Method for laser processing of metal material based on automatic determination of material or processing parameters

Abstract

A machine and associated method for laser processing of a metallic material, the method comprising the steps of: (a) controlling the emission of at least one pulse of a characterizing laser beam over a predetermined region of the material in a characterizing atmosphere in order to generate a metallic vapor and / or a plasma from the material; (b) acquiring spectral data representative of an optical emission spectrum of a metal vapor or plasma, the spectral data being indicative of the material being processed; (c) identifying one of a plurality of predetermined material classes or predetermined processing parameter classes corresponding to the acquired spectral data by means of an electronic processing and automatic discrimination device configured in a supervised learning phase by training a sample set of spectral data, the training spectral data sample set indicates a predetermined material category or a predetermined material processing parameter category; and (d) selecting a current processing parameter of the material according to the identified material category or processing parameter category.

Description

[0001] The present invention relates to laser processing of materials, more particularly metallic materials, and in particular to improvements in methods and machines for laser processing of materials, more particularly for laser cutting, drilling or welding of said materials or for additive manufacturing of three-dimensional structures from powders of said materials.

[0002] More particularly, the invention relates to a method for laser processing of metallic materials as specified in the preamble of claim 1 .

[0003] According to a further aspect, the invention relates to a machine for laser processing of metallic material according to the preamble of claim 18 .

[0004] In the following description and claims, the term "metal material" is used to identify any manufactured article, such as a sheet or an elongated profile, indiscriminately having a closed cross-section (e.g., hollow round, rectangular or square) or an open cross-section (e.g., flat cross-section or L-shaped, C-shaped, U-shaped, H-shaped, I-shaped cross-section, etc.).

[0005] In the industrial processing of materials, in particular of metal sheets and profiles, lasers are used as thermal tools for a wide variety of applications that depend on the interaction parameters of the laser beam with the processed material, specifically on the energy density of the incident volume of the laser beam on the material and on the interaction time interval.

[0006] Laser processing machines are well known and include industrial processing machines having a CO2 or ytterbium-doped fiber laser source, or even a direct diode laser source, adapted to emit a single-mode or multi-mode laser beam, the optical path of the laser beam being in air or an optical fiber. The laser beam is conducted from the emission source along the optical transmission path of the beam to a processing head, which includes an optical system for focusing the laser beam along an optical propagation axis incident on the material.

[0007] By placing low energy density (per mm 2 The laser beam with a power of tens of W (on the order of seconds) is irradiated directly on the metal material for a long time, and the hardening process is implemented. 2 The laser beam (on the order of tens of MW on the surface) is directed onto the same metal material for a duration of femtoseconds or picoseconds, implementing the ablation process. In the intermediate range of increasing energy density and reducing processing time, controlling these parameters allows the implementation of welding, cutting, perforating, etching and marking processes.

[0008] Therefore, the differences between the various types of processing that can be performed on a material are essentially attributable to the power of the laser beam used and the interaction time between the laser beam and the material being processed.

[0009] Likewise, in additive manufacturing, controlling the laser beam power and the interaction time between the laser beam and the material is critical to achieving processing goals, thereby ensuring the desired characteristics of shape, density, and stability of the resulting structure.

[0010] It is therefore obvious that, within the framework of a predetermined type of processing, the quality of the processing depends on the setting of the parameters of the interaction of the laser beam with the material being processed, which in turn depends on the type of material being processed and the conditions during processing (e.g. surface conditions and environmental conditions).

[0011] In fact, different metal alloys react differently to the machining laser beam, and additional differences are encountered during process execution due to the surface conditions of the material (such as its oxidation state or other surface modifications, such as the presence of metal surface coatings or protective oil layers) and the process atmosphere.

[0012] However, it is often impossible to know precisely the actual composition of a metal alloy, since it varies over time and due to storage conditions different from those indicated by the manufacturer. Furthermore, the processing of materials may use different types of assist gases, the parameters of which vary depending on the state of the material.

[0013] It is therefore desirable to improve the quality of the industrial process of laser machining by adapting the machining parameters as efficiently and quickly as possible to the machined material and the actual machining conditions.

[0014] The object of the present invention is to provide a method for laser processing of metal materials which is optimized with respect to materials and processing conditions.

