Abstract
Method for determining laser machining parameters for the machining of a material using a laser machining system comprises the following steps: a) providing said central unit with a learning machining function capable of learning on the basis of said plurality of machining data samples, said learning machining function comprising an algorithm capable of defining the following laser machining parameters for said machining result sought and for said machining system: a polarization, an pulse energy E.sub.p, a diameter at the focal point w, a Gaussian order p, a pulse repetition rate PRR of pulses n, a wavelength; b) making said learning machining function to learn so as to said laser machining system can machine said material to be machined according to the machining result sought.
Claims
1. A method for determining laser machining parameters for the machining of a material with a laser machining system and comprising the following steps: a) providing a learning database comprising a plurality of machining data samples comprising machining results obtained as a function of the laser machining parameters used; b) providing a central unit comprising: machine learning means; means for communicating with said learning database; c) defining at said central unit a machining result sought of a material to be machined; d) providing said central unit with a learning machining function capable of learning on the basis of said plurality of machining data samples, said learning machining function comprising an algorithm capable of defining for said machining result sought and for said machining system the following laser machining parameters: information regarding a polarization of said machining laser beam, a pulse energy of said machining laser beam E.sub.p, a diameter of said machining laser beam at the focal point w, an order of the Gaussian p of said machining laser beam between 1 and 20, a pulse repetition rate PRR of pulses n of the machining laser beam, a wavelength of the machining laser beam; e) making said learning machining function with said machine learning means of said central unit to learn on the basis of said plurality of machining data samples, so that said laser machining system can machine said material to be machined according to the machining result sought, by providing it with the machining parameters determined by said learning machining function when learning it.
2. The method according to claim 1 wherein the plurality of samples of the learning database at least partly comprises information in relation to the material being machined, this information comprises at least one of the following information: a delta related to a penetration depth of the machining laser beam into the material, a threshold fluence F.sub.th related to a minimum energy density of the laser beam allowing machining, an incubation coefficient S, between 0 and 1, related to the dependence of the threshold fluence based on a number of pulses of the laser beam at the same point, a complex refractive index n+ik.
3. The method according to claim 1 wherein: the laser machining system is capable of emitting a laser machining beam moving in an direction x and wherein, said algorithm of said learning function of step d) is capable of further defining, after its learning by the machine learning means, the following laser machining parameters: a distance of the focusing point of the laser machining beam from the surface of the material to be machined, a speed (v) of movement of the laser machining beam relative to said material to be machined, an angle of incidence of the machining laser beam with respect to said surface of the material to be machined, a number of lines to be machined, a distance between said lines previously defined, a number of passes of the machining laser beam on each line to be machined.
4. The method according to claim 1 wherein: the laser machining system is a system capable of emitting a laser machining beam in rotation about an axis of rotation, and wherein, said algorithm of said learning function of step d) is capable of further defining, after its learning by the machine learning means, the following laser machining parameters: a rotational speed co of said laser beam about said axis of rotation, a distance BFG from a surface of said material to be machined for which the rotating laser beam describes a fixed spot for all rotating laser machining beam positions, a distance BFI from a surface of said material to be machined at which the rotating laser beam is focused, an angle of incidence of the rotating machining laser beam for all the positions of the rotating laser beam in relation to the normal to the surface of the material, an activation time of the rotating machining laser beam.
5. The method according to claim 1 wherein the machining laser beam is a machining laser beam rotating about an axis of rotation and a point of rotation located at said distance BFG from said surface of said material to be machined.
6. The method according to claim 2 wherein the algorithm comprises a learning step based on a precession radius r.sub.p of a machining of said material by said rotating machining laser beam comprising the following formula:
7. (canceled)
8. The method according to claim 2 wherein the algorithm comprises a learning step based on an ablated crater radius r.sub.c for a pulse of the machining laser beam comprising the following formula:
9. The method according to claim 1 wherein a machining result sought is a two-dimensionally defined machining result sought profile according to: an axis y representing a direction substantially parallel to the surface of the material to be machined, and, an axis z representing a direction substantially perpendicular to the surface of the material to be machined, the axis z corresponding to an ablation depth z.sub.n in relation to the surface of the material.
