WO2023281681A1 - 加工条件探索装置および加工条件探索方法 - Google Patents
加工条件探索装置および加工条件探索方法 Download PDFInfo
- Publication number
- WO2023281681A1 WO2023281681A1 PCT/JP2021/025722 JP2021025722W WO2023281681A1 WO 2023281681 A1 WO2023281681 A1 WO 2023281681A1 JP 2021025722 W JP2021025722 W JP 2021025722W WO 2023281681 A1 WO2023281681 A1 WO 2023281681A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- machining
- evaluation value
- determination unit
- processing
- unit
- Prior art date
Links
- 238000003754 machining Methods 0.000 title claims abstract description 611
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000011156 evaluation Methods 0.000 claims abstract description 545
- 238000012545 processing Methods 0.000 claims description 449
- 238000004364 calculation method Methods 0.000 claims description 44
- 238000010801 machine learning Methods 0.000 claims description 23
- 238000009826 distribution Methods 0.000 claims description 14
- 239000006185 dispersion Substances 0.000 claims description 2
- 238000003860 storage Methods 0.000 description 72
- 230000006870 function Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 13
- 238000005520 cutting process Methods 0.000 description 4
- 206010000210 abortion Diseases 0.000 description 3
- 231100000176 abortion Toxicity 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 101000710013 Homo sapiens Reversion-inducing cysteine-rich protein with Kazal motifs Proteins 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 102100035353 Cyclin-dependent kinase 2-associated protein 1 Human genes 0.000 description 1
- 101001139126 Homo sapiens Krueppel-like factor 6 Proteins 0.000 description 1
- 102100029860 Suppressor of tumorigenicity 20 protein Human genes 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 239000000919 ceramic Substances 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013400 design of experiment Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010894 electron beam technology Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 108090000237 interleukin-24 Proteins 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000003534 oscillatory effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23Q—DETAILS, COMPONENTS, OR ACCESSORIES FOR MACHINE TOOLS, e.g. ARRANGEMENTS FOR COPYING OR CONTROLLING; MACHINE TOOLS IN GENERAL CHARACTERISED BY THE CONSTRUCTION OF PARTICULAR DETAILS OR COMPONENTS; COMBINATIONS OR ASSOCIATIONS OF METAL-WORKING MACHINES, NOT DIRECTED TO A PARTICULAR RESULT
- B23Q15/00—Automatic control or regulation of feed movement, cutting velocity or position of tool or work
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/18—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form
- G05B19/4155—Numerical control [NC], i.e. automatically operating machines, in particular machine tools, e.g. in a manufacturing environment, so as to execute positioning, movement or co-ordinated operations by means of programme data in numerical form characterised by programme execution, i.e. part programme or machine function execution, e.g. selection of a programme
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Definitions
- the present disclosure relates to a machining condition searching device and a machining condition searching method for searching for machining conditions.
- control parameters can be set for processing machines used in industrial applications.
- the processing result of the processing machine depends on processing conditions, which are combinations of respective parameter values of a plurality of control parameters. In other words, in order to obtain desired processing results, it is necessary to set appropriate processing conditions for the processing machine.
- each control parameter is a continuous value or can be set in multiple steps. For this reason, it takes an enormous amount of time if a person selects processing conditions that allow a processing machine to actually perform processing and obtain a desired processing result.
- processing conditions that allow a processing machine to actually perform processing and obtain a desired processing result.
- a person selects processing conditions that allow a processing machine to actually perform processing and obtain a desired processing result.
- Each control parameter is selected from one of multiple stages of values.
- the total number of combinations is 10 5 . At this time, if it takes 5 minutes to try one processing condition, it takes about 347 days to try 10 5 processing conditions.
- an evaluation corresponding to the processing conditions is performed based on the processing results obtained by having the processing machine perform processing under several processing conditions to be tried, which are generated from the processing conditions of combinations of control parameters that are assumed. value, predict the evaluation value corresponding to the processing condition that has not been tried using Gaussian process regression based on the calculated evaluation value and the processing condition corresponding to the evaluation value, and predict the calculated evaluation value and the prediction.
- a technique of searching for an optimum machining condition from among a huge number of combinations of machining conditions based on the evaluation values obtained for example, Patent Literature 1).
- a method of using Gaussian process regression to predict evaluation values corresponding to untried machining conditions is, for example, a probabilistic model generated by assuming that evaluation values for machining conditions are random variables following a specific distribution.
- a machining result obtained when a machine is caused to perform machining under certain machining conditions may vibrate during the process of machining.
- the machining speed obtained as a result of machining may appear to move at a constant speed over a long period of time, but it will vibrate in a short period of time.
- the evaluation value corresponding to the machining result also vibrates.
- the technique of searching for the optimum machining conditions represented by the technique disclosed in Patent Document 1
- the machining machine is operated for a certain amount of time until the vibrational change in the machining result settles down. Then, after waiting for the vibrational change in the machining result to settle down, the evaluation value corresponding to the machining conditions was calculated. Therefore, in the above-described search technique, it takes time to calculate the evaluation value corresponding to the tried processing conditions, and as a result, there is a problem that it takes time to search for the optimum processing conditions.
- the present disclosure solves the above problems, and when searching for the optimum processing conditions, for all processing conditions to be tried, processing under the processing conditions until the vibrational change in the processing result for the processing machine settles down. It is an object of the present invention to provide a machining condition searching device and a machining condition searching method capable of shortening the time required to search for optimum machining conditions, compared with the conventional technology that implements the above.
- a machining condition search device includes a machining condition calculation unit that generates machining conditions configured by a plurality of control parameters that can be set for a machining machine, and a machining condition that is provided to the machining machine according to the machining conditions generated by the machining condition calculation part.
- An actual machining command unit that executes machining a machining result collection unit that collects machining result information indicating the machining results of machining that the actual machining command unit causes the machine to execute, and the machining result information collected by the machining result collection unit.
- the convergence determination unit estimates the estimated convergence value that is the convergence destination of the provisional evaluation value, and if the convergence determination unit determines that the provisional evaluation value has not converged, the provisional evaluation A stop determination unit that determines whether or not to terminate machining under the machining conditions being trialed before the value converges, and if the stop determination unit determines to terminate machining under the machining conditions being trialed, the actual machining command part
- the estimated convergence value estimated by the convergence determination unit is determined as the evaluation value of the processing performed according to the processing conditions, and the stop determination unit performs processing under the processing conditions being trialed.
- the evaluation determination unit determines the convergence value of the provisional evaluation value as the evaluation value, and the evaluation value determined by the evaluation determination unit , a prediction unit that predicts a predicted value of an evaluation value corresponding to an untried processing condition based on the processing condition corresponding to the evaluation value; In this case, the optimum machining condition is determined based on the evaluation value determined by the evaluation determination unit, and if the search is not completed, the machining condition calculation unit should be tried next based on the predicted value predicted by the prediction unit.
- the processing machine when searching for the optimum processing conditions, for all processing conditions to be tried, the processing machine was made to perform processing under the processing conditions until the vibrational change in the processing result settled down Conventionally Compared to technology, it is possible to shorten the time required to find the optimum processing conditions.
- FIG. 1 is a diagram showing a configuration example of a machining condition search device according to Embodiment 1;
- FIG. 4 is a flow chart for explaining the operation of the machining condition search device according to Embodiment 1;
- the stop determination unit determines whether or not to discontinue machining under the machining conditions during trials by comparing the largest provisional evaluation value among the provisional evaluation values within the interquartile range with the discontinuation threshold value. It is an image diagram of an example of a method for determining.
- the stop determination unit determines whether or not to terminate machining under the machining conditions during trial by comparing the provisional evaluation value included in the interval ⁇ of the average value of the provisional evaluation values with the discontinuation threshold.
- FIG. 1 is a diagram showing a configuration example of a machining condition search device according to Embodiment 1;
- FIG. 4 is a flow chart for explaining the operation of the machining condition search device according to Embodiment 1;
- the stop determination unit determines whether or not to discontinue machining under the
- FIG. 10 is an image diagram of an example of a determination method; 4 is a graph conceptually showing a relationship between a predicted evaluation value and an index indicating uncertainty in Embodiment 1.
- FIG. 6A and 6B show the time until the optimal machining conditions are searched for in the conventional optimum machining condition searching technique and the time until the optimum machining conditions are searched for by the machining condition searching apparatus according to the first embodiment. It is a graph which shows an example of the result of having compared .
- FIG. 10 is a diagram for explaining an example of a method of setting a variable termination threshold value by a stop determination unit based on tried processing conditions and evaluation values corresponding to the processing conditions in the first embodiment
- 8A and 8B are diagrams showing an example of the hardware configuration of the machining condition search device according to Embodiment 1.
- FIG. 10 is a diagram for explaining an example of a method of setting a variable termination threshold value by a stop determination unit based on tried processing conditions and evaluation values corresponding to the processing conditions in the first embodiment
- 8A and 8B are diagrams showing an example of the hardware configuration of the machining condition search device according to Embodiment 1.
- FIG. 1 is a diagram showing a configuration example of a machining condition search device 1 according to Embodiment 1.
- a processing condition search device 1 according to Embodiment 1 is connected to a processing machine 2 and a display unit 3 .
- the machining condition search device 1 searches for optimum machining conditions (hereinafter referred to as "optimal machining conditions") from a large number of machining conditions that can be set for the machining machine 2.
- the optimum machining conditions are, for example, machining conditions that provide a machining result that satisfies the required machining specifications.
- the display unit 3 displays the processing conditions and the like searched by the processing condition search device 1 in accordance with requests from users such as processing workers.
- the display unit 3 displays processing conditions set in the processing machine 2 and evaluation values of processing performed by the processing machine 2 according to the processing conditions.
- the display unit 3 displays a processing condition that the processing machine 2 has not performed and a predicted evaluation value of the processing under the assumption that the processing machine 2 has performed processing according to this processing condition.
- the optimum machining conditions which are the search results of the search by the machining condition searching device 1, are displayed.
- the display unit 3 is provided outside the processing condition search device 1 and the processing machine 2 in FIG. 1, this is merely an example.
- the display unit 3 may be provided in the processing condition search device 1 or may be provided in the processing machine 2, for example.
- the processing machine 2 is an industrial device that performs processing according to processing conditions.
- the processing machine 2 forms a workpiece into a desired shape by, for example, removing unnecessary portions.
- the processing machine 2 can also perform additional processing, for example.
- the workpiece will be referred to as a workpiece.
- the work material is, for example, metal. Note that this is only an example, and the material of the workpiece is not limited to metal.
- the workpiece material may be ceramic, glass, or wood, for example.
- the processing machine 2 includes, for example, a laser processing machine, an electric discharge machine, a cutting machine, a grinding machine, an electrolytic processing machine, an ultrasonic processing machine, an electron beam processing machine, or an additional processing machine.
- the processing machine 2 shall be a laser processing machine. Note that this is merely an example, and in Embodiment 1, the processing machine 2 may be a processing machine other than a laser processing machine.
- the processing machine 2 is capable of performing normal processing for shaping the workpiece into a desired shape, and is also capable of performing experimental processing on the workpiece.
- the machining condition searching apparatus 1 according to Embodiment 1 generates trial machining conditions and causes the machining machine 2 to perform the experimental machining according to the machining conditions.
- the processing machine 2 performs preset experimental processing on the workpiece in accordance with the above processing conditions.
- the processing conditions are configured by combining a plurality of control parameters used for controlling the processing machine 2 .
- Control parameters are, for example, laser power, cutting speed, beam diameter, focus position, and gas pressure.
- the machining condition search device 1 generates trial machining conditions for searching from among such a huge number of combinations of machining conditions, and causes the machining machine 2 to perform experimental machining.
- the processing condition search device 1 collects information indicating processing results (hereinafter referred to as “processing result information”) from the processing machine 2 .
- the processing result information is, for example, information indicating the state of the processing machine 2 during processing, information indicating the state of the work being processed, or information indicating the state of the work after processing.
- the processing result information also includes information on the processing conditions that the processing machine 2 followed when performing processing.
- the processing machine 2 includes a sensor that detects sound, light, or processing speed generated during processing, and the processing condition search device 1 collects processing result information from the sensor.
- the sensor may be an imaging device that acquires an image of the workpiece after processing, or a measuring instrument that measures the unevenness of the cut surface of the workpiece.
- the sensor may be provided at a location different from the processing machine 2 . It is sufficient that the processing condition search device 1 can collect processing result information.
- the machining condition search device 1 determines the evaluation value of the machining carried out according to the machining conditions based on the machining result information collected from the machining carried out according to the machining conditions. Then, the machining condition search device 1 searches for the optimum machining condition while predicting the evaluation value corresponding to the untried machining condition based on the combination of the machining condition and the evaluation value. The details of how the machining condition search device 1 searches for the optimum machining conditions will be described later.
- the processing result obtained when the processing machine 2 performs processing under certain processing conditions may vibrate during the process of processing.
- the evaluation value corresponding to the machining result calculated based on the machining result also vibrates.
- the machining condition search device 1 causes the machining machine 2 to carry out machining for a certain amount of time until vibrational changes in the machining result of machining according to each machining condition settle down for all the machining conditions to be tried. If it is necessary to wait for the vibratory change in the machining result to settle down, it takes time to calculate the evaluation value corresponding to each machining condition.
