METHOD AND DEVICE FOR OPERATING A LASER MATERIAL PROCESSING MACHINE

Abstract

A computer-implemented method for operating a laser material processing machine. Process parameters are varied with the aid of Bayesian optimization until a result of the laser material processing is sufficiently good. The Bayesian optimization taking place with the aid of a data-based process model, and it being taken into consideration during the variation of the process parameters how probable it is that a variable which characterizes a quality of the result is within predefinable boundaries.

Claims

1. A computer-implemented method for operating a laser material processing machine, the method comprising: varying process parameters using Bayesian optimization until a result of the laser material processing is sufficiently good, the Bayesian optimization taking place using a data-based process model, wherein, during the variation of the process parameters, it is taken into consideration how probable it is that a variable which characterizes a quality of the result is within predefinable boundaries.

2. The method as recited in claim 1, wherein in an acquisition function, as a function of which the variation of the process parameters is ascertained, a probability is taken into consideration, which characterizes how probable it is that the result is within the predefinable boundaries.

3. The method as recited in claim 2, wherein the probability is ascertained based on a data-based quality model.

4. The method as recited in claim 3, wherein the data-based quality model is configured to output parameters which characterize a statistical prediction of the result to be expected.

5. The method as recited in claim 4, wherein it is taken into consideration during the variation of the process parameters how probable it is that further variables, which each characterize further qualities of the result, are within respective predefinable boundaries, each of the respective probabilities being ascertained using a respective further data-based quality model.

6. The method as recited in claim 5, wherein the data-based process model and/or the data-based quality model and/or one or multiple of the respective further data-based quality models, are trained as a function of the ascertained result which results during activation of the laser material processing machine and/or of the ascertained variable which results during the activation of the laser material processing machine and/or the further variables which result during the activation of the laser material processing machine.

7. The method as recited in claim 5, wherein the data-based model and/or the data-based quality model and/or one or multiple of the respective further data-based quality models is also trained as a function of an estimated result.

8. The method as recited in claim 7, wherein: in a first phase, the data-based model and/or the data-based quality model and/or one or multiple of the respective further data-based quality model(s) is trained as a function of estimated results, in a second phase, the data-based model and/or the data-based quality model and/or one or multiple of the respective further data-based quality models is trained as a function of the ascertained result which results during the activation of the laser material processing machine and/or the ascertained variable which results during the activation of the laser material processing machine and/or the ascertained further variables which result during the activation of the laser material processing machine.

9. The method as recited in claim 8, wherein the data-based model to be trained is given as a sum of a first regression model and a second regression model, the first regression model being trained in the first phase, and the second regression model being trained in the second phase.

10. The method as recited in claim 9, wherein the second regression model is a Gaussian process model.

11. The method as recited in claim 1, wherein, subsequent to setting the process parameters, the laser material processing machine is operated with the process parameters set.

12. A test bench for a laser material processing machine, the test bench configured to: vary process parameters using Bayesian optimization until a result of the laser material processing is sufficiently good, the Bayesian optimization taking place using a data-based process model, wherein, during the variation of the process parameters, it is taken into consideration how probable it is that a variable which characterizes a quality of the result is within predefinable boundaries.

13. A non-transitory machine-readable memory medium on which is stored a computer program for operating a laser material processing machine, the computer program, when executed by a computer, causing the computer to perform the following steps: varying process parameters using Bayesian optimization until a result of the laser material processing is sufficiently good, the Bayesian optimization taking place using a data-based process model, wherein, during the variation of the process parameters, it is taken into consideration how probable it is that a variable which characterizes a quality of the result is within predefinable boundaries.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0067] FIG. 1 schematically shows a configuration of a laser drilling machine in accordance with an example embodiment of the present invention.

[0068] FIG. 2 schematically shows a configuration of a laser welding machine in accordance with an example embodiment of the present invention.

[0069] FIG. 3 schematically shows a configuration of a test bench in accordance with an example embodiment of the present invention.