10. The method according to claim 2 wherein two successive pulses n of the rotating machining beam are separated by a distance dx and wherein the algorithm comprises a learning step based on an ablation depth z.sub.n comprising the following formula:
11. The method according to claim 10 wherein the delta is a constant parameter.
12-16. (canceled)
17. The method according to claim 1 wherein the method further comprises the following additional steps: f) providing a laser machining system comprising: said material to be machined; said laser machining device comprising: a laser source for emitting an ultra-short laser beam, less than 100 ps, onto said material to be machined; a control unit for controlling the emission of said laser beam from said laser source; said central unit for controlling said control unit; said database; communication means enabling communication from said central unit to said database; g) machining said material with said laser source configured with said laser machining parameters determined by said learning machining function in step e).
18. The method according to claim 17 wherein said laser machining device further comprises: a unit for analyzing the state of said material to be machined connected to said learning database, and wherein the following additional steps are implemented: h) acquiring machining results with said analysis unit after the step g); i) transmitting said machining results and said machining parameters used in machining in the step g) to said central unit, said central unit being configured to communicate a machining data pair comprising machining results obtained according to the laser machining parameters used; j) enriching said learning database with said machining data pair.
19. The method according to claim 1 wherein said central unit further comprises means for determining a difference obtained between information in relation to the material to be machined according to said variation of laser machining parameters; wherein the machine learning of the learning machining function is performed for a material to be machined according to a variation of the laser machining parameters from known laser machining parameters and when the difference determined by said determining means enables an improvement of said learning machining function, in particular the variation of the machining parameters is indicated by a predefined test.
20-25. (canceled)
26. The method according to claim 17 wherein the algorithm of the learning function is capable of determining laser machining parameters for feeding the learning database so as to allow a learning of the learning machining function and wherein the method further comprises the following additional steps: m) irradiating said material to be machined with said light beam with laser machining parameters defined by the learning function according to a range of predefined laser machining parameters; n) acquiring by the optical detection unit a result generated by irradiating said material to be machined with said light beam; o) transmitting said result to the central unit and accessing said result generated by irradiating said material to be machined with said light beam; p) extracting from said result by means of the central unit a pair of machining data comprising machining results obtained as a function of the laser machining parameters used in the step m); q) storing said pair of machining data in said learning database.
27. A laser machining device comprising a central unit capable of performing the steps of the method according to claim 1.
28. A device according to claim 27 wherein it comprises an optical unit for directing said laser beam towards said material to be machined.
29. A device according to claim 28 wherein said optical unit for directing said laser beam towards said material to be machined allows the precession of said laser beam.
30. A computer program for implementing a method according to claim 1.
31. (canceled)
32. A computer-readable medium on which the computer program according to claim 31 is recorded.
Description
BRIEF DESCRIPTION OF THE FIGURES
[0297] These and other aspects of the invention shall be clarified in the detailed description of particular embodiments of the invention, reference being made to the drawings of the figures, in which:
[0298] FIG. 1 represents the main characteristics of a beam having a Gaussian intensity profile;
[0299] FIG. 2 represents a matrix of craters with increasing numbers of pulses and pulse energies;
[0300] the FIG. 3a-FIG. 4b represent: FIG. 3a and FIG. 4a, a plurality of pulses along a line; FIG. 3b, a fluence profile obtained transversely to the line described in FIG. 3a; FIG. 3c represents a distribution of pulse number N transversely to the line described in FIG. 3a; FIG. 4b, the groove depth obtained transversely to the line described in FIG. 4a;
[0301] FIG. 5 shows simulation results obtained for a cross section of a line representing an ablation rate (left) and a maximum ablation rate as a function of the fluence of the laser;
[0302] FIGS. 6a and 6b show in FIG. 6a the movement of the sample in the direction of the axes Y and Z for the implementation of the D-Scan method and in FIG. 6b the ablation lobes resulting from the implementation of the experimental protocol of FIG. 6a;
[0303] FIGS. 7a and 7b show a schematic representation of the experiments to be performed to estimate the value of using the skin depth method;
[0304] FIGS. 7c and 7d show the experimental results compared to the models for the experiments shown in FIGS. 7a and 7b respectively;
[0305] FIGS. 8a and 8b show the results for the dimensions .sub.min and .sub.max from the experiments shown in FIGS. 7a and 7b for laser beam pulse frequencies of 5 KHz and 100 KHz respectively.