- the machining condition search device 1 has an evaluation value calculated in the process until the vibrational change of the machining result settles down, even if it is an evaluation value before the vibrational change settles down. If it is an evaluation value that is assumed to have no effect in searching for the optimum machining conditions, it is adopted for searching for the optimum machining conditions, and the machining under experiment according to the machining conditions under trial is terminated, and the search is performed. Switch the machining conditions for As a result, the machining condition search device 1 according to Embodiment 1 shortens the time required to search for the optimum machining conditions.
- the machining condition search device 1 includes a searched machining condition generation unit 11 , a machining result collection unit 12 , an evaluation value acquisition unit 13 , a convergence determination unit 14 , a stop determination unit 15 , an evaluation determination unit 16 and a machine learning unit 17 .
- the processing condition search device 1 includes a processing result storage unit 18A, an evaluation value storage unit 18B, a convergence result storage unit 18C, a stop determination storage unit 18D, a search result storage unit 18E, a prediction result storage unit 18F, and an uncertainty A storage unit 18G is provided. All or part of the storage units 18A to 18G may be provided in an external device provided separately from the processing condition searching apparatus 1.
- the search machining condition generation unit 11 generates machining conditions used in actual machining for experiments, and causes the processing machine 2 to perform machining according to the generated machining conditions. That is, the searched machining condition generation unit 11 generates machining conditions to be searched for by actual machining in a multi-dimensional space whose dimensions are control parameters constituting the machining conditions. As shown in FIG. 1 , the search machining condition generation unit 11 includes a machining condition calculation unit 111 , an actual machining command unit 112 , and a search end determination unit 113 .
- the machining condition calculation unit 111 of the search machining condition generation unit 11 generates machining conditions composed of a plurality of control parameters that can be set for the processing machine 2 .
- the machining condition calculator 111 generates machining conditions used in experimental machining.
- the processing condition calculation unit 111 selects a combination corresponding to the processing content from a combination of a plurality of control parameters of the processing machine 2 and the range of values that these control parameters can take, and generates processing conditions from the selected combination. do.
- Control parameters are, for example, laser power, cutting speed, beam diameter, focus position, and gas pressure.
- the machining condition calculation unit 111 outputs the generated machining conditions to the actual machining command unit 112 .
- the actual machining command unit 112 causes the processing machine 2 to perform machining according to the machining conditions generated by the machining condition calculation unit 111 .
- the actual machining command unit 112 causes the processing machine 2 to continue machining according to the machining conditions generated by the machining condition calculation unit 111 .
- the actual machining command unit 112 generates a command for operating the processing machine 2 according to the processing conditions output from the processing condition calculation unit 111 and outputs the generated command to the processing machine 2 .
- the processing machine 2 performs processing according to processing conditions based on commands output from the actual processing command unit 112 .
- the evaluation determining unit 16 when the evaluation determining unit 16 outputs an instruction to end machining under the machining conditions being trialed (hereinafter referred to as "machining end instruction"), the actual machining command unit 112 currently instructs the machining machine 2 to Terminate the processing for the experiment that is being performed. Details of the evaluation determining unit 16 will be described later.
- the search end determination unit 113 determines whether or not to end the search for the processing conditions based on the information stored in the prediction result storage unit 18F or the uncertainty storage unit 18G. When the search end determination unit 113 determines that there is no need to additionally search for processing conditions, the search completion determination unit 113 determines optimum processing conditions based on the evaluation values determined by the evaluation determination unit 16 . Specifically, the search end determination unit 113 sets the processing condition corresponding to the highest evaluation value among the evaluation values stored in the search result storage unit 18E as the optimum processing condition. Details of the evaluation determining unit 16 will be described later. Further, when the search end determination unit 113 determines that it is necessary to additionally search for processing conditions, it causes the processing condition calculation unit 111 to generate processing conditions for the search to be tried next.
- the machining result collection unit 12 collects, from the machining machine 2, machining result information indicating machining results of machining performed according to the machining conditions.
- the machining result collection unit 12 collects machining results each time the actual machining command unit 112 causes the machining to be performed. As described above, the actual machining command unit 112 continues machining according to the machining conditions. While the processing machine 2 performs processing, multiple steps of processing are performed. Therefore, when the processing machine 2 performs experimental processing according to certain processing conditions, a plurality of pieces of processing result information are collected.
- the processing result collection unit 12 stores the collected processing result information in the processing result storage unit 18A.
- the processing result collection unit 12 causes the processing result storage unit 18A to store the processing result information in association with, for example, the acquisition time of the processing result information.
- the processing result storage unit 18A stores processing result information in chronological order.
- the evaluation value acquisition unit 13 calculates an evaluation value for the processing performed by the processing machine 2 based on the processing result information collected by the processing result collection unit 12 .
- the evaluation value calculated by the evaluation value acquiring unit 13 based on the processing result information is also called "provisional evaluation value”.
- the evaluation value acquisition unit 13 calculates a provisional evaluation value for each piece of processing result information. That is, the evaluation value acquisition unit 13 calculates a provisional evaluation value for each processing step. Note that the evaluation value acquisition unit 13 acquires the processing result information collected by the processing result collection unit 12 from the processing result storage unit 18A.
- the evaluation value is a value that indicates the quality of machining, and is defined as a value that indicates that the larger the value, the better the machining.
- the evaluation value is indicated by a value from 0 to 1, for example. In this case, the evaluation value is 1 when the best processing is performed, and the evaluation value is 0 when the worst processing is performed.
- the evaluation value acquisition unit 13 stores information in which the acquisition time of the processing result information, the processing conditions, and the calculated provisional evaluation value (hereinafter referred to as “provisional evaluation value information”) are associated with each other in the evaluation value storage unit 18B.
- provisional evaluation value information it is assumed that the acquisition time of the processing result information is associated with the processing condition and the provisional evaluation value, but this is only an example.
- the calculation time of the provisional evaluation value may be associated with the processing condition and the provisional value.
- the evaluation value storage unit 18B stores provisional evaluation value information in chronological order.
- the convergence determination unit 14 determines whether or not the provisional evaluation values have converged based on the time-series provisional evaluation values calculated by the evaluation value acquisition unit 13 .
- "convergence” refers to the disappearance of oscillatory changes in values.
- the convergence determination unit 14 determines whether the provisional evaluation values have converged for each processing condition.
- the convergence determination unit 14 acquires the time-series provisional evaluation values calculated by the evaluation value acquisition unit 13 from the provisional evaluation value information stored in the evaluation value storage unit 18B.
- the convergence determination unit 14 determines that the provisional evaluation values have converged, the acquisition time of the processing result information, information indicating that the provisional evaluation values have converged, the processing conditions, the provisional evaluation values, the provisional Information associated with the convergence value of the evaluation value is stored in the convergence result storage unit 18C as post-convergence determination information.
- the calculation time of the provisional evaluation value may be associated instead of the acquisition time of the processing result information.
- the convergence determination unit 14 sets the latest provisional evaluation value as the convergence value of the provisional evaluation values. Note that this is only an example, and for example, information defining how to calculate the convergence value of the provisional evaluation value based on the time-series provisional evaluation value (hereinafter referred to as "convergence value calculation information”) is provided in advance.
- the convergence determination unit 14 may calculate the convergence value of the provisional evaluation value according to the convergence value calculation information. On the other hand, when determining that the provisional evaluation value has not converged, the convergence determination unit 14 estimates a value to which the provisional evaluation value converges (hereinafter referred to as "estimated convergence value"). Then, the convergence determination unit 14 determines the acquisition time of the processing result information, the information indicating that the provisional evaluation value has not converged, the processing condition, the provisional evaluation value, and the estimated convergence value information associated with each other. It is stored in the convergence result storage unit 18C as post-determination information. The calculation time of the provisional evaluation value may be associated instead of the acquisition time of the processing result information.
- the stop determination unit 15 determines whether or not to terminate machining under the machining conditions being trialed before the provisional evaluation value converges. The stop determination unit 15 determines whether or not to terminate the machining under the machining condition being trialed for each machining condition. Note that the stop determination unit 15 may determine from the post-convergence determination information stored in the convergence result storage unit 18C that the convergence determination unit 14 has determined that the provisional evaluation value has not converged. The stop determination unit 15 may directly acquire information from the convergence determination unit 14 indicating that the provisional evaluation value has not converged. Note that an arrow from the convergence determination unit 14 to the stop determination unit 15 is omitted in FIG.
- the stop determination unit 15 determines whether or not to discontinue machining under the machining conditions being trialed (hereinafter referred to as “discontinuation determination result") and the latest post-convergence determination information output from the convergence determination unit 14.
- the associated information (hereinafter referred to as “post-determination information”) is stored in the stop determination storage unit 18D.
- the stop determination storage unit 18D stores information after discontinuation determination.
- the evaluation determination unit 16 instructs the actual processing command unit 112 to end processing according to the processing conditions for the processing machine 2, and determines convergence.
- the estimated convergence value estimated by the unit 14 is determined as the evaluation value of the machining performed according to the machining conditions.
- the evaluation determination unit 16 determines that the convergence determination unit 14 determines that the provisional evaluation value has converged, and then determines the convergence value of the provisional evaluation value. is determined to be the evaluation value of the machining performed according to the machining conditions.
- the evaluation determining unit 16 determines an evaluation value for the machining performed according to the machining conditions for each machining condition.
- the evaluation determination unit 16 determines whether or not the stop determination unit 15 has determined to terminate machining under the processing conditions during trial, the estimated convergence value estimated by the convergence determination unit 14, or the convergence value of the provisional evaluation value. It may be specified from the post-determination information stored in the determination storage unit 18D. For example, the evaluation determination unit 16 may directly acquire post-discontinuation determination information from the stop determination unit 15 .
- the arrow from the stop determination part 15 to the evaluation determination part 16 is abbreviate
- the evaluation determination unit 16 causes the search result storage unit 18E to store the combination of the processing condition and the evaluation value as the search result.
- the search result storage unit 18E stores search results.
- the machine learning unit 17 uses the search results stored in the search result storage unit 18E to predict processing evaluation values corresponding to processing conditions that have not been tried (processing has not been performed). In addition, the machine learning unit 17 calculates the uncertainty with respect to the predicted value of the evaluation value, that is, how easily the prediction is off.
- the machine learning unit 17 has a prediction unit 171 and an uncertainty evaluation unit 172 .
- the prediction unit 171 predicts an evaluation value corresponding to an untried processing condition based on the evaluation value determined by the evaluation determination unit 16 and the processing condition corresponding to the evaluation value.
- the prediction unit 171 may acquire the evaluation value determined by the evaluation determination unit 16 and the processing conditions corresponding to the evaluation value from the search results stored in the search result storage unit 18E.
- the prediction unit 171 causes the prediction result storage unit 18F to store information (hereinafter referred to as "prediction result information") in which the predicted value of the evaluation value obtained by prediction is associated with the processing condition.
- the prediction result information is information in which an untried processing condition is associated with a corresponding predicted evaluation value.
- the prediction result storage unit 18F stores prediction result information.
- the uncertainty evaluation unit 172 calculates an index indicating the uncertainty of prediction of the evaluation value by the prediction unit 171 .
- the uncertainty evaluation unit 172 uses the search results stored in the search result storage unit 18E to calculate the uncertainty of the evaluation value with respect to the predicted value, that is, an index indicating how likely the prediction is to come off.
- the uncertainty evaluation unit 172 causes the uncertainty storage unit 18G to store information (hereinafter referred to as “uncertainty information”) in which the calculated index values are associated with the processing conditions.
- the uncertainty information is information in which an untried processing condition is associated with an index value indicating the uncertainty of prediction of the corresponding evaluation value.
- the uncertainty storage unit 18G stores uncertainty information.
- FIG. 2 is a flow chart for explaining the operation of the machining condition search device 1 according to the first embodiment.
- the machining condition calculator 111 of the searched machining condition generator 11 When the machining condition search process is started, first, the machining condition calculator 111 of the searched machining condition generator 11 generates initial machining conditions (step ST1).
- the machining condition calculator 111 generates initial machining conditions by selecting a predetermined number of machining conditions as initial machining conditions from among all combinations that can be set as machining conditions. Examples of methods for selecting the initial machining conditions by the machining condition calculator 111 include design of experiments, optimal design, optimal design, or random sampling.
- the machining condition calculator 111 may use the machining conditions input by the user as the initial machining conditions. Note that these methods are merely examples, and the machining condition calculation unit 111 may use any method to generate the initial machining conditions.
- the machining condition calculator 111 selects, for example, 10 kinds of machining conditions as initial machining conditions from this combination. Note that the number of control parameters constituting the machining conditions, the number of steps that can be set for each control parameter, or the number of machining conditions selected as initial machining conditions are not limited to these. Depending on the control parameter, the number of steps that can be set may differ.
- the machining condition search device 1 selects one initial machining condition from among the initial machining conditions generated by the machining condition calculator 111, and causes the machining machine 2 to carry out machining under the selected initial machining condition (step ST2). ).
- the machining condition calculation unit 111 selects one of the initial machining conditions and outputs the selected initial machining condition to the actual machining command unit 112 of the search machining condition generation unit 11 .