[0070] FIG. 4 shows one specific embodiment for operating the test bench in a flowchart in accordance with the present invention.

[0071] FIG. 5 shows one specific embodiment for operating the test bench in a flowchart in accordance with the present invention.

[0072] FIG. 6 shows one specific embodiment of a sub-aspect of one of the two aforementioned methods in a flowchart in accordance with the present invention.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

[0073] FIG. 1 schematically shows a configuration of a laser drilling machine 1. An activation signal A is provided by an activation logic 40 to activate a laser 10a. The laser beam strikes a material piece 12 where it generates a borehole 11.

[0074] FIG. 2 schematically shows a configuration of a laser welding machine 2. An activation signal A is also provided by an activation logic 40 here to activate a laser 10b. The laser beam strikes two material pieces 13, 14 and generates a weld seam 15 there.

[0075] A laser cutting machine (not shown) is also analogously possible.

[0076] FIG. 3 schematically shows a configuration of a test bench 3 for ascertaining optimal process parameters x. Instantaneous process parameters x are provided by a parameter memory P via an output interface 4 of the laser material processing machine, such as e.g. laser drilling machine 1 or laser welding machine 2. The machine carries out the laser material processing as a function of these provided process parameters ϕ. Sensors 30 ascertain sensor variables S, which characterize the result of the laser material processing. These sensor variables S are provided via an input interface 50 to a machine learning block 60 as quality properties y.sub.exp.

[0077] In the exemplary embodiment of the present invention, machine learning block 60 encompasses a Gaussian process model, which, as is illustrated in FIG. 4 or FIG. 5, is trained as a function of the provided quality properties y.sub.exp. As a function of the Gaussian process model, varied process parameters x′ may be provided, which are stored in parameter memory P.

[0078] As an alternative or in addition to the provision via output interface 4, process parameters x may also be provided to an estimation model 5, which provides estimated quality properties y.sub.sim to machine learning block 60 in place of the actual quality properties y.sub.exp.

[0079] In the exemplary embodiment of the present invention, the test bench includes a processor 45, which is configured to play back a computer program which is stored on a computer-readable memory medium 46. This computer program includes instructions which prompt processor 45 to execute the method illustrated in FIG. 4 or 5 when the computer program is being played back. This computer program may be implemented in software, or in hardware, or in a mixed form made up of hardware and software.

[0080] FIG. 4, in a flowchart, shows an exemplary method for operating test bench 3 in accordance with an example embodiment of the present invention. The method begins 100 in that initial process parameters x.sub.init are provided as process parameters x, and experimental data recorded thus far are initialized as an empty set. Optionally, process parameters x are predefined using a design-of-experiment method and, as is described in greater detail hereafter, laser material processing machine 1, 2 is activated with the aid of these process parameters x, variables y.sub.exp are ascertained, and Gaussian process model GP is trained using the experimental data thus ascertained.

[0081] In the case of laser drilling, these process parameters x in one exemplary embodiment encompass a pulse duration, a focus position resolved as a function of time via a characteristic map and/or a focus size and/or a pulse repetition frequency and/or a circular path diameter resolved as a function of time via a characteristic map as a function of time and/or a circular path frequency and/or an attack angle resolved as a function of time via a characteristic map and/or a drilling duration and/or a pulse energy resolved as a function of time via a characteristic map and/or a wavelength and/or parameters which characterize a process protective gas, such as, e.g., a process gas type or a process gas pressure. In the process, the described circular path is a conventional feature in many drilling methods, for example during spiral drilling or during trepanning drilling.

[0082] In the case of laser welding, these process parameters x encompass a laser power resolved as a function of time and/or location via characteristic maps and/or a focus diameter and/or a focus position and/or a welding speed and/or a laser beam inclination and/or a circular path frequency during laser wobbling and/or parameters characterizing a process protective gas.

[0083] Using the instantaneous process parameters x, laser material processing machine 1, 2 is activated 110, and variables y.sub.exp are ascertained 120, which characterize the actual result of the laser material processing.

