 23, and the difference 29 as state variables. The learning unit 65 learns the coefficient 25 for obtaining the position command value 23 that can compensate the moving speed according to the training data set created based on the state variables.
　学習部６５が用いる学習アルゴリズムはどのようなものを用いてもよい。一例として、強化学習（Ｒｅｉｎｆｏｒｃｅｍｅｎｔ　Ｌｅａｒｎｉｎｇ）を適用した場合について説明する。強化学習は、ある環境内におけるエージェントである行動主体が、現在の状態を観測し、取るべき行動を決定する、というものである。エージェントは行動を選択することで環境から報酬を得て、一連の行動を通じて報酬が最も多く得られるような方策を学習する。強化学習の代表的な手法として、Ｑ学習（Ｑ－ｌｅａｒｎｉｎｇ）およびＴＤ学習（ＴＤ－ｌｅａｒｎｉｎｇ）などが知られている。例えば、Ｑ学習の場合、行動価値関数Ｑ（ｓ，ａ）の一般的な更新式である行動価値テーブルは、次の式（１）で表される。行動価値関数Ｑ（ｓ，ａ）は、環境「ｓ」のもとで行動「ａ」を選択する行動の価値である行動価値Ｑを表す。 Any learning algorithm may be used as the learning algorithm used by the learning unit 65. As an example, a case where reinforcement learning (Reinforcement Learning) is applied will be described. Reinforcement learning is that an action subject who is an agent in a certain environment observes the current state and decides an action to be taken. Agents get rewarded from the environment by choosing an action and learn how to get the most reward through a series of actions. Q-learning and TD-learning are known as typical methods of reinforcement learning. For example, in the case of Q-learning, the behavior value table, which is a general update formula of the behavior value function Q (s, a), is represented by the following formula (1). The action value function Q (s, a) represents the action value Q, which is the value of the action of selecting the action “a” under the environment “s”.
Figure JPOXMLDOC01-appb-M000001
Figure JPOXMLDOC01-appb-M000001
　上記の式（１）において、「ｓｔ＋１」は、時刻「ｔ」における環境を表す。「ａｔ」は、時刻「ｔ」における行動を表す。行動「ａｔ」によって、環境は「ｓｔ＋１」に変わる。「ｒｔ＋１」は、その環境の変化によってもらえる報酬を表す。「γ」は、割引率を表す。「α」は、学習係数を表す。Ｑ学習を適用した場合、位置指令値２３が行動「ａｔ」となる。 In the above equation (1), " st + 1 " represents the environment at the time "t". "A t" represents the behavior in time "t". Action by "a t", the environment is changed to "s t + 1". "Rt + 1 " represents the reward received by the change of the environment. “Γ” represents the discount rate. “Α” represents a learning coefficient. If you apply the Q-learning, the position command value 23 becomes the action "a t".
　上記の式（１）により表される更新式は、時刻「ｔ＋１」における最良の行動「ａ」の行動価値が、時刻「ｔ」において実行された行動「ａ」の行動価値Ｑよりも大きければ、行動価値Ｑを大きくし、逆の場合は、行動価値Ｑを小さくする。換言すれば、時刻「ｔ」における行動「ａ」の行動価値Ｑを、時刻「ｔ＋１」における最良の行動価値に近づけるように、行動価値関数Ｑ（ｓ，ａ）を更新する。それにより、或る環境における最良の行動価値が、それ以前の環境における行動価値に順次伝播していくようになる。 In the update formula expressed by the above equation (1), if the action value of the best action "a" at the time "t + 1" is larger than the action value Q of the action "a" executed at the time "t". , The action value Q is increased, and in the opposite case, the action value Q is decreased. In other words, the action value function Q (s, a) is updated so that the action value Q of the action “a” at the time “t” approaches the best action value at the time “t + 1”. As a result, the best behavioral value in a certain environment is sequentially propagated to the behavioral value in the previous environment.