[0306] FIG. 9 shows the results of the 5 KHz and 100 KHz experiments of FIGS. 8a and 8b in a graph showing the fluence F.sub.th as a function of pulse number N;
[0307] FIG. 10 shows a simulation of the measured ablation as a function of the angle of incidence for a polarization S on the left and P on the right;
[0308] FIG. 11 shows the propagation of a Gaussian beam and represents the beam diameter w as a function of a propagation distance z;
[0309] FIGS. 12 and 13 represent experimental and simulation results for two values of . FIG. 12 represents the maximum depth as a function of the beam fluence. FIG. 13 represents the maximum depth as a function of scan speed of the beam;
[0310] FIGS. 14a, 14b, 15a, 15b, and 15c represent experimental and simulated profiles obtained on stainless steel 316L;
[0311] FIGS. 16a and 16b show in FIG. 16a a cross-section of a scan and the corresponding simulated cross section (FIG. 16b) obtained on stainless steel 316L;
[0312] FIGS. 17a and 17b show a schematic representation of a machining with a normal angle of incidence in FIG. 17a and a machining with a non-normal angle of incidence in FIG. 17b;
[0313] FIG. 18 shows a schematic representation of the angles of a light beam in experiments for the calculation of the local fluence at a non-normal incidence;
[0314] FIGS. 19a and 19b show a comparative example between a simulation of a machining profile according to the prior art (FIG. 19a) and a simulation according to a particular embodiment of the method of the invention for a machining carried out with a machining laser beam having a non-zero angle of attack with the normal to the surface of the material to be machined.
[0315] FIG. 20a shows a schematic representation of machining with a precessing laser machining beam;
[0316] FIGS. 20b, 20c, 21b and 21d show simulations of machining according to a particular embodiment of the method of the invention;
[0317] FIGS. 21a and 21c show machining results in the form of cross sectional images of a machining, the machining simulations illustrated in FIG. 21b and FIG. 21c are superimposed to the machining results respectively;
[0318] FIG. 22a shows a schematic representation of the points of impact of a precessing or rotating machining laser beam, the FIG. 22b shows a detail of the FIG. 22a;
[0319] FIGS. 23 to 31 show flow diagrams of the preferred embodiments of the invention.
[0320] The drawings of the figures are not to scale. Generally, similar elements are denoted by similar references in the figures. The presence of reference numbers in the drawings may not be considered as limiting, even when such numbers are indicated in the claims.
[0321] Example of an Experimental Device for Implementing the Method of the Invention
[0322] The experiments carried out to compare the models described above were performed on samples of polished stainless steel 316L and 316 and TiCr6Sn4. The tests were performed in air using a Satsuma HP2 (Amplitude Systems) femtosecond laser with a pulse duration of about 330 fs, a radiation wavelength of 1030 and a maximum power of 20 W at 500 kHz. The beam was focused on the surface of the samples using a telecentric lens with a focal length of 100 mm, producing a spot radius of about 10 m determined using the D.sup.2 method.
[0323] Morphological and topographical analysis of the processed samples was performed using a confocal optical microscope (Olympus LEXT OLS4100).
[0324] The D.sup.2 method was used to calculate the threshold fluence (F.sub.th) values. To determine the value of , several line scans with increasing pulse energy were produced and their depths measured. The value applied in the model was varied until the best match with the experimental results was obtained.
[0325] Preferably, the greater the amount of results provided to the database, the more accurately the laser machining parameters are determined.
DETAILED DESCRIPTION OF CERTAIN EMBODIMENTS OF THE INVENTION
[0326] FIG. 1 shows an intensity profile of a Gaussian laser beam. The center of the beam is defined by the value 0 along the abscissa axis defining a beam radius r and a maximum intensity I.sub.0 is observed at the center of the beam. A beam radius .sub.0 relative to the beam center is defined for a beam intensity value equal to I.sub.0/2.
