System and method for setting cutting parameter values of a laser cutting machine for sheet metal

Abstract

The invention relates to a method for setting values for cutting parameters (122) of a laser cutting machine (1) for sheet metal, wherein: at least one first cut edge (13) is produced by means of the laser cutting machine (1) using a plurality of first cutting parameter values (124); at least one first image (16) of the produced cut edge (13) is created; a plurality of second cutting parameter values (126) is determined automatically based on the first image (16), the plurality of second cutting parameter values (126) comprising the same cutting parameters (122) as the plurality of first cutting parameter values (124), the plurality of second cutting parameter values (126) having at least one cutting parameter value that differs from the plurality of first parameter values (124); and the plurality of second cutting parameter values (126) is set on the laser cutting machine (2).

Description

[0001] System and method for setting cutting parameter values ​​of a laser cutting machine for sheet metal

[0002] The invention relates to a method for setting cutting parameter values ​​of a laser cutting machine for sheet metal.

[0003] Furthermore, the invention relates to a system for setting cutting parameter values ​​of a laser cutting machine for sheet metal.

[0004] A method for setting cutting parameter values ​​is known from WO 2022 / 058113 A1.

[0005] The invention is based on the object of providing a system and a method as mentioned above which simplify the setting of cutting parameter values ​​of a laser cutting machine for sheet metal.

[0006] This object is achieved according to the invention in the method mentioned at the outset in that at least one first cutting edge with a plurality of first cutting parameter values ​​is produced using the laser cutting machine, wherein at least one first image of the produced cutting edge is created, wherein a plurality of second cutting parameter values ​​are determined automatically based on the first image, wherein the plurality of second cutting parameter values ​​comprise the same cutting parameters as the plurality of first cutting parameter values, wherein the plurality of second cutting parameter values ​​have at least one changed cutting parameter value compared to the plurality of first cutting parameter values, wherein the plurality of second cutting parameter values ​​are set on the laser cutting machine.

[0007] This process enables particularly simple adjustment of the cutting parameter values, as the cutting parameter values ​​are reset based on an image of a cutting edge. It is irrelevant whether the numerous initial cutting parameter values ​​are correctly implemented by the laser cutting machine, as the numerous second cutting parameter values ​​are determined based on the image of the first cutting edge and do not depend on the initial cutting parameter values. Since each sheet is unique, cutting edges can vary even with identical initial cutting parameter values. Therefore, even with identical cutting parameter values ​​and nominally identical sheet metal, a different change in the cutting parameter values ​​may be desired.

[0008] It is understood that the process can be repeated with the second cutting parameter values ​​to determine third cutting parameter values. In this way, the cutting parameter values ​​can be iteratively optimized.

[0009] In a preferred embodiment, a direction and an intensity of a change in the plurality of first cutting parameter values ​​are automatically determined based on the first image, wherein the change is applied to the plurality of first cutting parameter values ​​and thus the plurality of second parameter values ​​are determined. It is therefore not necessary for the current cutting parameter values ​​to be known, since only a relative change in the cutting parameters is determined. This is particularly advantageous since the actual cutting parameter values ​​can differ from the set cutting parameter values, be it due to wear or incorrect calibration of the laser cutting machine. The direction and intensity of the change can relate to several cutting parameters simultaneously. This can be understood as determining a change vector. The individual components of the vector are then applied to the corresponding cutting parameter values.

[0010] In a particularly preferred embodiment, the direction and magnitude of the change in the plurality of first cutting parameter values ​​is determined by means of a trained artificial intelligence based on the first image. The trained artificial intelligence can be, for example, an artificial neural network. The use of a trained artificial intelligence enables a particularly quick and simple determination of the direction and magnitude of the change in the plurality of first cutting parameter values. How an artificial intelligence can be trained is explained further below. In a preferred embodiment, the first image is broken down into a plurality of smaller images, and for each smaller image, a proportional change in the plurality of first cutting parameter values ​​is determined, wherein the direction and magnitude of the change in the plurality of first cutting parameter values ​​is determined from the proportional changes.By breaking the images down into smaller images, they can be analyzed more easily. This is particularly advantageous when using trained artificial intelligence, as the artificial intelligence model becomes simpler and requires less memory, for example. Analyzing the multitude of smaller images improves the quality of determining the second cutting parameters compared to using a single smaller image.