　学習部６５は、報酬計算部６６と関数更新部６７とを有する。報酬計算部６６は、状態変数に基づいて報酬を計算する。関数更新部６７は、報酬計算部６６によって計算される報酬に従って、係数２５を決定するための関数を更新する。 The learning unit 65 has a reward calculation unit 66 and a function update unit 67. The reward calculation unit 66 calculates the reward based on the state variable. The function update unit 67 updates the function for determining the coefficient 25 according to the reward calculated by the reward calculation unit 66.
　報酬計算部６６は、差分２９の絶対値を検出する。報酬計算部６６は、差分２９の絶対値の変化に基づいて報酬「ｒ」を計算する。例えば、位置指令値２３を変更した結果、差分２９の絶対値が小さくなる場合、報酬計算部６６は、報酬「ｒ」を増大させる。報酬計算部６６は、報酬の値である「１」を与えることによって報酬「ｒ」を増大させる。なお、報酬の値は「１」に限られない。また、位置指令値２３を変更した結果、差分２９の絶対値が大きくなる場合には、報酬計算部６６は、報酬「ｒ」を低減する。報酬計算部６６は、報酬の値である「－１」を与えることによって報酬「ｒ」を低減する。なお、報酬の値は「－１」に限られない。 The reward calculation unit 66 detects the absolute value of the difference 29. The reward calculation unit 66 calculates the reward "r" based on the change in the absolute value of the difference 29. For example, when the absolute value of the difference 29 becomes smaller as a result of changing the position command value 23, the reward calculation unit 66 increases the reward “r”. The reward calculation unit 66 increases the reward "r" by giving the reward value "1". The reward value is not limited to "1". Further, when the absolute value of the difference 29 becomes large as a result of changing the position command value 23, the reward calculation unit 66 reduces the reward “r”. The reward calculation unit 66 reduces the reward "r" by giving the reward value "-1". The reward value is not limited to "-1".
　関数更新部６７は、報酬計算部６６によって計算される報酬に従って、係数２５を決定するための関数を更新する。関数の更新は、訓練データセットに従って、例えば行動価値テーブルを更新することによって行うことができる。行動価値テーブルは、任意の行動と、その行動価値とを関連付けてテーブルの形式で記憶したデータセットである。例えばＱ学習の場合、上記の式（１）により表される行動価値関数Ｑ（ｓｔ，ａｔ）を係数２５の算出のための関数として用いる。 The function update unit 67 updates the function for determining the coefficient 25 according to the reward calculated by the reward calculation unit 66. The function can be updated according to the training data set, for example, by updating the action value table. The action value table is a data set stored in the form of a table in which an arbitrary action is associated with the action value. For example, in the case of Q-learning, it is used as a function for calculating the action value function Q (s t, a t) a coefficient 25 represented by the above formula (1).
　図９は、実施の形態２にかかる強化学習を用いた機械学習装置の動作フローを示すフローチャートである。図９のフローチャートを参照して、行動価値関数Ｑ（ｓ，ａ）を更新する強化学習方法について説明する。 FIG. 9 is a flowchart showing an operation flow of the machine learning device using the reinforcement learning according to the second embodiment. A reinforcement learning method for updating the action value function Q (s, a) will be described with reference to the flowchart of FIG.
　ステップＳ１１において、状態観測部６４は、状態変数を取得する。状態変数は、検出結果２４である放電パルス周波数と、位置指令値２３と、差分２９とである。ステップＳ１２において、報酬計算部６６は、位置指令値２３を変更させたことによる差分２９の絶対値の変化を検出する。 In step S11, the state observation unit 64 acquires a state variable. The state variables are the discharge pulse frequency which is the detection result 24, the position command value 23, and the difference 29. In step S12, the reward calculation unit 66 detects a change in the absolute value of the difference 29 due to the change of the position command value 23.