- the actual machining command unit 112 generates a command for operating the processing machine 2 based on the initial machining conditions output from the machining condition calculation unit 111 and outputs the generated command to the processing machine 2 .
- the processing machine 2 performs processing based on the initial processing conditions selected by the processing condition calculation unit 111 .
- the machining condition search device 1 first causes the machining machine 2 to carry out machining based on the initial machining conditions. Processing based on the initial processing conditions is hereinafter also referred to as “initial processing”.
- the machining result collection unit 12 collects machining result information indicating the machining result of the initial machining performed according to the initial machining conditions from the machining machine 2 (step ST3).
- the processing result collection unit 12 stores the collected processing result information in the processing result storage unit 18A.
- the evaluation value acquisition unit 13 calculates a provisional evaluation value for the processing performed by the processing machine 2 according to the initial processing conditions in step ST2 (step ST4).
- the evaluation value acquisition unit 13 causes the evaluation value storage unit 18B to store provisional evaluation value information in which the acquisition time of the processing result information, the processing condition (in this case, the initial processing condition) and the calculated provisional evaluation value are associated with each other.
- the convergence determination unit 14 determines whether or not the provisional evaluation values have converged based on the time-series provisional evaluation values calculated by the evaluation value acquisition unit 13 .
- the convergence determination unit 14 determines that the provisional evaluation values have converged, the acquisition time of the machining result information, the information indicating that the provisional evaluation values have converged, the machining conditions, here, the initial machining conditions, Information after convergence determination in which the provisional evaluation value and the convergence value of the provisional evaluation value are associated is stored in the convergence result storage unit 18C.
- the convergence determination unit 14 estimates the estimated convergence value, obtains the processing result information acquisition time, the information indicating that the provisional evaluation value has not converged, and the processing result information.
- the convergence result storage unit 18C stores information after convergence determination in which the conditions, here, the initial processing conditions, the provisional evaluation value, and the estimated convergence value are associated with each other (step ST5).
- step ST5 the convergence determination unit 14 determines whether or not the provisional evaluation values have converged based on the time-series provisional evaluation values, and when it is determined that the provisional evaluation values have not converged.
- a method for estimating the estimated convergence value of will be described with a specific example.
- the convergence determination unit 14 determines whether the provisional evaluation values have converged and estimates the determination estimated convergence value, for example, based on the degree of variation of the time-series provisional evaluation values. As a specific example, for example, the convergence determination unit 14 obtains the interquartile range of the provisional evaluation values from the time-series provisional evaluation values. Then, the convergence determination unit 14 determines whether or not the provisional evaluation values have converged based on the range of values in the interquartile range of the provisional evaluation values. For example, a range of values (hereinafter referred to as "first convergence determination range”) is determined in advance when it is determined that the provisional evaluation values have converged.
- first convergence determination range a range of values
- the convergence determination unit 14 determines that the provisional evaluation value has converged. The convergence determination unit 14 determines that the provisional evaluation value has not converged unless the interquartile range of the provisional evaluation value falls within the first convergence determination range. If the convergence determination unit 14 determines that the provisional evaluation values have not converged, then it estimates an estimated convergence value from the interquartile range of the provisional evaluation values obtained from the time-series provisional evaluation values. For example, the convergence determination unit 14 estimates the median value of the interquartile range of the provisional evaluation values as the estimated convergence value.
- the convergence determination unit 14 regards the time-series provisional evaluation values as a specific distribution, estimates the distribution, and calculates the average value ⁇ ⁇ of the provisional evaluation values in the distribution of the provisional evaluation values. It may be determined whether or not the provisional evaluation value has converged based on the value of the interval of . For example, a range of values (hereinafter referred to as "second convergence determination range”) is determined in advance when it is determined that the provisional evaluation values have converged. The convergence determination unit 14 determines that the provisional evaluation values have converged if the value in the interval of the average value ⁇ of the provisional evaluation values in the distribution of the provisional evaluation values is within the second convergence determination range.
- the convergence determination unit 14 determines that the provisional evaluation values have not converged if the value in the interval ⁇ of the average value of the provisional evaluation values in the distribution of the provisional evaluation values does not fall within the second convergence determination range. . If the convergence determination unit 14 determines that the provisional evaluation values have not converged, then it estimates an estimated convergence value from the distribution estimated from the time-series provisional evaluation values. For example, the convergence determination unit 14 estimates the average value of the provisional evaluation values as the estimated convergence value.
- the convergence determination unit 14 estimates an estimated convergence value based on a trained model (hereinafter referred to as a “first machine learning model”) that receives time-series evaluation values and outputs an estimated convergence value.
- the convergence determination unit 14 inputs the time-series provisional evaluation values to the first machine learning model to obtain an estimated convergence value.
- the first machine learning model may be a model that outputs information on the degree of variation of the provisional evaluation value in addition to the estimated convergence value.
- the convergence determination unit 14 determines whether or not the provisional evaluation values have converged based on information regarding the degree of variation in the provisional evaluation values obtained by inputting the time-series provisional evaluation values into the first machine learning model. good.
- the stop determination unit 15 determines whether or not to stop machining under the initial machining conditions during trials before the provisional evaluation value converges (step ST6).
- stop determination unit 15 determines whether or not to terminate machining under the machining conditions being trialed before the provisional evaluation value converges.
- the stop determination unit 15 determines the degree of variation in the time-series provisional evaluation values calculated by the evaluation value acquisition unit 13 and stored in the evaluation value storage unit 18B, and a threshold value (hereinafter referred to as “canceling threshold value”). Through the comparison, it is determined whether or not to terminate machining under the machining conditions being trialed before the provisional evaluation value converges.
- the abort threshold is specified in advance by the user and stored in the stop determination unit 15 .
- the user sets an evaluation value (hereinafter referred to as "reference evaluation value”) as a reference for discontinuation, which is to be discontinued under the processing conditions under trial if the value is not exceeded, as a threshold value for discontinuation. Specify.
- the user sets the reference evaluation value, for example, according to the performance required of the processing machine 2 .
- the stop determination unit 15 determines the interquartile range of the provisional evaluation values in the interquartile range. By comparing the largest provisional evaluation value and the discontinuation threshold value, it is determined whether or not to discontinue the processing under the processing conditions during the trial. In this case, if the largest provisional evaluation value among the provisional evaluation values within the interquartile range is less than the discontinuation threshold value, the stop determination unit 15 determines to terminate the processing under the processing conditions during trial. On the other hand, if the largest provisional evaluation value among the provisional evaluation values within the interquartile range is equal to or greater than the discontinuation threshold value, the stop determination unit 15 determines to continue processing under the processing conditions during trial.
- FIG. 3 shows that in the first embodiment, the stop determination unit 15 compares the largest provisional evaluation value among the provisional evaluation values in the interquartile range with the threshold value for discontinuation, and the processing under the processing conditions during the trial.
- FIG. 10 is an image diagram of an example of a method for determining whether or not to discontinue.
- the horizontal axis of FIG. 3 indicates the time width during which machining is performed according to certain machining conditions, and the vertical axis of FIG. 3 indicates the evaluation value (provisional evaluation value).
- the dots indicated by black circles in FIG. 3 indicate provisional evaluation values calculated based on the machining results of machining performed according to the machining conditions.
- FIG. 3 illustrates how the provisional evaluation values converge.
- FIG. 10 is an image diagram of an example of a method for determining whether or not to discontinue.
- the horizontal axis of FIG. 3 indicates the time width during which machining is performed according to certain machining conditions
- the vertical axis of FIG. 3 indicates the evaluation value (provisional evaluation
- 201a, 201b, and 201c indicate interquartile ranges of provisional evaluation values.
- the interquartile range of the provisional evaluation value is the range indicated by 201a at the time t1 hour has elapsed
- the interquartile range of the provisional evaluation value is the range indicated by 201b at the time t2 hours have elapsed.
- the largest provisional evaluation value among the provisional evaluation values within the interquartile range is equal to or greater than the discontinuation threshold. Therefore, in this case, the stop determination unit 15 determines to continue machining under the machining conditions being trialed.
- the stop determination unit 15 determines to terminate the machining under the machining conditions being trialed.
- the stop determination unit 15 determines the provisional It may be determined whether or not to discontinue machining under the machining conditions being trialed by comparing the evaluation value with the discontinuation threshold. In this case, if all the provisional evaluation values included in the interval ⁇ of the average value of the provisional evaluation values are less than the discontinuation threshold value, the stop determination unit 15 determines to terminate the processing under the processing conditions during the trial. On the other hand, the stop determination unit 15 determines to continue the machining under the machining conditions during trial if all the provisional evaluation values included in the interval ⁇ of the average value of the provisional evaluation values are not less than the discontinuation threshold.
- FIG. 4 shows that, in Embodiment 1, the stop determination unit 15 performs processing under the processing conditions during trial by comparing the provisional evaluation value included in the interval ⁇ of the provisional evaluation value with the threshold value for termination.
- FIG. 10 is an image diagram of an example of a method for determining whether or not to discontinue;
- the horizontal axis of FIG. 4 indicates the time width during which machining is performed according to certain machining conditions, and the vertical axis of FIG. 4 indicates the evaluation value (provisional evaluation value).
- the dots indicated by black circles in FIG. 4 indicate provisional evaluation values calculated based on the machining results of the machining performed according to the machining conditions.
- FIG. 4 illustrates how the provisional evaluation values converge.
- FIG. 4 illustrates how the provisional evaluation values converge.
- 301a, 301b, and 301c indicate the largest provisional evaluation value among the provisional evaluation values included in the interval of the mean value ⁇ of the provisional evaluation values.
- the largest provisional evaluation value among the provisional evaluation values included in the interval of the average provisional evaluation value ⁇ ⁇ is the value indicated by 301a
- the provisional evaluation value average ⁇ ⁇ is the value indicated by 301b. Both the value indicated by 301a and the value indicated by 301b are equal to or greater than the threshold for termination.
- the stop determination unit 15 determines to continue machining under the machining conditions being trialed.
- the largest provisional evaluation value among the provisional evaluation values included in the interval of the average provisional evaluation value ⁇ becomes the value indicated by 301c.
- the value indicated by 301c is less than the truncation threshold.
- the stop determination unit 15 determines to terminate the machining under the machining conditions being trialed.
- the stop determination unit 15 receives a time-series evaluation value and outputs information indicating whether or not to stop processing based on a learned model (hereinafter referred to as a “second machine learning model”). It is also possible to determine whether or not to terminate machining under the machining conditions being trialed before the provisional evaluation value converges.
- the stop determination unit 15 obtains information indicating whether to stop processing by inputting the time-series provisional evaluation values calculated by the evaluation value acquisition unit 13 to the second machine learning model.
- the stop determination unit 15 may acquire the time-series provisional evaluation values calculated by the evaluation value acquisition unit 13 from, for example, the post-convergence determination information stored in the convergence result storage unit 18C.
- the stop determination unit 15 causes the stop determination storage unit 18D to store post-discontinuation information in which the discontinuation determination result and the latest post-convergence determination information output from the convergence determination unit 14 are associated.
- the evaluation determination unit 16 sends the actual machining command unit At 112, the processing according to the initial processing conditions for the processing machine 2 is finished. Specifically, the evaluation determination unit 16 outputs a machining end instruction to the actual machining command unit 112 . When the evaluation determination unit 16 outputs a machining end instruction, the actual machining command unit 112 terminates the machining currently being performed by the processing machine 2 according to the initial machining conditions generated in step ST1. Let Also, the evaluation determination unit 16 determines the estimated convergence value estimated by the convergence determination unit 14 as the evaluation value of the machining performed according to the initial machining conditions.
- the evaluation determination unit 16 causes the search result storage unit 18E to store the combination of the processing condition and the evaluation value as the search result (step ST8). Specifically, the evaluation determination unit 16 causes the search result storage unit 18E to store the combination of the initial processing condition and the evaluation value, here the estimated convergence value, as the search result.
- the stop determination unit 15 determines that the machining under the initial machining conditions during trial is not terminated (“NO” in step ST6)
- the evaluation determination unit 16 determines that the convergence determination unit 14 has converged the provisional evaluation value. (step ST7).
- the convergence determination unit 14 determines that the provisional evaluation values have not converged (“NO" in step ST7), the operation of the processing condition search device 1 returns to the process of step ST2.
- the evaluation determination unit 16 determines the convergence value of the provisional evaluation value as the evaluation value. Then, the evaluation determination unit 16 causes the search result storage unit 18E to store the combination of the processing condition and the evaluation value as the search result (step ST8). Specifically, the evaluation determination unit 16 stores a combination of the initial processing condition and the evaluation value (here, the convergence value of the provisional evaluation value) as the search result in the search result storage unit 18E.
- the machining condition calculator 111 confirms whether or not the initial machining is completed for all the machining conditions selected as the initial machining conditions (step ST9). If there is an initial machining condition for which initial machining has not been completed (“NO” in step ST9), the processing from step ST1 to step ST8 is sequentially performed for the initial machining condition for which initial machining has not been completed. In step ST1 from the second time onward, the machining condition calculator 111 selects initial machining conditions that have not been selected in step ST1 so far. As a result, search results in which all initial machining conditions (for example, 10 initial machining conditions) and combinations of evaluation values are associated with each other are stored in the search result storage unit 18E.
- the prediction unit 171 calculates 99990 predicted evaluation values.
