Laser beam machine

Abstract

To automatically select an optimal cutting condition by using a learning model having learned learning selection capacities of the optimal cutting conditions.SOLUTION: A laser beam machine comprises: a cutting condition selection and training device including a selection unit for selecting a cutting condition set when lase-cutting a workpiece, and a training unit for training a learning model for selecting the cutting condition; and a determination device for determining a machining result of machining that is based on the cutting condition selected by the selection and training device. The selection and training device has a first learned model having learned first data including material quality, size of a machined shape and shape complication of the machined shape of the workpiece, second data including a cutting condition selected according to the first data and machined results of the machining based on these data.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a laser processing machine. [Background technology]

[0002] Conventionally, for processing devices such as laser processing machines that perform laser processing such as cutting and welding on workpieces, various proposals have been made regarding using machine learning to search for, determine, and confirm optimal processing conditions to be set in the processing device in accordance with the workpiece (see, for example, Patent Documents 1 and 2).

[0003] Generally, when performing laser cutting processing with a laser processing machine, cutting conditions suitable for the processing shape to be cut are selected from a plurality of preset cutting conditions according to the processing conditions selected for the workpiece. If the cutting conditions are selected inappropriately, it may cause processing defects even if the processing conditions are appropriate, so it is necessary to select the cutting conditions appropriately. As an example of the selection of cutting conditions, the cutting conditions are automatically selected using CAM (Computer Aided Manufacturing). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 6625914 [Patent Document 2] Patent No. 7126616 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the selection of cutting conditions using the above CAM is based only on the size of the shape to be cut by laser cutting, so cutting conditions that cause processing defects may be selected in rare cases. If such inappropriate cutting conditions are selected, an experienced worker will check the cutting conditions each time before starting processing and correct them to appropriate cutting conditions before starting laser cutting.

[0006] This causes an increase in the overall man-hours at the processing site, and for example, an inexperienced worker may start laser cutting without noticing that the selected inappropriate cutting conditions have been selected, resulting in processing defects and wasteful disposal of the processed material.The processing systems disclosed in the above Patent Documents 1 and 2 are designed to automatically select appropriate processing conditions, but do not take into consideration at all the selection of appropriate cutting conditions that match the processing conditions of the laser cutting process, and therefore may cause the same problems as those described above.

[0007] One aspect of the present invention is a laser processing machine that can automatically select optimal cutting conditions using a learning model that has learned the ability to select appropriate cutting conditions. [Means for solving the problem]

[0008] A laser processing machine according to one embodiment of the present invention comprises a cutting condition selection and learning device including a selection unit which selects cutting conditions to be set in a laser processing machine which laser cuts a workpiece, and a learning unit which learns a learning model used for the selection of the cutting conditions by the selection unit, and a judgment device which judges a processing result obtained by laser cutting the workpiece by the laser processing machine based on the cutting conditions selected by the cutting condition selection and learning device, and the cutting condition selection and learning device has a first learned model created by inputting, as the learning model in the learning unit, first data including the material, thickness, size of the processed shape, and shape complexity of the processed shape of the workpiece, second data including cutting conditions selected according to the first data, and processing results obtained by laser cutting based on the first data and the second data, and learning based on these.

[0009] According to a laser processing machine according to an aspect of the present invention, a cutting condition to be set in the laser processing machine is selected by a selection unit using a learning model that has been learned and re-learned by a learning unit of a cutting condition selection and learning device. The learning model has a first trained model. The first trained model is created by inputting first data including the material, plate thickness, size of the processed shape, and shape complexity of the processed shape of the workpiece, second data including cutting conditions selected according to the first data, and a processing result obtained by laser cutting processing based on the first data and the second data, and learning based on these. This makes it possible to automatically select optimal cutting conditions using a learning model that has learned the ability to select appropriate cutting conditions. Effect of the Invention

