Processing condition adjustment device

Abstract

This machining condition adjustment device comprises: a simulation unit for executing a simulation of a wire electrical discharge machine on the basis of predetermined machining details, required specifications, and predetermined machining conditions; an update unit for updating a calculation model of the simulation unit; and a calculation unit for calculating a machining state in the simulation. Further, the machining condition adjustment device stores a model learned by machine learning, in a predetermined machining state, an action of adjustment of the machining state suitable for making the machining state close to reference values of the machining state according to reference conditions, predicts an action of adjustment of the machining state using the stored model, and adjusts, on the basis of the prediction, the machining conditions to be used for a simulation process.

Description

[Technical field]

[0001] The present invention relates to a machining condition adjustment device, and more particularly to a machining condition adjustment device that adjusts the machining conditions of a wire electric discharge machine. [Background technology]

[0002] When machining with a wire EDM, machining conditions (voltage waveform, voltage polarity, current waveform, discharge time, pause time, machining fluid volume, machining fluid pressure, wire tension, wire feed control, etc.) are prepared according to the machining contents (workpiece material, workpiece thickness, machining shape, step, surrounding environment, etc.) and required specifications (machining speed, frequency of wire breakage, straightness accuracy, surface roughness, shape error, etc.) and machining is performed. Manufacturers perform actual machining many times while changing the machining conditions of the wire EDM, and determine the machining conditions that are considered appropriate for each combination of machining contents and required specifications, and use these as "standard conditions". However, due to individual differences in machines and deterioration over time, the output machining state data (machining voltage, machining current, machining speed, etc.) may unintentionally deviate from the expected values. Therefore, it is necessary to adjust the machining conditions presented by the manufacturer while evaluating the actual machining results at the manufacturing site.

[0003] As a conventional technique for adjusting machining conditions, for example, Patent Document 1 discloses a technique in which machining state data (machining voltage, machining current, machining speed, etc.) is acquired while actually performing machining, and the value of the machining state data that is assumed to be output under set reference conditions is set as a reference value, and a positive reward is given when the machining state data approaches the reference value, thereby adjusting the machining conditions to output the reference value. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2017-042882 A Summary of the Invention [Problem to be solved by the invention]

[0005] Such adjustment work must be performed by a skilled worker, or in some cases, a technician from the manufacturer must be called in.

[0006] In addition, although the technology described in Patent Document 1 makes it easy to adjust the machining conditions, it is necessary to perform actual machining while changing the machining conditions according to the machining contents and required specifications. However, the workpiece that has been actually machined in order to adjust the machining conditions cannot be used as a product. In particular, in the case of expensive workpieces, this results in a large cost. Therefore, there is a need for a technology that enables general workers to adjust machining conditions during actual machining, which involves a small amount of work. [Means for solving the problem]

[0007] The machining condition adjustment device according to the present invention solves the above problems by using simulation technology to calculate machining state data indicating the state of machining based on the adjusted machining conditions when adjusting the machining conditions. The machining condition adjustment device according to the present invention updates the calculation model used in the simulation process to be compatible with the on-site machine by performing a predetermined test actual machining. After the calculation model is updated by the test actual machining, the machining conditions can be adjusted according to the machining content and required specifications by simulation without using the on-site machine, i.e., without performing actual machining.

[0008] One aspect of the present invention is a machining condition adjustment device for adjusting machining conditions in a wire electric discharge machine, comprising: a machining information storage unit in which data relating to machining contents and data relating to required specifications are stored; a reference value storage unit for storing, as reference values, machining states corresponding to the machining contents and required specifications stored in the machining information storage unit and reference conditions which are predetermined machining conditions; a simulation unit for executing a simulation of the wire electric discharge machine based on the machining contents and required specifications stored in the machining information storage unit and predetermined machining conditions; and a calculation model update unit for updating a calculation model of the simulation unit by inputting, as a parameter, the machining states when a predetermined machining is performed. a processing condition calculation unit that calculates a processing state in the simulation based on a result of a simulation by the simulation unit; a model memory unit that stores a model learned by machine learning of an adjustment action of the processing state suitable for bringing the processing state closer to the reference value in a predetermined processing state; a processing condition prediction unit that predicts an adjustment action of the processing state using the model stored in the model memory unit based on the processing state calculated by the processing state calculation unit; and a processing condition update unit that adjusts the processing conditions used by the simulation unit for simulation processing based on the adjustment action of the processing state predicted by the processing condition prediction unit. Effect of the Invention