[0084] In the case of laser drilling, these variables y.sub.exp, in one exemplary embodiment, encompass variables which characterize the size of borehole 12 and/or the circularity of borehole 12 and/or the shape of a wall of borehole 12 and/or the presence of melt deposits and/or an amount of droplet expulsion during the drilling process and/or a rounding of the edges of borehole 12 and/or the productivity.

[0085] In the case of laser welding, these variables y.sub.exp, in one further exemplary embodiment, encompass variables which, along weld seam 15, characterize a minimal weld seam depth and/or a minimal weld seam width and/or the productivity and/or an extent of weld spatter and/or a number of pores and/or a welding distortion and/or residual welding stresses and/or welding cracks.

[0086] A cost function K is evaluated 130 as a function of these variables, as it may be given, for example, by equation 1, variables y.sub.exp being provided as quality properties q.sub.i and corresponding target values of these variables q.sub.i,Target.

[0087] A cost function K which penalizes deviations of the quality properties from the target values is also possible, in particular, if they exceed a predefinable tolerance distance, and which rewards a high productivity. The “penalizing” may, e.g., be implemented by a high value of cost function K, the “rewarding” accordingly by a low value.

[0088] Then, it is ascertained whether cost function K indicates that instantaneous process parameters x are sufficiently good, in the event that a penalization denotes a high value, and a reward denotes a low value, by checking whether cost function K falls below 140 a predefinable maximum cost value. If this is the case (“Yes”), the method ends 150 with the instantaneous process parameters x.

[0089] If this is not the case (“No”), data point x,y.sub.exp thus ascertained, made up of process parameters x and associated variables y.sub.exp characterizing the result, is added 160 to the ascertained experiment data, and hyperparameters θ.sub.0,θ.sub.1 of Gaussian process model GP are adapted in such a way that the probability that the experiment data result from Gaussian process model GP is maximized.

[0090] Then 170, an acquisition function is evaluated, as it is illustrated by way of example in formula 7, and in this way new process parameters x′ are ascertained. The method then branches back to step 110.

[0091] FIG. 5, in a flowchart, shows one further exemplary method for operating test bench 3. Steps 100, 110, 130, 140, 150 are the same as illustrated in FIG. 4; a separate description is thus dispensed with.

[0092] In step 120b, which replaces step 120 of the method illustrated in FIG. 4, several of variables y.sub.exp determined there are in each case provided separately as limited variables custom-character , , , . . . , which each are to be in a limited interval: ∈ [.sub.0−δ,.sub.0+δ], ∈ [.sub.0−ϵ,.sub.0+ϵ], ∈ [.sub.0−ϕ,.sub.0+ϕ], . . . .

[0093] In step 160b, which replaces step 160 of the method illustrated in FIG. 4, in addition to the step described in step 160, additionally for each of limited variables custom-character , , a respective data point x,, x,, x,, . . . is added to the respective ascertained test data, i.e., assigned to each of limited variables , , , and analogously to the training of Gaussian model GP, a dedicated Gaussian process model ,GP, is trained for each of limited variables , custom-character , .

[0094] In step 170b, which replaces step 170 of the method illustrated in FIG. 4, in addition to the evaluation of the acquisition function described for 170. For this purpose, as described above, a so-called “expected improvement” function may be evaluated and maximized, as illustrated in formula 7. Furthermore, a predefinable probability function custom-character (x) is provided, which characterizes a probability as to whether or not selected process parameters x of laser material processing machine 1, 2 result in a satisfactory result, i.e., whether limited variables D are actually in the associated interval, i.e., ∈ [.sub.0−δ,.sub.0+δ]. For the further limited variables custom-character , , . . . , corresponding probability functions p(x), (x) are provided. In the exemplary embodiment, probability function (x) is ascertained from a variance σ.sup.2 of Gaussian process model . For this purpose, lower boundaries .sub.0−δ and upper boundaries .sub.0+δ are provided for the variables ascertained by Gaussian process model custom-character and, for example with the aid of numerical integration, it is ascertained how great the probability is that the variables ascertained by Gaussian process model are between lower boundary .sub.0−δ and upper boundary .sub.0+δ, i.e., (x):=p(.sub.0−δ<(x)<.sub.0+δ). The procedure may be analogous for probability functions p custom-character (x), (x). The described acquisition function may now be additionally multiplied with the product of the ascertained probabilities (x).Math.p(x).Math.(x) . . . , and new process parameters (x′) are thus ascertained. The method then branches back to step 110.