[0010] According to one aspect of the present invention, optimal cutting conditions can be automatically selected using a learning model that has learned the ability to select appropriate cutting conditions. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a block diagram showing a basic configuration of a laser processing machine according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram for explaining the flow of data in the cutting condition selection and learning device 10 of the laser processing machine 100. As shown in FIG. [Diagram 3] FIG. 3 is a diagram for explaining the use priority order of learning models in the selection unit 11. As shown in FIG. [Figure 4] FIG. 4 is a diagram for explaining the processed shape. [Diagram 5] FIG. 5 is a diagram for explaining a method of calculating the shape complexity. [Figure 6] FIG. 6 is a flowchart showing an example of a cutting condition selection and learning process procedure of the laser processing machine 100. [Figure 7] FIG. 7 is a flowchart showing another example of the cutting condition selection and learning processing procedure of the laser processing machine 100. [Figure 8] FIG. 8 is an explanatory diagram showing an example of a display screen of the HMI (PC) 10A of the laser processing machine 100. [Figure 9] FIG. 9 is an explanatory diagram showing an example of a display screen of the HMI (PC) 10A of the laser processing machine 100. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, a laser processing machine according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings. However, the following embodiments do not limit the invention according to each claim, and all of the combinations of features described in the embodiments are not necessarily essential to the solution of the invention.

[0013] In the following embodiments, the same or corresponding components are denoted by the same reference numerals and redundant explanations are omitted. In the embodiments, the arrangement, scale, dimensions, etc. of each component may be exaggerated or minimized, and may not match the actual ones, and some components may be omitted.

[0014] [Basic configuration of a laser processing machine] FIG. 1 is a block diagram showing a basic configuration of a laser processing machine according to an embodiment of the present invention. As shown in Fig. 1, the laser processing machine 100 includes a cutting condition selection and learning device 10 including a selection unit 11 that selects cutting conditions to be set in the laser processing machine 100 that laser cuts a workpiece, and a learning unit 12 that learns a learning model 30 used for selecting cutting conditions by the selection unit 11, and a processing result judgment device (judgment device) 20 that judges a processing result obtained by laser cutting a workpiece by the laser processing machine 100 based on the cutting conditions selected by the cutting condition selection and learning device 10. The cutting condition selection and learning device 10 has a pre-trained model (first trained model) 31 that is created by inputting, as the learning model 30 in the learning unit 12, first data including processing data (material, plate thickness, size of the processed shape, etc. of the workpiece) and shape complexity of the processed shape, output data (second data) including cutting conditions selected according to the first data, and processing results obtained by laser cutting based on the processing data and the output data (second data), and learning based on these. The processing data can be used as data on processing conditions including various information such as the material of the workpiece, the plate thickness, and the size of the processed shape.

[0015] The cutting condition selection and learning device 10 is included in an HMI (Human Machine Interface) (PC: Personal Computer) 10A of the laser processing machine 100. The HMI (PC) 10A is configured with hardware such as a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Only Memory).

[0016] The HMI (PC) 10A further includes, for example, an operation unit 13 that receives operation input from an operator, and a display unit 14 that displays various information related to the laser processing machine 100. The operation unit 13 and the display unit 14 may be configured by a display device, a touch panel device arranged on the screen of the display device, or the like.

[0017] The laser processing machine 100 also includes an NC (Numerical Control) device 40 that controls the entire laser processing machine 100, and a laser oscillator 45 that generates laser light used in the laser processing machine 100. In Fig. 1, the processing result determination device 20, the NC device 40, and the laser oscillator 45 are mounted so as to be included in the laser processing machine 100, but are not limited to this and may be provided outside the laser processing machine 100.

[0018] The laser processing machine 100 performs processing such as laser cutting and laser drilling of a workpiece (hereinafter, "cutting" and "drilling" are collectively referred to as "laser cutting processing"). Although not shown in the figure, the laser processing machine 100 has the functions of a known laser processing machine including a processing table on which the workpiece is placed, a processing unit having a laser processing head that is movable relatively to the processing table, and an assist gas supply device that supplies assist gas.

[0019] The laser oscillator 45 emits, for example, a 1 μm band laser beam having a wavelength of 900 nm to 1100 nm. The laser oscillator 35 may be configured with a DDL (Direct Diode Laser) oscillator, a fiber laser oscillator, a multi-wavelength resonator, etc., but is not limited to these. The laser oscillator 35 may also emit a guide beam for confirming the position on the workpiece to which the laser beam is to be emitted.

[0020] The processing result judgment device 20 includes, for example, a light detection unit, and judges the quality of the processing result of the workpiece using a learning model of a pre-learned quality judgment standard (judgment information) based on time series data of the light intensity distribution of light from the processed part of the workpiece detected by this light detection unit. The processing part here includes not only the processing point of the workpiece irradiated with laser light, but also the cutting front, which is an inclined part that receives laser light in the processing direction in the cutting groove, and its neighboring area. The processing result judgment device 20 may also make judgment based on the reflected light of laser light in general, image information of the processed part of the workpiece, visual judgment information by an operator, etc.

