Method and device for recognizing a process state of a plasma arc method

Abstract

The present invention relates to a method for recognizing a process state of a plasma arc method, wherein a workpiece (1) is processed by a plasma torch (2) and measured values of a time curve of a physical variable of the plasma arc method are detected and a signal curve is determined from the measured values (S3-1). A pattern recognition is performed on the signal curve (S3-2) by extracting at least one feature from the signal curve and classifying said feature, and the signal curve is assigned to a specific state of the plasma arc method according to the classification of the feature (S3-4).

Description

[0001] Method and device for detecting a process state of a plasma arc process

[0002] The present invention relates to a method and a device for detecting a process state of a plasma arc process.

[0003] In plasma arc processes such as plasma cutting, events can occur that negatively impact the desired result. This affects, for example, the quality of a cut, damage to the plasma torch or the cutting material, particularly due to an unforeseen or unforeseeable failure, or the use of consumable parts beyond their recommended operating life. To counteract this, the plasma arc process is monitored with the aim of detecting such events in a timely manner and, if necessary, aborting the process or changing one or more process parameters. This not only serves to optimize the respective plasma arc process but also to protect the plasma torch used in it and the workpiece.

[0004] Several approaches are known in the prior art. Typically, a single parameter, such as the amplitude or rate of change of an electrical voltage, is determined, and appropriate action is initiated if a threshold is exceeded. For example, US Pat. No. 5,750,957 A discloses a method in which the standard deviation of an electrical signal is monitored and used to assess the quality of the process.

[0005] A disadvantage of this approach, however, is that the parameters used for quality control are often dependent on process parameters such as the speed of the plasma arc process and the material thickness of the workpiece, so the threshold value to be determined requires knowledge of these cutting parameters and is difficult to generalize.

[0006] Piercing the workpiece as a cutting material can also result in different signal curves for different material thicknesses, so a simple comparison with a threshold value generally yields insufficient results. The same applies to wear part failure, which produces different signal patterns that depend, among other things, on the cathode used.

[0007] The present invention is therefore based on the object of proposing a method and a device that avoid the aforementioned disadvantages and thus allow process states, in particular errors, of a plasma arc process to be quickly and reliably determined. This object is achieved according to the invention by a method according to claim 1 and a device according to the independent claim. Advantageous embodiments and further developments are described in the dependent claims.

[0008] In a method for detecting a process state of a plasma arc process, a workpiece is processed by a plasma torch, and measured values ​​of a temporal progression of a physical parameter of the plasma arc process are detected. A signal curve is determined from the measured values, and pattern recognition is performed on the signal curve. At least one feature is extracted from the signal curve and classified. Finally, depending on the classification of the feature, the signal curve is assigned to a specific state of the plasma arc process.

[0009] By recording the temporal progression of a physical quantity associated with the plasma arc process, a (temporal) signal progression of this physical quantity that is typical for the respective process is determined. By not only focusing on threshold exceedances, which can also trigger false alarms due to strong noise, deviating process parameters, or similar, but by using pattern recognition to look for similar patterns to already known process states, a more reliable statement about the process state is possible than with conventional methods. For this purpose, characteristics are extracted from the signal curves that can be assigned to a pattern to be detected and assigned to a common class (classification). Through classification, i.e.By assigning a specific state of the plasma arc process, a more reliable statement about the state of the plasma arc process can be made than with a pure threshold value consideration. The method can therefore be used in particular to monitor arc stability, detect failure of wearing parts, and terminate operation in order to protect the plasma torch from irreparable damage. Furthermore, conditions that could reduce cathode service life, such as the arc burning over a hole after a piercing process or passing over the end of a workpiece, can be determined in order to then take measures to maintain cathode service life by influencing process parameters, such as lowering the plasma torch or reducing the cutting current or cutting gas pressure.

[0010] The signal curve can be present directly in the time domain or also in the frequency domain by converting the temporal course of the measured values ​​into the frequency domain, for example by a Fourier transformation.

[0011] It can be provided that the at least one physical quantity is an electrical voltage, an electrical voltage drop, and / or an electrical current. These quantities are generally relevant to the plasma arc and can be reliably determined.