[0015] Another object of the present invention is to provide a method for laser processing of metal materials which is optimized and can be implemented efficiently and in a production line.

[0016] These objects are achieved according to the invention by a method for laser processing of metallic materials having the features claimed in claim 1 .

[0017] The detailed description forms the subject matter of the dependent claims, whose content is to be understood as an integral part of the present description.

[0018] Another subject matter of the invention is a machine for laser processing of metallic materials as claimed.

[0019] In summary, the present invention is inspired by the consideration that the type of material being processed (typically a metal alloy) can be qualitatively inferred from the optical emission phenomena of the metal vapor or plasma (or combination thereof) emitted from the material to be or has been processed and detected by acquiring the optical emission spectrum in the spatial region around and near the material. The term "material to be or has been processed" refers to the material exposed to the processing laser beam and therefore refers to the surface layer (finished, coated or formed due to unwanted oxidation) of the material or the sub-surface layer exposed during the processing. The term "optical emission" refers to detectable emission within a wide electromagnetic spectrum, also including the ultraviolet and infrared regions, caused by the ionization of the metal vapor or plasma, or in the case of the infrared indicating thermal emission. The collected spectrum is used to identify the material to be processed, rather than for process monitoring or 2D or 3D mapping.

[0020] Specifically, the present invention is inspired by a material spectroscopic characterization technique known as LIBS (Laser Induced Breakdown Spectroscopy), which allows the chemical composition of a sample to be identified at the atomic level qualitatively or quantitatively (relative proportions of different elements within the sample) by analyzing the optical spectrum obtained by generating plasma from the material using a laser.

[0021] exist Figure 1 The operation of the technique is illustrated in Figure 1. A short laser pulse L is focused on a sample (material) M and, by radiation-matter interaction, a plasma plume P is formed. During the recombination of ions and electrons, radiation R is emitted in the optical band from the UV to the near infrared: part of it is a continuous emission and the other part is characteristic of the chemical elements that make up the sample. The radiation is then collected by an optical line and conducted to a spectrometer S where it is separated into its various wavelengths. This generates a spectrum U that is characteristic of the emission of the material, but is also affected by the laser used and the atmosphere in which the process is carried out. The collected spectra are then analyzed and processed using appropriate algorithms, which also allow the chemical composition of the material to be traced quantitatively.

[0022] Lasers used for characterizing materials by LIBS are typically Q-switched pulsed lasers with irradiances up to 10 10 -10 15 W / cm 2 , whose pulse duration is in the nanosecond range. Although lasers with femtosecond pulses can be used in a vacuum in the laboratory, the most common portable systems are based on Nd:YAG lasers, which emit at 1064 nm in air and can only function in an inert atmosphere (such as argon) in certain circumstances.

[0023] The collected spectra are analyzed and processed by comparison with pre-existing data sets or standard samples to achieve a fine calibration of the instrument, allowing a quantitative assessment of the chemical composition of the material to be obtained.

[0024] The method of the present invention uses LIBS technology in an innovative way to control laser processing of materials. Specifically, controlling the laser processing of materials is accomplished by selecting processing parameters of the material based on a material class or processing parameter class associated with characteristic spectral data of optical emissions of metal vapor or plasma (or a combination thereof) generated by the material (i.e., identified by automatic discrimination from such spectral data, wherein the automatic discrimination is based on supervised learning of the correspondence between a training spectral data sample set and the relevant material class or processing parameter class).

[0025] The method covered by the invention can be implemented by a system adaptable to a laser processing machine, on which it can be easily integrated, even sharing part of the physical hardware (i.e. the laser source and part of the optical path intended to perform the processing) or already present on the machine and provided with processing resources for overall control of the laser beam and the laser processing. Advantageously, in fact, the method covered by the invention can be implemented by using in an innovative way the laser beam generated by the laser source as a processing tool to generate the emission of metal vapor or plasma from the material, which is necessary for the characterization of the material.

[0026] Advantageously, the quality of the laser process is improved due to the on-line (i.e., not in a vacuum, but possibly in a protected atmosphere) identification of optimal processing parameters that best approximate (or are consistent with) machining parameters employed in a previous process applied to a material having the same (or closest) spectral characteristics (i.e., the same (or closest) type of material to be processed). On-line identification of processing parameters (whether directly or by material identification) is advantageously achieved immediately prior to processing, so as to enable automatic and more accurate calibration of the parameters required for the process, correct any errors in the advance indication of the material, and adapt the calibration to bulk material or specific surface treatments.