[0327] FIG. 2 shows a crater matrix to be machined in order to implement the ablation threshold determination by the D.sup.2 method. FIG. 2 describes a preferred embodiment for producing a crater matrix with the laser machining machine of the invention. The matrix shown in FIG. 2 shows a number of pulses for each of the columns between 10 and 110 and beam energies between 10% and 100% of a maximum beam value that can be provided by the laser machining machine of the invention. The realization of such a matrix of craters while respecting the beam energies and the numbers of pulses indicated, followed by the topological and/or dimensional characterization of the craters of this matrix thus enables to define an ablation threshold, an incubation coefficient or even a beam radius. Preferably the production of the crater matrix should be performed with increasing pulse numbers and pulse energies, as illustrated by the arrows in FIG. 2. In order to have sufficient experimental points, at least 20 energy values and several pulse number values less than 100 pulses are required. The diameter of each crater can be measured from images obtained with an optical or electron microscope or a profilometer.
[0328] FIG. 3a shows a plurality of pulses along a line. A pulse is represented by a circle with its center on a line. The centres of each circle, i.e. the points where the maximum intensity of each pulse is maximum I.sub.0, are spaced along a line of: x=v/PRR. v being the speed of movement of the beam and PRR being a number of pulses per unit time. FIG. 3a shows that for pulses along a line, with a distance between each circle less than the radius of each circle, then statistically more pulses reach the centre of the line than the edges of the line.
[0329] FIG. 3b shows a fluence profile obtained transverse to the line described in FIG. 3a. This fluence profile shows that the fluence at the center of the line is much higher than at the edges of the line.
[0330] FIG. 3c shows a distribution of the pulse number N transversaly to the line described in FIG. 3a. As expected from the description in FIG. 3a, statistically, a greater number of pulses reach the center of the line than the edges of the line.
[0331] FIG. 4a shows as in FIG. 3a a plurality of pulses along a line. FIG. 4b, shows the groove depth obtained transversaly to the line described in FIG. 4a with the following experimental parameters: [0332] material: steel; [0333] w.sub.0=12.5 m; [0334] F=3 J/cm.sup.2; [0335] V=50 mm/s; [0336] PRR=200 kHz.
[0337] FIGS. 7c and 7d represent the experimental results compared to the corresponding models obtained by the experiments described in FIGS. 7a and 7b. FIG. 7c shows the maximum scan depths obtained on a steel substrate as a function of the maximum fluence. FIG. 7d shows the maximum depths of scans obtained on a steel substrate as a function of the offset speed or the scan speed.
[0338] FIGS. 8a and 8b show point clouds obtained for experiments described in FIGS. 6a and 6b for which a regression according to an algorithm of the invention allows to determine the parameter .sub.0.
[0339] FIG. 9 shows the experimental results of the D-scan method for two pulse frequencies of 5 KHz and 100 KHz. The regressions for these two sets of experimental results are also illustrated by the lines. The parameters of these two regressions are indicated by Gama, Beta and Alfa.
[0340] FIG. 10 shows a parameter for the interaction of a polarized light beam with a steel substrate having the following properties: n=0.9+2.25i. FIG. 10 shows the interraction parameter for incidence angles between 0 and 90, for a polarization s and for a polarization p. The curve representing the polarization p is the one with a maximum at 100% for an incidence angle of 90.
[0341] FIG. 11 shows a simulation of the beam width w.sub.z at its minimum width. The minimum beam width is shown at a distance z in a propagation direction z having a value of 0. The simulated beam profile shows the divergence of the beam in the direction z.
[0342] FIG. 12 shows experimental results and simulations for two values for the maximum depth as a function of beam fluence. For peak fluence values less than 2 J/cm.sup.2 and greater than 4.5 J/cm.sup.2, the experimental results vary from the two simulations having for 6 value 32 nm and 34 nm. This figure illustrates very well the need for a learning algorithm in order to obtain a better simulation of the parameters or a better prediction of the machining parameters when fluences lower than 2 J/cm.sup.2 and higher than 4.5 J/cm.sup.2 are used.
[0343] FIG. 13, showing the maximum depth as a function of beam scanning speed, shows a good correlation between the experimental ablation results and the beam scanning speed. The experimental results also show a relatively good robustness when varying the parameter from 32 nm to 34 nm.