- steps ST15 to ST22 selection of machining conditions, execution of machining, collection of machining results, calculation of provisional evaluation values, prediction of convergence values of the provisional evaluation values, A decision is made as to whether or not to terminate the machining immediately, and an evaluation value is determined, and the process of step ST10 is performed after the process of step ST22.
- step ST10 is performed via the processing of steps ST15 to ST22, the machining conditions set in step ST15 are excluded from the untried machining conditions.
- An example of a method for the prediction unit 171 to calculate a predicted value of an evaluation value corresponding to an untried processing condition is a method using Gaussian process regression.
- the prediction unit 171 predicts an evaluation value corresponding to an untried processing condition using Gaussian process regression, the following calculations are performed.
- a method using Gaussian process regression is an example of a method using a probabilistic model for processing conditions of evaluation values generated on the assumption that evaluation values for processing conditions are random variables following a specific distribution.
- N be the number of observation values, that is, the number of processing conditions for which processing is performed and evaluation values are calculated
- CN the Gram matrix, and let this in each processing condition stored in the search result storage unit 18E.
- the predicted value m(x N+1 ) of the evaluation value for the untried machining condition x N+1 can be calculated by the following equation (1).
- k is a vector in which the values of the kernel function are arranged when each of the searched machining conditions x 1 , . .
- the superscript T represents transposition, and the superscript -1 represents an inverse matrix.
- m(x N+1 ) kT ⁇ (C N ⁇ 1 ) ⁇ t (1)
- the evaluation value prediction method used by the prediction unit 171 is not limited to this.
- the prediction unit 171 may predict the evaluation value using supervised learning such as decision tree, linear regression, boosting, and neural network.
- the prediction unit 171 When the prediction unit 171 predicts an evaluation value corresponding to an untried processing condition, the prediction unit 171 stores the predicted evaluation value (step ST11). Specifically, the prediction unit 171 causes the prediction result storage unit 18F to store prediction result information in which the predicted value of the evaluation value predicted in step ST10 is associated with the processing condition.
- the uncertainty evaluation unit 172 of the machine learning unit 17 uses the search results stored in the search result storage unit 18E to obtain an index indicating the uncertainty of the prediction of the evaluation value corresponding to the untried processing conditions. is calculated (step ST12).
- An example of an indicator of uncertainty is a standard deviation calculated using Gaussian process regression, which is an example of a stochastic model.
- Gaussian process regression uses Gaussian process regression to provide an indicator of uncertainty, for example, the following calculations are performed. Let N be the number of observed values, that is, the number of processing conditions for which evaluation values are calculated by processing, let CN be the Gram matrix, and let the vector in which the processing conditions stored in the search result storage unit 18E are arranged.
- k be k
- c be the scalar value obtained by adding the accuracy parameter of the prediction model to the value of the kernel between untried processing conditions x N+1 .
- one of the control parameters constituting the machining conditions is set to x i (i is a natural number), and the values of this control parameter for each machining condition stored in the search result storage unit 18E are x 1 to x N.
- the standard deviation ⁇ (x N+1 ) which is an index indicating the uncertainty of the prediction of the evaluation value for the untried machining condition x N+1 , can be calculated by the following equation (3).
- Equation (3) the standard deviation ⁇ (x N+1 ) can be obtained by calculating the square root of the variance.
- ⁇ 2 (x N+1 ) c ⁇ k T ⁇ (C N ⁇ 1 ) ⁇ k (3)
- the uncertainty evaluation unit 172 calculates the index indicating the uncertainty of prediction using Gaussian process regression, but the method of calculating the index indicating uncertainty is not limited to this.
- the uncertainty evaluation unit 172 may calculate the index using techniques such as density estimation and mixture density network.
- FIG. 5 is a graph conceptually showing the relationship between the predicted value of the evaluation value and the index indicating uncertainty in the first embodiment.
- FIG. 5 shows an example in which a Gaussian process regression is used to calculate a predicted value and an indicator of uncertainty.
- the horizontal axis of FIG. 5 indicates the value x of the control parameter, which is the machining condition, and the vertical axis of FIG. 5 indicates the evaluation value.
- Points indicated by black circles in FIG. 5 indicate evaluation values calculated based on actual machining using initial machining conditions (hereinafter also referred to as evaluation values of actual machining).
- the evaluation value is predicted assuming that the evaluation value follows a Gaussian distribution. Therefore, if the predicted value of the evaluation value is the average m(x) of the Gaussian distribution and the index indicating the uncertainty of the prediction is the standard deviation ⁇ (x) of the Gaussian distribution, the actual evaluation value is about 95%. It is statistically shown that the probability falls within the range of m(x) ⁇ 2 ⁇ (x) or more and m(x)+2 ⁇ (x) or less.
- the solid curve indicates m(x), which is the predicted value of the evaluation value.
- the dashed curves represent the m(x)-2 ⁇ (x) curve and the m(x)+2 ⁇ (x) curve. As shown in FIG. 5, the index indicating uncertainty is small at locations close to the evaluation value of actual machining, and the index indicating uncertainty is large at locations distant from the evaluation value of actual machining.
- the uncertainty evaluation unit 172 stores an index indicating the uncertainty of the predicted value (step ST13). Specifically, the uncertainty evaluation unit 172 causes the uncertainty storage unit 18G to store uncertainty information in which the calculated index values are associated with the processing conditions.
- the search end determination unit 113 of the search processing condition generation unit 11 determines the predicted value of the evaluation value of the processing condition stored in the prediction result storage unit 18F and the predicted value of the evaluation value stored in the uncertainty storage unit 18G. It is determined whether or not to end the search for machining conditions using an index indicating uncertainty (step ST14). For example, the search end determination unit 113 compares the value of the index, which is stored in the uncertainty storage unit 18G and indicates the uncertainty of prediction of the evaluation values of all the processing conditions searched so far, with a threshold value, If the value of the index is equal to or less than the threshold, it is determined that the optimum machining conditions have been found, and the search for the machining conditions ends.
- the search end determination unit 113 uses the processing condition x, the predicted value m(x) of the evaluation value for this processing condition x, and the index (standard deviation) ⁇ (x) indicating the uncertainty of the prediction of this evaluation value.
- ⁇ is a parameter that is determined before searching for machining conditions. As the value of ⁇ is smaller, machining conditions with higher predicted evaluation values are selected. As for the value of ⁇ , the same value may be used continuously, or the value may be changed on the way.
- the search end determination unit 113 selects the highest evaluation value among the evaluation values of all the processing conditions stored in the search result storage unit 18E.
- the machining conditions associated with the values are determined as the optimum machining conditions.
- the search end determination unit 113 extracts the optimum machining conditions and outputs the extracted machining conditions to the actual machining command unit 112 .
- the actual machining command unit 112 outputs a command including the machining conditions output from the search end determination unit 113 to the processing machine 2 and sets the machining conditions in the processing machine 2 .
- the actual machining command unit 112 causes the processing machine 2 to perform normal machining according to the set machining conditions.
- the search end determination unit 113 may store the determined optimum machining conditions in a storage unit (not shown).
- the search end determination unit 113 determines that the processing condition The calculation unit 111 is instructed to generate machining conditions to be tried next.
- the processing condition calculation unit 111 uses the predicted value of the evaluation value of the processing condition stored in the prediction result storage unit 18F to perform the next processing. (step ST15). Specifically, the machining condition calculation unit 111 selects a machining condition to be tried next, that is, a new machining condition from among all the machining conditions. The machining conditions to be tried next generated by the machining condition calculation unit 111 are output to the actual machining command unit 112.
- the actual machining command unit 112 outputs to the processing machine 2 the command generated by the machining condition calculation unit 111 in step ST15 and including the machining conditions to be tried next, and causes the machining machine 2 to perform machining under the machining conditions.
- the processing result collection unit 12 collects processing result information (step ST17).
- the evaluation value acquiring unit 13 calculates a provisional evaluation value for the processing performed in step ST16 (step ST18).
- the convergence determination unit 14 determines whether the provisional evaluation values have converged and estimates an estimated convergence value based on the degree of variation of the time-series provisional evaluation values (step ST19).
- the stop determination unit 15 determines whether or not to terminate the machining under the machining conditions being trialed (step ST20).
- the evaluation determination unit 16 determines the estimated convergence value as the evaluation value, and the stop determination unit 15 determines that the machining under the machining conditions under trial is terminated. If it is determined not to terminate, the convergence determination unit 14 determines that the provisional evaluation values have converged (step ST21), and then determines the convergence value of the provisional evaluation values as the evaluation value. Then, the evaluation determination unit 16 stores the search result (step ST22). Next, the process proceeds to steps ST10 and ST12, and the processes described above are executed.
- the display unit 3 displays the information obtained in the course of the processing described above, the optimum processing conditions obtained as a result of the processing, and the like. For example, the display unit 3 displays the machining conditions obtained during the search for the machining conditions by the machining condition search device 1 and the evaluation values corresponding to the machining conditions. The display unit 3 also displays the machining conditions and the predicted values of the evaluation values corresponding to the machining conditions. Moreover, the display unit 3 displays the optimum machining conditions of the search result. That is, the display unit 3 displays the machining conditions read from the search result storage unit 18E and the evaluation values corresponding to the machining conditions, the machining conditions read from the prediction result storage unit 18F and the evaluation values corresponding to the machining conditions. or the optimum machining condition of the search result output from the machining condition calculator 111 is displayed. Accordingly, by referring to the information displayed on the display unit 3, the user can recognize the search status and search results of the processing conditions.
- the processing condition searching device 1 calculates a provisional evaluation value for the processing that has been performed based on the processing result information collected by causing the processing machine 2 to perform processing according to the generated processing conditions. Based on the calculated time-series provisional evaluation values, the processing condition search device 1 determines whether or not the provisional evaluation values have converged. Before convergence, it is determined whether or not to terminate machining under the machining conditions being trialed. The machining condition search device 1, for example, compares the degree of variation in the time-series provisional evaluation values (for example, the interquartile range of the provisional evaluation values or the distribution of the provisional evaluation values) with the discontinuation threshold to continue the processing as it is.
- the degree of variation in the time-series provisional evaluation values for example, the interquartile range of the provisional evaluation values or the distribution of the provisional evaluation values
- the processing condition search device 1 determines whether or not to end the search for the processing condition.
- the processing condition search device 1 repeats the above-described processing until it determines to end the search for the processing conditions. Thereby, the machining condition search device 1 determines the optimum machining conditions.
- the machining is performed by the machining machine 2 for a certain amount of time until the vibrational changes in the machining results settle down for all the machining conditions to be tried, and the vibrational changes in the machining results are performed.
- the evaluation value corresponding to the processing conditions was calculated after waiting for the sudden change to settle down. Therefore, the conventional technique for searching for optimum machining conditions has poor time efficiency until the optimum machining conditions can be searched.
- the machining condition search device 1 calculates the evaluation value ( Before the provisional evaluation value) converges, the machining under the machining conditions under trial is terminated, and the estimated convergence value is set as the evaluation value corresponding to the machining conditions under trial.
- the machining condition search device 1 determines that the machining result of machining according to a certain machining condition, which has been determined that a high evaluation value cannot be obtained, is discontinued from the point of time until the machining result of the machining converges. can omit the time until convergence. That is, the machining condition search device 1 can shorten the total time required to search for the optimum machining conditions by the time omitted above.
- FIGS. 6A and 6B show the time until the optimal machining conditions are searched for in the conventional optimum machining condition searching technique and the time until the optimum machining conditions are searched by the machining condition searching apparatus 1 according to the first embodiment. It is a graph which shows an example of the result of having compared with.
- FIG. 6A is a graph showing evaluation values until optimum machining conditions are searched for in a conventional technique for searching for optimum machining conditions
- FIG. It is a graph which shows the evaluation value until it is searched.
- the black circle points indicate the evaluation values calculated based on the machining results of the actual machining performed until the machining results converge.
- 6A and 6B are results of searching for optimum machining conditions for obtaining the same desired machining result for the same machining machine 2.
- the conventional optimum machining condition search technique continues machining until the machining result, in other words, the evaluation value converges, regardless of whether the evaluation value is good or bad. It takes time to In the example shown in FIG. 6A, it takes 21 minutes to find the optimum machining conditions.
- the machining condition searching apparatus 1 according to Embodiment 1 as shown in FIG. 6B, machining is terminated when the machining result, in other words, the evaluation value is expected to be low. Conditions can be explored. In the example shown in FIG. 6B, the optimal machining conditions are searched for in 14 minutes.
- the time required to search for the optimum machining condition by the machining condition searching apparatus 1 according to Embodiment 1 is the time required to search for the optimum machining condition by the conventional optimum machining condition search technique shown in FIG. 6A. 7 minutes less than the time.
- the stop determination unit 15 determines whether or not to terminate the machining under the machining condition being trialed before the provisional evaluation value converges.
- a reference evaluation value specified in advance by the user was used as the threshold value for termination. That is, the threshold value for termination was set to a fixed value. Then, the stop determination unit 15 determines whether or not to terminate the machining under the machining conditions during the trial by comparing the degree of dispersion of the time-series provisional evaluation values with the discontinuation threshold value before the provisional evaluation values converge. I was doing However, this is only an example.