[0012] The at least one feature can be determined from amplitudes of the measured values ​​and / or an amplitude spectrum, wherein the feature is preferably determined from a sum of the amplitude spectrum over at least one predetermined frequency range, particularly preferably at least two predetermined frequency ranges. In particular, when using certain relevant frequency ranges of the amplitude spectrum, process states in which frequencies lying within these frequency ranges often occur can be determined with high accuracy. The term "signal curve" can therefore describe both the temporal progression of the measured values ​​and an amplitude spectrum derived from them.

[0013] However, it can also be provided that the at least one feature is determined from a parameterization of at least one sub-section or several sub-sections or a parameterization of the entire signal curve. This enables, in particular, process states that can be demonstrated by a change over time, a rapid and reliable detection of these states. Preferably, the feature is determined from a parameterization of one or more individual sub-sections of the signal curve by forming a sequence of mean values ​​and / or rates of change and / or regression values ​​of a regression analysis. The number of recorded measured values ​​that form a signal curve can be at least 500 measured values, preferably 10,000 measured values, particularly preferably 15,000 measured values, in order to achieve a sufficiently high accuracy and thus reliability of the method.

[0014] The time between two individual measured values ​​forming a signal curve can be a maximum of 1 ms, preferably 0.05 ms, and particularly preferably 0.03 ms, to ensure sufficiently fast determination of process states. Pattern recognition is typically performed in real time, i.e., within a maximum of 1.5 s, or is available in the form of a look-up table in which pattern recognition results for any combination of features are calculated and stored in a table.

[0015] The time for measuring and forming a signal curve can be a maximum of 1500 ms, preferably a maximum of 500 ms, and particularly preferably a maximum of 350 ms, in order to obtain timely information about the process status. The measurement time is preferably longer than the time required for obtaining the features or pattern recognition.

[0016] To increase the reliability of the method, it is usually also intended to determine at least two features and use them to determine the process state, i.e., to extract and classify at least two features from the signal curve. However, the number of features is typically smaller than the number of individual measured values.

[0017] Classification classes typically include at least one stable process state and / or at least one unstable process state. If the stable process state is present, the plasma arc process continues unchanged; however, if the unstable process state is present, the plasma arc process is preferably aborted to avoid damage. The unstable process state can be caused, in particular, by damaged or worn consumables or indicate these. Other states can also be classified that indicate defects such as hafnium ejection from the cathode or double arcs between the cathode and nozzle, but do not require intervention in the process.

[0018] The plasma arc process is typically plasma cutting with a plasma cutting torch. In principle, however, all processes in which material is processed using a plasma arc can be considered a plasma arc process. In this case, the process states classes can include idle, ignition of a pilot arc, burning of the pilot arc, piercing the workpiece, piercing the workpiece, cutting the workpiece, passing over a kerf, passing over a workpiece edge, the presence of a damaged wearing part and / or the presence of one or more worn wearing parts. Alternatively or additionally, the consumption of a wear part or parts can bein particular the classification into the classes presence of a damaged wearing part, presence of a new wearing part, presence of a used wearing part and / or presence of a worn wearing part based on the burnback of the cathode (measured in millimetres) and / or based on the burnback in relation to a predetermined service life. In the first case, for example, a classification can be made according to the remaining length of the cathode, i.e. a length in the range 0 mm to 1 mm, a length in the range 1 mm to 1.8 mm and a remaining length greater than 1.8 mm. In relation to a maximum service life, which is manifested in the maximum possible burnback, a classification can also be made based on percentages between 0 percent, i.e. no burnback, and 100 percent, i.e. maximum burnback and achievement of the maximum service life.

[0019] Depending on the classification result, an optical, acoustic, electrical, and / or electronic signal, or an optical, acoustic, electrical, and / or electronically analyzable signal, can be generated or emitted to stabilize or terminate the plasma arc process. The signal can also be a warning signal to prompt a human intervention.

[0020] A device for detecting a process state of a plasma arc process has a plasma torch configured to process a workpiece. The device also includes a measured value recording device configured to record measured values ​​of a temporal profile of a physical variable of the plasma arc process and to determine a signal profile from the measured values. Furthermore, an evaluation unit is part of the device, which is configured to extract and classify at least one feature from the signal profile and, depending on the classification of the feature, to assign the signal profile to a specific state of the plasma arc process.