[0021] A selection unit 11 of the cutting condition selection and learning device 10 inputs processing data of processing conditions including the material, thickness and size of the processed shape of the workpiece, and the shape complexity of the processed shape calculated from the processing data, into a learning model 30 of a learning unit 12, and selects optimal cutting conditions for the laser processing machine 100. The optimal cutting conditions here refer to cutting conditions appropriately selected in accordance with the processing conditions set in the laser processing machine 100. The cutting conditions include, for example, shaft speed, laser output (peak power, repetition frequency, pulse width), machining velocity, focus position, etc.

[0022] The cutting condition selection and learning device 10 further includes, as a learning model 30 in the learning unit 12, an operator reselection trained model (second trained model) 32 that is created by inputting the processing data and the shape complexity of the processed shape (first data), output data (third data) including cutting conditions obtained by modifying the cutting conditions selected by the selection unit 11, and the processing results obtained by the laser cutting processing based on the processing data and the output data (third data) and performing re-learning (additional learning) based on these. By having the operator reselection trained model 32, it becomes possible to select appropriate cutting conditions selected by the operator.

[0023] Furthermore, the selection unit 11 of the cutting condition selection / learning device 10 preferentially uses the operator reselection trained model 32 as the learning model 30 used to select cutting conditions over the pre-trained model 31. This makes it possible to preferentially use the learning model 30 that has learned the ability of a skilled operator to select more appropriate cutting conditions, for example, and makes it easier to automatically select optimal cutting conditions.

[0024] The cutting condition selection and learning device 10 further includes, as a learning model 30 in the learning unit 12, a post-operation trained model (third trained model) 33 created by inputting the processing data and the shape complexity of the processing shape (first data), output data (fourth data) including the cutting conditions selected by inputting the first data into a pre-trained model (first trained model) 31 by the selection unit 11, and a processing result obtained by laser cutting processing based on the processing data and the output data (fourth data) and performing re-learning (additional learning) based on these. By having the post-operation trained model 33, it becomes possible to select appropriate cutting conditions selected at the processing site.

[0025] In this case, the selection unit 11 of the cutting condition selection / learning device 10 uses, as the learning model 30 to select cutting conditions, the worker reselection trained model (second trained model) 32, the post-operation trained model (third trained model) 33, and the pre-trained model (first trained model) 31 in that order. This allows, for example, the learning model 30 that has learned the ability of a skilled worker to select more appropriate cutting conditions to be used as the top priority, and then the learning model 30 capable of selecting appropriate cutting conditions selected at the machining site to be used, making it easier to automatically select optimal cutting conditions.

[0026] [Data flow for cutting condition selection and learning device] FIG. 2 is a diagram for explaining the flow of data in the cutting condition selection and learning device 10 of the laser processing machine 100. As shown in FIG.

[0027] 2, processing data 15, which is an explanatory variable, and a shape complexity 15d of the processed shape are input to the selection unit 11 of the cutting condition selection and learning device 10. The processing data 15 includes, for example, information on the material quality 15a of the workpiece, the thickness 15b of the workpiece, and shape data (G code) 15c of the processed shape.

[0028] The shape data (G code) 15c included in the machining data 15 includes information on the machining width 15ca of the machining shape, the machining height 15cb of the machining shape, and the machining outer shape or machining inner shape (hole) 15cc which is the machining shape. The machining shape represented by this shape data (G code) 15c will be described later.

[0029] The shape complexity 15d of the machining shape is an index obtained by integrating the curvature of the trajectory of the machining shape of the shape data 15c or the angle of the curved portion, and is therefore substantially included in the input machining data 15. Note that the shape complexity 15d of the machining shape is, for example, input to the selection unit 11 and then used as teacher data of the explanatory variables when the learning unit 12 learns the learning model 30, but is not limited to this. Also, the shape complexity 15d may be calculated in advance from the machining data 15 by, for example, an external device, and then input to the selection unit 11.