- the stop determination unit 15 can also set the abort threshold based on the tried processing conditions and the evaluation values corresponding to the processing conditions.
- the tried processing conditions and evaluation values corresponding to the processing conditions are stored in the search result storage unit 18E by the evaluation determination unit 16 as search results.
- the termination threshold set by the stop determination unit 15 based on the determined evaluation value is also referred to as a “variable termination threshold”.
- the stop determination unit 15 compares the estimated convergence value estimated by the convergence determination unit 14 with the variable termination threshold, for example, before the provisional evaluation value converges. Determine whether or not to terminate machining under the machining conditions being trialed.
- the estimated convergence value estimated by the convergence determination unit 14 is the estimated convergence value in the latest post-convergence determination information stored in the convergence result storage unit 18C.
- the stop determination unit 15 for example, based on the tried processing condition and the evaluation value corresponding to the processing condition, according to a preset condition (hereinafter referred to as "variable abort threshold setting condition"), Sets the variable censoring threshold.
- conditions such as ⁇ Condition (1)>, ⁇ Condition (2)>, or ⁇ Condition (3)> below are set as the variable cut-off threshold setting conditions.
- ⁇ Condition (1)> If the number of trials is less than X times, the value for not terminating machining is set as a variable abortion threshold value, and if the number of trials is X times or more, the Xth among the evaluation values corresponding to all the tried machining conditions Use the evaluation value of the rank as a variable censoring threshold
- the evaluation value of the top Y rank is set as the threshold value for variable discontinuation.
- ⁇ Condition (3)> The lowest evaluation value among the top Z% evaluation values among the evaluation values corresponding to all the tried processing conditions is set as the variable abort threshold.
- ⁇ Condition (1)> The value of X, Y, or Z in ⁇ Condition (1)>, ⁇ Condition (2)>, or ⁇ Condition (3)> can be set as appropriate.
- the “value for not terminating processing” is, for example, “0”. It should be noted that this is only an example, and it is sufficient if a value that does not exceed the conceivable estimated convergence value is set for the "value for not terminating machining".
- FIG. 7 is a diagram for explaining an example of a method in which the stop determination unit 15 sets the variable abort threshold value based on the tried processing conditions and the evaluation values corresponding to the processing conditions in the first embodiment. It is a diagram.
- FIG. 7 shows that the stop determination unit 15 sets the variable abortion threshold according to the variable abortion threshold setting condition of ⁇ Condition (1)> above based on the trialed machining conditions and the evaluation values corresponding to the machining conditions.
- FIG. 10 is a diagram for explaining an example of a method for setting a variable termination threshold in a case.
- X in ⁇ Condition (1)> is set to "5".
- the horizontal axis in FIG. 7 indicates the number of trials of the machining conditions.
- the number of trials is the number of processing conditions that have been tried.
- the vertical axis in FIG. 7 indicates the evaluation value corresponding to each processing condition. It should be noted that when the machining condition is being tested, the evaluation value on the vertical axis in FIG. 7 is the estimated convergence value. Points indicated by black circles in FIG. 7 are evaluation values or estimated convergence values corresponding to each processing condition.
- the stop determination unit 15 sets the evaluation value corresponding to the processing conditions tried for the third time as the variable termination threshold.
- the stop determination unit 15 determines that the machining under the machining condition being trialed is determined to be discontinued.
- the ninth processing condition is being tested. That is, in this case, in FIG. 7, the evaluation value corresponding to the ninth trial is the estimated convergence value. In this case, according to FIG. 7, when the eight trials are completed, the fifth evaluation value among the evaluation values corresponding to the processing conditions that have been tried eight times is the fourth trial. It is an evaluation value corresponding to the processing conditions. Therefore, the stop determination unit 15 sets the evaluation value corresponding to the processing conditions tried for the fourth time as the variable termination threshold.
- the stop determination unit 15 determines that the machining under the machining condition being trialed is determined to be discontinued.
- the stop determination unit 15 can change the criteria used when determining whether or not to terminate machining under the machining conditions being trialed before the provisional evaluation value converges, in other words, the threshold for termination. can be done. For example, if the discontinuation threshold value is too high, the processing condition search device 1 may discontinue the processing condition for processing that should wait for the convergence of the processing result, and the deviation between the predicted evaluation value and the predicted value may increase. have a nature. As a result, the machining condition searching device 1 may not be able to search for the optimum machining conditions.
- the processing condition search device 1 requires time to determine that processing under the processing conditions corresponding to the evaluation value that is not high is discontinued before the provisional evaluation value converges.
- the machining condition search device 1 may take time to search for the optimum machining conditions.
- the machining condition searching device 1 by enabling the stop determination unit 15 to change the discontinuation threshold value, the machining condition searching device 1 maintains the possibility of searching for the optimum machining condition, and continues until the optimum machining condition can be searched. time can be shortened.
- a step of performing processing for setting a threshold is added.
- a hardware configuration for realizing the functions of the machining condition search device 1 is as follows.
- the functions of the search processing condition generation unit 11, the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, the evaluation determination unit 16, and the machine learning unit 17 in the processing condition search device 1 are as follows. It is implemented by a processing circuit. That is, the processing condition search device 1 has a processing circuit that executes the processing from step ST1 to step ST22 in FIG.
- the processing circuit may be dedicated hardware, or may be a CPU (Central Processing Unit) that executes a program stored in memory.
- CPU Central Processing Unit
- FIG. 8A is a block diagram showing the hardware configuration that implements the functions of the processing condition search device 1.
- FIG. 8B is a block diagram showing a hardware configuration for executing software that implements the functions of the processing condition search device 1.
- the input interface device 102 relays the processing result information output from the processing machine 2 to the processing condition searching device 1, and the storage units 18A to 18G output to the processing condition searching device 1. Relay information.
- the output interface device 103 relays information output from the processing condition search device 1 to the display unit 3 or information output from the processing condition search device 1 to each of the storage units 18A to 18G.
- the processing circuit 101 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an Application Specific Integrated Integrated Circuit (ASIC), Circuit), FPGA (Field-Programmable Gate Array), or a combination thereof.
- ASIC Application Specific Integrated Integrated Circuit
- FPGA Field-Programmable Gate Array
- the functions of the search processing condition generation unit 11, the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, the evaluation determination unit 16, and the machine learning unit 17 in the processing condition search device 1 are as follows. Separate processing circuits may be implemented, or these functions may be collectively implemented by one processing circuit.
- the processing circuit is the processor 104 shown in FIG. 4B
- the functions of the determination unit 16 and the machine learning unit 17 are implemented by software, firmware, or a combination of software and firmware.
- Software or firmware is written as a program and stored in the memory 105 .
- the processor 104 reads out and executes a program stored in the memory 105, thereby controlling the searched machining condition generation unit 11, the machining result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop
- the functions of the determination unit 15, the evaluation determination unit 16, and the machine learning unit 17 are realized.
- the processing condition search device 1 includes a memory 105 for storing a program that, when executed by the processor 104, results in the processing from step ST1 to step ST22 in the flowchart shown in FIG. .
- These programs are the processing procedures of the search machining condition generation unit 11, the machining result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, the evaluation determination unit 16, and the machine learning unit 17 or Run the method on a computer.
- the memory 105 causes the computer to function as the search processing condition generation unit 11, the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, the evaluation determination unit 16, and the machine learning unit 17. It may be a computer-readable storage medium storing a program for.
- the memory 105 includes, for example, non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically-EPROM), magnetic Discs, flexible discs, optical discs, compact discs, mini discs, DVDs, and the like are applicable.
- RAM Random Access Memory
- ROM Read Only Memory
- flash memory EPROM (Erasable Programmable Read Only Memory)
- EEPROM Electrically-EPROM
- magnetic Discs flexible discs, optical discs, compact discs, mini discs, DVDs, and the like are applicable.
- One of the functions of the search processing condition generation unit 11, the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, the evaluation determination unit 16, and the machine learning unit 17 in the processing condition search device 1 Parts may be implemented in dedicated hardware and parts in software or firmware.
- the search processing condition generation unit 11, the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, and the evaluation determination unit 16 are operated by the processing circuit 101, which is dedicated hardware.
- the function is realized, and the function of the machine learning unit 17 is realized by the processor 104 reading and executing the program stored in the memory 105 .
- the processing circuitry may implement the above functions through hardware, software, firmware, or a combination thereof.
- the processing condition search device 1 may be installed in the processing machine 2, or may be provided in a server connected to the processing machine 2 via a network.
- the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, the stop determination unit 15, the evaluation determination unit 16, and the machine learning unit 17, a part of the processing machine 2, and others may reside on servers.
- the machining condition searching apparatus 1 includes the machining condition calculation unit 111 that generates machining conditions composed of a plurality of control parameters that can be set for the machining machine 2, and the machining machine 2, An actual machining instruction unit 112 that causes the machining condition generated by the machining condition calculation unit 111 to perform machining, and a machining result that collects machining result information indicating the machining result of the machining that the actual machining command unit 112 causes the machine 2 to carry out.
- a collection unit 12 for calculating a provisional evaluation value for the processed processing based on the processing result information collected by the processing result collection unit 12; a convergence determination unit 14 for determining whether or not the provisional evaluation value has converged based on the evaluation value, and estimating an estimated convergence value to which the provisional evaluation value converges when it is determined that the provisional evaluation value has not converged; , if the convergence determination unit 14 determines that the provisional evaluation value has not converged, a stop determination unit 15 that determines whether or not to terminate machining under the machining conditions being trialed before the provisional evaluation value converges; When the determination unit 15 determines to terminate the machining under the machining conditions being trialed, it causes the actual machining command unit 112 to end the machining according to the machining conditions for the machining machine 2 and the estimated convergence value estimated by the convergence determination unit 14.
- the evaluation value of the machining performed according to the machining conditions is determined, and when the stop determination unit 15 determines not to terminate the machining under the machining conditions during trial, the convergence determination unit 14 determines that the provisional evaluation values have converged. After that, an evaluation determination unit 16 that determines the convergence value of the provisional evaluation values as the evaluation value, and determines whether or not to end the search for the machining conditions.
- the search end determination unit causes the processing condition calculation unit 111 to generate processing conditions to be tried next based on the predicted value predicted by the prediction unit 171 113, until the search end determination unit 113 determines to end the search, the processing condition calculation unit 111, the actual processing command unit 112, the processing result collection unit 12, the evaluation value acquisition unit 13, the convergence determination unit 14, and the stop Each process by the determination unit 15, the evaluation determination unit 16, the prediction unit 171, and the search end determination unit 113 is repeated.
- the machining condition search device 1 performs machining under the machining conditions for all the machining conditions to be tried until vibrational changes in the machining results of the machining machine 2 settle down. Compared to the conventional technology that has been implemented, it is possible to shorten the time until the optimum processing conditions can be searched.
- a processing condition search device can be used, for example, to search for processing conditions for a laser processing machine.