　ステップＳ１３において、報酬計算部６６は、差分２９の絶対値の変化に基づいて報酬「ｒ」を算出する。ステップＳ１４において、関数更新部６７は、ステップＳ１３において算出された報酬「ｒ」に基づいて行動価値関数Ｑ（ｓ，ａ）を更新する。関数更新部６７は、上記の式（１）に従って行動価値関数Ｑ（ｓ，ａ）を更新する。 In step S13, the reward calculation unit 66 calculates the reward "r" based on the change in the absolute value of the difference 29. In step S14, the function update unit 67 updates the action value function Q (s, a) based on the reward “r” calculated in step S13. The function update unit 67 updates the action value function Q (s, a) according to the above equation (1).
　ステップＳ１５において、関数更新部６７は、行動価値関数Ｑ（ｓ，ａ）が収束したか否かを判定する。関数更新部６７は、ステップＳ１４における行動価値関数Ｑ（ｓ，ａ）の更新が行われなくなることによって行動価値関数Ｑ（ｓ，ａ）が収束したと判定する。 In step S15, the function update unit 67 determines whether or not the action value function Q (s, a) has converged. The function update unit 67 determines that the action value function Q (s, a) has converged because the action value function Q (s, a) in step S14 is not updated.
　行動価値関数Ｑ（ｓ，ａ）が収束していないと判定された場合（ステップＳ１５，Ｎｏ）、機械学習装置６２は、動作手順をステップＳ１１へ戻す。行動価値関数Ｑ（ｓ，ａ）が収束したと判定された場合（ステップＳ１５，Ｙｅｓ）、学習部６５による学習が終了する。これにより、機械学習装置６２は、図９に示す手順による動作を終了する。なお、機械学習装置６２は、ステップＳ１５による判定を行わず、ステップＳ１４からステップＳ１１へ動作手順を戻すことによって学習を継続させることとしても良い。 When it is determined that the action value function Q (s, a) has not converged (steps S15, No), the machine learning device 62 returns the operation procedure to step S11. When it is determined that the action value function Q (s, a) has converged (steps S15, Yes), the learning by the learning unit 65 ends. As a result, the machine learning device 62 ends the operation according to the procedure shown in FIG. The machine learning device 62 may continue learning by returning the operation procedure from step S14 to step S11 without performing the determination in step S15.
　意思決定部６３は、学習部６５による学習の結果、すなわち更新された行動価値関数Ｑ（ｓ，ａ）に基づいて、報酬が最も多く得られる係数２５を選択する。乗算器１７は、意思決定部６３によって決定された係数２５を検出結果２４へ乗算することによって、推定値２７を算出する。加工面移動速度推定部２０は、算出された推定値２７を加算器１３へ出力する。ＮＣ装置１は、機械学習装置６２による係数２５の学習によって、加工面５ａの移動速度を補償可能とする位置指令値２３を得ることが可能となる。これにより、ＮＣ装置１は、高い周波数の放電発生を維持可能とする位置指令値２３を得ることができる。 The decision-making unit 63 selects the coefficient 25 that gives the most reward based on the result of learning by the learning unit 65, that is, the updated action value function Q (s, a). The multiplier 17 calculates the estimated value 27 by multiplying the detection result 24 by the coefficient 25 determined by the decision-making unit 63. The machined surface moving speed estimation unit 20 outputs the calculated estimated value 27 to the adder 13. The NC device 1 can obtain a position command value 23 capable of compensating for the moving speed of the machined surface 5a by learning the coefficient 25 by the machine learning device 62. As a result, the NC device 1 can obtain the position command value 23 that can maintain the discharge generation at a high frequency.
　なお、報酬計算部６６は、放電パルス周波数の検出結果２４に基づいて報酬「ｒ」を算出しても良い。実際の極間距離が最良の極間距離に近くなると、放電パルス周波数は大きくなる。このことから、報酬計算部６６は、放電パルス周波数が大きくなる場合に報酬「ｒ」を増大させ、放電パルス周波数が小さくなる場合には報酬「ｒ」を低減しても良い。 Note that the reward calculation unit 66 may calculate the reward "r" based on the detection result 24 of the discharge pulse frequency. The discharge pulse frequency increases as the actual pole-to-pole distance approaches the best pole-to-pole distance. For this reason, the reward calculation unit 66 may increase the reward "r" when the discharge pulse frequency increases, and decrease the reward "r" when the discharge pulse frequency decreases.