[0027] In another advantageous manner, the technology covered by the present invention enables the detection of processed metal material on the machine during laser processing of the material and continuously during the process without the need for stopping the machine or requiring the prior acquisition (particularly visual acquisition by an operator) of data on the material received from the supplier.

[0028] Ultimately, the invention enables the process of laser processing of metallic materials to be improved due to direct identification of the material in the machine, since it can identify any errors in the advance indication of the material being processed, thus avoiding production waste, controlling production costs, and optimizing the expected quality of the finished product.

[0029] Further characteristics and advantages of the invention will appear in more detail in the following detailed description of its embodiments given by way of non-limiting example with reference to the accompanying drawings, in which:

[0030] As mentioned above, Figure 1 An operation diagram of the LIBS technique is shown;

[0031] Figure 2a to Figure 2c Three different configurations of a machine for laser processing of metallic materials operated according to the method covered by the present invention are schematically shown, wherein identical or functionally equivalent elements or components are indicated with the same reference numerals;

[0032] Figure 3 is a flow chart of operations performed in preparation for laser processing of a metallic material according to a method covered by the present invention;

[0033] Figures 4a to 4e is the optical plasmon emission spectrum of an exemplary material;

[0034] Figure 5 is a schematic diagram of a preferred classification method employed in the methods encompassed by the present invention;

[0035] Figure 6 It is the projection of the spectral data in the two-dimensional classification space defined by two principal components, representing four material classes;

[0036] Figure 7 Shows Figure 5 a subdivision of the classification space in that maximizes the margins between classes; and

[0037] Figure 8 Shows the abnormal situation in the characterization Figure 6 spectral data for new categories within the segmentation.

[0038] As described above Figure 1 , and its contents are hereby intended to be considered common to the construction of a controlled processing machine for implementing the method according to the teachings of the present invention.

[0039] Figure 2a is a schematic diagram of a machine for laser processing of metallic material according to the invention in a first configuration.

[0040] Figure 2a The machine comprises at least one laser emitting source 10 adapted to emit a laser beam B in a transmission device (e.g. an optical fiber 12) adapted to conduct the processing laser beam emitted from the source toward a processing head 14 located near the material WP. The processing laser beam B is, for example, a continuously emitted laser beam having a peak power of 1 kW or more.

[0041] A laser emission control unit 16 may be coupled to the source 10, the laser emission control unit being arranged to control the power and emission duration parameters of the laser beam B and / or to select one of the plurality of laser beams in case of a plurality of sources. In one embodiment, the control unit 16 is therefore arranged to control the emission of at least one pulse of a characterizing laser beam on a predetermined region of the material WP, the predetermined region being suitable for generating metal vapor and / or plasma, and the subsequent optical and thermal emission representing a "signature" of the material. Due to the power and / or pulse duration of the characterizing laser beam, the material undergoes a process of gradual heating, which partially causes vaporization of the material and may subsequently cause excitation and ionization to form a plasma. In the spatial volume around the point of incidence of the characterizing laser beam on the material, an atmosphere is formed which may be in the form of a combination of metal vapor and plasma (also called a "plume") or only one of them. The characterizing laser beam may be generated from a source (not shown) separate from the source 10 and conducted separately to the machining head, wherein said characterizing laser beam is advantageously guided along an optical path common to the machining laser beam.

[0042] In a currently preferred embodiment, a single source 10 is used, and the characterization laser beam is obtained by modulating a continuously emitted machining laser beam, such as by modulating the machining laser beam to emit one or more laser pulses of duration between 10 microseconds and 200 microseconds. In this way, the additional cost of a second source in the machine is eliminated.

[0043] The processing head 14 includes an optical system that generally includes a set of optical elements, such as a collimating lens 20 , a beam shaping lens 22 , a focusing lens 24 and protective optical elements 26 .

[0044] An optical combiner device 28 is associated with the source and is arranged to combine and emit multiple laser beams on a transmission optical fiber and to guide the optical signal emitted from the metal vapor or plasma of the material and collected by the optical system of the processing head to a spectral data acquisition device 30 (such as a spectrometer), which is suitable for acquiring spectral data representing the optical emission spectrum of the metal vapor or plasma generated by the characterizing laser beam. These spectral data indicate the material WP subjected to processing. Although not shown in the figure, the path of the optical signal downstream of the combiner device 28 toward the spectrometer 30 may include an optical collimation and filtering system.