[0344] FIGS. 14a and 14b show experimental and simulated profiles obtained on stainless steel 316L with a maximum fluence of 2.14 J/cm.sup.2, scanning speed 230 mm/s (FIG. 14a) and 1.25 J/cm.sup.2 and scanning speed 50 mm/s (FIG. 14b, polarization p and 250 kHz frequency). A good correspondence between experimental and simulated results is observed with respect to the maximum depth of ablation as well as the ablation profile. On the other hand, the redeposition of material at the edges of the ablation is not taken into account by the simulation. The prediction and learning method of the invention allows better consideration of the redeposition of material at the edges of the ablation performed experimentally.
[0345] FIGS. 15a, 15b and 15c represent experimental and simulated profiles obtained on stainless steel 316L with a maximum fluence of 1.92 J/cm.sup.2, scanning speed 100 mm/s, frequency 100 kHz, the FIG. 15a is obtained after 17 scans, the FIG. 15b after 33 scans and the FIG. 15c after 65 scans. The same observations as for FIGS. 14a and 14b can be observed.
[0346] FIGS. 16a and 16b show in FIG. 16a a cross section of a scan and the simulated correspondent (FIG. 16b) obtained on stainless steel 316L with a maximum fluence of 2.9 J/cm.sup.2, scanning speed of 500 mm/s, frequency of 100 kHz and 500 passes with the beam tilted of 15 in relation to the normal to the surface. The model of the invention allows a good taking into account of the angle of incidence of the beam incident on the material to be machined. As shown in FIG. 16a, there is a good correspondence between the micrography of the machining (ablation) and the simulation which is superimposed.
[0347] FIGS. 17a and 17b show machining profiles obtained with an angle of incidence normal to the surface of the workpiece and with an angle of incidence describing an angle 90- in relation to the surface of the workpiece. For example, these schematic representations with different angles of incidence in relation to the surface of the workpiece enable to calculate the local fluence.
[0348] FIG. 18 shows a schematic representation of the angles of a light beam in experiments for the calculation of the local fluence under a non-normal incidence. In particular, this schematic representation illustrates the divergence of the beam having a non-normal incidence with a substrate to be machined. The peculiarities caused by this divergence of the focused beam are thus illustrated.
[0349] FIGS. 19a and 19b show a comparative example between a simulation of a machining profile according to the prior art (FIG. 19a) and a simulation according to a particular embodiment of the invention (FIG. 19b) for a machining performed with a machining laser beam having a non-zero angle of attack with the normal to the surface of the material to be machined. Both FIGS. 19a and 19b are multipass profiles obtained on stainless steel (in both cases =32 nm, F.sub.th(1)=0.1 J/cm.sup.2, S=0.8, n=0.9-2.25i) for w.sub.0=14 m, PRR=200 kHz, x=0.75 m, F.sub.0=8 J/cm.sup.2 and an angle of incidence of 8. The numbers in the figures show the number of passes of the machining laser beam for each of the simulated two-dimensional machining profiles. The simulations were therefore carried out for 50, 100, 200, 300, 402 passes. The particular embodiment of the invention which allows the processing of machining laser beams having a non-zero angle in relation to the surface of the material allows a very representative simulation of how the machining would be using the same laser parameters. In the case of the simulation using prior art methods, it is observed that a high number allows a machining flank to be approached perpendicular to the surface of the material but without ever reaching it, which is problematic for the simulation of machining with a conicity of the flanks in relation to the surface of the material which is well controlled. Indeed, the precession or other machining laser beam modules that allow machining with a well-controlled beam angle require suitable simulation tools. The particular embodiment of the invention seems to be one of them. The ordinate axis in FIGS. 19a and 19b represents a simulated ablation depth expressed in m. The abscissa axis of FIGS. 19a and 19b represents a dimension y corresponding to a dimension parallel to the surface of the material to be machined.
[0350] FIG. 20a shows a schematic representation of a machining with a precessional laser machining beam. It shows a material to be machined whose surface corresponds to the coordinate z=0 in relation to the axis z essentially perpendicular to the surface of the material. The laser beam is rotating, especially during precession as indicated by the circular arrow indicating a direction of rotation. The beam comes from the side of this circular arrow and is directed towards the material to be machined. The machining laser beam is shown as passing through the material so as to illustrate the trajectory of the precession laser beam. Thus the laser beam describes a point BFG located on a plane BFG essentially parallel to the surface of the material. As can be seen, the precession beam always passes through the same point at the point BFG. Furthermore, the plane BFI is the plane in which the precessional laser beam is focused. The precession laser beam rotates around an axis represented as being normal to the surface of the material. The precession laser beam therefore describes an angle for all precession positions of the beam with respect to the axis of rotation of the beam.
