Processing condition adjustment device

Abstract

The machining condition adjustment device includes a simulation unit that simulates a wire electrical discharge machine based on predetermined machining content, required specifications, and predetermined machining conditions; an updating unit that updates a calculation model of the simulation unit; and a calculation unit that calculates a machining state in the simulation. The device stores a model, obtained by machine learning, of an adjustment behavior of the machining state, suitable for bringing the machining state closer to a reference value of a machining state based on reference conditions, for the predetermined machining state. The device predicts the adjustment behavior of the machining state using the stored model, and adjusts the machining conditions used for the simulation based on the prediction.

Description

Technical Field

[0001] The present invention relates to a processing condition adjusting device, and in particular to a processing condition adjusting device for adjusting the processing conditions of a wire electric discharge machine. Background Art

[0002] When machining with a wire EDM, machining is performed using machining conditions (voltage waveform, voltage polarity, current waveform, discharge time, rest time, machining fluid volume, machining hydraulic pressure, wire tension, wire feed control, etc.) that are prepared to match the machining content (workpiece material, workpiece thickness, machining shape, step difference, surrounding environment, etc.) and required specifications (machining speed, wire breakage frequency, straightness accuracy, surface roughness, shape error, etc.). Manufacturers perform actual machining multiple times while changing the machining conditions of the wire EDM to determine the machining conditions that are considered appropriate for each combination of machining content and required specifications, and use these as "baseline conditions." However, due to individual differences in the machines and aging degradation, the output machining status data (machining voltage, machining current, machining speed, etc.) may unexpectedly deviate from the assumed values. Therefore, it is necessary to evaluate the actual machining results at the manufacturing site and adjust the machining conditions provided by the manufacturer.

[0003] As an existing technology involved in adjusting processing conditions, for example, the following technology is disclosed in Patent Document 1: processing status data (processing voltage, processing current, processing speed, etc.) is obtained while actually processing is performed, the value of the processing status data originally output under the set reference conditions is used as the reference value, and reinforcement learning is performed to provide positive rewards when the processing status data is close to the reference value, thereby adjusting the processing conditions to output the reference value.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 2017-042882 Summary of the Invention

[0007] Problems to be solved by the invention

[0008] Such adjustments need to be performed by a skilled operator, or the manufacturer's technicians may be called upon as needed.

[0009] Furthermore, while the technique described in Patent Document 1 facilitates adjustment of machining conditions, it requires actual machining while adjusting machining conditions according to the machining content and required specifications. However, workpieces that have undergone actual machining to adjust machining conditions cannot be used as commercial products. This increases costs, especially for high-priced workpieces.

[0010] Therefore, a technology that allows general operators to adjust machining conditions with less actual machining is desired.

[0011] Means for solving problems

[0012] The processing condition adjustment device of the present invention solves the aforementioned problems by using simulation technology to calculate processing status data representing the processing status based on the adjusted processing conditions when adjusting processing conditions. In the processing condition adjustment device of the present invention, the calculation model used for simulation processing is updated to be suitable for on-site machinery by performing a predetermined actual test process. Furthermore, after the calculation model is updated through the actual test process, the processing conditions can be adjusted according to the processing content and required specifications through simulation, even without using on-site machinery, that is, without performing actual processing.

[0013] Furthermore, one embodiment of the present invention is a processing condition adjustment device that adjusts the processing conditions in a wire electric discharge machine, the processing condition adjustment device comprising: a processing information storage unit that stores data related to processing content and data related to required specifications; a reference value storage unit that stores, as reference values, processing states corresponding to the processing content and required specifications stored in the processing information storage unit, and predetermined processing conditions, i.e., reference conditions; a simulation unit that performs simulation of the wire electric discharge machine based on the processing content and required specifications stored in the processing information storage unit, and predetermined processing conditions; and a calculation model updating unit that inputs the processing state when a predetermined processing is performed as a parameter, and This updates the calculation model of the simulation unit; a processing state calculation unit, which calculates the processing state in the simulation based on the simulation result of the simulation unit; a model storage unit, which is used to store a model, which is obtained by learning the adjustment behavior of the processing state suitable for making the processing state close to the reference value through machine learning under a predetermined processing state; a processing condition prediction unit, which predicts the adjustment behavior of the processing state using the model stored in the model storage unit based on the processing state calculated by the processing state calculation unit; and a processing condition updating unit, which adjusts the processing conditions used by the simulation unit in the simulation processing based on the adjustment behavior of the processing state predicted by the processing condition prediction unit.