[0011] In a preferred embodiment, a plurality of first images are generated, and for each first image, a proportional change in the plurality of first cutting parameter values ​​is determined, wherein the direction and magnitude of the change in the plurality of first cutting parameter values ​​are determined from the proportional changes. By generating a plurality of first images and determining proportional changes, the determination of the second cutting parameters is further improved. It is understood that the plurality of first images can also be broken down into a plurality of smaller images.

[0012] In a particularly preferred embodiment, a plurality of first cutting edges are each generated with a plurality of first cutting parameter values, wherein the plurality of first cutting parameter values ​​differ in pairs in at least one parameter value when generating the first cutting edges, wherein at least one first image is created of each first cutting edge, wherein a plurality of second cutting parameter values ​​are generated automatically based on the first images.

[0013] Each plurality of first cutting parameter values ​​therefore differs from all other plurality of first cutting parameters in at least one cutting parameter value. The cutting parameter having a different cutting parameter value can be different. Thus, a first plurality of first cutting parameters can differ from a second plurality of first cutting parameters in the gas pressure value, differ from a third plurality of first cutting parameters in the feed rate value, and differ from a fourth plurality of first cutting parameters in both the gas pressure value and the feed rate value.

[0014] The invention comprises a method for training an artificial intelligence for setting cutting parameter values ​​of a laser cutting machine for sheet metal, wherein a plurality of cutting edges are generated, each with different cutting parameter values, wherein at least one training image is created for each cutting edge, wherein a quality parameter value is determined for at least one quality parameter for each cutting edge, wherein the quality parameter value is determined by means of one or more sensors, wherein the quality parameter values ​​of the plurality of cutting edges are entered into a matrix based on the associated cutting parameter values, wherein a preferred direction and strength of a change in the cutting parameter values ​​is determined for each entry in the matrix, wherein the artificial intelligence is trained for each image that is assigned to an entry in the matrix,to output the preferred direction and strength of the change in the cutting parameter values, wherein the artificial intelligence after training is suitable, in particular in a method described above, to be used to determine the direction and strength of the change in the first cutting parameter values.

[0015] For training, the multitude of training images is divided into reference images and test images. The reference images are used to adjust the parameters of the artificial intelligence. The test images are used to check the quality of the trained artificial intelligence. In a first training step, the reference images are presented to the artificial intelligence, and the artificial intelligence parameters are algorithmically adjusted so that the corresponding preferred direction and magnitude of a change in the cutting parameter values ​​is output for each reference image. In a second training step, the test images are presented to the artificial intelligence, and the preferred direction and magnitude of a change in the cutting parameter values ​​output by the artificial intelligence are compared with the preferred direction and magnitude of a change in the cutting parameter values ​​associated with the respective test image.This process is repeated until the preferred direction and magnitude of a change in the cutting parameter values ​​output by the artificial intelligence essentially matches the preferred direction and magnitude of a change in the cutting parameter values ​​associated with the respective test image. If necessary, additional cutting edges and images of the cutting edges can be generated for further training data.

[0016] The cutting parameter values ​​used for the multitude of cutting edges are preferably different in pairs. A separate set of cutting parameter values ​​is used for each cutting edge. This creates more different images and increases the matrix size. The denser data space makes the method more robust and improves the training of the artificial intelligence.

[0017] In a preferred embodiment, a gradient of the quality parameter value is determined for each entry in the matrix, wherein the preferred direction and magnitude of the change in the cutting parameter values ​​is the gradient of the quality parameter value at the respective entry in the matrix. The gradient can, for example, be determined for each entry in the matrix from the differences to neighboring entries. Preferably, a mathematical function is determined for the gradient, wherein the values ​​of the matrix serve as support points for the mathematical function. For each point in the matrix, the gradient of the mathematical function is assumed to be the gradient of the quality parameter value. The preferred direction of change in the cutting parameter values ​​therefore points in the direction of the greatest change in the quality parameter value. The preferred magnitude of the change depends on the magnitude of the change in the quality parameter value.The mathematical function can be any suitable function for approximation or

[0018] Interpolation of the matrix values. The mathematical function is preferably a polynomial. This method trains a particularly preferred change in the cutting parameter values.