- 1 machining condition search device 1 machining condition search device, 2 machining machine, 3 display unit, 11 search machining condition generation unit, 111 machining condition calculation unit, 112 actual machining command unit, 113 search end determination unit, 12 machining result collection unit, 13 evaluation value acquisition unit , 14 convergence determination unit, 15 stop determination unit, 16 evaluation determination unit, 17 machine learning unit, 171 prediction unit, 172 uncertainty evaluation unit, 18A processing result storage unit, 18B evaluation value storage unit, 18C convergence result storage unit, 18D stop judgment storage unit, 18E search result storage unit, 18F prediction result storage unit, 18G uncertainty storage unit, 101 processing circuit, 102 input interface device, 103 output interface device, 104 processor, 105 memory.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Human Computer Interaction (AREA)
- Manufacturing & Machinery (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Mechanical Engineering (AREA)
- Numerical Control (AREA)
- Electrical Discharge Machining, Electrochemical Machining, And Combined Machining (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
Description
そこで、従来、想定される制御パラメータの組み合わせの加工条件のうちから生成した、試行するいくつかの加工条件で加工機に加工を実施させて得られた加工結果に基づいて加工条件に対応する評価値を算出し、算出した評価値と当該評価値に対応する加工条件とに基づき、ガウス過程回帰を用いて、試行されていない加工条件に対応する評価値を予測し、算出した評価値と予測した評価値とに基づいて、膨大な組み合わせの数の加工条件の中から最適な加工条件を探索する技術が知られている(例えば、特許文献1)。試行されていない加工条件に対応する評価値を予測するのにガウス過程回帰を用いる方法としては、例えば、加工条件に対する評価値が特定の分布に従う確率変数であると仮定して生成された確率モデルを用いる方法が挙げられる。
特許文献1に開示されている技術に代表される最適な加工条件の探索技術では、試行する全ての加工条件に対し、それぞれ、加工結果の振動的な変化が落ち着くまでのある程度の時間、加工機に加工を継続して実施させ、加工結果の振動的な変化が落ち着くのを待って、加工条件に対応する評価値を算出していた。
そのため、上述の探索技術では、試行した加工条件に対応する評価値を算出するのに時間を要し、その結果、最適な加工条件を探索するまでに時間を要するという課題があった。
図1は、実施の形態1に係る加工条件探索装置1の構成例を示す図である。
実施の形態1に係る加工条件探索装置1は、加工機2および表示部3と接続される。加工条件探索装置1は、加工機2に設定可能な多数の加工条件から最適な加工条件(以下「最適加工条件」という。)を探索する。最適加工条件は、例えば、加工の要求仕様を満足する加工結果が得られる加工条件である。また、表示部3は、加工作業者等のユーザからの要求に従って、加工条件探索装置1によって探索された加工条件等を表示する。例えば、表示部3は、加工機2に設定された加工条件と、この加工条件に従って加工機2が実施した加工の評価値を表示する。また、例えば、表示部3は、加工機2が実施していない加工条件と、この加工条件に従って加工機2が加工を実施したと想定した場合の当該加工の評価値の予測値を表示する。また、例えば、加工条件探索装置1による探索の探索結果である最適加工条件を表示する。なお、図1では、表示部3は、加工条件探索装置1および加工機2の外部に備えられているが、これは一例に過ぎない。表示部3は、例えば、加工条件探索装置1に備えられてもよいし、加工機2に備えられてもよい。
加工機2には、例えば、レーザ加工機、放電加工機、切削加工機、研削加工機、電解加工機、超音波加工機、電子ビーム加工機、または、付加加工機がある。以下の実施の形態1では、一例として、加工機2は、レーザ加工機であるものとする。なお、これは一例に過ぎず、実施の形態1において、加工機2は、レーザ加工機以外の加工機であってもよい。
実験用の加工では、実施の形態1に係る加工条件探索装置1は、試行用の加工条件を生成し、当該加工条件に従って、加工機2に実験用の加工を実施させる。加工機2は、上記加工条件に従って、ワークに、事前に設定した実験用の加工を施す。
ここで、加工条件は、加工機2の制御に用いられる複数の制御パラメータの組み合わせによって構成される。制御パラメータは、例えば、レーザ出力、切断速度、ビーム径、焦点位置、ガス圧である。加工条件に含まれる各制御パラメータは、調整可能である。例えば、レーザ加工機の加工において調整可能な制御パラメータが5つあり、各制御パラメータの値を10段階で選択できる場合、各制御パラメータの組み合わせによって構成される加工条件は、105=100000通りある。
例えば、加工機2は、加工中に発生した音、光、または、加工速度を検出するセンサを備えており、加工条件探索装置1は、当該センサから加工結果情報を収集する。例えば、センサは、加工後のワークを撮像した画像を取得する撮像装置、または、ワークの切断面の凹凸を計測する計測器であってもよい。また、センサは、加工機2とは別の場所に設けられていてもよい。加工条件探索装置1が、加工結果情報を収集可能になっていればよい。
そこで、実施の形態1に係る加工条件探索装置1は、加工結果の振動的な変化が落ち着くまでの過程にて算出された評価値について、振動的な変化が落ち着く前の評価値であっても、最適な加工条件を探索するにあたり影響がないと想定される評価値であれば、最適な加工条件の探索にこれを採用して試行中の加工条件に従った実験中の加工を打ち切り、探索のための加工条件を切り替える。これにより、実施の形態1に係る加工条件探索装置1は、最適な加工条件が探索できるまでの時間を短縮する。
加工条件探索装置1は、探索加工条件生成部11、加工結果収集部12、評価値取得部13、収束判定部14、停止判定部15、評価決定部16、および、機械学習部17を備える。また、加工条件探索装置1は、加工結果記憶部18A、評価値記憶部18B、収束結果記憶部18C、停止判定記憶部18D、探索結果記憶部18E、予測結果記憶部18F、および、不確実性記憶部18Gを備える。なお、記憶部18A~18Gの全てまたはその一部は、加工条件探索装置1とは別に設けられた外部装置が備えてもよい。
加工条件計算部111は、生成した加工条件を実加工指令部112に出力する。
また、実加工指令部112は、評価決定部16から、試行中の加工条件での加工を終了させる指示(以下「加工終了指示」という。)が出力された場合は、現在、加工機2に対して実施させている実験用の加工を終了させる。評価決定部16の詳細については、後述する。
探索終了判定部113は、加工条件の探索を追加で行う必要がないと判定した場合、評価決定部16が決定した評価値に基づいて、最適な加工条件を決定する。具体的には、探索終了判定部113は、探索結果記憶部18Eに記憶された評価値のうち、最も高い評価値に対応する加工条件を、最適加工条件とする。評価決定部16の詳細については、後述する。
また、探索終了判定部113は、加工条件の探索を追加で行う必要があると判定した場合、加工条件計算部111に、次に試行すべき、探索のための加工条件を生成させる。
加工結果収集部12は、実加工指令部112が加工を実施させる毎に加工結果を収集する。上述のとおり、実加工指令部112は、加工条件に従った加工を継続して実施させる。加工機2が加工を実施する間、複数ステップの加工が実施される。したがって、加工機2がある加工条件に従って実験用の加工を実施するとき、複数の加工結果情報が収集される。
加工結果収集部12は、収集した加工結果情報を、加工結果記憶部18Aに記憶させる。加工結果収集部12は、加工結果情報を、例えば、当該加工結果情報の取得時刻と対応付けて、加工結果記憶部18Aに記憶させる。
加工結果記憶部18Aは、加工結果情報を時系列で記憶する。
実施の形態1において、評価値は、加工の良否を示す値であり、その値が大きいほど良い加工であることを示す値として定義される。評価値は、例えば、0から1までの値で示される。この場合、最も良い加工が行われた場合に評価値は1となり、最も悪い加工が行われた場合の評価値は0となる。
評価値取得部13は、加工結果情報の取得時刻と、加工条件と、算出した暫定評価値とを対応付けた情報(以下「暫定評価値情報」という。)を、評価値記憶部18Bに記憶させる。なお、ここでは、暫定評価値情報において、加工結果情報の取得時刻が加工条件および暫定評価値と対応付けられるものとするが、これは一例に過ぎない。例えば、暫定評価値情報において、暫定評価値の算出時刻と、加工条件および暫定用価値とが対応付けられてもよい。
評価値記憶部18Bは、暫定評価値情報を時系列で記憶する。
収束判定部14は、暫定評価値が収束していると判定した場合、加工結果情報の取得時刻と、暫定評価値が収束している旨の情報と、加工条件と、暫定評価値と、暫定評価値の収束値とを対応付けた情報を、収束判定後情報として、収束結果記憶部18Cに記憶させる。加工結果情報の取得時刻に代えて、暫定評価値の算出時刻が対応付けられていてもよい。収束判定部14は、例えば、最新の暫定評価値を暫定評価値の収束値とする。なお、これは一例に過ぎず、例えば、時系列の暫定評価値に基づきどのように暫定評価値の収束値を算出するかを定義した情報(以下「収束値算出用情報」という。)が予め決められており、収束判定部14は、収束値算出用情報に従って、暫定評価値の収束値を算出してもよい。
一方、収束判定部14は、暫定評価値が収束していないと判定した場合、暫定評価値の収束先となる値(以下「推定収束値」という。)を推定する。そして、収束判定部14は、加工結果情報の取得時刻と、暫定評価値が収束していない旨の情報と、加工条件と、暫定評価値と、推定収束値とを対応付けた情報を、収束判定後情報として、収束結果記憶部18Cに記憶させる。加工結果情報の取得時刻に代えて、暫定評価値の算出時刻が対応付けられていてもよい。
停止判定記憶部18Dは打ち切り判定後情報を記憶する。
評価決定部16は、停止判定部15が試行中の加工条件での加工を打ち切ると判定したか否か、収束判定部14が推定した推定収束値、または、暫定評価値の収束値を、停止判定記憶部18Dに記憶されている打ち切り判定後情報から特定すればよい。例えば、評価決定部16は、停止判定部15から直接、打ち切り判定後情報を取得してもよい。なお、図1において、停止判定部15から評価決定部16への矢印は省略している。
探索結果記憶部18Eは、探索結果を記憶する。
予測部171は、評価決定部16により決定された評価値と、当該評価値に対応する加工条件とに基づいて、未試行の加工条件に対応する評価値を予測する。予測部171は、評価決定部16により決定された評価値と当該評価値に対応する加工条件を、探索結果記憶部18Eに記憶されている探索結果から取得すればよい。
予測部171は、予測により得られる評価値の予測値を加工条件と対応付けた情報(以下「予測結果情報」という。)を予測結果記憶部18Fに記憶させる。予測結果情報は、未試行の加工条件と、これに対応する評価値の予測値とが対応付けられた情報である。
予測結果記憶部18Fは、予測結果情報を記憶する。
不確実性記憶部18Gは、不確実性情報を記憶する。
図2は、実施の形態1に係る加工条件探索装置1の動作について説明するためのフローチャートである。
加工条件探索処理が開始されると、まず、探索加工条件生成部11の加工条件計算部111が、初期加工条件を生成する(ステップST1)。加工条件計算部111は、加工条件として設定可能な全ての組み合わせの中から、予め定められた数の加工条件を初期加工条件として選択することにより、初期加工条件を生成する。加工条件計算部111による初期加工条件の選択方法としては、例えば、実験計画法、最適計画法、最適計画法、または、ランダムサンプリングが挙げられる。また、ユーザが過去の利用実績などから最適だと思われる加工条件の見当がついている場合は、加工条件計算部111は、ユーザから入力された加工条件を初期加工条件として用いてもよい。なお、これらの方法は一例に過ぎず、加工条件計算部111は、どのような方法を用いて初期加工条件を生成してもよい。
加工結果収集部12は、収集した加工結果情報を、加工結果記憶部18Aに記憶させる。
評価値取得部13は、加工結果情報の取得時刻と、加工条件、ここでは初期加工条件と、算出した暫定評価値とを対応付けた暫定評価値情報を、評価値記憶部18Bに記憶させる。
具体例を挙げると、例えば、収束判定部14は、時系列の暫定評価値から、暫定評価値の四分位範囲を求める。そして、収束判定部14は、暫定評価値の四分位範囲がどれぐらいの値の範囲となっているかに基づいて、暫定評価値が収束しているか否かを判定する。例えば、予め、暫定評価値が収束していると判定する場合の値の範囲(以下「第1収束判定用範囲」という。)が決められている。収束判定部14は、暫定評価値の四分位範囲が第1収束判定用範囲内におさまっていれば、暫定評価値が収束していると判定する。収束判定部14は、暫定評価値の四分位範囲が第1収束判定用範囲内におさまっていなければ、暫定評価値が収束していないと判定する。
収束判定部14は、暫定評価値が収束していないと判定すると、次に、時系列の暫定評価値から求めた暫定評価値の四分位範囲から、推定収束値を推定する。例えば、収束判定部14は、暫定評価値の四分位範囲の中央値を、推定収束値と推定する。
収束判定部14は、暫定評価値が収束していないと判定すると、次に、時系列の暫定評価値から推定した分布から、推定収束値を推定する。例えば、収束判定部14は、暫定評価値の平均値を、推定収束値と推定する。
また、例えば、第1機械学習モデルは、推定収束値に加え、暫定評価値のばらつき度合いに関する情報を出力するモデルであってもよい。収束判定部14は、時系列の暫定評価値を第1機械学習モデルに入力して得た暫定評価値のばらつき度合いに関する情報に基づき、暫定評価値が収束しているか否かを判定してもよい。
打ち切り用閾値は、例えば、予め、ユーザによって指定され、停止判定部15に記憶されている。例えば、ユーザは、予め、その値を超えない場合は試行中の加工条件での加工を打ち切るものとする、打ち切りの基準となる評価値(以下「基準評価値」という。)を打ち切り用閾値に指定しておく。ユーザは、例えば、加工機2に求める要求性能に応じて基準評価値を設定する。
図3の横軸はある加工条件に従った加工を実施している時間幅を示し、図3の縦軸は評価値(暫定評価値)を示す。図3の黒丸で示した点は、加工条件に従って実施された加工の加工結果に基づいて算出された暫定評価値を示す。なお、図3では、わかりやすさのため、暫定評価値が収束していく様子を図示するようにしている。図3において、201a、201b、および、201cは、暫定評価値の四分位範囲を示している。
t1時間経過時点では暫定評価値の四分位範囲は201aで示す範囲であり、t2時間経過時点では暫定評価値の四分位範囲は201bで示す範囲である。201aおよび201bに示す四分位範囲について、四分位範囲内の暫定評価値のうち最も大きい暫定評価値は、打ち切り用閾値以上となっている。よって、この場合、停止判定部15は、試行中の加工条件での加工を継続させると判定する。