[0014] Effects of the Invention

[0015] According to one embodiment of the present invention, when adjusting processing conditions to satisfy processing content and required specifications input by an operator, simulation technology is used to obtain a processing state based on the adjusted processing conditions, thereby enabling adjustment of processing conditions with less actual processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1This is a schematic hardware configuration diagram of the control device according to the first embodiment.

[0017] Figure 2 This is a schematic block diagram showing the functions of the control device according to the first embodiment. DETAILED DESCRIPTION

[0018] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0019] Figure 1 This is a schematic hardware configuration diagram showing the main components of a machining condition adjustment device according to one embodiment of the present invention. The machining condition adjustment device 1 of the present invention can be implemented, for example, as a control device for controlling a wire-type electrical discharge machine 8. Alternatively, it can be implemented as a personal computer installed in parallel with the control device, or as a computer such as a fog computer 6 or a cloud server 7. The machining condition adjustment device 1 of this embodiment is implemented as a computer connected to the wire-type electrical discharge machine 8 (or the control device for controlling the wire-type electrical discharge machine 8) via a wired / wireless network 5.

[0020] The CPU 11 included in the processing condition adjustment device 1 of this embodiment is a processor that controls the processing condition adjustment device 1 as a whole. The CPU 11 reads a system program stored in the ROM 12 via the bus 22 and controls the entire processing condition adjustment device 1 according to the system program. The RAM 13 temporarily stores temporary calculation data, display data, and various data input from the outside.

[0021] The non-volatile memory 14 is composed of, for example, a battery-backed memory or an SSD (Solid State Drive), and maintains its stored state even when the processing condition adjustment device 1 is powered off. The non-volatile memory 14 stores programs and data read from an external device 72 via the interface 15, programs and data input from the input device 71 via the interface 18, and programs and data obtained from other devices such as the fog computer 6 and the cloud server 7 via the network 5. The data stored in the non-volatile memory 14 may include, for example, data related to the processing content of the wire EDM 8, data related to required specifications, and other data related to various physical quantities detected by sensors (not shown) installed in the wire EDM 8. The programs and data stored in the non-volatile memory 14 may also be expanded into the RAM 13 during execution / use. Furthermore, various system programs, such as well-known analysis programs, are pre-written in the ROM 12.

[0022] The interface 15 is an interface for connecting the CPU 11 of the processing condition adjustment device 1 to an external device 72 such as an external storage medium. For example, programs and data used for simulation are read from the external device 72. In addition, programs and data edited in the processing condition adjustment device 1 can be stored in an external storage medium (not shown) such as a CF card or USB memory via the external device 72.

[0023] Interface 20 is used to connect the CPU of the processing condition adjustment device 1 to a wired or wireless network 5. Network 5 can communicate using, for example, serial communication such as RS-485, Ethernet (registered trademark), optical communication, wireless LAN, Wi-Fi (registered trademark), or Bluetooth (registered trademark). Higher-level management devices such as control devices for other machines, fog computers 6, and cloud servers 7 are connected to network 5, exchanging data with the processing condition adjustment device 1.

[0024] The display device 70 outputs and displays data read into the memory and data obtained as a result of executing a program, etc., via the interface 17. Furthermore, the input device 71, which includes a keyboard, a pointing device, etc., transmits instructions and data based on operator operations to the CPU 11 via the interface 18.

[0025] The interface 21 is an interface for connecting the CPU 11 and the machine learning machine 2. The machine learning machine 2 includes: a processor 201 for unified control of the entire machine learning machine 2; a ROM 202 storing system programs, etc.; a RAM 203 for temporary storage of various processes involved in machine learning; and a non-volatile memory 204 for storing models learned for data groups that become learning objects. The machine learning machine 2 can observe the data generated by the processing condition adjustment device 1 (for example, data related to processing content, data related to required specifications, data related to processing conditions, etc.) via the interface 21. In addition, the processing condition adjustment device 1 obtains the processing results output from the machine learning machine 2 via the interface 21, stores or displays the obtained results, or sends them to other devices via the network 5, etc. In addition, in Figure 1 The machine learning machine 2 is built into the processing condition adjustment device 1 , but may be externally connected to the processing condition adjustment device 1 via a predetermined interface.