[0019] In an alternative embodiment, an optimum value of the quality parameter value is determined from all entries in the matrix, wherein a direction and a distance to the optimum value of the quality parameter value are determined for each entry in the matrix, wherein the preferred direction and strength of the change in the cutting parameter values ​​is the direction and the distance to the optimum value of the quality parameter value at the respective entry in the matrix. The optimum value can be the best quality parameter value of all entries in the matrix. Preferably, a mathematical function, in particular a polynomial, is determined for the optimum value, wherein the values ​​of the matrix serve as support points of the mathematical function. In this case, the optimum value is an extreme value of the mathematical function. Depending on the quality parameter, the optimum value can be a minimum or a maximum of the mathematical function.The preferred direction of change in the cutting parameter values ​​therefore points from the entry in the matrix toward the best possible value for the quality parameter, which is interpolated by the mathematical function. The preferred magnitude of the change depends on the distance of the entry in the matrix from the best possible quality parameter value. This procedure trains an alternative, particularly favorable change in the cutting parameter values.

[0020] In a preferred embodiment, the values ​​of a first cutting parameter and a second cutting parameter are varied, and all other cutting parameters have constant cutting parameter values. The matrix has two dimensions, with the first cutting parameter being varied in a first dimension and the second cutting parameter being varied in the second dimension. By varying only two cutting parameters, a very simple two-dimensional matrix can be created. For this two-dimensional matrix, a mathematical function with two variables and an output value as an interpolation function can be very easily adapted to the matrix values, thus enabling one of the embodiments described above.

[0021] It is understood that a multitude of two-dimensional matrices can be created, each with a different set of cutting parameters. A mathematical function can be adapted for each of these matrices. For each matrix, a separate artificial intelligence can be trained to output the preferred direction and magnitude of the change in the cutting parameter values. For example, one artificial intelligence can output the change in feed rate and distance between the gas nozzle and the sheet metal, while another artificial intelligence can output the change in the focus position of the cutting laser and the pressure of the cutting gas. The overall preferred direction and magnitude of the change in the cutting parameter values ​​can then be calculated from the individual outputs of the artificial intelligences.

[0022] Alternatively, a multidimensional matrix with many varied cutting parameters can be created. The mathematical function for interpolation then requires a corresponding number of variables. The artificial intelligence is then trained to output the preferred direction and magnitude of the change in the cutting parameter values ​​for a corresponding number of cutting parameters. Preferably, the artificial intelligence outputs a vector with a corresponding number of components.

[0023] In a preferred embodiment, training is performed for a specified process, wherein the process is specified by one or more values ​​of the following process parameters: a sheet metal type, a sheet metal thickness, a process type, a nozzle used, and a laser power. The sheet metal type is, for example, stainless steel, mild steel, aluminum, or copper. The sheet metal thickness is, for example, 1 mm, 2 mm, 3 mm, 5 mm, 10 mm, 15 mm, 20 mm, or a value in between. The process type is, for example, oxygen cutting, nitrogen cutting, or another laser cutting process. The laser power is, for example, 2 kW, 5 kW, 10 kW, 15 kW, 20 kW, or a value in between. Training for a specified process significantly improves the training result. When using artificial intelligence, the artificial intelligence trained for the corresponding process is used. The trained artificial intelligence can also be used for processes that do not exactly correspond to the trained process.For processes with differing process parameter values, similar quality characteristics lead to similar gradients or differences. For processes with differing process parameter values, it is preferable to use several trained artificial intelligences, and the direction and magnitude of the change determined by the trained artificial intelligences are offset against each other. For example, for a sheet thickness of 12 mm, the artificial intelligences trained for 10 mm and 15 mm can be used. The quality parameter is preferably one of the following: a burr on the generated cutting edge, discoloration on the generated cutting edge, roughness on the generated cutting edge, pitting on the cutting edge, or melt tipping on the cutting edge. These quality parameters are easily recognizable in images, especially from a trained artificial intelligence, and can be easily influenced by adjusting the cutting parameters.

[0024] Preferably, the plurality of cutting parameters is selected from the following: a cutting gas pressure, a feed rate, a focus position of the cutting laser, and a distance between the gas nozzle and the sheet metal. A laser cutting machine cuts sheet metal by directing a laser beam and a cutting gas from a cutting head onto the sheet metal. The laser melts the sheet metal, and the gas expels the melt. The cutting head is moved across the sheet metal at the feed rate along the contour to be cut. The focus position of the cutting laser determines whether the laser focus is above the sheet metal, within the sheet metal, or below the sheet metal. These cutting parameters each have a significant influence on at least one of the aforementioned quality parameters.