　実施の形態２では、学習部６５が強化学習を利用して機械学習を実行する場合について説明した。学習部６５は、他の公知の方法、例えばニューラルネットワーク、遺伝的プログラミング、機能論理プログラミング、サポートベクターマシンなどに従って機械学習を実行してもよい。 In the second embodiment, the case where the learning unit 65 executes machine learning by using reinforcement learning has been described. The learning unit 65 may perform machine learning according to other known methods such as neural networks, genetic programming, functional logic programming, support vector machines, and the like.
　実施の形態２によると、ＮＣ装置１は、加工面５ａの移動速度を補償可能とする位置指令値２３を得るための係数２５を学習する機械学習装置６２を有することによって、高い周波数の放電発生を維持可能とする位置指令値２３を求めることができる。これにより、ＮＣ装置１は、高い周波数の放電発生の維持によって速い加工速度での放電加工を放電加工装置１００に行わせることができるという効果を奏する。 According to the second embodiment, the NC device 1 has a machine learning device 62 that learns a coefficient 25 for obtaining a position command value 23 capable of compensating for the moving speed of the machined surface 5a, thereby generating a high frequency discharge. The position command value 23 that makes it possible to maintain the above can be obtained. As a result, the NC device 1 has an effect that the discharge processing device 100 can perform the discharge processing at a high processing speed by maintaining the generation of the discharge at a high frequency.
　以上の実施の形態に示した構成は、本発明の内容の一例を示すものであり、別の公知の技術と組み合わせることも可能であるし、本発明の要旨を逸脱しない範囲で、構成の一部を省略、変更することも可能である。 The configuration shown in the above-described embodiment shows an example of the content of the present invention, can be combined with another known technique, and is one of the configurations without departing from the gist of the present invention. It is also possible to omit or change the part.
　１　ＮＣ装置、２　加工部、３　加工電源、４　加工電極、５　被加工物、５ａ　加工面、６　極間電圧測定器、７　電流計、８　サーボアンプ、１１　差分器、１２　フィードバック速度指令演算部、１３　加算器、１４　積分器、１５　放電状態検出部、１６　係数設定部、１７　乗算器、２０　加工面移動速度推定部、２１　極間電圧指令値、２２　フィードバック制御量、２３　位置指令値、２４　検出結果、２５　係数、２６，３０　測定値、２７　推定値、２８　速度指令値、２９　差分、５１　処理回路、５２　インタフェース、５３　プロセッサ、５４　メモリ、６１　係数算出部、６２　機械学習装置、６３　意思決定部、６４　状態観測部、６５　学習部、６６　報酬計算部、６７　関数更新部、１００　放電加工装置。 1 NC device, 2 machining unit, 3 machining power supply, 4 machining electrode, 5 workpiece, 5a machined surface, 6 pole voltage measuring instrument, 7 ammeter, 8 servo amplifier, 11 differential device, 12 feedback speed command calculation unit , 13 adder, 14 integrator, 15 discharge state detector, 16 coefficient setting unit, 17 multiplier, 20 machined surface movement speed estimation unit, 21 pole voltage command value, 22 feedback control amount, 23 position command value, 24 Detection result, 25 coefficient, 26,30 measured value, 27 estimated value, 28 speed command value, 29 difference, 51 processing circuit, 52 interface, 53 processor, 54 memory, 61 coefficient calculation unit, 62 machine learning device, 63 decision Unit, 64 state observation unit, 65 learning unit, 66 reward calculation unit, 67 function update unit, 100 discharge processing device.