[0045] Associated with the machining head 14 are optical system actuator means, which are identified in the figure by way of example as actuator means 32 for the movement of the beam shaping mirror 22 and actuator means 34 for the adjustment of the focus of the beam acting on the focusing lens 24, both controlled by a control unit, generally indicated by 36, connected to a field bus 38 to which are also connected processing and control means 40, arranged to control the application of the machining laser beam along a predetermined machining trajectory. Controlling the application of the machining laser beam along the predetermined machining path comprises controlling the delivery of an auxiliary gas flow (in cutting applications) and controlling the irradiation of the laser beam with a predetermined power distribution towards a predetermined working area by reference to a predetermined working model or program (i.e., according to working trajectory information and working parameters obtained in the form of movement instructions for the machining head and / or the material being machined, and physical machining parameters indicating the power distribution of the beam, the beam power intensity and the activation time of the laser beam as a function of the machining trajectory). The processing means 40 can be integrated in a single processing unit on board the machine, or realized in a distributed form, whereby they comprise processing modules located in different parts of the machine, including, for example, the machining head.

[0046] The main advantages of this configuration are the non-intrusive installation of the spectral data acquisition device, no modification or special design of the processing head is required to accommodate the spectrometer, and the system is generally more robust because there are no delicate instruments near the processing area. These considerations make this configuration preferred in industrial environments. Emissions with wavelengths shorter than 400nm may not usually be observed in this configuration because the optical fiber 12 as a transmission device does not transmit light in the ultraviolet region. Further, the emission peak will appear at the emission wavelength of the characterization and processing laser beams.

[0047] In an alternative embodiment to the described embodiment including the optical combiner device 28 (which is Figure 2b As shown in Figure 2a ), the spectral data acquisition device 30 is separated from the laser source and integrated into the processing head 14.

[0048] In this configuration, the optical signal emitted from the metal vapor or plasma of the material is collected by the optical system of the processing head (including the protective optical element 26 and the focusing lens 24) and transmitted through the dichroic beam shaping mirror 22 towards the spectral data acquisition device 30 integrated in the processing head 14. The path of the optical signal downstream of the dichroic mirror 22 to the spectrometer 30 may include a collimation and filtering optical system 30'.

[0049] In each case, the collection circuits allowing the acquisition of the optical signal emitted from the metal vapor or plasma include a spectrometer using a very wide band (ultraviolet to near infrared) or using only a part of it. Alternatively, sensor devices such as a plurality of photodiodes can be used that are intended to determine the intensity of the signal only at a predetermined wavelength in the near infrared to ultraviolet optical band.

[0050] Advantageously, in both configurations described, the collection of the optical signal to the spectral data acquisition device 30 is performed according to a coaxial configuration, wherein the optical signal follows, at least over a certain distance, the same optical propagation path as the characterization and processing laser beams, passing through the processing head and even more advantageously reaching the optical beam combiner device 28 associated with the source. The use of the same optical system for focusing the characterization (and processing) laser beams and acquiring the optical emission of the plasma ensures that the emission is collected from the same region of the material where the plasma is generated.

[0051] Alternatively, if Figure 2c As shown, Figure 2a Elements or components identical to those shown in or functionally equivalent to those shown in are indicated with the same reference numerals, and the collection of optical signals can be performed off-axis, outside the processing head, by positioning the spectral data acquisition device 30 laterally at a certain distance and angle relative to the defined propagation direction of the characterization and processing laser beam, depending on the space available in the machine. This configuration allows the collection of spectral data that are neither distorted nor filtered by the optical shaping system of the characterization and processing laser, thereby maintaining the natural shape of the emission, and it is also possible to acquire spectral data in the ultraviolet region. In any case, it is preferred to place a cutoff filter of the reflected wavelength of the laser beam on the material in order not to saturate the emission spectrum. Disadvantageously, because the optical emission of the metal vapor or plasma from the material is not collected by the optical focusing system of the characterization (and processing) laser beam, any change in the position of the processing head relative to the material (e.g., a change in the processing distance from the material) requires adjustment of the distance and positioning angle of the spectral data acquisition device relative to the propagation direction of the characterization and processing laser beam.