[0025] The invention further relates to a method for setting cutting parameter values ​​of a laser cutting machine with a trained artificial intelligence, wherein the trained artificial intelligence was trained according to one of the methods described above.

[0026] The invention further relates to a system for setting cutting parameter values ​​of a laser cutting machine for sheet metal, comprising a laser cutting machine for sheet metal, a camera, and a computing unit, wherein the computing unit is provided and configured to carry out a method according to the invention. It is understood that the computing unit can consist of several parts. For example, a computing unit can be provided and configured for training the artificial intelligence, a computing unit for controlling the laser cutting machine, and a computing unit for utilizing the trained artificial intelligence. The following description of preferred embodiments, in conjunction with the drawings, serves to explain the invention in more detail. They show:

[0027] Fig. 1 a system;

[0028] Fig. 2 shows a training procedure;

[0029] Fig. 3 shows a method;

[0030] Fig. 4 a training matrix;

[0031] Fig. 5 another training matrix;

[0032] Fig. 6 an iterative setting of cutting parameter values; and

[0033] Fig. 7 a setting of cutting parameter values.

[0034] Identical or functionally equivalent elements are designated by the same reference numerals in all embodiments.

[0035] Fig. 1 shows a system 100 for setting cutting parameter values ​​of a sheet metal laser cutting machine 1. In addition to the sheet metal laser cutting machine 1, the system 100 includes a camera 14 and a processing unit 10.

[0036] The laser cutting machine 1 has a laser beam generator 4 for generating the laser beam 3, a laser cutting head 5, and a workpiece support 6 on which the sheet metal 2 is arranged. The laser beam 3 is generated in the laser beam generator 4 and is guided from the laser beam generator 4 to the laser cutting head 5 by means of a fiber optic cable (not shown) or deflecting mirrors (not shown). The laser beam 3 is directed onto the sheet metal 2 by means of a focusing optics arranged in the laser cutting head 5. The laser beam 3 exits the laser cutting head 5 in a z-direction and strikes the workpiece 2 in the z-direction. The laser cutting head 5 can be moved above the sheet metal 2 in the x- and y-directions by means of a drive 7. The laser cutting machine 1 is also supplied with cutting gases 8, for example oxygen or nitrogen.The cutting gas 8 is fed to a cutting nozzle 9 of the laser cutting head 5, from whose nozzle opening it exits together with the laser beam 3.

[0037] The laser cutting machine 1 further comprises a computing unit 10 for controlling laser cutting parameters of the laser beam 3 and the cutting gas 8 and for controlling the drive 7 in order to move the laser cutting head 5 together with the emerging laser beam 3 according to a predetermined cutting contour 11 relative to the sheet 2 at a feed rate v.

[0038] Using the laser beam 3 and the cutting gas 8, the laser cutting machine 1 can cut a sheet metal part 12 from the sheet metal 2. Cutting parameter values ​​set on the laser cutting machine 1 are used for the cutting process. The sheet metal part 12 is given a cutting edge 13 by cutting along the cutting contour 11. An image 16 of the cutting edge 13 can be created using the camera 14. The image 16 can be transmitted from the camera 14 to the computing unit 10. The computing unit 10 is provided and configured to automatically set the laser cutting parameter values ​​on the laser cutting machine 1 based on the first image 16. To set the laser cutting parameter values, the computing unit contains predefined algorithms, e.g., a trained artificial intelligence.

[0039] Fig. 2 shows a training method 200 for an artificial intelligence for setting cutting parameter values ​​of a laser cutting machine 1 for sheet metal. The following describes a training method for an artificial intelligence for setting cutting parameter values ​​of a laser cutting machine 1 for sheet metal in connection with the system shown in Fig. 1.

[0040] In a first training step 210, the laser cutting machine 1 generates a plurality of cutting edges 13, each with different cutting parameter values. The values ​​of the process parameters, e.g., sheet metal type, sheet thickness, process type, nozzle used, and / or laser power, remain unchanged. The process parameter values ​​define the process, and the artificial intelligence is trained for the thus-defined process. It is understood that an artificial intelligence can be trained in the same way for a differently defined process, i.e., a process with different process parameter values.