[0030] The selection unit 11 of the cutting condition selection and learning device 10 outputs selected cutting conditions 16, which are the objective variables, and processing data 15. The output cutting conditions 16 are optimal cutting conditions 16 for the processing data 15, selected by the selection unit 11 using the learning model 30 of the learning unit 12 based on the input processing data 15. The output processing data 15 is the processing data 15 corresponding to the time when the cutting conditions 16 were selected.

[0031] The learning unit 12 of the cutting condition selection and learning device 10 receives the output cutting conditions 16, the output processing data 15, and the processing result 17 of the laser cutting process performed by the laser processing machine 100 based on these cutting conditions 16 and processing data 15 as teacher data for machine learning and re-learning of the learning model 30. Note that these input data may be input and learned only when the judgment result by the processing result judgment device 20 indicates that the processing result 17 is good, but is not limited to this.

[0032] [Priority of learning model usage] FIG. 3 is a diagram for explaining the use priority order of learning models in the selection unit 11. As shown in FIG. As shown in Figure 3, the usage priority order table 101 shows the correspondence between the pre-trained model 31, the worker reselection trained model 32, the post-operation trained model 33, and the usage priority order items 102 for these learning situations.

[0033] As shown in row 101a of the usage priority order table 101, when the pre-trained model 31 has been trained and the worker reselection trained model 32 and the post-operation trained model 33 have not yet been trained, the selection unit 11 uses the pre-trained model 31. Also, as shown in row 101b of the usage priority order table 101, when the pre-trained model 31 and the post-operation trained model 33 have been trained and the worker reselection trained model 32 has not yet been trained, the selection unit 11 prioritizes the use of the post-operation trained model 33 and the pre-trained model 31, in that order.

[0034] In addition, as shown in row 101c of the usage priority order table 101, when the pre-trained model 31 and the worker reselection trained model 32 have each been trained and the post-operation trained model 33 has not yet been trained, the selection unit 11 prioritizes the use of the worker reselection trained model 32 and the pre-trained model 31, in that order.

[0035] Then, as shown in row 101d of the usage priority order table 101, when all of the pre-trained model 31, the operator reselection trained model 32, and the post-operation trained model 33 have been trained, the selection unit 11 uses the operator reselection trained model 32 with the highest priority by prioritizing the use of the operator reselection trained model 32, the post-operation trained model 33, and the pre-trained model 31 in that order as described above. The determination of the usage priority order of the pre-trained model 31, the operator reselection trained model 32, and the post-operation trained model 33 becomes effective, for example, after the cutting condition selection and learning device 10 has determined that the models have been trained.

[0036] The learning unit 12 of the cutting condition selection / learning device 10 performs machine learning and re-learning (additional learning) of the learning model 30 thus configured using a learning method including various algorithms such as regression, classification, clustering, discrimination, interpolation, feature extraction, and time series modeling.

[0037] Specifically, the learning unit 12 may use, for example, ensemble learning (random forest) as machine learning, but may also use a rule base such as simple conditional branching to perform learning, and the learning method is not limited to these. In addition, re-learning (additional learning) may include techniques such as transfer learning, distillation, and fine tuning.

[0038] The pre-trained model 31, the worker reselection trained model 32, and the post-operation trained model 33 may be created by performing training using the same or different algorithms. The pre-trained model 31, the worker reselection trained model 32, and the post-operation trained model 33 may be provided as separate training models, or may be provided as a single training model combined into the training model 30.

[0039] In particular, since the operator reselection trained model 32 is a trained model including cutting conditions obtained by modifying the cutting conditions selected by the selection unit 11, for example, when an operator modifies the cutting conditions via the operation unit 13 of the HMI (PC) 10A, the operator reselection trained model 32 may be created and provided individually for each operator who modifies the cutting conditions. This makes it possible to have individual trained models that have learned the ability to select cutting conditions according to the proficiency of each operator.

[0040] The processing data used by the learning unit 12 of the cutting condition selection / learning device 10 to create the pre-trained model 31, the worker re-selection trained model 32, and the post-operation trained model 33 is data on processing conditions including information such as the material of the workpiece, the plate thickness, and the size of the processed shape as described above, as teaching data for explanatory variables, and may include the shape complexity of the processed shape calculated based on this data.

[0041] [Description of processed shape] Here, the machining shape as the explanatory variable can be defined by shape data (for example, a known G code) 15c shown in FIG. 2, and includes information such as machining width, machining height, and machining outer or inner shape (hole).