[0021] The plasma torch typically has at least one electrode and one nozzle and is supplied with electrical energy by an electrical current source or an electrical voltage source, which may also be part of the device. The workpiece serves as the second electrode. A control unit can also be provided that controls the current or voltage source, in particular based on signals received from the evaluation unit.

[0022] The device can also comprise an inductive component that is electrically connected in series with the plasma torch. In this case, the measured value recording device is also designed to measure an electrical current flowing through the inductive component or an electrical voltage drop across the inductive component. Typically, at least a portion of the electrical current that also flows through the electrode of the plasma torch flows through the inductive component.

[0023] The electrical voltage to be measured can be the electrical voltage between the electrode of the plasma torch and the workpiece, the electrical voltage between the electrode and the nozzle and / or the voltage drop across the inductive component, which is typically designed as a choke.

[0024] The electrical current to be measured can, in particular, be the electrical current flowing from the electrode to the nozzle and / or from the electrode to the workpiece. The described device is typically configured to perform the described method, i.e., the described method can be performed with this device.

[0025] A computer program product comprises a computer program having software means for carrying out the described method and / or for controlling the described device when the computer program is executed in an automation system or computing unit.

[0026] Embodiments of the invention are illustrated in the drawings and are explained below with reference to Figures 1 to 11.

[0027] They show:

[0028] Fig. 1 is a schematic side view of a device for plasma arc processing;

[0029] Fig. 2 shows a diagram with measured values ​​of physical quantities recorded over a certain period of time;

[0030] Fig. 3 shows a schematic process flow of pattern recognition;

[0031] Fig. 4 is a view corresponding to Fig. 3 of a detailed process flow for pattern recognition;

[0032] Fig. 5 is a schematic representation of a signal curve consisting of voltage and current for determining the condition of wearing parts;

[0033] Fig. 6 is a schematic representation of a signal curve in the case of a defective wearing part;

[0034] Fig. 7 is a schematic representation corresponding to Fig. 6 of a signal curve when passing over a joint or leaving a sheet; Fig. 8 is a schematic representation of the summation of an amplitude spectrum;

[0035] Fig. 9 is a schematic diagram for pattern recognition using parameterization;

[0036] Fig. 10 is a schematic representation of a feature space and

[0037] Fig. 11 a real representation of a feature space.

[0038] Figure 1 shows a schematic side view of a device for carrying out and monitoring a plasma arc process, in the illustrated embodiment, plasma cutting. A plasma torch 2 or plasma cutting torch, which has a nozzle and an electrode (as well as other components such as coolant inlets, etc., which are known from the prior art and are therefore not listed here for reasons of clarity), is electrically connected to a power source 4 and emits a plasma arc 3 to cut a workpiece 1, which is also electrically connected to the power source 4. In the illustrated embodiment, a choke 7 is electrically connected in series with the plasma torch 2 as an inductive component; however, in other embodiments, the choke 7 can also be omitted.A measured value recording device 5 records measured values ​​of a temporal progression of a physical quantity such as electrical current, electrical voltage or electrical voltage drop, in particular during the cutting process, and evaluates them in an evaluation unit 6.

[0039] The evaluation unit 6 is configured to perform pattern recognition on the signal curve from the obtained measured values ​​and a signal curve determined by the measured value recording device 5. At least one feature, but typically two features, are extracted and classified by the evaluation unit 6 from the signal curve, with the various classes of the classification representing a specific state of the plasma arc process. Depending on the determined class, the evaluation unit 6 can output an electrical signal to the current source 4, which accordingly controls or regulates the electrical current. In further embodiments, a voltage source can also be used instead of the current source 4, which accordingly regulates an electrical voltage.

[0040] In the embodiment shown in Figure 1, the measured value recording device 5 is also designed to measure a voltage drop across the choke 7. The measured electrical voltage can be the voltage between the electrode and the workpiece 1, the voltage between the electrode and the nozzle, and / or the aforementioned voltage drop across the choke 7. The measured electrical current can be the electrical current flowing from the electrode to the nozzle, and / or the electrical current flowing from the electrode to the workpiece 1.