[0041] In a second training step 220, at least one image of each cutting edge 13 is generated using the camera 14. For each image, at least one quality parameter value is determined and assigned to the image. The quality parameter value can be determined, for example, by the computing unit using an image processing algorithm. Alternatively, the quality parameter value can be determined using a suitable sensor directly at the respective cutting edge. Alternatively, the quality parameter value can be determined by an operator. In an optional third training step 230, the images are broken down into smaller images. By breaking them down into smaller images, the artificial intelligence model can be kept simpler, and more training images are easily generated. In a fourth training step 240, the images are divided into reference images and test images.The preferred division is 70% to 30%, but any other division is possible as long as sufficient images are available for both groups. It is particularly preferred to divide the images in such a way that at least one reference image is obtained for each of the cutting parameter values ​​used. This ensures that the matrix is ​​particularly well-populated in the next step.

[0042] In a fifth training step 250, at least one matrix is ​​created. The reference images are assigned to a position in the matrix according to the respective plurality of cutting parameter values ​​used. Example matrices 140 are shown in Fig. 4 and Fig. 5. The gas pressure is plotted on the horizontal axis 141 as the cutting parameter. The focal position is plotted on the vertical axis 142 as the cutting parameter. The quality parameter is the roughness of the cutting edge.

[0043] In a sixth training step 260, a mathematical function is determined based on the cutting parameter values ​​used in the reference images and the associated quality parameter values. The mathematical function is determined such that its variables correspond to the cutting parameters on the axes of matrix 140 and its output value corresponds to the quality parameter. When cutting parameter values ​​from reference images are inserted into the mathematical function, the output value should correspond as closely as possible to the quality parameter value of the reference image. The mathematical function thus interpolates the quality parameter values ​​between the values ​​of matrix 140. In Fig. 4 and Fig.

[0044] 5 Cutting parameter pairs with the same quality parameter values ​​are shown in dashed lines.

[0045] In a seventh training step 270, the preferred direction and strength of the change in the cutting parameters is determined for each reference image and test image. In the example from Fig. 4, the gradient 143 is determined from the mathematical function for each point in the matrix 140 and assigned to the reference image corresponding to that point. For reasons of clarity, three examples of points and gradients are shown in Fig. 4. Likewise, for the test images, the gradient 143 is determined at the corresponding points in the matrix 140 using the mathematical function, and the gradient 143 is assigned to the test images. In the example from Fig. 5, the optimal value 144 of the quality parameter is first determined, i.e., those cutting parameter values ​​at which the quality parameter value assumes the best possible value. The optimal value 144 can be the highest quality parameter value of all reference images. Preferably, and as shown here, the optimal value 144 is determined from the mathematical function.The preferred direction and magnitude of the change in the cutting parameter values ​​for each reference image is determined by the vector 145 from the reference image point in the matrix 140 to the optimum value 144. For clarity, three examples of points and corresponding vectors to the optimum value are shown in Fig. 5. Each test image can also be assigned a point in the matrix and a corresponding vector 145 based on the cutting parameter values ​​used.

[0046] In an eighth training step 280, an artificial intelligence model, preferably a convolutional neural network (CNN), is trained. For this purpose, the reference images are used as input data, and the model parameters are adjusted so that the output corresponds to the preferred direction and magnitude of the change in the cutting parameters determined in the seventh training step. For each reference image, the artificial intelligence should therefore output the corresponding vector. To verify whether the model training was successful, the test images are used as input data, and the artificial intelligence output is compared with the preferred direction and magnitude of the change in the cutting parameters determined in the seventh training step.

[0047] Fig. 3 shows a method 300 for setting values ​​for cutting parameters of a laser cutting machine 1 for sheet metal. In a first application step 310, a cutting edge 13 is generated. For this purpose, the laser cutting machine 1 cuts a sheet metal part 12 from the sheet metal 2. First cutting parameter values ​​124 set on the laser cutting machine 1 are used for the cutting process. The sheet metal part 12 receives a cutting edge 13 as a result of the cutting process.

[0048] In a second application step 320, a first image 16 of the cutting edge 13 is created using the camera 14. The first image 16 is transmitted from the camera 14 to the computing unit 10. In an optional third application step 330, the first image 16 is broken down into a plurality of smaller images.

