JP2010052009A - Abnormality sensing method of laser beam welding system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an abnormality sensing method of a laser beam welding system capable of simply sensing any abnormality element with excellent accuracy. <P>SOLUTION: A weld part is formed on a reference workpiece, the reference state data on the welding state is acquired, the reference state data is normalized to obtain the reference normalized data, the SN ratio of the MT (Mahalanobis-Taguchi) system is calculated based on the reference normalized data, and the SN ratio is applied to a neural network as the teacher data (S11-S15). By applying the SN ratio to a plurality of reference workpieces, a neural network model is constructed (S16). Successively, a weld part is formed on a welding workpiece, the welding state data on the welding condition is acquired, the welding state data is normalized to obtain the welding normalized data, the SN ratio of the MT system is calculated based on the welding normalized data, and any abnormality element is sensed based on the SN ratio and the constructed neural network model. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an abnormality detection method for a laser welding system for detecting an abnormality in a laser welding system.

As a method for detecting an abnormality in a conventional laser welding system, a workpiece is irradiated with a laser beam to form a welded portion, and state data relating to a welding state at that time is acquired, and based on this state data, a welding abnormality is detected. A device for determining the presence or absence is known (for example, see Patent Document 1).
JP 2007-253220 A

  Here, in the above-described abnormality detection method of the laser welding system, although the presence or absence of welding abnormality can be determined as described above, the elements constituting the laser welding (for example, the installation posture of the laser transmitter, the gas nozzle, the gas component, and the gas) It is difficult to accurately detect which of the flow rate) is an abnormal element. Further, in recent laser welding system abnormality detection methods, those that can easily detect abnormal elements are particularly desired.

  Then, this invention makes it a subject to provide the abnormality detection method of the laser welding system which can detect an abnormal element simply and accurately.

  In order to solve the above problems, an abnormality detection method for a laser welding system according to the present invention is an abnormality detection method for detecting an abnormality in a laser welding system in which a welded portion is formed on a workpiece by laser beam irradiation. A pre-weld process for constructing a neural network model for detecting an abnormal element, and a welding main process for detecting an abnormal element based on the neural network model constructed in the pre-weld process. A reference process is performed by irradiating a reference workpiece with a laser beam to form a welded portion, and acquiring reference state data relating to the welding state at that time, and normalizing the acquired reference state data The second step of obtaining normalized data for the MT, and calculating the SN ratio of the MT system based on the obtained normalized data for reference, A third process that is applied to the network as teacher data, and a fourth process that constructs a neural network model by performing the first to third processes on a plurality of reference workpieces, Is to form a welded portion by irradiating a workpiece to be processed with a laser beam, and to normalize the acquired processing state data and a fifth step of acquiring processing state data relating to the welding state at that time The 6th step of obtaining the machining normalization data in step 6 and calculating the SN ratio of the MT system based on the obtained machining normalization data, and abnormalities based on this SN ratio and the neural network model constructed in the pre-welding process And a seventh step of detecting the element.

  In this laser welding system abnormality detection method, an abnormal element is detected by suitably cooperating the MT system and the neural network. That is, in general, in the detection of abnormality by MD (Mahalanobis distance) of the MT system, for example, if there are a large number of elements that should be detected as abnormal elements, quantitative evaluation becomes difficult and detection accuracy decreases. . In this regard, in the present invention, while the MT system is applied, abnormal elements are detected by a neural network regardless of the MD of the MT system. Therefore, an abnormal element can be detected with high accuracy. On the other hand, in the detection of anomaly by a neural network, usually, a very large number (for example, about 100 to 1000) of teacher data is required when constructing a neural network model. In this regard, in the present invention, the SN ratio of the MT system is applied to the teacher data. Further, the SN ratio is calculated based on normalized data characterized by normalization (a difference between normal and abnormal is emphasized). Therefore, the influence of variations in teacher data can be minimized, the number of necessary teacher data can be minimized, and an accurate neural network model can be constructed. Therefore, according to the present invention, an abnormal element can be detected easily and accurately.