[0009] According to one aspect of the present invention, when adjusting machining conditions to satisfy the machining content and required specifications input by the operator, simulation technology is used to obtain the machining state under the adjusted machining conditions, making it possible to adjust the machining conditions with minimal actual machining. [Brief description of the drawings]

[0010] [Figure 1] FIG. 2 is a schematic hardware configuration diagram of a control device according to the first embodiment. [Diagram 2] FIG. 2 is a schematic block diagram showing the functions of a control device according to the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. 1 is a schematic hardware configuration diagram showing the main parts 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 that controls a wire electric discharge machine 8, or as a personal computer attached to the control device, or can also be implemented as a computer such as a fog computer 6 or a cloud server 7. The machining condition adjustment device 1 according to this embodiment is implemented as a computer connected to the wire electric discharge machine 8 (or a control device that controls the wire electric discharge machine 8) via a wired / wireless network 5.

[0012] The CPU 11 provided in the machining condition adjustment device 1 according to this embodiment is a processor that controls the entire machining condition adjustment device 1. The CPU 11 reads out a system program stored in the ROM 12 via the bus 22, and controls the entire machining condition adjustment device 1 according to the system program. The RAM 13 temporarily stores temporary calculation data, display data, various data input from outside, and the like.

[0013] The non-volatile memory 14 is composed of, for example, a memory backed up by a battery (not shown) or an SSD (Solid State Drive), and the memory state is maintained even when the power supply of the machining condition adjustment device 1 is turned off. The non-volatile memory 14 stores programs and data read from the external device 72 via the interface 15, programs and data input from the input device 71 via the interface 18, and programs and data acquired 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 machining content in the wire electric discharge machine 8, data related to the required specifications, and data related to each physical quantity detected by a sensor (not shown) attached to the wire electric discharge machine 8. The programs and data stored in the non-volatile memory 14 may be expanded in the RAM 13 when executed / used. In addition, various system programs such as known analysis programs are written in the ROM 12 in advance.

[0014] The interface 15 is an interface for connecting the CPU 11 of the machining condition adjustment device 1 to an external device 72 such as an external storage medium. For example, programs and data used in a simulation are read from the external device 72. In addition, the programs and data edited in the machining condition adjustment device 1 can be stored in an external storage medium (not shown) such as a CF card or a USB memory via the external device 72.

[0015] The interface 20 is an interface for connecting the CPU of the machining condition adjustment device 1 to a wired or wireless network 5. The network 5 may communicate using a technology such as serial communication such as RS-485, Ethernet (registered trademark), optical communication, wireless LAN, Wi-Fi (registered trademark), Bluetooth (registered trademark), etc. The network 5 is connected to a control device that controls other machines, a fog computer 6, a cloud server 7, and other higher-level management devices, and exchanges data with the machining condition adjustment device 1.

[0016] The display device 70 displays various data loaded into the memory, data obtained as a result of executing programs, etc., outputted via the interface 17. An input device 71 consisting of a keyboard, pointing device, etc., transfers commands, data, etc. based on operations by an operator to the CPU 11 via the interface 18.

[0017] The interface 21 is an interface for connecting the CPU 11 and the machine learning device 2. The machine learning device 2 includes a processor 201 for controlling the entire machine learning device 2, a ROM 202 for storing a system program, etc., a RAM 203 for temporarily storing each process related to machine learning, and a non-volatile memory 204 used for storing a model obtained by learning a group of data to be learned. The machine learning device 2 can observe data (e.g., data related to the processing content, data related to the required specifications, data related to the processing conditions, etc.) generated by the processing condition adjustment device 1 through the interface 21. In addition, the processing condition adjustment device 1 acquires the processing result output from the machine learning device 2 through the interface 21, and stores or displays the acquired result, or transmits it to another device through the network 5 or the like. Note that, although the machine learning device 2 is built into the processing condition adjustment device 1 in FIG. 1, it may be externally connected to the processing condition adjustment device 1 through a predetermined interface.