t3時間が経過すると、暫定評価値の四分位範囲は201cで示す範囲となり、四分位範囲内の暫定評価値のうち最も大きい暫定評価値は打ち切り用閾値未満となる。この場合、停止判定部15は、試行中の加工条件での加工を打ち切ると判定する。
図4の横軸はある加工条件に従った加工を実施している時間幅を示し、図4の縦軸は評価値(暫定評価値)を示す。図4の黒丸で示した点は、加工条件に従って実施された加工の加工結果に基づいて算出された暫定評価値を示す。なお、図4では、わかりやすさのため、暫定評価値が収束していく様子を図示するようにしている。図4において、301a、301b、および、301cは、暫定評価値の平均値±κσの区間に含まれる暫定評価値のうち最も大きい暫定評価値を示している。
t4時間経過時点では暫定評価値の平均値±κσの区間に含まれる暫定評価値のうち最も大きい暫定評価値は301aで示す値であり、t5時間時点で暫定評価値の平均値±κσの区間に含まれる暫定評価値のうち最も大きい暫定評価値は301bで示す値である。301aで示す値および301bで示す値はいずれも打ち切り用閾値以上である。すなわち、301aで示す値を含む暫定評価値の平均値±κσの区間に含まれる全ての暫定評価値が打ち切り用閾値未満となっていない。また、301bで示す値を含む暫定評価値の平均値±κσの区間に含まれる全ての暫定評価値が打ち切り用閾値未満となっていない。よって、この場合、停止判定部15は、試行中の加工条件での加工を継続させると判定する。
t6時間が経過すると、暫定評価値の平均値±κσの区間に含まれる暫定評価値のうち最も大きい暫定評価値は301cで示す値となる。301cで示す値は、打ち切り用閾値未満である。すなわち、301cで示す値を含む暫定評価値の平均値±κσの区間内の全ての暫定評価値が打ち切り用閾値未満となる。この場合、停止判定部15は、試行中の加工条件での加工を打ち切ると判定する。
評価決定部16は、停止判定部15が試行中の初期加工条件での加工を打ち切らないと判定した場合(ステップST6の“NO”の場合)は、収束判定部14が暫定評価値は収束したと判定したか否かを判定する(ステップST7)。収束判定部14が暫定評価値は収束していないと判定した場合(ステップST7の“NO”の場合)、加工条件探索装置1の動作は、ステップST2の処理に戻る。収束判定部14が暫定評価値は収束したと判定すると(ステップST7の“YES”の場合)、評価決定部16は、暫定評価値の収束値を評価値に決定する。そして、評価決定部16は、加工条件と評価値の組み合わせを探索結果として探索結果記憶部18Eに記憶させる(ステップST8)。詳細には、評価決定部16は、初期加工条件と、評価値、ここでは暫定評価値の収束値、の組み合わせを探索結果として探索結果記憶部18Eに記憶させる。
初期加工が終了していない初期加工条件がある場合(ステップST9の“NO”の場合)、初期加工が終了していない初期加工条件について、ステップST1からステップST8までの処理が順に実施される。2回目以降のステップST1では、加工条件計算部111は、これまでのステップST1にて選択していない初期加工条件を選択する。これにより、探索結果記憶部18Eには、全ての初期加工条件(例えば、10通りの初期加工条件)と評価値の組み合わせとが対応付けられた探索結果が記憶される。
m(xN+1)=kT・(CN -1)・t ・・・(1)
σ2(xN+1)=c-kT・(CN -1)・k ・・・(3)
図5は、実施の形態1における、評価値の予測値と、不確実性を示す指標との関係を概念的に示すグラフである。
図5には、ガウス過程回帰を用いて予測値と不確実性を示す指標とが算出される例が示されている。図5の横軸は加工条件である制御パラメータの値xを示し、図5の縦軸は評価値を示す。図5の黒丸で示した点は、初期加工条件を用いた実加工に基づいて算出された評価値(以下、実加工の評価値ともいう)を示す。ガウス過程回帰を用いた予測では、評価値がガウス分布に従うとして評価値を予測する。このため、評価値の予測値をガウス分布の平均m(x)とし、予測の不確実性を示す指標をガウス分布の標準偏差σ(x)とすると、実際の評価値は、約95%の確率で、m(x)-2σ(x)以上、かつ、m(x)+2σ(x)以下の範囲に入ることが統計的に示される。図5において、実線で示された曲線は、評価値の予測値であるm(x)を示す。また、図5において、破線で示された曲線は、m(x)-2σ(x)の曲線、および、m(x)+2σ(x)の曲線を示す。
図5に示すように、実加工の評価値に近い箇所では不確実性を示す指標は小さくなり、実加工の評価値から離れた箇所では不確実性を示す指標は大きくなる。
不確実性評価部172は、予測値の不確実性を示す指標を記憶する(ステップST13)。詳細には、不確実性評価部172は、算出した指標の値を加工条件と対応付けた不確実性情報を、不確実性記憶部18Gに記憶させる。
これに対し、実施の形態1に係る加工条件探索装置1は、上述のとおり、評価値を算出するにあたり、このまま加工を継続させても高い評価値が得られないと判定した場合、評価値(暫定評価値)が収束する前に試行中の加工条件での加工を打ち切り、推定収束値を、試行中の加工条件に対応する評価値とする。これにより、加工条件探索装置1は、高い評価値が得られないと判定した、ある加工条件に従った加工について、その加工の加工結果が収束するまでの時間のうち、打ち切った時点から加工結果が収束するまでの間の時間を省略することができる。すなわち、加工条件探索装置1は、上記省略した時間の分だけ、最適加工条件を探索するまでに要するトータルの時間を短縮することができる。
図6Aは、従来の最適加工条件の探索技術において最適加工条件が探索されるまでの評価値を示すグラフであり、図6Bは、実施の形態1に係る加工条件探索装置1によって最適加工条件が探索されるまでの評価値を示すグラフである。
図6Aおよび図6Bにおいて、黒丸で示す点は、加工結果が収束するまで実施した実加工の加工結果に基づいて算出された評価値を示している。図6Bにおいて、白丸で示す点は、加工結果が収束する前に打ち切った実加工の加工結果に基づいて算出された推定収束値を示している。
なお、図6Aおよび図6Bは、同じ加工機2に対して、同じ所望の加工結果を得られるような最適加工条件を探索した結果としている。
これに対し、実施の形態1に係る加工条件探索装置1では、図6Bに示すように、加工結果、言い換えれば、評価値が低いと予想される場合は加工を打ち切るため、短時間で最適加工条件を探索することができる。図6Bに示す例では、14分で最適加工条件が探索されることになる。実施の形態1に係る加工条件探索装置1で最適加工条件が探索されるまでに要した時間は、図6Aに示す従来の最適加工条件の探索技術で最適加工条件が探索されるまでに要した時間よりも7分短縮されている。
例えば、停止判定部15は、試行済みの加工条件および当該加工条件に対応する評価値に基づいて打ち切り用閾値を設定することもできる。試行済みの加工条件および当該加工条件に対応する評価値は、探索結果として、評価決定部16によって探索結果記憶部18Eに記憶されている。停止判定部15が決定済みの評価値に基づいて設定する打ち切り用閾値を、「可変打ち切り用閾値」ともいう。なお、この場合、停止判定部15は、可変打ち切り用閾値を設定すると、例えば、収束判定部14が推定した推定収束値と、可変打ち切り用閾値との比較によって、暫定評価値が収束する前に試行中の加工条件での加工を打ち切るか否かを判定する。収束判定部14が推定した推定収束値は、収束結果記憶部18Cに記憶された最新の収束判定後情報における推定収束値である。
詳細には、停止判定部15は、例えば、試行済みの加工条件および当該加工条件に対応する評価値に基づき、予め設定されている条件(以下「可変打ち切り用閾値設定条件」という。)に従って、可変打ち切り用閾値を設定する。
<条件(1)>
試行回数がX回未満の場合は加工を打ち切らないための値を可変打ち切り用閾値とし、試行回数がX回以上である場合、試行済みの全ての加工条件に対応する各評価値のうち第X位の評価値を可変打ち切り用閾値とする
<条件(2)>
試行済みの全ての加工条件に対応する各評価値のうち、上位Y位の評価値を可変打ち切り用閾値とする
<条件(3)>
試行済みの全ての加工条件に対応する各評価値のうちの上位Z%の評価値のうち、最下位の評価値を可変打ち切り用閾値とする
また、<条件(1)>において、「加工を打ち切らないための値」は、例えば、「0」とする。なお、これは一例に過ぎず、「加工を打ち切らないための値」には、想定され得る推定収束値を超えない値が設定されるようになっていればよい。
図7は、停止判定部15が、試行済みの加工条件および当該加工条件に対応する評価値に基づき、上述の<条件(1)>の可変打ち切り用閾値設定条件に従って可変打ち切り用閾値を設定した場合の、可変打ち切り用閾値の設定方法の一例を説明するための図としている。図7では、一例として、<条件(1)>におけるXを「5」としている。
図7の横軸は加工条件の試行回数を示す。試行回数とは、すなわち、試行済みの加工条件の数である。図7の縦軸は各加工条件に対応する評価値を示す。なお、加工条件が試行中であるとき、図7の縦軸の評価値は推定収束値である。図7において黒丸で示す点が、各加工条件に対応する評価値、または、推定収束値である。
この場合、図7によると、5回の試行を終えた時点で、当該5回試行した試行済みの加工条件に対応する各評価値のうち、第5位の評価値は、3回目に試行された加工条件に対応する評価値である。そこで、停止判定部15は、3回目に試行された加工条件に対応する評価値を、可変打ち切り用閾値に設定する。なお、試行中の加工条件、言い換えれば、6回目に試行されている加工条件に対する推定収束値は、この可変打ち切り用閾値未満であるので、停止判定部15は、試行中の加工条件での加工を打ち切ると判定することになる。
この場合、図7によると、8回の試行を終えた時点で、当該8回試行した試行済みの加工条件に対応する各評価値のうち、第5位の評価値は、4回目に試行された加工条件に対応する評価値である。そこで、停止判定部15は、4回目に試行された加工条件に対応する評価値を、可変打ち切り用閾値に設定する。なお、試行中の加工条件、言い換えれば、9回目に試行されている加工条件に対する推定収束値は、この可変打ち切り用閾値未満であるので、停止判定部15は、試行中の加工条件での加工を打ち切ると判定することになる。
例えば、打ち切り用閾値が高すぎた場合、加工条件探索装置1は、加工結果の収束を待つべき加工の加工条件までも途中で打ち切ってしまい、予測した評価値の予測値のズレが大きくなる可能性がある。その結果、加工条件探索装置1は、最適加工条件を探索できなくなる可能性がある。逆に、例えば、打ち切り用閾値が低すぎた場合、加工条件探索装置1は、高くない評価値に対応する加工条件での加工を暫定評価値の収束前に打ち切ると判定するまでに時間を要してしまう、または、当該暫定評価値が収束するまで加工を打ち切ることなく待ってしまうことになる可能性がある。その結果、加工条件探索装置1は、最適加工条件が探索できるまでに時間を要してしまう可能性がある。
加工条件探索装置1において、停止判定部15が打ち切り用閾値を変更可能とすることで、加工条件探索装置1は、最適加工条件を探索できる可能性を保ちつつ、当該最適加工条件が探索できるまでの時間を短縮できる。
なお、この場合、図2のフローチャートを用いて説明した加工条件探索装置1の動作について、ステップST5とステップST6の間、および、ステップST19とステップST20の間に、停止判定部15が可変打ち切り用閾値を設定する処理を行うステップが追加になる。
加工条件探索装置1における探索加工条件生成部11、加工結果収集部12、評価値取得部13、収束判定部14、停止判定部15、評価決定部16、および、機械学習部17の機能は、処理回路によって実現される。すなわち、加工条件探索装置1は、図2のステップST1からステップST22までの処理を実行する処理回路を備えている。処理回路は、専用のハードウェアであってもよいし、メモリに記憶されたプログラムを実行するCPU(Central Processing Unit)であってもよい。
Claims (11)
- 加工機に設定可能な複数の制御パラメータで構成される加工条件を生成する加工条件計算部と、
前記加工機に、前記加工条件計算部により生成された前記加工条件に従って加工を実施させる実加工指令部と、
前記実加工指令部が前記加工機に実施させた前記加工の加工結果を示す加工結果情報を収集する加工結果収集部と、
前記加工結果収集部が収集した前記加工結果情報に基づいて、実施された前記加工に対する暫定評価値を算出する評価値取得部と、
前記評価値取得部が算出した時系列の前記暫定評価値に基づき前記暫定評価値は収束しているか否かを判定し、前記暫定評価値は収束していないと判定した場合、前記暫定評価値の収束先となる推定収束値を推定する収束判定部と、
前記収束判定部が前記暫定評価値は収束していないと判定した場合、前記暫定評価値が収束する前に試行中の前記加工条件での前記加工を打ち切るか否かを判定する停止判定部と、
前記停止判定部が試行中の前記加工条件での前記加工を打ち切ると判定した場合、前記実加工指令部に前記加工機に対する前記加工条件に従った前記加工を終了させるとともに前記収束判定部が推定した前記推定収束値を前記加工条件に従って実施された前記加工の評価値に決定し、前記停止判定部が試行中の前記加工条件での前記加工を打ち切らないと判定した場合は、前記収束判定部が前記暫定評価値は収束したと判定した後、前記暫定評価値の収束値を前記評価値に決定する評価決定部と、
前記評価決定部が決定した前記評価値と、前記評価値に対応する前記加工条件に基づいて、未試行の前記加工条件に対応する前記評価値の予測値を予測する予測部と、
前記加工条件の探索を終了するか否かを判定し、前記探索を終了する場合は、前記評価決定部が決定した前記評価値および前記予測部が予測した前記評価値に基づいて最適な前記加工条件を決定し、前記探索を終了しない場合は、前記加工条件計算部に、前記予測部が予測した前記予測値に基づいて次に試行すべき前記加工条件を生成させる探索終了判定部とを備え、
前記探索終了判定部により前記探索を終了すると判定されるまで、前記加工条件計算部、前記実加工指令部、前記加工結果収集部、前記評価値取得部、前記収束判定部、前記停止判定部、前記評価決定部、前記予測部、前記探索終了判定部による各処理を繰り返し行う加工条件探索装置。 - 前記収束判定部は、前記評価値取得部が算出した時系列の前記暫定評価値のばらつき度合いに基づき、前記推定収束値を推定する
ことを特徴とする請求項1記載の加工条件探索装置。 - 前記収束判定部は、前記評価値取得部が算出した時系列の前記暫定評価値と、時系列の前記評価値を入力とし前記推定収束値を出力する第1機械学習モデルとに基づき、前記推定収束値を推定する
ことを特徴とする請求項1記載の加工条件探索装置。 - 前記停止判定部は、前記評価値取得部が算出した時系列の前記暫定評価値のばらつき度合いと打ち切り用閾値との比較によって、前記暫定評価値が収束する前に試行中の前記加工条件での前記加工を打ち切るか否かを判定する
ことを特徴とする請求項1記載の加工条件探索装置。 - 前記停止判定部は、試行済みの前記加工条件および当該加工条件に対応する前記評価値に基づいて可変打ち切り用閾値を設定し、前記収束判定部が推定した前記推定収束値と設定した前記可変打ち切り用閾値との比較によって、前記暫定評価値が収束する前に試行中の前記加工条件での前記加工を打ち切るか否かを判定する
ことを特徴とする請求項4記載の加工条件探索装置。 - 前記停止判定部は、前記評価値取得部が算出した時系列の前記暫定評価値と、時系列の前記評価値を入力とし加工を停止させるか否かを示す情報を出力する第2機械学習モデルとに基づき、前記暫定評価値が収束する前に、試行中の前記加工条件での前記加工を打ち切るか否かを判定する
ことを特徴とする請求項1記載の加工条件探索装置。 - 前記予測部による予測の不確実性を示す指標を算出する不確実性評価部を備え、
前記加工条件計算部は、前記予測部が予測した前記評価値の前記予測値と予測の不確実性を示す前記指標とに基づいて、次に試行すべき前記加工条件を生成する
ことを特徴とする請求項1記載の加工条件探索装置。 - 前記探索終了判定部は、前記評価値の前記予測値および前記評価値の不確実性を示す前記指標を用いて前記加工条件の前記探索を終了するか否かを判定し、前記加工条件の前記探索を終了すると判定した場合、前記評価決定部が決定した前記評価値のうち最高となる前記評価値に対応する前記加工条件を、前記最適な前記加工条件とする