  Here, specifically, the state data may relate to at least one of laser output, plasma, reflected light, temperature, sound, vibration, assist gas pressure, and supply power.

  Moreover, it is preferable that SN ratio is a SN ratio of a desired size characteristic. In this case, an abnormal element can be detected more easily while ensuring sufficient accuracy.

  In addition, there are a plurality of elements detected as abnormal elements, and the SN ratio is preferably calculated as SN ratio data including an SN ratio group divided for each of the plurality of elements. As a result, when there are a plurality of elements detected as abnormal elements, it is not necessary to calculate the SN ratio for each element and construct a neural network model. That is, it is possible to identify and detect a plurality of abnormal elements with one neural network model.

  According to the present invention, an abnormal element can be detected easily and accurately.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In the present embodiment, terms such as “upper” and “lower” are based on the state shown in the drawings and are for convenience.

  FIG. 1 is a schematic diagram showing a configuration of a laser welding system according to an embodiment of the present invention. As shown in FIG. 1, the laser welding system 1 according to the present embodiment has an abnormality or a sign of abnormality among elements (laser oscillator, installation position of a gas nozzle, gas component, gas flow rate, etc.) constituting laser welding. Is detected and identified as an abnormal element. Specifically, in the laser welding system 1, an abnormal element is detected by suitably cooperating an MT system and a neural network. First, each of the MT system and the neural network will be described.

[MT system]
In general, the MT (Mahalanobis-Taguchi) system is a multivariate analysis method for discriminating a change in state using a basic concept of MD. That is, the MT system handles a plurality of measured quantities (multivariables) by expressing them in MD (one variable). In this MT system, for example, the algorithm is constructed in the following order (1) to (3).

(1) Construction of unit space First, as shown in FIG. 2, data necessary for detecting an abnormal element is measured (detected) and assumed to be normal and abnormal data. Then, characterization, which is numerical processing that emphasizes the difference between normal and abnormal, is performed on the measured measurement amount. By characterizing the measurement amount in this way, it is possible to suppress normal data from becoming abnormal data due to various factors, and information used for abnormality detection is accurately extracted. Thereafter, a unit space is formed by the measured amount and the characterized items (A ′, B ′, etc. in the figure).

(2) Effectiveness analysis Next, an effectiveness analysis is performed to determine the validity of the data. The evaluation at this time can be performed based on the S / N ratio that is a ratio of information useful for detecting abnormality and harmful information. Specifically, as shown in FIG. 3, the SN ratio is calculated for each item by assigning the data characterized in (1) above to the orthogonal table. Here, as the orthogonal table, and using the L ₈ orthogonal table. In the orthogonal table, 1 and 2 indicate the level, "1" means that the item is used (included in the unit space), and "2" does not use the item (excluded from the unit space). Means. The column numbers (A, B... In the figure) correspond to elements detected as abnormal elements.

  Thereby, as shown in FIG. 4, as a result of the effectiveness analysis, SN ratio data in which the SN ratio is shown for each element is obtained. In addition, the vertical axis | shaft in a figure means that there exists an effect for abnormality detection, so that it is large.

(3) Calculation of MD Next, an item with a large SN ratio (factor effect) is selected from the SN ratio data obtained in (2) above, and MD is calculated. For details of MD calculation, refer to Japanese Patent Application Laid-Open No. 2006-160153.

[neural network]
In general, a neural network is a mathematical model that imitates the information processing mechanism of the cranial nervous system, and is a method of constructing an algorithm by giving various patterns as teacher data and using the pattern difference as an abnormal state.

  As shown in FIG. 5, the neural network 50 here is of a hierarchical type, and is expressed as a hierarchical structure in which arithmetic elements called units 51 are combined in layers. The neural network 50 includes an input layer in which an input unit 52 for receiving data from the outside is arranged, an output layer in which an output unit 53 for outputting data to the outside is arranged, and a hidden layer (intermediate layer) between these layers. It is equipped with. In addition, an arrow 55 in the drawing represents a signal flow.

  According to this neural network, unlike the MT system, the result is binarized like “◯”, “×”, “Yes”, “No”. Therefore, it is possible to clearly determine the occurrence of abnormality.