[0018] 2 is a schematic block diagram showing functions of the machining condition adjustment device 1 according to one embodiment of the present invention. Each function of the machining condition adjustment device 1 according to this embodiment is realized by the CPU 11 of the machining condition adjustment device 1 and the processor 201 of the machine learning device 2 shown in FIG. 1 executing system programs and controlling the operation of each part of the machining condition adjustment device 1 and the machine learning device 2.

[0019] The machining condition adjustment device 1 of this embodiment includes a simulation unit 100, a machining state calculation unit 110, and a machining condition update unit 120. The simulation unit 100 includes a calculation model update unit 112. The machine learning device 2 includes a machining condition prediction unit 205. The RAM 13 to the nonvolatile memory 14 of the machining condition adjustment device 1 include a machining information storage unit 200, which is an area for storing data related to a predetermined machining content and data related to a required specification, and a reference value storage unit 210, which is an area for storing a reference value, which is data related to a machining state observed when a basic condition is set as a machining condition, and the RAM 203 to the nonvolatile memory 204 of the machine learning device 2 include a model storage unit 209, which is an area for storing a model created in the machining condition prediction unit 205.

[0020] The simulation unit 100 is realized by the CPU 11 included in the machining condition adjustment device 1 shown in FIG. 1 executing a system program read from the ROM 12, and the CPU 11 mainly performing arithmetic processing using the RAM 13 and the non-volatile memory 14. The simulation unit 100 performs a simulation of the wire electric discharge machine based on data related to the machining contents and data related to the required specifications stored in the machining information storage unit 200, and data related to the initial machining conditions (basic conditions) set by an operator. The simulation unit 100 sequentially outputs the machining state during the simulation as a simulation result. The simulation of the wire electric discharge machine performed by the simulation unit 100 may use a known method disclosed in, for example, Japanese Patent Application Laid-Open No. 2002-160127 or Masanori Kunieda, "Electric Discharge Machining Simulation," Journal of the Japan Society for Precision Engineering, Vol. 76, No. 8, 2010, pp. 861-866, etc.

[0021] The calculation model update unit 112 updates the calculation model used in the simulation process by the simulation unit 100. When updating the calculation model, actual machining for testing is performed. The actual machining for testing is performed by executing a machining program for testing. The machining program for testing is created in advance by a manufacturer or the like to allow a machine to perform a predetermined machining operation. The actual machining for testing is a simple task of placing a test workpiece at a predetermined position and executing the machining program for testing. Therefore, even a general worker can perform it. In the actual machining for testing, for example, a test workpiece is placed on a wire electric discharge machine at the site, and predetermined machining is performed on a path including movement in the positive and negative directions of each axis at a plurality of predetermined locations. At that time, machining is performed while setting a plurality of predetermined conditions for the machining power source. Then, data indicating the machining state, such as the moving speed of the wire during each machining, the number of discharges, the number of short circuits, and the number of breaks, is obtained. After that, the parameters of the calculation model used in the simulation process are calculated based on the data indicating the acquired machining state. When calculating the parameters of the calculation model, for example, a simulation process based on a test machining program is executed using an initial calculation model, and data indicating the machining state obtained therefrom is used as basic data. A calculation is performed to correct the parameters of the calculation model based on the difference between this basic data and data indicating the machining state obtained from a wire electric discharge machine on-site. Then, the calculation model is updated with the calculated parameters. For other types of machines, the calculation model used in the simulation process can also be updated based on basic data indicating the machining state obtained by performing actual machining on at least one workpiece under multiple conditions. After updating the calculation model used in the simulation process in this way, it becomes possible to adjust the machining conditions according to the machining content and required specifications using the updated calculation model.

[0022] The machining state calculation unit 110 is realized by the CPU 11 included in the machining condition adjustment device 1 shown in Fig. 1 executing a system program read from the ROM 12, and the CPU 11 mainly performing arithmetic processing using the RAM 13 and the non-volatile memory 14. The machining state calculation unit 110 calculates and outputs data related to the machining state, such as the machining voltage, machining current, machining speed, number of discharges, machining fluid amount, machining fluid pressure, whether or not a short circuit has occurred, and whether or not a break has occurred, based on the result of the simulation by the simulation unit 100. In addition, the machining state calculation unit 110 stores the data related to the machining state calculated based on the result of the simulation performed in the simulation unit 100 under the condition in which the basic conditions are set as the machining conditions, as a reference value in the reference value storage unit 210 in response to an instruction from an operator.