[0041] Typically, the evaluation unit 6 and the measured value acquisition unit 5 are configured as a single processing unit, on which a computer program product can generally be executed. The computer program product comprises software means that control the device shown in Figure 1 and carry out the method described in more detail below when the computer program product is implemented in this processing unit as an automation unit.

[0042] Figure 2 shows a diagram in which a typical curve of quantities recorded by the data recording device 5, such as electrical voltage or electrical current, is plotted over time in seconds. In this diagram, the cutting current is plotted in amperes (top curve), the voltage between electrode and workpiece 1 in volts (middle curve), and the electrical voltage between electrode and nozzle in volts (bottom curve). In the range from 1.5 s onwards, increased fluctuations occur, and from 2.9 s onwards there is a voltage dip. Any of these deviations indicates errors in the process and requires the cutting process to be stopped quickly to protect the plasma torch. The fluctuations in the range around 0.7 s are caused, among other things, by hafnium in the arc or double arcs between the cathode and nozzle.However, since, particularly with the fluctuations at 0.7 s, a threshold value which, according to the state of the art, is intended to serve as an error indicator, can be exceeded (and the process can be interrupted if necessary) and the process can then be continued normally, pattern recognition is carried out in the described process in order to more reliably detect actual error events which make it necessary to abort the plasma arc process.

[0043] Figure 3 shows a schematic view of a corresponding pattern recognition process. In step S3-1, the signal curve determined by the measurement recording device 5 from the measured values ​​is stored and, if necessary, transferred to the evaluation unit 6. In step S3-2, at least one feature is determined from the signal curve (or, if signal curves of several physical quantities are present, from each of the signal curves or at least from at least two signal curves). In step S3-3, pattern recognition is performed using a similarity comparison to determine which known pattern or which state of the plasma arc process the present signal curve resembles.

[0044] Pattern recognition is understood as the ability to recognize regularities, repetitions, similarities, or laws in a set of data, such as measured values ​​or a signal curve. The features assigned to a pattern serve to distinguish it from the content of other classes. In addition to the acquisition of data by sensors (as described above), preprocessing for data reduction can also be provided to improve data quality. By extracting the features as part of pattern recognition, the patterns are then transformed into a feature space during feature extraction. In the feature space, the patterns are represented as points, whereby the features can typically be represented mathematically as vectors, so-called feature vectors. Here, the features are measured as numerical values ​​and combined into a (multidimensional) feature vector.If necessary, an intermediate step of feature reduction can be included in the process flow to limit the patterns to their essentials. Classification assigns a feature vector to a class that provides the greatest agreement or similarity with the feature vector to be examined, i.e., whose features provide the greatest agreement or similarity with the feature vector to be examined. This can be done using a mathematical function, such as the scalar product between the feature vector to be examined and one or more vectors representing the respective class, or a classifier that assigns the features to different classes and thereby maps a process state.If, for example, a faulty condition of a wearing part is detected in the subsequent step S3-4 (by classifying the pattern into the corresponding class), the power source 4 can be switched off in step S3-5, for example.

[0045] During preprocessing, the signal curve can be subjected to a high-pass filter, a low-pass filter, or a bandpass filter, for example. Alternatively or additionally, the measured values ​​can also be normalized. The actual features can vary in their characteristics and include, for example, key figures of a distribution function, moments such as the expected value and variance, or correlation and convolution.

[0046] It is also possible to convert the measured values ​​into the frequency domain using a transformation such as a discrete Fourier transform in order to obtain a more manageable feature space.

[0047] Figure 4 also shows a detailed flowchart of pattern recognition. After starting in step S4-1, a memory for a total of M samples is created in step S4-2 in the measured value recording device 5. The number of samples is at least 5000 for the selected embodiment, and the time between the recording of two samples is a maximum of 0.1 ms. The total measurement period can be a maximum of 1 s in the example shown.