[0351] FIGS. 20b, 20c, 21b and 21d show machining simulations according to a particular embodiment of the invention for a precession laser beam as shown schematically in FIG. 20a. FIGS. 20b, 20c, 21b and 21d show two-dimensional machining profiles along the axes y, z, with the ordinate axis representing the dimension z (depth of ablation) and the abscissa axis representing the dimension y. FIG. 20b corresponds to a simulation according to the particular embodiment of the invention allowing the simulation of a precession laser machining beam with BFG=150 m and BFI=50 m. Similarly, FIG. 20c was obtained with BFG=100 m and BFI=500 m. The simulations in FIGS. 20b and 20c were obtained with the laser parameters: w.sub.0=10 m, PRR=500 kHz, E.sub.p=20 J, =30 000 rpm, =4.5, machining time=50 ms. FIG. 21b corresponds to a simulation according to the particular embodiment of the invention allowing the simulation of a precession laser machining beam with BFG=0 m and BFI=400 m. Similarly, FIG. 21d was obtained with BFG=100 m and BFI=500 m. The simulations in FIGS. 21b and 21d were obtained with the laser parameters: w.sub.0=12 m, PRR=100 kHz, E.sub.p=27 J, =30 000 rpm, =4.5, machining time=50 ms.
[0352] FIGS. 21a and 21c show machining results in the form of cross-sectional cutting images of a machining, the machining simulations shown in FIG. 21b and FIG. 21c are superimposed on the machining results respectively. It is thus observed that the machining simulations allow a good anticipation of the machining obtained on the basis of the same parameters for a precession laser machining beam.
[0353] FIG. 22a shows a schematic representation of the points of impact of a precessing or rotating machining laser beam, FIG. 22b shows a detail of FIG. 22a. FIG. 22a, shows the pulse superposition of a precession laser machining beam. Around an axis represented here by a point because it is a projection of the laser pulses in a plane, preferably on the surface of the material. The axis of rotation of the precession laser beam is therefore perpendicular to the surface of the material and therefore represented by a point. Each circle represents one pulse. The center of each pulse is located on a precession circle centered on the axis of rotation of the machining laser beam. The precession circle has a radius r.sub.p. The distance between the center of each adjacent circle on the precession circle is dx. Preferably dx is constant during machining. The center of each adjacent circle describes an angle in relation to the precession axis. FIG. 22b represents a detailed view of FIG. 22a in which, the center of the circles corresponding to a pulse is named n.sub.0, n.sub.1, n.sub.2, n.sub.3, . . . . The distance between n.sub.0 and n.sub.1 is d.sub.1, the distance between n.sub.0 and n.sub.2 is d.sub.2, the distance between n.sub.0 and n.sub.3 is d.sub.3.
[0354] FIGS. 23 to 31 show flow diagrams describing the individual components of a laser machining system and the possible steps or decisions that can be taken. In particular, these flow diagrams show the means allowing the flow of information and allowing to achieve the machining of a material with optimum parameters or allowing to enrich a material database and/or functions learned by a machine learning.
[0355] FIG. 23 represents a Functional Graph for Step and Transition Control (Grafcet) or functional flowchart of the different actions available according to the invention. In particular, this flowchart shows that the method of the invention may decide, in the event that the parameters of a material are not available in a database, to resort to the launching of a predefined test machining in order to allow the acquisition of data in relation to the said material to be machined.
[0356] FIG. 23 shows the following procedure: a material to be machined is communicated to the laser machining system, which queries whether information about this material is present in a material database or in a database of learned functions. In the case of a positive answer, it is then possible to choose between manual or automated determination of the machining parameters.