Claims (7)

  1. 　加工電極と被加工物との間隙において放電を発生させることにより前記被加工物を加工する放電加工装置における前記加工電極と前記被加工物との相対速度を制御する数値制御装置であって、
    　前記放電の状態を検出して、前記被加工物のうち前記加工電極と対向する面である加工面が加工の進行によって前記加工電極から離れる方向へ移動する移動速度を前記放電の状態の検出結果に基づいて推定する加工面移動速度推定部と、
    　前記移動速度の推定値に基づいて、前記相対速度の指令値であって前記移動速度を補償する速度指令値を算出する速度指令値補償部と、
    　を備えることを特徴とする数値制御装置。     A numerical control device that controls the relative speed between the machining electrode and the workpiece in an electric discharge machine that processes the workpiece by generating a discharge in the gap between the machining electrode and the workpiece.
    The discharge state is detected, and the moving speed at which the machined surface, which is the surface of the workpiece facing the machined electrode, moves away from the machined electrode as the work progresses, is the detection result of the discharge state. The machined surface movement speed estimation unit that estimates based on
    A speed command value compensating unit that calculates a speed command value that is a command value of the relative speed and compensates for the moving speed based on the estimated value of the moving speed.
    A numerical control device characterized by comprising.
  2. 　前記加工面移動速度推定部は、前記放電の状態の検出結果に係数を乗算することによって前記推定値を求めることを特徴とする請求項１に記載の数値制御装置。 The numerical control device according to claim 1, wherein the machined surface moving speed estimation unit obtains the estimated value by multiplying the detection result of the discharge state by a coefficient.
  3. 　前記係数は、加工時に取得された情報に基づいて設定された係数であることを特徴とする請求項２に記載の数値制御装置。 The numerical control device according to claim 2, wherein the coefficient is a coefficient set based on information acquired at the time of processing.
  4. 　前記加工電極と前記被加工物との間に印加される極間電圧の指令値と前記極間電圧の測定値との差分を算出する差分器と、
    　前記加工電極と前記被加工物との相対位置を変化させるための位置指令値を前記速度指令値に基づいて算出する位置指令値算出部と、
    　前記係数を算出する係数算出部と、を備え、
    　前記係数算出部は、
    　前記移動速度を補償可能とする前記位置指令値を得るための前記係数を学習する機械学習装置と、
    　前記機械学習装置が学習した結果に基づいて前記係数を決定する意思決定部と、
    　を有し、
    　前記機械学習装置は、
    　前記放電の状態の検出結果と前記位置指令値と前記差分とを状態変数として観測する状態観測部と、
    　前記状態変数に基づいて作成される訓練データセットに従って前記係数を学習する学習部と、
    　を有することを特徴とする請求項２に記載の数値制御装置。     A diffifier that calculates the difference between the command value of the electrode voltage applied between the processed electrode and the workpiece and the measured value of the electrode voltage.
    A position command value calculation unit that calculates a position command value for changing the relative position between the processing electrode and the work piece based on the speed command value, and
    A coefficient calculation unit for calculating the coefficient is provided.
    The coefficient calculation unit
    A machine learning device that learns the coefficient for obtaining the position command value that can compensate the moving speed, and
    A decision-making unit that determines the coefficient based on the result learned by the machine learning device,
    Have,
    The machine learning device
    A state observing unit that observes the detection result of the discharge state, the position command value, and the difference as state variables.
    A learning unit that learns the coefficients according to the training data set created based on the state variables, and
    The numerical control device according to claim 2, wherein the numerical control device has.