  Next, the laser welding system 1 of this embodiment will be described.

  Returning to FIG. 1, the laser welding system 1 forms a linear welded portion W at a substantially central portion of the workpieces (workpieces) 10 </ b> A and 10 </ b> B, and lap welds the workpieces 10 </ b> A and 10 </ b> B. As the workpieces 10A and 10B, plate-like outer plate panels and bone members used for a railway vehicle structure are used. The laser welding system 1 includes a laser welding apparatus 2 that forms a welded portion W by irradiating a laser beam, and an evaluation apparatus 3 that evaluates whether or not laser welding has been normally performed.

  The laser welding apparatus 2 includes a feeding device 21, a workpiece fixing device 22, a laser irradiation device 23, and a gas supply device 24. Each of these devices 21 to 24 is connected to a host control device (not shown), and automatically executes each operation in accordance with operation instruction information output from this control device.

  The feeding device 21 scans the irradiation position of the laser beam on the workpieces 10A and 10B. Specifically, the feeding device 21 moves the workpieces 10 </ b> A and 10 </ b> B placed on the movable stage 25 relative to the laser beam by the laser irradiation device 23 along the planned welding region R.

  The workpiece fixing device 22 fixes the workpieces 10A and 10B to the movable stage 25. In the work fixing device 22, both end portions of the works 10 </ b> A and 10 </ b> B sandwiching the planned welding region R are pressed against the movable stage 25 by a long press plate 26 a. Thereby, the adhesion of the workpieces 10A and 10B in the vicinity of the planned welding region R is improved.

The laser irradiation device 23 irradiates a laser beam toward the planned welding region R of the workpieces 10A and 10B. Specifically, the laser irradiation device 23 emits a laser beam such as a YAG laser or a CO ₂ laser for a predetermined time from the tip of the laser head 27 above the workpieces 10A and 10B. The laser irradiation device 23 includes an output switching mechanism (not shown) inside, and can be switched between continuous irradiation with the laser beam and irradiation with the laser beam in pulses. .

  The gas supply device 24 supplies assist gas to the planned welding region R of the workpieces 10A and 10B. In the gas supply device 24 here, the supply nozzle 28 is disposed so as to be inclined by about 30 degrees with respect to the thickness direction of the workpieces 10A and 10B. The gas supply device 24 supplies assist gas to laser beam irradiation positions (hereinafter referred to as “processing points”) of the workpieces 10A and 10B with a predetermined supply amount. As the assist gas, argon gas or the like is used for the purpose of preventing oxidation and sputtering of the workpieces 10A and 10B.

  On the other hand, the evaluation device 3 is physically a computer system including a CPU, a memory, a communication interface, a storage unit such as a hard disk, a display unit such as a display, and the like. A light intensity detection sensor 4 is connected to the evaluation device 3 via a filter 5.

  The light intensity detection sensor 4 is a photosensor for detecting the light intensity at the processing point. The light intensity detection sensor 4 is disposed in the vicinity of the irradiation position of the laser beam. The filter 5 is an optical filter. The filter 5 here selectively allows light intensity related to plasma (wavelength 1100 nm or less) generated at the processing point and light intensity related to the temperature of processing point (wavelength in the infrared region) to pass through. Accordingly, an output signal is input from the light intensity detection sensor 4 to the evaluation device 3 via the filter 5, and state data regarding the welding state (here, plasma and temperature) is input to the evaluation device 3.

  The evaluation device 3 includes an abnormal element detection unit 31 and a storage unit 32 as functional components. The abnormal element detector 31 detects an abnormal element based on the input state data (details will be described later). The storage unit 32 receives and stores detection result information output from the abnormal element detection unit 31. The storage unit 32 stores a neural network model for detecting an abnormal element.

  It can be said that the MT system and the neural network are difficult to cooperate for the following reasons, for example. In other words, the time when each method was born is different. Specifically, neural networks are studied with a peak in the 1980s-1990s, while MT systems are studied with a peak in the 2000s. In addition, since a neural network performs calculations from a large amount of data, it is very difficult to perform calculations using a general-purpose computer before the 2000s. Therefore, conventionally, the neural network has been regarded as a practically difficult technique in terms of calculation cost, slow abnormality detection speed estimated from the calculation speed, and the like.