[0042] Fig. 4 is a diagram for explaining the processed shape. The following description will be given with reference to Fig. 2 as well. As shown in Fig. 4(a), the machining width 15ca included in the machining shape of the workpiece W represents the size (dimension) (mm) indicated by the arrow X in the figure, and the machining height 15cb represents the size (dimension) (mm) indicated by the arrow Y in the figure, and can define the size of the machining shape. And, for example, as shown in Fig. 4(b), when the machining shape C of the workpiece W is in the state shown in the figure, the machining outer shape or machining inner shape (hole) 15cc can represent the shape of the machining outer shape Ca and the machining inner shape (hole) Cb as shown in Fig. 4(c).

[0043] The outer machining shape Ca and inner machining shape (hole) Cb of these machining shapes C have different cutting condition selection criteria (e.g., information that serves as a basis for selecting cutting conditions according to standard machining conditions) that are preset in the laser machining machine 100, and therefore selecting cutting conditions using the learning model 30 that has learned machining shapes including these is extremely useful for reducing labor hours and avoiding machining defects.

[0044] [Geometric complexity of processed shape] In addition, the learning unit 12 of the cutting condition selection and learning device 10 includes the shape complexity 15d of the machining shape as the first data together with the machining data 15 as described above when machine learning (including re-learning) the learning model 30 (see FIG. 2, the same below). This shape complexity is an index obtained by integrating the curvature of the trajectory of the machining shape or the angle of the curved portion as described above. By performing learning using the shape complexity 15d of the machining shape together with the machining data 15 as teacher data of explanatory variables, it is possible to further improve the selection accuracy of the cutting conditions using the learning model 30. In addition, the shape complexity can be easily calculated by integrating the curvature or the angle of the curved portion.

[0045] FIG. 5 is a diagram for explaining a method of calculating the shape complexity. 5, for example, cutting is performed along a trajectory T from a processing start point P1 to a processing end point P2. The shape complexity of the processed shape can be calculated, for example, by the sum of the curvatures (integral of the curvatures) of the trajectory T. That is, since it is more difficult to process an arc portion in the trajectory T as the distance is longer and the radius (R) is smaller, the shape complexity of the trajectory T increases as the number of corners and arcs increases.

[0046] The shape complexity can also be calculated as follows. For example, the curvature is 1 / R, and the angle of the arc portion is θ(°). Here, the corners formed by the straight line and the direct line shown in FIG. 5 are considered to be arc portions with a small radius R, and the angle θ of the arc portion is considered to be equivalent to the angle between the tangent at the start and end points of the arc portion and the straight line that intersects at right angles. If the length of the curve is 2πR(θ / 360), the shape complexity can be calculated from the angle of the arc portion alone, by the length of the curve × curvature = 2πR(θ / 360) × (1 / R) = 2π(θ / 360). Therefore, the shape complexity can be calculated by [the sum of the angles of the curved portions (arc portions, corner portions) with respect to the direction of travel of the trajectory T] / (360 × 2π).

[0047] Since the shape complexity obtained in this manner represents the complexity of the machining shape, by using a learning model 30 trained on this shape complexity 15d and shape data (G code) 15c representing the size of the machining shape, such as the above-mentioned machining width 15ca and machining height 15cb, as explanatory variables, it is possible to select cutting conditions more accurately.

[0048] That is, even if the shape complexity is the same (the processed shape is the same), if the size of the processed shape is different, the degree of difficulty of the laser cutting process will be different. If the degree of difficulty of the laser cutting process is different, the cutting conditions to be selected must be appropriate for each case. A learning model that simply learns the shape complexity of the processed shape, and a learning model that learns only the size of the processed shape as in the past, cannot accurately select appropriate cutting conditions like the learning model 30 of this embodiment.

[0049] [Cutting condition selection and learning process procedure] 6 is a flow chart showing an example of a cutting condition selection and learning process procedure of the laser processing machine 100. In the following, explanations that overlap with those already explained will be omitted. In the above-mentioned laser processing machine 100, the processing results obtained by the laser cutting processing based on the processing data and the output data are used as teacher data of explanatory variables in the learning model 30, regardless of whether the processing is good or bad.

[0050] For this reason, the learning model 30 can be configured as a learning model that can select not only cutting conditions that result in good machining, but also cutting conditions that result in poor machining. In the following cutting condition selection and learning processing procedure, a process is described for using the machining result when the machining result is determined to be good by the machining result determination device 20.