[0052] In each of the described configurations, an electronic processing and automatic identification device 42 is connected to the field bus 38, the electronic processing and automatic identification device being coupled to the data storage memory DB, being arranged to process the signal emitted from the spectral data acquisition device 30 and to identify one of a plurality of predetermined material classes or predetermined processing parameter classes corresponding to the spectral data. The electronic processing and automatic identification device 42 is configured in a supervised learning phase using a training spectral data sample set indicating a predetermined material class or a predetermined material processing parameter class. The training spectral data sample set may be obtained based on nameplate data of a material that is the object of previous processing and is regularly updated, or by a preliminary acquisition of spectral data samples, such as a few dozen spectra for each available material. The electronic processing and automatic identification device 42 is configured to select a processing parameter corresponding to a processing parameter class or material class identified by the data storage memory DB, which includes a reference model indicating a nominal relationship between the spectral data and the processing parameter class or material class.

[0053] In general, the term "processing parameters" refers to physical parameters of the processing laser beam and the processing environment (for example, at least one of the repetition rate of the processing laser beam pulses, the duration of the processing laser beam pulses, the power of the processing laser beam pulses, the power density distribution of the processing laser beam, and the pressure of the auxiliary gas), or more generally, refers to a processing strategy including the physical parameters of the processing laser beam and the processing environment and the control parameters of the movement of the processing beam.

[0054] The processing and control device 40 is arranged for controlling the application of the machining laser beam along a predetermined machining trajectory in dependence on the selected machining parameters.

[0055] exist Figure 3 A flow chart of the method involved in the present invention is briefly shown in FIG.

[0056] At step 100, the control unit 16 is triggered by the processing and control device 40 or the electronic processing and automatic recognition device 42 to control the emission of at least one pulse of a characterization laser beam in a characterization atmosphere on a predetermined area of ​​the material WP in order to generate metal vapor and / or plasma from the material. In the specific case, the characterization atmosphere is the same as the atmosphere of the working process.

[0057] The control unit 16 controls the emission of at least one pulse characterizing the laser beam in an early stage of processing the material, such as on an area of ​​the material intended to be subjected to cutting or drilling or in a calibration step before processing the material, on an unprocessed zone of the material.

[0058] The control unit 16 may also control the emission of at least one pulse of the characterizing laser beam during processing of the material, advantageously in case the characterizing laser beam is obtained by modulation of the processing laser beam, possibly by modifying the process atmosphere to temporarily provide a different characterizing atmosphere.

[0059] At step 200 , the spectral data acquisition device 30 acquires spectral data representing an optical emission spectrum of metal vapor or plasma, and the spectral data indicates the processed material WP.

[0060] By way of example, in Figures 4a to 4c The aluminum alloy Al 6060 ( Figure 4a ), galvanized steel( Figure 4b ) and stainless steel AISI 304( Figure 4c Depending on the plasma generation conditions of the material, such as due to the power of the characteristic laser beam or the process atmosphere and more generally due to the laser processing machine used, different intensity values ​​of the optical emission can be detected, such as Figure 4d , which shows the spectra of aluminum alloy Al 6060 in an oxygen atmosphere obtained from characterizing laser beams emitted by different laser processing machines with different powers and diameters. Depending on the surface conditions of the material (such as its oxidation state or its coating), different spectra will also be obtained, such as Figure 4e , which shows a comparison between the spectra of carbon steel S235JR under normal conditions (curve A) and rusted carbon steel S235JR (curve B).

[0061] At step 300, the electronic processing and automatic discrimination device 42, which has been previously configured in a supervised learning step (not shown), identifies one of a plurality of predetermined material classes or predetermined processing parameter classes corresponding to the acquired spectral data.

[0062] Before applying the automatic discrimination algorithm to the spectral data, it is useful to normalize the data in order to obtain comparable data, since even with the same parameters characterizing the laser beam, there may be a certain degree of variability in the overall intensity of the optical emission of the plasma. Due to the spectral normalization, any fluctuations in the intensity of the acquired spectral data can be compensated.

[0063] Each spectrum includes N emission lines I(λ) acquired within a predetermined wavelength range between ultraviolet and infrared (e.g., between 181 nm and 1100 nm, with an average resolution of 0.56 nm). i )(i=1,...,N).