  Next, processing for detecting an abnormal element in the laser welding system 1 having the above-described configuration will be described with reference to the flowcharts of FIGS. In the following description, abnormal elements are detected for a plurality of elements. Specifically, the first element is a gas flow rate, the second element is a gas component, and the third element is a gas nozzle angle. Yes.

  In this laser welding system 1, laser welding is performed on reference workpieces 10A and 10B (hereinafter referred to as “reference workpieces 10A and 10B”), and a neural network model for detecting abnormal elements is constructed in advance ( Pre-welding process: S1). Then, laser welding is performed on the workpieces for processing (hereinafter referred to as “processing workpieces 10A and 10B”), and these are welded to each other, and abnormal elements are detected based on the constructed neural network model to perform laser welding. Is evaluated (inspected) (welding main process: S2).

[Pre-welding process]
First, the pre-welding process will be described. The reference workpieces 10A and 10B are placed on the movable stage 25, and the pressurizing jig 26 is lowered from above to fix the reference workpieces 10A and 10B to the movable stage 25. Subsequently, the laser head 27 emits a laser beam, and the movable stage 25 is moved at a speed of about 10 m / s to scan the reference workpieces 10A and 10B in the direction of arrow A (see FIG. 1). At the same time, the assist gas is supplied at a flow rate of 15 L / min by the gas supply device 24. As a result, a welded portion W is formed along the planned welding region R of the reference workpieces 10A and 10B (S11).

  Here, when forming the welded portion W, the light intensity detection sensor 4 detects the light intensity at the processing point via the filter 5. Thereby, for each of the plasma and temperature, an output signal of a waveform pattern is acquired as state data (hereinafter referred to as “reference state data”) (S12).

  Subsequently, in order to characterize the acquired reference state data, the reference state data is normalized to obtain reference normalized data (S13). Specifically, the reference state data is normalized, the number of the normalized reference state data crossing the normalized value (hereinafter referred to as “change amount”) is obtained, and the normalized reference state data is The number exceeding the normalized value (hereinafter referred to as “abundance”) is obtained. Note that the normalization value is set in advance for each element, and five normalization values here are set for each element.

Subsequently, the effectiveness analysis is performed based on the obtained reference normalization data, and the SN ratio of the MT system is calculated (S14). Specifically, by assigning the reference normalized data to the orthogonal table, the SN ratio of the MT system is calculated as the reference SN ratio data including the SN ratio group divided for each element. The signal-to-noise ratio is preferable from the viewpoint that an abnormal element can be detected. The orthogonal table, suitable one according to the number of data of the state data are used, for example L ₄ orthogonal table, L ₈ orthogonal table, L ₁₂ orthogonal table and L ₁₆ orthogonal table and the like.

  Subsequently, the calculated reference signal-to-noise ratio data is applied to the neural network as teacher data (S15). And said S11-S15 are repeatedly implemented with respect to several reference | standard workpiece | work 10A, 10B. Here, for each of the first to third elements, the reference SN ratio data when it is an abnormal element is repeatedly applied as teacher data 3 to 4 times. As a result, a neural network model that can detect and identify an abnormal element from the first to third elements based on a difference in the SN ratio pattern of the SN ratio data is constructed (S16).

  FIG. 9 is a diagram illustrating an example of the reference SN ratio data. The reference signal-to-noise ratio data D1 in the figure shows data when the second element is an abnormal element. The reference SN ratio data D1 includes a plurality of SN ratios divided for each of the first to third elements. Further, in the reference SN ratio data D1, the SN ratio of the reference state data, the SN ratio of a plurality of changes, and the SN ratio of a plurality of abundances are calculated for each of plasma and temperature. Note that the SN ratio data may not calculate the SN ratio of the state data (the same applies hereinafter).