[0051] That is, the processing result determination device (determination device) 20 determines whether the processing result is good or bad, and the cutting condition selection and learning device 10 trains and re-learns the learning model 30 by machine learning using the learning unit 12 when the processing result determination device (determination device) 20 determines that the processing result is good. This makes it possible to more accurately select cutting conditions that provide good processing results in laser cutting processing using the learning model 30.

[0052] In addition, since the cutting condition selection and learning process is performed by the cutting condition selection and learning device 10, the description will be omitted unless otherwise specified below. In addition, it is assumed that the pre-trained model 31 has already been trained, but the worker re-selection trained model 32 has not yet been trained.

[0053] 6, first, for example, in accordance with an operation instruction via the operation unit 13 by an operator of the laser processing machine 100, data on processing conditions including the material of the workpiece, the plate thickness, and the size of the processed shape (processing data) is read (step S1), and the read data is analyzed (step S2). In the analysis process of step S2, a shape complexity is calculated based on the shape data (processed shape) of the read processing data. The calculated shape complexity may be displayed as information representing the shape complexity on the display screen of the display unit 14, for example.

[0054] Then, according to the usage priority of the learning model 30, the pre-trained model 31 is selected from the pre-trained model 31 and the worker reselected trained model 32, and the selection unit 11 inputs the analyzed data into the selected pre-trained model 31 and selects cutting conditions that may be optimal in accordance with this (step S3).

[0055] Then, it is determined whether the selected cutting conditions have been reselected, for example, by an operator making changes to the cutting conditions (step S4). If it is determined that the cutting conditions have not been reselected (No in step S4), the laser processing machine 100 starts laser cutting processing based on the cutting conditions selected by the selection unit 11 together with the processing data (step S5).

[0056] Thereafter, the processing result determination device 20 determines whether the processing result is good or bad (step S6). If it determines that the processing result is not good (No in step S6), the cutting condition selection and learning process is terminated, and the series of processes according to this flowchart is terminated.

[0057] On the other hand, if it is determined that the processing result is good (Yes in step S6), the learning unit 12 assumes that cutting conditions that can produce good processing results have been selected, and performs machine learning based on the above processing data, output data including the selected cutting conditions, and the processing results to create a post-operation trained model 33 (step S7), terminating the series of processes according to this flowchart. Note that, if a post-operation trained model 33 already exists in step S7 and the training is incomplete, re-learning (additional training) of the post-operation trained model 33 is performed based on the above data and processing results, assuming that cutting conditions that can produce good processing results have been selected.

[0058] If it is determined in step S4 that the cutting conditions have been reselected (Yes in step S4), the laser cutting process based on the reselected cutting conditions together with the processing data is started by the laser processing machine 100 (step S8). Thereafter, the processing result determination device 20 determines whether the processing result is good or bad (step S9), and if it determines that the processing result is not good (No in step S9), the cutting condition selection and learning process is terminated, and the series of processes according to this flowchart is terminated.

[0059] On the other hand, if it is determined that the processing result is good (Yes in step S9), the learning unit 12 regards this as a case in which cutting conditions that can obtain good processing results have been reselected, and re-learning (additional learning) of the worker reselection trained model 32 is performed based on the above processing data, output data including the reselected cutting conditions, and the processing results (step S10), and a series of processes according to this flowchart is terminated. By performing the cutting condition selection and learning process in this manner, it is possible to gradually train a learning model that can select cutting conditions that can obtain good processing results, and make it trained.

[0060] 7 is a flowchart showing another example of the cutting condition selection and learning process procedure of the laser processing machine 100. Here, it is assumed that a pre-trained model 31, an operator reselection trained model 32, and a post-operation trained model 33 are prepared as training models 30, and the pre-trained model 31 has already been trained.

[0061] As shown in Fig. 7, the data (processed data) is read by an operator (step S11) in the same manner as in Fig. 6, and the read data is analyzed (step S12). The shape complexity is also calculated in the analysis process of step S12. Then, based on the use priority of the learning model 30, it is first determined whether the operator reselection trained model 32 has been trained (step S13).

[0062] Here, if it is determined that the worker reselection trained model 32 has been trained (Yes in step S13), the worker reselection trained model 32 is selected, and the selection unit 11 uses the selected worker reselection trained model 32 to input the analyzed data into this worker reselection trained model 32, and selects cutting conditions that may be optimal based on this (step S17).