[0064] Two normalization procedures can be advantageously employed. The first normalization procedure is to normalize the spectrum by dividing all emission line intensities by the maximum intensity (in other words, using a value of 1 to represent the maximum peak and scaling other values ​​accordingly), according to the following expression:

[0065]

[0066] The second procedure is to divide each intensity of the emission line by the sum of all intensities (essentially scaling the corresponding value relative to the total emission represented by the area encompassed by the spectral curve), according to the following expression:

[0067]

[0068] I norm Information about the shape of the spectrum representing the material's signature is retained, and therefore useful information enabling an automated discrimination algorithm to perform classification.

[0069] At step 400, the electronic processing and automatic identification device 42 selects current processing parameters for the material according to the material category or the identified processing parameter category.

[0070] Finally, in step 500 , the processing and control device 40 controls the irradiation of the processing laser beam in a predetermined area of ​​the metal material or along a predetermined processing trajectory of the metal material according to the current processing parameters selected in the previous step.

[0071] Conveniently, in order to be able to discern the material over a large surface area or in a deep volume, where the material condition may vary, step 100 is repeated successively or at predetermined processing intervals, whereby the control unit 16 controls the emission of a plurality of pulses of the characterization laser beam over a series of predetermined areas of the two-dimensional surface scanning area of ​​the material or over a series of predetermined areas of the three-dimensional scanning volume of the material, so as to generate corresponding metal vapor or plasma from the material present in such scanning areas. For example, the control unit 16 may control the emission of a plurality of pulses of the characterization laser beam over a series of depths of predetermined areas of the three-dimensional scanning volume of the material, so as to focus the characterization laser beam on different planes of the material, so as to include a surface layer of the material and at least one sub-surface layer exposed due to the characterization of the covering layer.

[0072] More particularly, the electronic processing and automatic discrimination means 42 can advantageously be implemented as a discrimination and classification module based on principal component analysis (PCA) techniques for extracting characteristics and artificial intelligence classification systems, such as neural networks or support vector machines (SVM). The spectral data can be processed by various statistical techniques, not only by PCA but also by normalization techniques, such as by using spectral data acquired over the entire detection spectrum.

[0073] exist Figure 5 A schematic diagram of the above classification method is shown in FIG.

[0074] In a currently preferred embodiment, the identification of a material class or a processing parameter class corresponding to the acquired spectral data is accomplished by: transforming the acquired spectral data into a classification space defined by predetermined orthogonal latent variables, the predetermined orthogonal latent variables including a subset of predetermined significant latent variables indicating the variance of the spectral data; and comparing the values ​​of the n-tuple of the significant latent variables (calculated from the detected spectral data) with a set of reference values ​​of the n-tuple of the significant latent variables (indicative of a training spectral data sample set). Selecting a parameter corresponding to a predetermined processing parameter as a processing parameter of the material, the predetermined processing parameter being associated with a reference value of the n-tuple of latent variables, the reference value having a predetermined metric relationship with the value of the n-tuple of significant latent variables calculated from the detected spectral data.

[0075] Advantageously, the predetermined metric relationship is a minimum distance relationship.

[0076] Figure 6 The projection diagram of the spectral data acquired for a specific characterization laser beam and a specific characterization atmosphere in the two-dimensional classification space defined by the principal components identified as PC2 and PC3 is shown, wherein the training spectral data samples include four categories, namely, material A16060 (aluminum, magnesium, silicon family alloy), AISI304 (stainless steel), E220 (steel) characterized in an O2 atmosphere, and E220 (steel) characterized in a N2 atmosphere. As can be seen from the figure, the space of the two principal components PC2 and PC3 clearly distinguishes the four categories of training spectral data samples.

[0077] Figure 7 The spatial partitioning of principal components PC2 and PC3 obtained by applying the SVM classifier is shown. This partitioning is the result of using a linear kernel and a partitioning function that maximizes the margin between classes.