  5. 　前記加工面移動速度推定部は、前記放電が発生した回数を検出することによって前記放電の状態を検出し、前記回数の検出結果に基づいて前記移動速度を推定することを特徴とする請求項１から４のいずれか１つに記載の数値制御装置。 The machined surface moving speed estimation unit detects the state of the discharge by detecting the number of times the discharge has occurred, and estimates the moving speed based on the detection result of the number of times. The numerical control device according to any one of 4 to 4.
  6. 　加工電極と被加工物との間隙において放電を発生させることによって前記被加工物を加工する放電加工装置であって、
    　前記放電の状態を検出して、前記被加工物のうち前記加工電極と対向する面である加工面が加工の進行によって前記加工電極から離れる方向へ移動する移動速度を前記放電の状態の検出結果に基づいて推定する加工面移動速度推定部と、
    　前記移動速度の推定値に基づいて、前記加工電極と前記被加工物との相対速度の指令値であって前記移動速度を補償する速度指令値を算出する速度指令値補償部と、
    　を備えることを特徴とする放電加工装置。     An electric discharge machine that processes an workpiece by generating an electric discharge in the gap between the workpiece and the workpiece.
    The discharge state is detected, and the moving speed at which the machined surface, which is the surface of the workpiece facing the machined electrode, moves away from the machined electrode as the work progresses, is the detection result of the discharge state. The machined surface movement speed estimation unit that estimates based on
    Based on the estimated value of the moving speed, a speed command value compensating unit that calculates a speed command value that is a command value of the relative speed between the machining electrode and the workpiece and compensates for the moving speed.
    An electric discharge machine characterized by being equipped with.
  7. 　加工電極と被加工物との間隙において放電を発生させることによって前記被加工物を加工する放電加工方法であって、
    　前記放電の状態を検出して、前記被加工物のうち前記加工電極と対向する面である加工面が加工の進行によって前記加工電極から離れる方向へ移動する移動速度を前記放電の状態の検出結果に基づいて推定する工程と、
    　前記移動速度の推定値に基づいて、前記加工電極と前記被加工物との相対速度の指令値であって前記移動速度を補償する速度指令値を算出する工程と、
    　を含むことを特徴とする放電加工方法。     An electric discharge machining method for machining the workpiece by generating an electric discharge in the gap between the machining electrode and the workpiece.
    The discharge state is detected, and the moving speed at which the machined surface, which is the surface of the workpiece facing the machined electrode, moves away from the machined electrode as the work progresses, is the detection result of the discharge state. And the process of estimating based on
    A step of calculating a speed command value for compensating for the moving speed, which is a command value of the relative speed between the processing electrode and the workpiece based on the estimated value of the moving speed.
    An electric discharge machining method characterized by including.
PCT/JP2019/013886 2019-03-28 2019-03-28 Numerical control device, electric discharge machining device, and electric discharge machining method WO2020194751A1 (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CN201980094602.9A CN113646119B (en) 2019-03-28 2019-03-28 Numerical control device, electric discharge machining device, and electric discharge machining method
PCT/JP2019/013886 WO2020194751A1 (en) 2019-03-28 2019-03-28 Numerical control device, electric discharge machining device, and electric discharge machining method
JP2019552298A JP6647467B1 (en) 2019-03-28 2019-03-28 Numerical control device, electric discharge machine and electric discharge machine method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2019/013886 WO2020194751A1 (en) 2019-03-28 2019-03-28 Numerical control device, electric discharge machining device, and electric discharge machining method