[0049] In a fourth application step 340, the computing unit 10 determines a plurality of second cutting parameter values ​​126 based on the first image 16. Preferably, a trained artificial intelligence is used for this purpose, the training of which was described above. Based on the process parameters, an artificial intelligence trained for the corresponding process is selected. The first image 16 or the smaller images generated from the first image 16 are used as input data for the trained artificial intelligence. For each image, trained as described above, the trained artificial intelligence outputs the preferred direction and magnitude of a change in the cutting parameter values. The preferred direction and magnitude of the change are applied to the plurality of first cutting parameter values ​​124, thus determining the plurality of second cutting parameter values ​​126.The plurality of second cutting parameter values ​​126 are set on the laser cutting machine 1, and the method 300 is completed. For better results, the method can be repeated, starting with the plurality of second cutting parameter values. Thus, as indicated in Fig. 4, a preferred direction and intensity of a change can be determined based on a plurality of first cutting parameter values ​​124. After the second cutting parameter values ​​have been set, a new cutting edge 13 is generated. A new image 16 is generated from the new cutting edge 13, and based on the new image 16, a further preferred direction and intensity of a change is determined. By applying the further preferred direction and intensity of a change to the second cutting parameter values ​​126, third cutting parameter values ​​128 are determined, and the third cutting parameter values ​​128 are set on the laser cutting machine 1.When using an iterative approach, it is advantageous to create a cutting edge at a time and to determine the next preferred change based on an image, since each sheet is unique and the same changes to the cutting parameter values ​​can have different effects on the cutting edge, even if the sheet is nominally the same.

[0050] Additionally or alternatively, for better results, multiple cutting edges 13 can be created, each with a different plurality of first cutting parameter values. For example, a sheet metal part 12 with multiple cutting edges 13 can be cut out by the laser cutting machine 1, with a different plurality of first cutting parameter values ​​124 being used for each cutting edge 13. Thus, as indicated in Fig. 5, a preferred direction and magnitude of a change can be determined based on each plurality of first cutting parameter values ​​124. The respectively determined preferred direction and magnitude of a change is applied to each plurality of first cutting parameter values ​​124, and a plurality of second cutting parameter values ​​is determined from the changed plurality of first cutting parameter values, e.g., by averaging. The plurality of second cutting parameter values ​​is then set on the laser cutting machine. List of Reference Symbols

[0051] system

[0052] Laser cutting machine for sheet metal

[0053] sheet metal

[0054] laser beam

[0055] Laser beam generator

[0056] Cutting head

[0057] Workpiece support

[0058] drive

[0059] Cutting gas

[0060] Cutting nozzle

[0061] Computing unit

[0062] Cutting contour

[0063] Sheet metal part

[0064] cutting edge

[0065] camera

[0066] Picture

[0067] Quality parameters

[0068] Cutting parameters first cutting parameter values ​​second cutting parameter values ​​third cutting parameter values ​​matrix

[0069] Gas pressure

[0070] Focus position

[0071] gradient

[0072] Optimal value

[0073] vector

[0074] Training procedures

[0075] First training step

[0076] Second training step

[0077] Third training step

[0078] Fourth training step

[0079] Fifth training step Sixth training step

[0080] Seventh training step

[0081] Setting procedure

[0082] First application step

[0083] Second application step

[0084] Third application step

[0085] Fourth application step

Claims

Patent claims 1. A method for setting values ​​for cutting parameters of a laser cutting machine (1) for sheet metal, wherein at least one first cutting edge (13) having a plurality of first cutting parameter values ​​(124) is produced with the laser cutting machine (1), wherein at least one first image (16) of the produced cutting edge (13) is created, wherein a plurality of second cutting parameter values ​​(126) are determined automatically based on the first image (16), wherein the plurality of second cutting parameter values ​​(126) comprise the same cutting parameters as the plurality of first cutting parameter values ​​(124), wherein the plurality of second cutting parameter values ​​(126) have at least one changed cutting parameter value compared to the plurality of first cutting parameter values ​​(124), wherein the plurality of second cutting parameter values ​​(126) are set on the laser cutting machine (2).

2. Method according to claim 1, characterized in that a direction and a strength of a change (143, 145) of the plurality of first cutting parameter values ​​(124) is determined automatically based on the first image (16), wherein the change (143, 145) is applied to the plurality of first cutting parameter values ​​(124) and thus the plurality of second parameter values ​​(126) is determined.