[0095] FIG. 6 illustrates, in a flowchart, the sequence of a method as it may be used for training Gaussian process models GP,GP custom-character ,GP,. It is illustrated hereafter by way of example based on Gaussian process model GP and may be applied accordingly to any of the others.

[0096] The method is made up of the first Gaussian process GP.sub.0 and the second Gaussian process GP.sub.1, which together additively yield Gaussian process model GP, i.e.,

GP(x)=GP.sub.0(x)+GP.sub.1(x).

[0097] The method is made up of a first phase A and a second phase B. In each of the phases, the method illustrated in FIG. 4 or that illustrated in FIG. 5 may be used, in first phase A the simulatively ascertained estimated variables y.sub.sim taking the place of actual variables y.sub.exp.

[0098] In first phase A, initially process parameters x are predefined 6000. Then, estimated variables y.sub.sim are ascertained simulatively 6010.

[0099] In the case of laser drilling, this may take place, for example, as follows with the aid of a physical model: For a radius r of borehole 11 along a depth coordinate z, r(z) is numerically ascertained as the solution to equation

[1−R(r,z,a,θ)].Math.cos θ.Math..Math.F.sub.0(r,z)−{tilde over (F)}.sub.th=0 (8)

with

[00001] $\begin{matrix} 1 - R = \frac{1}{2} .Math. (\frac{4 n \cos θ}{(n^{2} + k^{2}) + 2 n \cos θ + \cos^{2} θ} + \frac{4 n \cos θ}{(n^{2} + k^{2}) \cos^{2} θ + 2 n \cos θ + 1}) & (9) \\ F_{0} (r, z) = \frac{2 Q}{π w^{2} (z)} .Math. \exp (- \frac{2 r^{2}}{w^{2} (z)}) & (10) \\ w (z) = \frac{d_{F o k}}{2} \sqrt[]{1 + {(\frac{z}{l_{Rayleigh}})}^{2}} & (11) \\ \tan α = \frac{r}{w (z)} \frac{d w (z)}{d z} & (12) \end{matrix}$

Where:

[0100] n=n+ik is a predefinable complex refractive index of material piece 12, with refractive index n and coefficient of extinction k; [0101] {tilde over (F)}.sub.th is a predefinable ablation threshold fluence of material piece 12; [0102] Q is a predefinable pulse energy of laser 10a; [0103] d.sub.Fok is a predefinable focus diameter of laser 10a; [0104] l.sub.Rayleigh is a predefinable Rayleigh length of laser 10a; [0105] R is an ascertained reflectivity of material piece 12; [0106] α is an ascertained angle of the local beam propagation direction; [0107] θ is a predefinable relative angle between the incident laser beam and the surface normal of material piece 12; [0108] F.sub.0 is an ascertained irradiated fluence of laser 10a; and [0109] w(z) is an ascertained local beam radius.

[0110] The prediction of several quality properties, such as a presence of melt deposits and/or an amount of droplet expulsion during the drilling process, is not possible using this physical model. For the ascertainment of these quality properties, an empirical model may be predefined in the process, for example.

[0111] As an alternative or in addition, it is possible that at least several of the quality properties cannot be reliably calculated for all process parameters x. It is possible that it is checked whether the instantaneous process parameters x are within a predefinable area, and that, when this is not the case, the quality properties are ascertained with the aid of one of the aforementioned approaches.