[0026] Figure 2 This is a diagram showing the functions of the processing condition adjustment device 1 according to one embodiment of the present invention as a schematic block diagram. Figure 1This is achieved by the CPU 11 included in the machining condition adjustment device 1 and the processor 201 included in the machine learning machine 2 respectively executing system programs to control the operations of the respective components of the machining condition adjustment device 1 and the machine learning machine 2 .

[0027] The processing condition adjustment device 1 of this embodiment includes a simulation unit 100, a processing state calculation unit 110, and a processing condition update unit 120. The simulation unit 100 includes a calculation model update unit 112. Furthermore, the machine learning device 2 includes a processing condition prediction unit 205. Furthermore, the RAM 13 or non-volatile memory 14 of the processing condition adjustment device 1 includes a processing information storage unit 200, which is an area for storing data related to the planned processing content and data related to the required specifications; and a reference value storage unit 210, which is an area for storing reference values, which are data related to the processing state observed when setting basic conditions as processing conditions. A model storage unit 209 is pre-prepared in the RAM 203 or non-volatile memory 204 of the machine learning device 2 as an area for storing the model generated by the processing condition prediction unit 205.

[0028] The simulation unit 100 is Figure 1 The machining condition adjustment device 1 shown is provided with a CPU 11 that executes a system program read from a ROM 12. The CPU 11 primarily performs computational processing using the RAM 13 and nonvolatile memory 14. The simulation unit 100 simulates a wire electrical discharge machine based on data related to machining details and required specifications stored in a machining information storage unit 200, as well as data related to initial machining conditions (basic conditions) set by the operator. The simulation unit 100 sequentially outputs the machining status during the simulation as simulation results. The simulation unit 100 can simulate the wire electrical discharge machine using known methods such as those disclosed in Japanese Patent Application Laid-Open No. 2002-160127, Masanori Kunieda, and "Electrical Discharge Machining Simulation," Journal of the Society of Precision Engineering, Vol. 76, No. 8, 2010, pp. 861-866.

[0029] The calculation model update unit 112 updates the calculation model used in the simulation processing of the simulation unit 100. While the calculation model is being updated, actual test machining is performed. This actual test machining is performed by executing a test machining program. Test machining programs are pre-generated by manufacturers, etc., to cause the machine to perform predetermined machining operations. This simple operation involves placing a test workpiece at a predetermined location and executing the test machining program. Therefore, even a typical operator can perform this operation. During actual test machining, for example, a test workpiece is placed in an on-site wire EDM machine, and the predetermined machining is performed on multiple predetermined locations along a path that includes movement in both the positive and negative directions of each axis. Machining is performed while setting the machining power supply to multiple predetermined conditions. Data representing the machining status, such as the wire travel speed, number of discharges, number of short circuits, and number of wire breaks, is then acquired during each machining operation. Parameters for the calculation model used for the simulation are then calculated based on this acquired machining status data. When calculating the parameters of the calculation model, for example, an initial calculation model is used to perform a simulation based on a test machining program, and the data representing the machining state obtained thereby is used as the basic data. Based on the difference between this basic data and the data representing the machining state obtained from the on-site wire EDM, the parameters of the calculation model can be corrected. The calculation model is then updated using the calculated parameters. In other types of machinery, the calculation model used for simulation processing can also be updated based on the basic data representing the machining state obtained by actually machining at least one workpiece under multiple conditions. After the calculation model for simulation processing is updated in this way, the updated calculation model can be used to adjust the machining conditions according to the machining content and required specifications.

[0030] The machining state calculation unit 110 calculates the machining state by Figure 1 The machining condition adjustment device 1 shown is implemented primarily by CPU 11 executing a system program read from ROM 12, with the CPU 11 performing computational processing using RAM 13 and nonvolatile memory 14. Based on the simulation results of the simulation unit 100, the machining state calculation unit 110 calculates and outputs data related to the machining state, such as machining voltage, machining current, machining speed, number of discharges, machining fluid volume, machining fluid pressure, and the presence or absence of short circuits and wire breaks. Furthermore, in response to instructions from the operator, the machining state calculation unit 110 stores data related to the machining state calculated based on the results of a simulation performed with basic conditions set as machining conditions in the simulation unit 100 as reference values in the reference value storage unit 210.