  As shown in FIG. 9, it can be seen that the reference SN ratio data D1 has the following tendency. That is, it can be seen that the SN ratio tends to be positive in the item of the second element which is an abnormal element. Among them, the SN ratio of the temperature item tends to be particularly large. It can also be seen that the SN ratio tends to be particularly large in one item of temperature in the first element that is not an abnormal element.

  FIG. 10 is a diagram illustrating another example of the reference SN ratio data. The reference signal-to-noise ratio data D2 in the figure shows data when the third element is an abnormal element. The reference SN ratio data D2 has the same configuration as the reference SN ratio data D1. That is, it is configured to include a plurality of SN ratios divided for each of the first to third elements, and for each of the plasma and temperature, the SN ratio of the reference state data, the SN ratio of the plurality of changes, The SN ratio of the abundance is calculated.

  As shown in FIG. 10, it can be seen that the reference SN ratio data D2 has the following tendency. That is, in the item of the third element that is an abnormal element, there is no particular tendency in the SN ratio. On the other hand, it can be seen that the S / N ratio tends to increase in one item of plasma in the first and second elements which are not abnormal elements.

[Welding main process]
Next, the welding main process will be described. First, similarly to S11 described above, the processing workpieces 10A and 10B are placed on the movable stage 25, and the pressing jig 26 is lowered from above to fix the processing workpieces 10A and 10B to the movable stage 25. Subsequently, the laser head 27 emits a laser beam, and the movable stage 25 is moved at a speed of about 10 m / s to scan the workpieces 10A and 10B in the direction of arrow A (see FIG. 1). At the same time, the assist gas is supplied at a flow rate of 15 L / min by the gas supply device 24. Thereby, the welding part W is formed along the welding planned area | region R of the workpiece | work 10A, 10B for a process (S21).

  Here, when the welded portion W is formed, the light intensity detection sensor 4 detects the light intensity at the processing point via the filter 5. Thereby, for each of the plasma and temperature, an output signal of a waveform pattern is acquired as state data (hereinafter referred to as “processing state data”) (S22).

  Subsequently, in order to characterize the acquired machining state data as in S13 described above, the machining state data is normalized to obtain normalized processing data (S23). Specifically, the machining state data is normalized, and the change amount and the existence amount of the normalized machining state data are obtained.

  Subsequently, as in S14 described above, the effectiveness analysis is performed based on the processing normalization data, and the SN ratio of the MT system is calculated (S24). Specifically, by assigning the processing normalization data to the orthogonal table, the SN ratio of the MT system is calculated as the processing SN ratio data including the SN ratio group divided for each element.

  Subsequently, based on the calculated processing SN ratio data and the neural network model constructed in S16, an element in which welding abnormality has occurred is detected as an abnormal element (S25). That is, by applying the processing SN ratio data to the neural network model as unknown data, an abnormal element is identified and specified from the difference in the pattern of each SN ratio included in the processing SN ratio data.

  When an abnormal element is detected, the detection result is stored in the storage unit 32, and the laser welding system 1 is completely stopped (S26 → S27). After that, maintenance of abnormal elements is performed. On the other hand, if no abnormal element is detected, the above-described processes of S21 to S27 are repeated until laser welding is completed (S28 → S21).

  Here, in general, in the detection of abnormality by MD of the MT system, for example, if there are a large number of elements that should be detected as abnormal elements, quantitative evaluation becomes difficult and detection accuracy decreases. In this regard, in the present embodiment, as described above, the MT system is applied, but the neural network does not depend on the MD of the MT system (without performing the above “(3) MD calculation”). An abnormal element is detected. Therefore, for example, even when a plurality of abnormal elements are detected, the abnormal elements can be detected with high accuracy.

  On the other hand, in general, in detecting anomalies using a neural network, a very large number (for example, about 100 to 1000) of teacher data is required when a neural network model is constructed. This number of data is several tens of times the number of data required in the MT system. The fact that the neural network requires a large number of teacher data indicates that the variation of the teacher data is large. In this regard, in this embodiment, the SN ratio of the MT system is applied to the teacher data. Further, the SN ratio is calculated based on normalized data characterized by normalization. Therefore, the influence of variations in teacher data can be minimized, the number of necessary teacher data can be minimized (in this case, 3 to 4), and an accurate neural network model can be constructed.