3. The method according to claim 2, characterized in that the direction and the strength of the change (143, 145) of the plurality of first cutting parameter values ​​(124) is determined by means of a trained artificial intelligence based on the first image (16).

4. Method according to claim 2 or 3, characterized in that the first image (16) is divided into a plurality of smaller images and for each smaller Image a proportional change (143, 145) of the plurality of first cutting parameter values ​​is determined, wherein the direction and the strength of the change (143, 145) of the plurality of first cutting parameter values ​​is determined from the proportional changes.

5. Method according to one of the preceding claims 2 to 4, characterized in that a plurality of first images (16) are generated and for each first image (16) a proportional change (143, 145) of the plurality of first cutting parameter values ​​(124) is determined, wherein the direction and the strength of the change (143, 145) of the plurality of first cutting parameter values ​​(124) is determined from the proportional changes.

6. Method according to one of the preceding claims, characterized in that a plurality of first cutting edges (13) are each generated with a plurality of first cutting parameter values ​​(124), wherein the plurality of first cutting parameter values ​​(124) differ in pairs in at least one parameter value when generating the first cutting edges (13), wherein at least one first image (16) is created of each first cutting edge (13), wherein a plurality of second cutting parameter values ​​(126) are generated automatically based on the first images (16).

7. A method for training an artificial intelligence for setting cutting parameter values ​​of a laser cutting machine (1) for sheet metal, wherein a plurality of cutting edges (13) are generated, each with different cutting parameter values, wherein at least one training image is created for each cutting edge (13), wherein for each cutting edge (13) a quality parameter value is determined for at least one quality parameter, wherein the quality parameter values ​​are determined by means of one or more sensors, in particular a camera, wherein the quality parameter values ​​of the plurality of cutting edges are entered into a matrix (140) based on the associated cutting parameter values, wherein for each entry of the matrix (140) a preferred direction and strength of a change (143, 145) of the cutting parameter values ​​is determined, wherein the artificial intelligence is trained to output the preferred direction and strength of the change of the cutting parameter values ​​for each image which is assigned to an entry of the matrix (140), wherein the artificial intelligence is suitable after training, in particular in a method according to one of the preceding claims, to be used for determining a direction and strength of a change of cutting parameter values.

8. The method according to claim 7, characterized in that a gradient (143) of the quality parameter value is determined for each entry of the matrix (140), wherein the preferred direction and strength of the change (143) of the cutting parameter values ​​is the gradient of the quality parameter value at the respective entry of the matrix (140).

9. The method according to claim 7, characterized in that an optimum value (144) of the quality parameter value is determined from all entries of the matrix (140), wherein for each entry of the matrix (140) a direction and a distance to the optimum value (144) of the quality parameter value are determined, wherein the preferred direction and strength of the change (145) of the cutting parameter values ​​is the direction and the distance to the optimum value (144) of the quality parameter value at the respective entry of the matrix (140).

10. Method according to one of claims 7 to 9, characterized in that values ​​of a first cutting parameter and a second cutting parameter are varied and all other cutting parameters have constant cutting parameter values, wherein the matrix (140) has two dimensions (141, 142), wherein the first cutting parameter is varied in a first dimension (141) and the second cutting parameter is varied in the second dimension (142).

11. Experience according to one of claims 7 to 10, characterized in that the training is carried out for a specified process, wherein the process is specified by one or more values ​​of the following process parameters: a sheet metal type, a sheet metal thickness, a process type, a gas nozzle used, a laser power.

12. Experience according to one of claims 7 to 11, wherein the quality parameter is one of the following: a burr of the produced cut edge, a discoloration of the produced cut edge, a roughness of the produced cut edge, a pitting at the cut edge, or a folding melt at the cut edge.

13. Method according to one of the preceding claims, wherein the plurality of cutting parameters are selected from the following: a pressure of a cutting gas, a feed rate, a focus position of the cutting laser, a distance between the gas nozzle and the sheet metal.

14. A method for setting cutting parameter values ​​of a laser cutting machine (1) according to claim 3, wherein the trained artificial intelligence has been trained according to one of claims 9 to 13.

15. System for setting cutting parameter values ​​of a laser cutting machine (1) for sheet metal, comprising a laser cutting machine (1) for sheet metal, a camera (14) and a computing unit (10), characterized in that the computing unit (10) is provided and configured to carry out a method according to one of claims 1 to 8.

Images