ことを特徴とする請求項7記載の加工条件探索装置。 - 前記予測部は、前記加工条件に対する前記評価値が特定の分布に従う確率変数であると仮定して生成された、評価値の前記加工条件に対する確率モデルを用いて前記予測値を予測し、
前記不確実性評価部は、前記確率モデルを用いて前記予測の不確実性を示す前記指標を算出する
ことを特徴とする請求項7記載の加工条件探索装置。 - 前記加工条件および当該加工条件に対応する前記評価値、前記加工条件および当該加工条件に対応する前記評価値の前記予測値、または、探索結果の前記加工条件のうち少なくとも一つを表示する表示部を備えた
ことを特徴とする請求項1記載の加工条件探索装置。 - 加工条件計算部が、加工機に設定可能な複数の制御パラメータで構成される加工条件を生成するステップと、
実加工指令部が、前記加工機に、前記加工条件計算部により生成された前記加工条件に従って加工を実施させるステップと、
加工結果収集部が、前記実加工指令部が前記加工機に実施させた前記加工の加工結果を示す加工結果情報を収集するステップと、
評価値取得部が、前記加工結果収集部が収集した前記加工結果情報に基づいて、実施された前記加工に対する暫定評価値を算出するステップと、
収束判定部が、前記評価値取得部が算出した時系列の前記暫定評価値に基づき前記暫定評価値は収束しているか否かを判定し、前記暫定評価値は収束していないと判定した場合、前記暫定評価値の収束先となる推定収束値を推定するステップと、
停止判定部が、前記収束判定部が前記暫定評価値は収束していないと判定した場合、前記暫定評価値が収束する前に試行中の前記加工条件での前記加工を打ち切るか否かを判定するステップと、
評価決定部が、前記停止判定部が試行中の前記加工条件での前記加工を打ち切ると判定した場合、前記実加工指令部に前記加工機に対する前記加工条件に従った前記加工を終了させるとともに前記収束判定部が推定した前記推定収束値を前記加工条件に従って実施された前記加工の評価値に決定し、前記停止判定部が試行中の前記加工条件での前記加工を打ち切らないと判定した場合は、前記収束判定部が前記暫定評価値は収束したと判定した後、前記暫定評価値の収束値を前記評価値に決定するステップと、
予測部が、前記評価決定部が決定した前記評価値と、前記評価値に対応する前記加工条件に基づいて、未試行の前記加工条件に対応する前記評価値の予測値を予測するステップと、
探索終了判定部が、前記加工条件の探索を終了するか否かを判定し、前記探索を終了する場合は、前記評価決定部が決定した前記評価値に基づいて最適な前記加工条件を決定し、前記探索を終了しない場合は、前記加工条件計算部に、前記予測部が予測した前記予測値に基づいて次に試行すべき前記加工条件を生成させるステップとを備え、
前記探索終了判定部により前記探索を終了すると判定されるまで、前記加工条件計算部、前記実加工指令部、前記加工結果収集部、前記評価値取得部、前記収束判定部、前記停止判定部、前記評価決定部、前記予測部、前記探索終了判定部による各処理を繰り返し行う加工条件探索方法。
Priority Applications (7)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE112021007612.4T DE112021007612T5 (de) | 2021-07-08 | 2021-07-08 | Bearbeitungsbedingungs-sucheinrichtung und bearbeitungsbedingungs-suchverfahren |
CN202180099531.9A CN117500635A (zh) | 2021-07-08 | 2021-07-08 | 加工条件探索装置及加工条件探索方法 |
KR1020237044767A KR102677644B1 (ko) | 2021-07-08 | 2021-07-08 | 가공 조건 탐색 장치 및 가공 조건 탐색 방법 |
PCT/JP2021/025722 WO2023281681A1 (ja) | 2021-07-08 | 2021-07-08 | 加工条件探索装置および加工条件探索方法 |
JP2021569281A JP7109689B1 (ja) | 2021-07-08 | 2021-07-08 | 加工条件探索装置および加工条件探索方法 |
TW110143466A TWI854169B (zh) | 2021-07-08 | 2021-11-23 | 加工條件探索裝置及加工條件探索方法 |
US18/522,354 US20240094688A1 (en) | 2021-07-08 | 2023-11-29 | Machining condition search device and machining condition search method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2021/025722 WO2023281681A1 (ja) | 2021-07-08 | 2021-07-08 | 加工条件探索装置および加工条件探索方法 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US18/522,354 Continuation US20240094688A1 (en) | 2021-07-08 | 2023-11-29 | Machining condition search device and machining condition search method |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023281681A1 true WO2023281681A1 (ja) | 2023-01-12 |
Family
ID=82652267
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2021/025722 WO2023281681A1 (ja) | 2021-07-08 | 2021-07-08 | 加工条件探索装置および加工条件探索方法 |
Country Status (6)
Country | Link |
---|---|
US (1) | US20240094688A1 (ja) |
JP (1) | JP7109689B1 (ja) |
KR (1) | KR102677644B1 (ja) |
CN (1) | CN117500635A (ja) |
DE (1) | DE112021007612T5 (ja) |
WO (1) | WO2023281681A1 (ja) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003162706A (ja) * | 2001-11-27 | 2003-06-06 | Matsushita Electric Works Ltd | 遺伝的アルゴリズムを用いた最適化装置およびその方法 |
JP2008036812A (ja) * | 2006-07-11 | 2008-02-21 | Mitsubishi Electric Corp | 加工条件探索装置 |
JP2010042499A (ja) * | 2008-07-18 | 2010-02-25 | Mitsubishi Electric Corp | 加工条件探索装置 |
US20160187874A1 (en) * | 2013-07-22 | 2016-06-30 | Texas State University | Autonomous performance optimization in robotic assembly process |
JP2019159864A (ja) * | 2018-03-14 | 2019-09-19 | 株式会社日立ハイテクノロジーズ | 探索装置、探索方法及びプラズマ処理装置 |
WO2020261572A1 (ja) * | 2019-06-28 | 2020-12-30 | 三菱電機株式会社 | 加工条件探索装置および加工条件探索方法 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6599956B2 (ja) * | 2017-10-23 | 2019-10-30 | ファナック株式会社 | 工作機械の加工条件選定装置 |
-
2021
- 2021-07-08 DE DE112021007612.4T patent/DE112021007612T5/de active Granted
- 2021-07-08 JP JP2021569281A patent/JP7109689B1/ja active Active
- 2021-07-08 KR KR1020237044767A patent/KR102677644B1/ko active IP Right Grant
- 2021-07-08 CN CN202180099531.9A patent/CN117500635A/zh active Pending
- 2021-07-08 WO PCT/JP2021/025722 patent/WO2023281681A1/ja active Application Filing
-
2023
- 2023-11-29 US US18/522,354 patent/US20240094688A1/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003162706A (ja) * | 2001-11-27 | 2003-06-06 | Matsushita Electric Works Ltd | 遺伝的アルゴリズムを用いた最適化装置およびその方法 |
JP2008036812A (ja) * | 2006-07-11 | 2008-02-21 | Mitsubishi Electric Corp | 加工条件探索装置 |
JP2010042499A (ja) * | 2008-07-18 | 2010-02-25 | Mitsubishi Electric Corp | 加工条件探索装置 |
US20160187874A1 (en) * | 2013-07-22 | 2016-06-30 | Texas State University | Autonomous performance optimization in robotic assembly process |
JP2019159864A (ja) * | 2018-03-14 | 2019-09-19 | 株式会社日立ハイテクノロジーズ | 探索装置、探索方法及びプラズマ処理装置 |
WO2020261572A1 (ja) * | 2019-06-28 | 2020-12-30 | 三菱電機株式会社 | 加工条件探索装置および加工条件探索方法 |
Also Published As
Publication number | Publication date |
---|---|
KR20240005995A (ko) | 2024-01-12 |
JP7109689B1 (ja) | 2022-07-29 |
DE112021007612T5 (de) | 2024-03-14 |
US20240094688A1 (en) | 2024-03-21 |
KR102677644B1 (ko) | 2024-06-24 |
JPWO2023281681A1 (ja) | 2023-01-12 |
CN117500635A (zh) | 2024-02-02 |
TW202302269A (zh) | 2023-01-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6063013B1 (ja) | びびり或いは工具摩耗/破損の発生を抑制する加工条件調整機能を有する数値制御装置 | |
US20180018408A1 (en) | Material generation apparatus and material generation method | |
JP6880335B1 (ja) | 加工条件探索装置および加工条件探索方法 | |
CN114025912B (zh) | 加工条件搜索装置以及加工条件搜索方法 | |
KR101544457B1 (ko) | 최적 설계 파라미터 탐색을 위한 최적화 방법 | |
JP7124768B2 (ja) | パラメータ推定装置、方法、及びプログラム | |
JP2014102578A (ja) | 情報処理装置、制御方法、プログラム、及び記録媒体 | |
US11619926B2 (en) | Information processing device, program, process treatment executing device, and information processing system | |
US11550274B2 (en) | Information processing apparatus and information processing method | |
WO2023281681A1 (ja) | 加工条件探索装置および加工条件探索方法 | |
CN111160459A (zh) | 超参数的优化装置和方法 | |
TWI854169B (zh) | 加工條件探索裝置及加工條件探索方法 | |
CN110893515B (zh) | 加工条件调整装置以及机器学习装置 | |
JP2021074816A (ja) | 工具条件決定装置、学習装置、および方法 | |
JP7538258B2 (ja) | 機械学習装置 | |
JP2022108560A (ja) | 解析装置、パラメータ最適化方法及びパラメータ最適化プログラム | |
JP2016024713A (ja) | パラメータ選定方法、パラメータ選定プログラム及びパラメータ選定装置 | |
JP2011145905A (ja) | 予測関数生成装置、方法、及び、プログラム | |
WO2022168864A1 (ja) | 情報処理装置、シミュレータシステム、ニューラルネットワークシステム、引数値決定方法および記録媒体 | |
CN117950323B (zh) | 基于机械手臂加工控制的自适应调节方法及系统 | |
JP7415100B1 (ja) | パラメータ調整装置およびパラメータ調整方法 | |
CN117532625B (zh) | 焊接机器人的路径优化方法及系统 | |
US20240045387A1 (en) | Transfer learning device and transfer learning method | |
JP2024014330A (ja) | 影響因子推定装置、影響因子推定方法、およびコンピュータプログラム | |
JP6550899B2 (ja) | モデル推定システム、モデル推定方法およびモデル推定プログラム |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2021569281 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: 21949314 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202180099531.9 Country of ref document: CN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 112021007612 Country of ref document: DE |
|
ENP | Entry into the national phase |
Ref document number: 20237044767 Country of ref document: KR Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1020237044767 Country of ref document: KR |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202427000448 Country of ref document: IN |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 21949314 Country of ref document: EP Kind code of ref document: A1 |