Publications (1)

Publication Number Publication Date
WO2020194751A1 true WO2020194751A1 (en) 2020-10-01

Family

ID=69568123

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2019/013886 WO2020194751A1 (en) 2019-03-28 2019-03-28 Numerical control device, electric discharge machining device, and electric discharge machining method

Country Status (3)

Country Link
JP (1) JP6647467B1 (en)
CN (1) CN113646119B (en)
WO (1) WO2020194751A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2021164971A (en) * 2020-04-07 2021-10-14 株式会社ソディック Electric discharge machining device

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP6854990B1 (en) * 2020-05-29 2021-04-07 三菱電機株式会社 Electric discharge machine, learning device, inference device, and electric discharge machining method

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6458427A (en) * 1987-05-08 1989-03-06 Mitsubishi Electric Corp Electric discharge machining
JPH0523917A (en) * 1991-07-17 1993-02-02 Hitachi Seiko Ltd Electric discharge machining device
JPH0911043A (en) * 1995-06-29 1997-01-14 Nec Corp Electric discharge machining method and electric discharge machining device
WO2015145484A1 (en) * 2014-03-27 2015-10-01 三菱電機株式会社 Control device for wire electric discharge machine and control method for wire electric discharge machine

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2946666B2 (en) * 1990-07-20 1999-09-06 日本電気株式会社 EDM control device
JP5372252B2 (en) * 2010-06-22 2013-12-18 三菱電機株式会社 Electric discharge machining control device
CN104853872B (en) * 2013-12-10 2017-07-14 三菱电机株式会社 Wire discharge processing apparatus, line electrode discharge-treating method and control device
JP6227599B2 (en) * 2015-08-25 2017-11-08 ファナック株式会社 Wire electrical discharge machine with constant distance between poles
JP6076577B1 (en) * 2016-01-25 2017-02-08 三菱電機株式会社 Wire electric discharge machine

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6458427A (en) * 1987-05-08 1989-03-06 Mitsubishi Electric Corp Electric discharge machining
JPH0523917A (en) * 1991-07-17 1993-02-02 Hitachi Seiko Ltd Electric discharge machining device
JPH0911043A (en) * 1995-06-29 1997-01-14 Nec Corp Electric discharge machining method and electric discharge machining device
WO2015145484A1 (en) * 2014-03-27 2015-10-01 三菱電機株式会社 Control device for wire electric discharge machine and control method for wire electric discharge machine

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2021164971A (en) * 2020-04-07 2021-10-14 株式会社ソディック Electric discharge machining device
JP7026159B2 (en) 2020-04-07 2022-02-25 株式会社ソディック Electric discharge machine

Also Published As

Publication number Publication date
CN113646119A (en) 2021-11-12
CN113646119B (en) 2022-07-29
JP6647467B1 (en) 2020-02-14
JPWO2020194751A1 (en) 2021-04-30

Similar Documents

Publication Publication Date Title
JP6680748B2 (en) Control device and machine learning device
EP3382475B1 (en) Wire electric discharge machine comprising control device and machine learning device
JP6923484B2 (en) Machining condition adjustment device and machine learning device
CN110914770B (en) Numerical control device
JP6767416B2 (en) Machining condition adjustment device and machine learning device
WO2020194751A1 (en) Numerical control device, electric discharge machining device, and electric discharge machining method
JP2017030067A (en) Control device-added machining apparatus with machining time measuring function and on-machine measuring function
JP6557285B2 (en) Control device and machine learning device
JP3621278B2 (en) Servo control device
JP6841852B2 (en) Control device and control method
JP2019168973A (en) Drive device and machine learning device
JP6050865B1 (en) Servo control device with function to optimize control gain online with evaluation function
KR20150139519A (en) Motor drive device
Wu et al. Fast and stable electrical discharge machining (EDM)
JP7205012B1 (en) Control device for wire electric discharge machine and control method for wire electric discharge machine
CN110893515B (en) Machining condition adjustment device and machine learning device
CN110677094B (en) Motor control device
CN110893518B (en) Machining condition adjustment device and machine learning device
JP5307696B2 (en) Wire cut electric discharge machining method and wire cut electric discharge machining apparatus
WO2021186545A1 (en) Wire electrical discharge machining machine and machine learning apparatus
JP2008117295A (en) Process control device

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2019552298

Country of ref document: JP

Kind code of ref document: A

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

Ref document number: 19921899

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19921899

Country of ref document: EP

Kind code of ref document: A1