[0357] If the machining parameters are determined in an automated manner, the central unit sends a request for knowing the machining result sought. This result to be achieved can also be communicated when the material to be machined is communicated. On the basis of the result to be achieved (target result) defined by the user, the laser machining system and in particular the central unit collects information about the characteristics of the machining device as well as the information available in one of the databases about the information of the material to be machined. The learning central unit then enables the exploitation of the means for modelling and/or machine learning by taking into account the information of the material to be machined, the characteristics of the machining device and the machining result sought. The learning central unit can then generate optimal machining parameters. These optimal machining parameters are then transmitted to the laser machining device in order to start the laser machining of the material to be machined according to the result to be achieved. It is possible that the learning central unit is connected to the laser machining device via a network connection so that the learning central unit can be relocated in relation to the laser machining device.
[0358] If the machining parameters are determined manually, the learning central unit sends a request to obtain parameters from the user for a result to be achieved. Preferably the parameters transmitted by the user are those parameters which the user believes to be the optimum results. However, several iterations are often necessary for the operator of the laser machining system to specify parameters that will achieve the result to be achieved. When parameters are communicated by the user, then modeling means are implemented. The modelling means preferably comprises a model including an algorithm allowing an estimation of the machining on the basis of the parameters provided. Thus simulation means coupled to the modeling means enable to carry out a simulation of the result expected to be obtained on the basis of the parameters communicated. The operator can then compare the simulation of the result on the basis of the parameter with the result to be achieved. If the operator considers that the simulation does not conform (or sufficiently conforms) to the result to be achieved and communicates this to the learning central unit, then the learning central unit offers a choice between manual or automated determination of the machining parameters. The operator can then decide to test several different machining parameters manually until a simulation of the result expected on the basis of the communicated parameters conforms to the result to be achieved. When the operator is satisfied with the simulation of the result (machining) expected on the basis of the communicated parameters, then he can decide to start machining on the basis of the parameters used in the last modeling and simulation. The operator can choose at any time to use the automated parameter search so that the laser machining system uses the automated mode as described above. The learning central unit then sends a request to know the machining result sought and determines the optimal machining parameters as described above and as represented by the flowchart. During the modeling/simulation step in manual mode, the modeling/simulation means have access to the material database as well as to the characteristics of the machining device.
[0359] If the query as to whether information on the material to be machined is present in a material database or in a database of learned functions is answered in the negative, the laser machining system then defines predefined machining tests. These predefined machining tests enable the laser machining system to produce a predefined machining on a material, preferably the material in the material database being identical to the material to be machined, the machining being analysed by the analysis unit and the analysis results being transmitted to the learning central unit.
[0360] The results of the analysis are either communicated to modeling means for extracting physical or material light interaction parameters specific to the material to be machined. Modeling means for receiving information about the characteristics of the machining device. The parameters determined by the modelling means are then machining parameters which are communicated to the material database in order to enrich it. When the material to be machined is known in the database, it is then possible to continue the machining process and in particular towards the stage proposing a manual (this invention) or automated parameter search.
[0361] When the analysis results are reported to the learning central unit, the learning central unit having access to the characteristics of the machining device enables to generate machining parameters in the form of machining data comprising machining results obtained according to the laser machining parameters used. This machining data is communicated to the learning database. In a preferred embodiment, the material database and the learning database are common and all machining parameters are then accessible from either database. Thus, when the question arises as to whether the material is known in the database, having the characteristics of the material to be machined in the form of physical parameters or in the form of a learning database enables to answer this question in the affirmative.
[0362] The embodiment described in the previous paragraph is shown in FIG. 24. Thus, the same database allows to gather for the same material, the physical characteristics of the materials as well as the information in relation to the laser machining system so as to form a learning database for the implementation of an machine learning algorithm. Preferably the physical characteristics of the materials as well as the information for the learning represent the learning database. The operation of the flowchart described for the FIG. 23 applies to the FIG. 24 mutatis mutandis.
[0363] FIG. 25 shows an embodiment of the invention for determining the optimum machining parameters in a deterministic regime for a laser machining system. In particular, this embodiment is based primarily on the use of known algorithms as described by the invention. This embodiment also allows the simulation of ablations which would be obtained for laser parameters obtained by the algorithms of the invention or defined by a user. This figure describes the embodiments detailed in FIGS. 26 to 28. These embodiments are derived from the flowcharts shown in FIGS. 23 and 24.