[0048] In step S4-3, N samples of the voltage between electrode and nozzle are sampled and these samples are added to the memory (where N <M ist). In Schritt S4-4 wird geprüft, ob der Speicher voll ist, sich also insgesamt M Samples im Speicher befinden. Falls nein, wird der Schneidprozess weiter aufgenommen, falls ja, wird das Mustererkennungsverfahren in den Schritten S4-6, S4-7 und S4-8 durchgeführt. Hierzu wird in Schritt S4-6 in dem dargestellten Ausführungsbeispiel zunächst eine Vorverarbeitung der M Samples durchgeführt und in Schritt S4-7 die Merkmale aus dem Signalverlauf gewonnen. In Schritt S4-8 wird geprüft, ob ein vorab definierter Klassifikator eine größte Übereinstimmung mit einer bestimmten Klasse, beispielsweise der Klasse „Kathodenversagen" oder „Verschleißteilversagen", liefert. Falls nein, werden in Schritt S4-5 die N ältesten Samples aus dem Speicher entfernt und der Prozess fortgeführt.If so, the cutting process is terminated in step S4-9 and, for example, the error message "wear part failure" is displayed on a display unit before the end is reached in step S4-10. In the illustrated embodiment, further classes can be defined as "regular process flow" (here the plasma arc process continues) and "irregular process flow" (here the process is also continued, but a warning message is displayed to a user, for example, in the case of the fluctuations at 0.7 s shown in Figure 2).

[0049] In general, a "learning phase" can also be provided, in which the evaluation unit 6 stores a database with several patterns and the associated classes or learns them using a classification process. For example, an artificial neural network or a "support vector machine" can also be used here.

[0050] Figure 5 shows an electrical voltage plotted against an electrical current to illustrate different classes. As a result of the classification, the signal curve is divided into a "new wear parts" class, a "used wear parts" class, or a "worn wear parts" class. Further classes can be made based on the cathode burnback in the range from 0 percent to 100 percent relative to a maximum possible burnback. Figures 6 and 7 show signal curves in which an electrical voltage in V was plotted against time in s, and example signal curves are shown for the "defective wear parts" class (Figure 6), for the "movement out of the sheet" class (solid line in Figure 7), "joint overrun" class (dashed lines in Figure 7), and for the "piercing process completed" class (alternating dashed and dotted lines in Figure 7).For example, the classes "idling", "ignition of a pilot arc", "maintenance of the pilot arc", "piercing into the workpiece", "piercing through the workpiece", "cutting the workpiece" or "traversing a workpiece edge" can also be defined and the characteristics assigned accordingly.

[0051] Figure 8 shows the result of a Fourier transform of a signal curve obtained with a sampling rate of 50 kHz. In the gray-shaded frequency ranges, the amplitudes are summed (LI and L2) and the sums are used as features for classification, i.e., based on the obtained numbers, a feature vector can be created and assigned to a specific class. By using the Fourier transform, the signal curve can be evaluated independently of the magnitude of the voltage amplitudes or similar and the rates of change. Figure 8 shows a signal curve of a regular cutting process (solid curve) and a signal curve of a defect in a wearing part (dashed curve). Between 50 Hz and 100 Hz, beats are present; these are caused by the current source 4 and are therefore neglected in the evaluation.The two resulting sums correspond to two features, and the resulting numerical values ​​allow for a clear distinction. In the illustrated embodiment, the last 327 ms of the process were buffered in a memory, resulting in 16,384 samples, which can be reduced to just two digits for calculating class membership by determining the features in the amplitude spectrum. A "k nearest neighbor" classifier was used as the classifier in the illustrated embodiment.In further embodiments, however, depending on the application, a classifier such as a "linear support vector machine (linear SVM)", "radial basis function support vector machine (RBF SVM)", "Gaussian process", "random forest", "neural network", "adaptive boosting", "naive Bayes" or "quadratic (quadratic discriminant analysis, QDA)" or a calculation of a distance between the feature vectors belonging to the classes and the feature vector to be examined can also be used. The calculation of the class membership for a feature combination is carried out by determining the k nearest objects classified from the training phase. A feature or object is assigned to the class that occurs most frequently among the k nearest neighbors.