[0112] In the case of laser welding, the ascertainment of estimated variables y.sub.sim may, for example, take place as follows using a physical model:

[00002] $\begin{matrix} T (x, y, z) - T_{0} = \frac{1}{2 π λ h} \exp - \frac{v (x - x_{0})}{2 a} (q_{net} K_{0} (\frac{vr}{2 a}) + 2 Σ_{m = 1} \cos \frac{m_{p i} z}{h} K_{0} (\frac{vr}{2 a} \sqrt{1 + {(\frac{2 m_{π} a}{v h})}^{2}}) I_{m}) & (13) \end{matrix}$

with

[00003] $\begin{matrix} r = \sqrt{{(x - x_{0})}^{2} + y^{2}} & (14) \\ I_{m} = \int_{0}^{h} q_{1 net} (z) \cos \frac{m_{π} z}{h} dz & (15) \end{matrix}$

and the parameters
T.sub.0—a predefinable ambient temperature;
x.sub.0—a predefinable offset of the beam of laser 10b in relation to the origin of a coordinate system movable with laser 10b;
λ—a predefinable thermal conductivity of material pieces 13, 14;
a—a predefinable temperature conductivity of material pieces 13, 14;
q.sub.net—a predefinable power of laser 10b;
q.sub.1net—a predefinable power distribution of laser 10b along a depth coordinate of material piece 10b;
ν—a predefinable velocity of laser 10b;
h—a predefinable thickness of material pieces 13, 14; and Bessel function

[00004] $K_{0} (ω) = \frac{1}{2} \int_{- \infty}^{\infty} \frac{e^{i ω t}}{\sqrt{t^{2} + 1}} d t$

as well as an ascertained temperature distribution T(x,y,z). A width or a depth of the weld seam may be ascertained (e.g., via the ascertainment of isotherms at a melting temperature of a material of material pieces 13, 14) from the temperature distribution.

[0113] Then, first Gaussian process GP.sub.0, which trains 6020 with the aid of estimated results y.sub.sim.

[0114] Thereafter, it is checked whether an abort criterion is reached 6030. For example, it may be checked whether a cost function K ascertained as a function of estimated result y.sub.sim falls below a predefinable threshold value. If the abort criterion is met, second phase B follows, otherwise the method branches back to 6000.

[0115] In second phase B, the method illustrated in FIG. 4 or FIG. 5 is executed 6040, in place of Gaussian process model GP second Gaussian process GP.sub.1 being trained with the aid of actual results y.sub.exp and the associated process parameters x, and in particular, in that actual results y.sub.exp are ascertained by the difference of the actual results y.sub.exp and the prediction of first Gaussian process GP.sub.0(x) at the associated process parameters x.sub.exp, i.e.,

y.sub.exp.fwdarw.y.sub.exp−GP.sub.0=y.sub.exp*. (16)

[0116] Using the actual results y.sub.exp* thus transformed and the associated process parameters x.sub.exp, second Gaussian process GP.sub.1 is then trained.

[0117] In place of first Gaussian process GP.sub.0, it is also possible to use another suitable regression model. For example, it is possible to use a suitable (possibly defined piecewise) polynomial or a spline instead.

METHOD AND DEVICE FOR OPERATING A LASER MATERIAL PROCESSING MACHINE

Inventors

Cpc classification

Classification Explorer

G06N7/01

PHYSICS

Classification Explorer

B23K26/40

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

B23K26/21

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

G06N20/00

PHYSICS

Classification Explorer

B23K26/382

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

B23K26/0622

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

G06F17/18

PHYSICS

Classification Explorer

B23K31/006

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

B23K26/38

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

B23K26/702

PERFORMING OPERATIONS; TRANSPORTING

International classification

Classification Explorer

B23K26/40

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

B23K26/382

PERFORMING OPERATIONS; TRANSPORTING

Classification Explorer

G06F17/18

PHYSICS

Classification Explorer

G06N20/00

PHYSICS

Abstract

Claims

Description