[0063] On the other hand, if it is determined that the worker reselection trained model 32 has not been trained (training is incomplete) (No in step S13), it is determined whether the post-operation trained model 33 has been trained or not (step S14) based on the usage priority of the training model 30. Here, if it is determined that the post-operation trained model 33 has been trained (Yes in step S14), the post-operation trained model 33 is selected, and the selected post-operation trained model 33 is used by the selection unit 11 to input analyzed data into this post-operation trained model 33 and select cutting conditions that may be optimal for this (step S16).

[0064] In addition, if it is determined that the post-operation trained model 33 has not been trained (training is incomplete) (No in step S14), the pre-trained model 31 is selected, and the selection unit 11 uses the selected pre-trained model 31 to input the analyzed data into this pre-trained model 31 and selects the cutting conditions that may be optimal for this (step S15).

[0065] Then, for the cutting conditions selected in the above steps S15 to S17, it is determined whether the cutting conditions have been reselected, for example, by an operator making changes to the cutting conditions (step S18). If it is determined that the cutting conditions have not been reselected (No in step S18), the laser cutting process based on the cutting conditions selected by the selection unit 11 together with the processing data is started by the laser processing machine 100 (step S19).

[0066] Thereafter, the processing result judgment device 20 judges whether the processing result is good or bad (step S20), and if it is judged that the processing result is good (Yes in step S20), the learning unit 12 re-learns (additional learning) the post-operation trained model 33 based on the processing data, the output data including the selected cutting conditions, and the processing result, and the series of processes according to this flowchart is terminated.

[0067] On the other hand, if it is determined in step S18 that the cutting conditions have been reselected (Yes in step S18), the laser cutting process based on the reselected cutting conditions together with the processing data is started by the laser processing machine 100 (step S22). Thereafter, the processing result determination device 20 determines whether the processing result is good or bad (step S23), and if it is determined that the processing result is good (Yes in step S23), the learning unit 12 re-learns (additional learning) the operator reselection trained model 32 based on the processing data, the output data including the reselected cutting conditions, and the processing result, and the series of processes according to this flowchart is terminated.

[0068] If it is determined in the above steps S20 and S23 that the processing results are not good (No in step S20 and No in step S23), the cutting condition selection and learning process is terminated and the series of processes according to this flowchart is terminated. By performing such cutting condition selection and learning process, it becomes possible to more easily and accurately automatically select optimal cutting conditions using a learning model that has learned the ability to select more appropriate cutting conditions.

[0069] [Display screen example] 8 and 9 are explanatory diagrams showing an example of a display screen of the HMI (PC) 10A of the laser processing machine 100. The example shown in Fig. 8 shows an example of a display screen after processing data has already been read by the cutting condition selection / learning device 10 and cutting conditions have been selected by the selection unit 11 using the learning model 30 of the learning unit 12.

[0070] 8, the display screen 110 has an information display area 111. The information display area 111 is provided with a shape display area 112 that displays the machining shape of the workpiece. On the left side of the shape display area 112, for example, a cutting condition display field 113 that displays each item of the selected cutting conditions may be displayed.

[0071] In addition, on the right side of the shape display area 112, an automatic cutting condition selection button 114 for instructing the cutting condition selection / learning device 10 to automatically select cutting conditions, and a processing start button 115 for instructing the laser processing machine 100 to start laser cutting processing may be displayed.

[0072] Here, for example, when the worker aligns the pointer 116 with an arbitrary machining shape (e.g., hole Cb), a pop-up screen 117 visibly displaying the shape complexity of the machining shape where the pointer 116 is aligned is displayed, for example, above the cutting condition display field 113. With reference to the shape complexity displayed on this pop-up screen 117, the worker can easily determine whether each item of the cutting condition displayed on the cutting condition display field 113 is appropriate or whether it should be reselected. Note that the display mode for displaying the shape complexity is not limited to the pop-up screen 117.

[0073] When reselecting the cutting conditions, a setting screen (not shown) is displayed on the display screen 110, and each item of the cutting conditions can be modified, changed, etc. After reselecting the cutting conditions, for example, by pressing the processing start button 115, laser cutting processing based on the reselected cutting conditions together with the processing data is performed, and the learning unit 12 re-learns the reselected operator learned model 32.