[0031] The processing condition prediction unit 205 Figure 1The machine learning device 2 shown in FIG. 2 is implemented primarily by processor 201 executing a system program read from ROM 202. The processor 201 performs computational processing using RAM 203 and nonvolatile memory 204. A processing condition prediction unit 205 performs machine learning based on processing state data output from the processing state calculation unit 110 and reference values stored in the reference value storage unit 210 to predict processing conditions deemed appropriate for the current processing state. The processing condition prediction unit 205 primarily uses known reinforcement learning techniques to predict processing conditions. The processing condition prediction unit 205 includes a learning unit 206, a reward calculation unit 207, and an estimation unit 208.

[0032] The learning unit 206 calculates the processing state data calculated by the processing state calculation unit 110 and the reference value stored in the reference value storage unit 210, and observes the predetermined time point t i Processing status s i When the execution learns to take the adjustment action of the predetermined processing conditions a i Reinforcement learning of the good or bad results under the circumstances. Adjustment behavior a as a processing condition i , it is possible to add a predetermined variable, for example, to each item of the machining condition. In the machining condition adjustment device 1 of this embodiment, the adjustment behavior of the machining condition is used as a parameter for the simulation operation of the wire electric discharge machine by the simulation unit 100. Basically, in the simulation, machining conditions that cannot be actually machined due to the specifications of the actual machine can also be set. Therefore, the adjustment behavior of the machining condition learned by the learning unit 206 is a i It is also possible to adjust the processing conditions beyond the range of general setting values and the range of combinations.

[0033] The learning unit 206 can also learn the quality of the result when a predetermined behavior is performed for the state of the learning object at a predetermined time point as a function representing the total reward r obtained by taking the behavior. By repeatedly performing the learning cycle of reinforcement learning by trial and error, a strategy that maximizes the total reward r (adjustment behavior for the processing conditions of the processing state of the wire discharge machine) is learned. As a reinforcement learning method, Q learning can be cited. In this case, the learning unit 206 generates a total reward r in the processing state s i Return to the state of adjustment behavior a iThe behavior value function Q of the value is used as a model. The behavior value function Q is expressed as a function with state and behavior as independent variables, and is updated in the learning process of repeated behavior based on the reward obtained for the behavior in a certain state, the value of the behavior in the future state transferred by the behavior, etc. The update formula of the behavior value function is defined according to the reinforcement learning algorithm. For example, in Q learning, which is one of the representative reinforcement learning algorithms, the behavior value function is defined by the following mathematical formula 1. In addition, in mathematical formula 1, α is called the learning coefficient, and γ is called the discount rate, which is defined in the range of 0<α≤1, 0<γ≤1. In addition, when the model is fully learned, the learning unit 206 is not a necessary structure.

[0034] [Mathematical formula 1]

[0035]

[0036] The reward calculation unit 207 observes the predetermined time point t i Processing status s i When the predetermined processing condition adjustment action a is taken i The quality of the result is calculated as the reward r i The reward calculation unit 207 observes the processing state s i Take adjustment actions when i As a result, it is transferred to the processing state s i+1 In the case of processing state s i+1 The high reward is calculated by comparing the data of each processing state with the reference value stored in the reference value storage unit 210. For example, it is also possible to set the processing state s to be represented by a vector. i+1 The reward calculation unit 207 calculates the reward based on the reward calculation rule that a high reward is obtained when the distance between the data of each processing state and the reference value is close to 0. The reward r calculated by the reward calculation unit 207 is output to the learning unit 206 for learning.

[0037] The estimating unit 208 uses the model stored in the model storage unit 209 to estimate the adjustment behavior of the wire electric discharge machine based on the data related to the machining state calculated by the machining state calculating unit 110. The estimating unit 208 uses the model generated by the reinforcement learning of the learning unit 206. The data related to the machining state calculated by the machining state calculating unit 110 is observed as the machining state s i In the case of the current situation, the adjustment behavior a of multiple processing conditions is desirable j (j is 1 to n) calculate the rewards respectively, and adjust the processing conditions that calculate the maximum reward a jThe processing condition adjustment behavior estimated by the estimating unit 208 is output to the processing condition updating unit 120 .

[0038] Furthermore, during learning by the learning unit 206, the estimation unit 208 may select a random behavior a with a fixed probability, aiming for the progress of learning (ε-greedy method). In particular, in the present invention, actual machining using the wire EDM is not performed, but the machining state is determined based on simulation results. Therefore, the machining condition adjustment device of the present invention selects a random behavior a with a high probability, particularly in the initial stages of learning. This allows for the derivation of favorable machining conditions that are outside the machine specifications or unexpected by humans.