[0078] As with any type of classifier, each new measurement is classified into one of the known categories (the ones used to train the model), even if it is a new material or a new experimental condition. Figure 7 It is evident from the figure that there are regions far from the clusters observed in training where any new observation will be forced to be classified into one of the known classes, despite being very different from the training data. Therefore, an improvement to the method can be introduced by extending the classification to report whether a given classification result is reliable. This can be done, for example, by automatically determining whether the projection of a new observation in the space of principal components is within (or near) the cluster of the training data belonging to the assigned class or deviates from that cluster by a distance greater than a predetermined threshold, and reporting this characterization anomaly to the operator. Consideration should be given to Figure 8 An example is shown, where a new observed class N is characterized by a score in a region of space where the material is classified as being in the A16060 series, despite the acquired spectral data being significantly deviated from the corresponding cluster. If the distance of each representation of the acquired spectral data, or the distance of the center of the representation of the spectral data acquired in a plurality of observations, to the center of the training spectral data cluster (or clusters of all known training spectral data classes) is greater than a predetermined threshold, then the characterized anomaly is verified. The principle behind the calculation of the distance to a reference cluster in multidimensional space is based on the same mechanism as the SVM method, but its variant is called single-class SVM or support vector data description (SVDD), as described in the following literature: Tax, DM, Duin, RP, "Support vector data description". Machine Learning, 54(1), pp. 45-66, 2004; Ning, X., Tsung, F., "Improved design of kernel distance-based charts using support vector methods", IIE transactions, 45(4), pp. 464-476, 2013; and Grasso M., Colosimo BM, Semeraro Q., Pacella M., "A Comparison Study of Distribution-Free Multivariate SPC Methods for Multimode Data". Data), Quality & Reliability Engineering International, 31(1), pp. 75-96, 2015.

[0079] Characteristic anomalies can be tracked by the operator investigating whether it is a new material or a new combination of material and process atmosphere. If so, such data can be added as new training spectral data samples to the reference model and the classifier can be retrained once a sufficient number of observations attributable to this class are available. If not, it may still be useful to include the new observations in the training sample set for this material to account for additional variability that was not observed in the first training phase. This will be able to improve future classification performance. Again, in this case, if the training spectral data sample set is extended, the classification algorithm needs to be retrained and the distance threshold that limits the characterization anomaly reporting situation also needs to be redefined.

[0080] Of course, without prejudice to the principle of the invention, the embodiments and the execution details may vary widely with respect to what has been described and illustrated purely by way of non-limiting example, without thereby departing from the scope of protection of the invention as defined by the appended claims.

Claims

1. A method for laser processing of a metal material, in particular for laser cutting, drilling or welding of a certain volume of the material or additive manufacturing of a three-dimensional structure from powder of the material, the method comprising irradiating a processing laser beam in a predetermined area or along a predetermined trajectory in a predetermined process atmosphere and according to current processing parameters selected based on the material to process the metal material, characterized in that The method comprises the following steps: - controlling the emission of at least one pulse of a characterizing laser beam in a characterizing atmosphere on a predetermined area of ​​said material in order to generate metal vapor and / or plasma from said material; - acquiring spectral data representing an optical emission spectrum of said metal vapor or plasma, said spectral data being indicative of said material being processed in said characterizing atmosphere; - identifying one of a plurality of predetermined material classes or predetermined processing parameter classes corresponding to the spectral data by means of electronic processing and automatic identification means configured in a supervised learning phase using a training set of spectral data samples indicative of a predetermined material class or a predetermined material processing parameter class; and - selecting said current processing parameters of said material according to the identified material class or processing parameter class.

2. The method according to claim 1, wherein: The predetermined processing parameters include at least one of a repetition frequency of the pulses of the processing laser beam, a duration of the pulses of the processing laser beam, a power of the pulses of the processing laser beam, a power density distribution of the processing laser beam, and an auxiliary gas pressure.

3. Method according to any of the preceding claims, comprising controlling the emission of at least one pulse of said characterizing laser beam at an early stage of said processing of said material, on a zone of said material intended to be cut or drilled.

4. The method according to any one of claims 1 or 2, comprising controlling the emission of at least one pulse of the characterizing laser beam on an unprocessed zone of the material in a calibration step prior to processing the material.

5. The method according to claim 3 or 4 comprises controlling the emission of multiple pulses of the characterizing laser beam over a series of predetermined areas of a two-dimensional scanning area of ​​the material or over a series of predetermined areas of a three-dimensional scanning volume of the material so as to generate corresponding metal vapor or plasma from the material present in the scanning area.