[0074] Such an operator reselection trained model 32 can be used and saved, for example, for each operator, in the cutting condition selection and learning device 10. To switch the operator reselection trained model 32 used for selecting cutting conditions, a switching screen 120 is displayed on the display screen 110 as shown in FIG. 9 by an instruction via the operation unit 13 by the operator, etc.

[0075] 9, a model profile selection item 121 is displayed at the top. By checking a check box 122 of this model profile selection item 121, for example, a profile display field 123 is displayed below the model profile selection item 121, where the profile of the worker reselection trained model 32 for each worker can be selected.

[0076] In the profile display field 123, items such as "user name," "level," "display language," "date format," "department," "nickname," and "memo" can be displayed as a profile for each user (worker). The worker can select any user from the profile of each user displayed in this profile display field 123. For example, when the worker selects item 123a in the profile display field 123, the worker reselection trained model 32 corresponding to the profile of the user name "USER01" is selected.

[0077] In this way, the cutting condition selection and learning device 10 can use the selected user's worker reselection trained model 32 for selecting cutting conditions as appropriate. Note that the display screen 110 shown in Fig. 8 and Fig. 9 is an example, and the display screen in the cutting condition selection and learning device 10 of this embodiment is not limited to this.

[0078] Although the embodiment of the present invention has been described above, this embodiment is presented as an example and is not intended to limit the scope of the invention. This new embodiment can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the gist of the invention. For example, the learning model 30 in the learning unit 12 may use at least one of the pre-trained model 31, the worker reselection trained model 32, and the post-operation trained model 33. These learning models may also be rule-based. Furthermore, pre-training and re-training are not limited to machine learning, and may be rule-based or statistical processing. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the scope of the invention and its equivalents described in the claims. [Explanation of symbols]

[0079] 10 Cutting condition selection and learning device 11 Selection Department 12 Learning Department 13 Control section 14 Display section 20 Processing result judgment device 30 Learning Model 31 Pre-trained Models (First Pre-trained Models) 32 Worker reselection trained model (second trained model) 33 Post-operation trained model (third trained model) 40 NC device 45 Laser Oscillator 100 Laser processing machine

Claims

1. A cutting condition selection and learning device including a selection unit that selects cutting conditions to be set in a laser processing machine that laser cuts a workpiece, and a learning unit that learns a learning model used for the selection of the cutting conditions by the selection unit; a determination device for determining a processing result obtained by laser cutting the workpiece by the laser processing machine based on the cutting conditions selected by the cutting condition selection and learning device; Equipped with The cutting condition selection and learning device includes, as the learning model in the learning unit, The first data includes the material, thickness, size and complexity of the processed shape of the workpiece, second data includes cutting conditions selected according to the first data, and a processing result obtained by laser cutting based on the first data and the second data. The first trained model is created by training based on these data. Laser processing machine.

2. The shape complexity is an index obtained by integrating the curvature of the trajectory of the processed shape or the angle of the curved portion.

2. The laser processing machine according to claim 1.

3. The cutting condition selection and learning device includes, as the learning model in the learning unit, The apparatus further includes a second trained model that is created by inputting the first data, third data including cutting conditions obtained by modifying the cutting conditions selected by the selection unit, and a processing result obtained by laser cutting based on the first data and the third data, and performing re-learning based on these.

2. The laser processing machine according to claim 1.

4. The selection unit of the cutting condition selection and learning device uses the second trained model in preference to the first trained model as the learning model used for selecting the cutting condition.

4. The laser processing machine according to claim 3.

5. The cutting condition selection and learning device includes, as the learning model in the learning unit, The method further includes inputting the first data, fourth data including cutting conditions selected by inputting the first data into the first trained model by the selection unit, and a processing result obtained by laser cutting processing based on the first data and the fourth data, and performing re-learning based on these to create a third trained model.

4. The laser processing machine according to claim 3.

6. The selection unit of the cutting condition selection and learning device uses the second trained model, the third trained model, and the first trained model in this order as the learning model to be used for selecting the cutting condition.

6. The laser processing machine according to claim 5.

7. The judgment device judges whether the processing result is good or bad, The cutting condition selection and learning device, when the determining device determines that the processing result is good, learns and re-learns the learning model by the learning unit.

6. The laser processing machine according to claim 1, 3 or 5.

Images