[0039] Figure 1 The CPU 11 included in the illustrated processing condition adjustment device 1 executes a system program read from the ROM 12. The processing condition updating unit 120 is implemented primarily by the CPU 11 performing arithmetic processing using the RAM 13 and the nonvolatile memory 14. The processing condition updating unit 120 updates the processing conditions in the simulation unit 100 by applying the processing condition adjustment behavior a predicted by the processing condition prediction unit 205 to the processing conditions in the simulation performed by the simulation unit 100.

[0040] The processing condition adjustment device 1 of this embodiment, having the above-described structure, uses simulation technology to calculate processing status data representing the processing status based on the adjusted processing conditions when adjusting processing conditions. Therefore, except for the initial update of the simulation model, processing conditions can be adjusted to suit the processing content and required specifications without actual processing. Processing condition adjustment can be performed without actual processing and without using on-site machinery. This eliminates the need to tie up machinery for processing condition adjustment and allows efficient use of the machinery for other operations. Processing condition adjustment can also be performed in the manufacturer's laboratory, etc. Therefore, once the user updates the simulation model, the manufacturer's technicians do not need to travel to the user's site to perform the adjustment. Furthermore, by using high-speed computing resources for simulation, processing condition adjustment can be performed in a much shorter time than actual processing. Simulations also allow evaluation of processing conditions that would not be practical due to the specifications of the actual machine. Consequently, simulations can be performed for processing conditions in unimagined ranges and combinations, enabling the evaluation of processing conditions outside of machine specifications and unexpected processing conditions, which can serve as a guideline for machine specification improvements and new research and development.

[0041] As mentioned above, although one embodiment of the present invention has been described, the present invention is not limited to the example of the above embodiment, and can be implemented in various forms by adding appropriate changes.

[0042] Explanation of symbols

[0043] 1. Processing condition adjustment device;

[0044] 5. Network;

[0045] 6. Fog computer;

[0046] 7 cloud servers;

[0047] 8-wire EDM;

[0048] 11CPU;

[0049] 12ROM;

[0050] 13 RAM;

[0051] 14 non-volatile memory;

[0052] Interfaces 15, 17, 18, 20, and 21;

[0053] 22 buses;

[0054] 70 display device;

[0055] 71 input device;

[0056] 72 external devices;

[0057] 100 simulation department;

[0058] 110 processing state calculation unit;

[0059] 112 Computational Model Update Department;

[0060] 120 Processing Condition Update Department;

[0061] 200 processing information storage unit;

[0062] 210 reference value storage unit;

[0063] 2. Machine Learning Machine;

[0064] 201 processor;

[0065] 202ROM;

[0066] 203RAM;

[0067] 204 non-volatile memory;

[0068] 205 Processing Condition Prediction Department;

[0069] 206 Learning Department;

[0070] 207 Return Calculation Department;

[0071] 208 Presumption Department;

[0072] 209 Model storage unit.

Claims

1. A processing condition adjustment device for adjusting processing conditions in a wire electric discharge machine, characterized in that: The processing condition adjustment device has: a processing information storage unit storing data related to processing contents and data related to required specifications; a reference value storage unit that stores, as reference values, a processing state corresponding to the processing contents and required specifications stored in the processing information storage unit and predetermined processing conditions, namely, reference conditions; a simulation unit that performs a simulation of the wire electrical discharge machine based on the machining content and required specifications stored in the machining information storage unit and predetermined machining conditions; a calculation model updating unit that updates the calculation model of the simulation unit by inputting a machining state when a predetermined machining is performed as a parameter; a machining state calculation unit that calculates a machining state in the simulation based on a result of the simulation by the simulation unit; a model storage unit for storing a model obtained by learning, under a predetermined machining state, an adjustment behavior suitable for bringing the machining state close to the reference value through machine learning; a machining condition prediction unit that predicts an adjustment behavior of the machining state using a model stored in the model storage unit based on the machining state calculated by the machining state calculation unit; as well as A machining condition updating unit adjusts the machining conditions used by the simulation unit in the simulation process based on the adjustment behavior of the machining state predicted by the machining condition prediction unit.

2. The processing condition adjustment device according to claim 1, characterized in that: The machine learning is reinforcement learning.

Images