[0364] FIG. 26 shows an embodiment of the invention allowing the feeding of physical material characteristics to the material database, for example in the case where a predefined test has been machined and for which the results are communicated to the modeling means as described for FIGS. 23 and 24. The arrow connecting the material database to the modelling means can be interpreted in either direction. The purpose of this embodiment is the determination and the storage of experimental parameters in a database. From a machining performed by a laser machining device, the results of this machining are communicated to a central unit which allows the experimental data to be processed in order to extract physical parameters of materials.
[0365] FIG. 27 shows an embodiment of the invention allowing the simulation without necessarily requiring the use of a laser machining device. Physical characteristics of materials previously stored in the material database allow a simulation of the results that could be obtained for machining according to user parameters and characteristics of the machining device.
[0366] FIG. 28 shows an embodiment of the invention for determining optimal machining parameters from the results to be achieved, the characteristics of the laser machining device and from a database comprising physical characteristics of materials, for example observed and stored in the material database during previous machinings.
[0367] FIG. 29 shows an embodiment of the invention and in particular an embodiment of the invention based on machine learning means for taking into account physical characteristics of materials or functions learned by the machine learning. A learned function is the result of the learning of a learning algorithm itself included in a learning function. This information is preferably stored in a database or in several databases. This embodiment is preferably based on a non-deterministic optimal parameter determination regime, i.e. not necessarily using modelling means comprising an algorithm defining deterministic functions.
[0368] Preferably, the embodiment detailed in FIG. 29 includes a learning function whose learned functions (resulting from the learning) are stored in a database. This embodiment allows the method of the invention to learn and refine the optimal parameters based on past experiences and observations. For example, experiments conducted on other laser machining devices are taken into account in learning the algorithm for determining optimal machining parameters by accessing the learning database. This embodiment also allows the performance of a specific experiment for collecting experimental data in order to obtain optimal machining parameters for a target machining. For example, predefined tests are available in the laser machining system of the invention and are applicable when it is considered necessary to obtain machining with high fidelity in relation to a target machining.
[0369] FIG. 30 shows a particular embodiment of the FIG. 29 allowing machine learning and recording the learning functions as learned functions. FIG. 30 shows the feeding of the database of learned functions by the machine learning in non-deterministic regime. Preferably, this embodiment implements a predefined machining test, which can be started on a selected material. For example, a predefined machining test is defined in FIGS. 2, 3a, 6a, 6b, 7a, 7b. These tests allow, after characterization of the machinings obtained by the analysis unit, the machine learning and the determination of a learned function. The results of these tests and of the learned functions generated by taking these tests into account are then stored in a learning database.
[0370] FIG. 31 shows an embodiment for machining a workpiece using the optimum parameters generated by the central unit or by the learning means on the basis of the learned functions corresponding to the material to be machined. This embodiment allows to obtain machining results very close to the desired target machining and very quickly. The FIG. 31 shows the prediction of the optimum machining parameters in non-deterministic regime, i.e. by learning the learning function from the learning database on the basis of the learning algorithm.
[0371] The present invention has been described in relation to specific embodiments, which have a purely illustrative value and should not be considered as limiting. In general, the present invention is not limited to the examples illustrated and/or described above. The use of the verbs comprise, include, comsist, or any other variant, as well as their conjugations, can in no way exclude the presence of elements other than those mentioned. The use of the indefinite article a, an, or the definite article the to introduce an element does not exclude the presence of a plurality of such elements. Reference numbers in claims shall not limit their scope.
[0372] In summary, the invention may also be described as follows.
[0373] A method for determining laser machining parameters for the machining of a material with a laser machining system comprising inter alia the steps of:
[0374] a) providing said central unit with a learning machining function capable of learning on the basis of said plurality of machining data samples, said learning machining function comprising an algorithm capable of defining for said machining result sought and for said machining system the following laser machining parameters: [0375] a polarization, [0376] an E.sub.p pulse energy, [0377] a diameter at the focal point w, [0378] an order of the gaussian p, [0379] a pulse repetition rate PRR n, [0380] a wavelength;
[0381] b) making said learning machining function to learn so that said laser machining system can machine said material to be machined according to the machining result sought.