6. The method according to claim 5, wherein: Controlling the emission of multiple pulses of the characterizing laser beam over a series of predetermined areas of the three-dimensional scanning volume of the material includes focusing the laser beam on different planes of the material, the different planes including a surface layer and at least one sub-surface layer of the material, a first pulse generating metal vapor or plasma from the material in the surface layer and at least a second pulse generating metal vapor or plasma from the material in the sub-surface layer.

7. A method according to any one of the preceding claims, wherein: The spectral data representing the optical emission spectrum of the metal vapor or plasma of the material is acquired in an optical band between near infrared and ultraviolet.

8. The method according to claim 7, wherein: Spectral data representing the optical emission spectrum of the metal vapor or plasma of the material is acquired at a predetermined wavelength in the optical band between near infrared and ultraviolet.

9. A method according to any one of the preceding claims, wherein: The characterizing atmosphere is the process atmosphere.

10. A method according to any one of the preceding claims, wherein: The characterization laser beam is obtained by modulating the processing laser beam.

11. The method according to claim 10, wherein: The at least one pulse of the characterization laser beam is emitted with a duration between 10 microseconds and 200 microseconds by modulating a continuously emitted machining laser beam having a peak power of 1 kW or more.

12. The method according to any one of claims 1 to 9, wherein: The characterizing laser beam is guided in a processing head of a machine for carrying out the laser processing process along a common optical path with the processing laser beam.

13. The method according to claim 10 or 11, comprising detecting the optical emission spectrum of the metal vapour or plasma coaxially with the propagation direction of the machining laser beam in a machining head of a machine for carrying out a laser machining process.

14. The method according to claim 10 or 11, comprising detecting the optical emission spectrum of the metal vapour or plasma at a predetermined angle to the propagation direction of the machining laser beam outside a machining head of a machine for carrying out a laser machining process.

15. A method according to any one of the preceding claims, comprising controlling the emission of at least one pulse of the characterising laser beam during processing of the material.

16. A method according to any one of the preceding claims, wherein: Identifying, by electronic processing and automatic identification means, one of a plurality of predetermined material categories or process parameter categories corresponding to the spectral data comprises: - transforming the spectral data into a classification space defined by predetermined orthogonal latent variables, the latent variables comprising a subset of predetermined significant latent variables indicative of the variance of the spectral data; and - comparing the values ​​of the n-tuple of the significant latent variables calculated from the spectral data with a set of reference values ​​of the n-tuple of the significant latent variables indicative of the training spectral data sample set, Thereby, the material processing parameter corresponding to the predetermined processing parameter is selected, the predetermined processing parameter is associated with the reference value of the n-tuple of the latent variable, and the reference value has a predetermined metric relationship with the value of the n-tuple of the significant latent variable calculated based on the collected spectral data.

17. The method according to claim 16, wherein: The predetermined metric relationship is a minimum distance relationship.

18. A machine for laser processing of metallic material, in particular for laser cutting, drilling or welding of a volume of said material or for additive manufacturing of three-dimensional structures from powder of said material, said machine comprising: - a source (10) for emitting a machining laser beam (B); - means for conducting said machining laser beam (B) emitted from said source (10) along an optical path for transmitting said beam (B) to a machining head (14) located in the vicinity of the material (WP); and - a processing and control device (40) arranged to control the application of the processing laser beam (B) on the material (WP) along a predetermined processing trajectory in a predetermined process atmosphere and according to current processing parameters selected based on the material, Characterized in that the machine also includes: - a source (10) for emitting at least one pulse of a characterizing laser beam in a characterizing atmosphere onto a predetermined area of ​​said material in order to generate metal vapor and / or plasma from said material; - means (30) for acquiring spectral data representative of an optical emission spectrum of a metal vapor or plasma, said spectral data being indicative of said material being processed in said characterizing atmosphere; and - electronic processing and automatic discrimination means (42) configured in a supervised learning phase by a set of training spectral data samples indicating predetermined material classes or predetermined material processing parameter classes, and arranged to identify one of a plurality of predetermined material classes or predetermined processing parameter classes corresponding to the acquired spectral data, Thereby, the processing and control device (40) is arranged for controlling the application of the processing laser beam (B) according to current processing parameters selected in dependence on the identified material class or processing parameter class.

Images