[0023] The machining condition prediction unit 205 is realized by the processor 201 included in the machine learning device 2 shown in FIG. 1 executing a system program read from the ROM 202, and the processor 201 mainly performing arithmetic processing using the RAM 203 and the non-volatile memory 204. The machining condition prediction unit 205 performs machine learning processing based on the data related to the machining state output from the machining state calculation unit 110 and the reference value stored in the reference value storage unit 210, and predicts machining conditions that are considered to be suitable for the current machining state. The machining condition prediction unit 205 mainly predicts the machining conditions using a known reinforcement learning technique. The machining condition prediction unit 205 includes a learning unit 206, a reward calculation unit 207, and an estimation unit 208.

[0024] The learning unit 206 learns the machining state at a predetermined time t based on the data related to the machining state calculated by the machining state calculation unit 110 and the reference value stored in the reference value storage unit 210. i Processing state s i When a given processing condition is observed, adjustment action a i Execute reinforcement learning to learn the results when the processing conditions are adjusted. iFor example, a predetermined variable may be added to each item of the machining conditions. In the machining condition adjustment device 1 according to this embodiment, the adjustment action of the machining conditions is used as a parameter of the simulation operation of the wire electric discharge machine by the simulation unit 100. Basically, in the simulation, it is possible to set machining conditions that cannot be actually used for machining due to the specifications of the actual machine. Therefore, the adjustment action of the machining conditions that the learning unit 206 learns is i may adjust processing conditions beyond the range of typical settings or combinations.

[0025] The learning unit 206 may learn the outcome of performing a predetermined action for the state of the learning target at a predetermined time as a function indicating the total reward r obtained by taking that action. By repeating the learning cycle of reinforcement learning by trial and error, a measure (adjustment action of machining conditions for the machining state of the wire electric discharge machine) that maximizes the total reward r is learned. Examples of reinforcement learning methods include Q-learning. In this case, the learning unit 206 learns the outcome of performing a predetermined action for the state s i In the state of i An action value function Q that returns a value that takes a state and an action as arguments is created as a model. The action value function Q is expressed as a function with a state and an action as arguments, and is updated based on the reward obtained for an action in a certain state and the value of an action in a future state to which the action will lead during learning. The update formula for the action value function is defined according to the algorithm of reinforcement learning. For example, in Q-learning, which is one of the representative reinforcement learning algorithms, the action value function is defined by the following formula 1. In formula 1, α is called a learning coefficient and γ is called a discount rate, and they are defined in the ranges of 0<α≦1, 0<γ≦1. The learning unit 206 is not necessarily an essential component when the model has been sufficiently learned.

[0026]

number

[0027] The remuneration calculation unit 207 calculates the remuneration at a predetermined time t i Processing state s i When a given processing condition is observed, adjustment action a i The reward is based on the result of the i The remuneration calculation unit 207 calculates the processing state s i When observed, corrective action a i As a result of taking i+1 When the processing state transitions to s i+1 A high reward is calculated according to how close the data relating to each processing state in the processing state s is to the reference value stored in the reference value storage unit 210. For example, i+1 A rule may be set for reward calculation such that a high reward is obtained when the distance between the data relating to each processing state and the reference value, when expressed as a vector, is close to 0, and the reward calculation unit 207 may calculate the reward based on the rule. The reward r calculated by the reward calculation unit 207 is output to the learning unit 206 and used for learning.

[0028] The estimation unit 208 executes estimation of adjustment actions for the wire electric discharge machine using a model stored in a model storage unit 209 based on data relating to the machining state calculated by the machining state calculation unit 110. The estimation unit 208 estimates adjustment actions for the wire electric discharge machine using a model generated by the reinforcement learning of the learning unit 206 based on data relating to the machining state calculated by the machining state calculation unit 110. i When the processing conditions are observed as j (j is 1 to n) calculate the reward for each, and select the processing condition adjustment action a that produces the largest reward. j is estimated as a solution suited to the current situation. The adjustment action of the machining conditions estimated by the estimation unit 208 is output to the machining condition update unit 120.