[0052] Figure 9 shows a parameterization of individual sub-areas. In the diagrams (the left diagram represents new wear parts, the right diagram represents wear parts approaching the end of their service life), the electrical voltage between the electrode and nozzle is plotted against the current intensity, and the ranges between 75 A and 150 A, 170 A and 250 A, and between 250 A and 300 A are used for a regression calculation. The coefficients a1, b1, c1, a2, b2, c2 and a3, b3, c3, d3, obtained from the regression calculation, can be used as features for classification. In general, however, the amplitude, a concave or convex curve, and / or a slope can also be used as features.

[0053] Figure 10 shows a schematic two-dimensional feature space in which two different features are plotted on the axes and a feature vector formed from these features is directed to a specific point in the feature space. To create the feature space shown in Figure 10, in a first step, the features (shown as points in Figure 10) are collected from a training data set and assigned to classes. By training a classifier, a function is obtained that automatically assigns a class to a new feature vector. The classifier thus assigns the class with the "highest" match to a feature combination, as exemplified by the two arrows. In the embodiment shown in Figure 10, therefore, three classes are formed. Finally, Figure 11 shows a real feature space in which the endpoints of the feature vectors are plotted.The three classes correspond to "destruction of wear parts", "regular cutting area" and "irregular process flow".

[0054] Only features of the various embodiments disclosed in the exemplary embodiments can be combined with one another and claimed individually.

Claims

Patent claims Method for detecting a process state of a plasma arc process, in which a workpiece (1) is processed by a plasma torch (2) and measured values ​​of a temporal profile of a physical variable of the plasma arc process are detected and a signal profile is determined from the measured values ​​(S3-1), wherein pattern recognition is carried out on the signal profile (S3-2) by extracting and classifying at least one feature from the signal profile, and depending on the classification of the feature, the signal profile is assigned to a specific state of the plasma arc process (S3-4). Method according to claim 1, characterized in that the physical variable is / are an electrical voltage, an electrical voltage drop and / or an electrical current.Method according to one of the preceding claims, characterized in that the at least one feature is determined from amplitudes of the measured values ​​and / or an amplitude spectrum. Method according to claim 3, characterized in that the at least one feature is determined from a sum of the amplitude spectrum over at least one predetermined frequency range. Method according to one of the preceding claims, characterized in that the at least one feature is determined from a parameterization of the signal curve. Method according to claim 5, characterized in that the at least one feature is determined from a parameterization of individual subsections of the signal curve by forming a sequence of mean values. and / or rates of change and / or regression values ​​of a regression analysis.

7. Method according to one of the preceding claims, characterized in that at least two features are extracted from the signal curve and classified.

8. Method according to one of the preceding claims, characterized in that classes of the classification comprise at least one stable process state and at least one unstable process state.

9. Method according to claim 8, characterized in that if the process state is classified as stable, the plasma arc process is continued unchanged and if the process state is classified as unstable, the plasma arc process is discontinued.

10. Method according to one of the preceding claims, characterized in that the plasma arc process is plasma cutting and the classes used are idle, ignition of a pilot arc, burning of the pilot arc, piercing the workpiece, piercing the workpiece, cutting the workpiece, passing over a kerf, passing over a workpiece edge, presence of a damaged wear part, presence of a new wear part, presence of a used wear part and / or presence of a worn wear part.

11. Method according to claim 10, characterized in that a classification into the classes presence of a damaged wear part, presence of a new wear part, presence of a used wear part and / or presence of a worn wear part is carried out based on a burn-back of the cathode or based on a maximum service life.

12. Device for detecting a process state of a plasma arc process, with a plasma torch (2) for processing a Workpiece (1), a measured value recording device (5) for recording measured values ​​of a temporal profile of a physical quantity of the plasma arc process and for determining a signal profile from the measured values, an evaluation unit (6) for carrying out pattern recognition on the signal profile, which is designed to extract and classify at least one feature from the signal profile and, depending on the classification of the feature, to assign the signal profile to a specific state of the plasma arc process.

13. Device according to claim 12, characterized by an inductive Component (7) connected in series with the plasma torch, wherein the measured value recording device (5) is configured to measure an electrical current or an electrical voltage drop as the physical quantity.

14. A computer program product comprising a computer program, comprising software means for carrying out a method according to one of claims 1 to 11 and / or for controlling a device according to one of claims 12 or 13, when the computer program is executed in an automation system.

Images