[0029] Note that while learning is being performed by the learning unit 206, the estimation unit 208 may select random action a with a certain probability for the purpose of progressing the learning (ε-greedy method). In particular, in the present invention, the machining state is determined based on the results of a simulation without performing actual machining using a wire electric discharge machine. Therefore, the machining condition adjustment device according to the present invention can be used to derive good machining conditions that are outside the machine specifications or good machining conditions that are not anticipated by humans, by selecting random action a with a high probability, particularly in the early stages of learning.

[0030] The machining condition update unit 120 is realized by the CPU 11 included in the machining condition adjustment device 1 shown in Fig. 1 executing a system program read from the ROM 12, and the CPU 11 mainly performing calculation processing using the RAM 13 and the non-volatile memory 14. The machining condition update unit 120 updates the machining conditions in the simulation unit 100 by applying the adjustment action a of the machining conditions predicted by the machining condition prediction unit 205 to the machining conditions in the simulation performed by the simulation unit 100.

[0031] The machining condition adjustment device 1 according to the present embodiment having the above configuration calculates machining state data showing the machining state based on the adjusted machining conditions using a simulation technique when adjusting the machining conditions. Therefore, except for the first update of the simulation model, it is possible to adjust the machining conditions according to the machining contents and the required specifications without performing actual machining. Since it is no longer necessary to perform actual machining and the machining conditions can be adjusted without using the machine on-site, the machine is not occupied for the adjustment work of the machining conditions and the machine can be used for other work. Since the machining conditions can be adjusted in the manufacturer's laboratory, as long as the user updates the simulation model, it is not necessary for the manufacturer's engineer to visit the user's site to make adjustments. In addition, by making the computational resources used in the simulation high-speed, it is possible to adjust the machining conditions in a much shorter time than performing actual machining. In the simulation, it is possible to evaluate machining conditions that cannot be actually processed due to the specifications of the actual machine. Therefore, it is possible to perform simulations for machining conditions whose ranges and combinations were not expected, so it is possible to evaluate machining conditions outside the machine specifications and machining conditions that people cannot think of, which can serve as a guideline for improving machine specifications and new research and development.

[0032] Although one embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and can be embodied in various forms by making appropriate modifications. [Explanation of symbols]

[0033] 1 Processing condition adjustment device 5. Network 6. Fog Computer 7. Cloud Server 8. Wire Electric Discharge Machine 11 CPU 12 ROM 13 RAM 14 Non-volatile memory 15,17,18,20,21 Interface 22 Bus 70 Display device 71 Input Devices 72 External equipment 100 Simulation Division 110 Machining state calculation unit 112 Calculation model update section 120 Machining condition update section 200 Processing information storage section 210 Reference value storage unit 2 Machine learning 201 Processor 202 ROM 203 RAM 204 Non-volatile memory 205 Machining Condition Prediction Unit 206 Learning Department 207 Remuneration Calculation Department 208 Estimation Department 209 Model Memory Section

Claims

1. A machining condition adjustment device for adjusting machining conditions in a wire electric discharge machine, A processing information storage unit in which data related to processing contents and data related to required specifications are stored; a reference value storage unit that stores, as a reference value, a machining state corresponding to the machining content and required specifications stored in the machining information storage unit and reference conditions that are predetermined machining conditions; a simulation unit that executes a simulation of a wire electric discharge machine based on the machining contents and required specifications stored in the machining information storage unit and on predetermined machining conditions; a calculation model update unit that updates a 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 that is learned by machine learning of an adjustment action of a machining state suitable for bringing the machining state closer to the reference value in a predetermined machining state; a machining condition prediction unit that predicts an adjustment action 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; a machining condition update unit that adjusts the machining conditions used in the simulation process by the simulation unit based on the adjustment action of the machining state predicted by the machining condition prediction unit; A processing condition adjustment device equipped with the above.

2. The machine learning is reinforcement learning. The processing condition adjusting device according to claim 1 .

Images