  Therefore, according to the present embodiment, the MT system and the neural network can preferably cooperate with each other, and an abnormal element can be automatically detected (abnormal diagnosis). That is, in order to compensate for the shortcomings of the MT system and the shortcomings of the neural network, the respective elements are emphasized with each other and the abnormal elements are automatically detected. As a result, an abnormal element can be detected easily and accurately. Furthermore, since the abnormal elements can be detected with high accuracy, the main parts of the laser welding system 1 can be monitored, and preventive maintenance can be realized. As a result, it is possible to suppress a decrease in welding quality due to a failure of the laser welding system 1 and to improve the yield.

  FIG. 11 is a chart showing abnormal element detection results by the laser welding system of FIG. In the detection result in the figure, the abnormal element is known as a determination value for confirmation of detection accuracy. In the results shown in the first to seventh stages in the figure, the SN ratio data is applied to the neural network as teacher data to construct a neural network model. In addition, the SN ratio data is applied to the neural network model as unknown data to confirm the detection of abnormal elements. In the result shown in the eighth row in the figure, the S / N ratio data is applied to the neural network model as unknown data, and abnormal elements are detected. As shown in FIG. 11, according to the present embodiment, it is understood that an abnormal element is detected with a correct answer rate of 100% by applying the SN ratio data to the neural network 3-4 times as teacher data. Therefore, it is possible to confirm the above effect of detecting the abnormal element easily and accurately.

  In the present embodiment, as described above, the S / N ratio is the S / N ratio having the desired characteristics. Therefore, it is possible to detect an abnormal element more easily while ensuring sufficient accuracy. Note that the S / N ratio may be the S / N ratio of the dynamic characteristics or the like in some cases.

  In the present embodiment, as described above, there are a plurality of elements detected as abnormal elements, and the S / N ratio is calculated as S / N ratio data composed of S / N ratio groups divided for each of the plurality of elements. (See FIGS. 9 and 10). In this case, it is not necessary to calculate the SN ratio for each element and construct a neural network model for each element. That is, a plurality of abnormal elements can be specified and detected with one model.

  As described above, the calculated S / N ratio may show a characteristic tendency in the item of an element that is not an abnormal element (see FIG. 10). In this respect, in this embodiment, the SN ratio is not calculated for each element, but is calculated as SN ratio data, and this SN ratio data is applied to the neural network. Therefore, even when the S / N ratio shows a characteristic tendency in the item of an element that is not an abnormal element, a neural network model can be constructed in consideration of the characteristic. Therefore, it is possible to detect an abnormal element with higher accuracy.

  The preferred embodiment of the present invention has been described above, but the present invention is not limited to the above embodiment.

  For example, in the above-described embodiment, the light intensity detection sensor 4 detects the light intensity at the processing point via the filter 5 to acquire the state data related to plasma and temperature. In addition, status data regarding the reflected light of the laser beam may be detected. The state data to be acquired is appropriately selected depending on the abnormal element (abnormal part) to be detected, the type of laser beam, the material of the workpiece, and the like.

  Further, a processing sound detection sensor for detecting the sound of the processing point or a vibration detection sensor for detecting the magnitude of vibration of the workpieces 10A and 10B may be further provided, and state data relating to the sound or vibration may be acquired. . In this case, FTT (Fast Fourier Transform) processing is performed on the output signal of the sensor.

  The apparatus further includes a supply power detection sensor for detecting the amount of power supplied to each device, or a gas pressure detection sensor for detecting the pressure of the assist gas, and obtains status data relating to the supply power or the assist gas pressure. May be. In this case, moving average processing is performed on the output signal of the sensor.

  In the above embodiment, abnormalities in these elements are detected using the gas flow rate, gas component, and gas nozzle angle as elements, but in addition to these, for example, abnormalities in the laser head 27 that is a laser oscillator, and around the laser irradiation device 23 Abnormalities and the like at various sites may be detected. Further, any one or more of these may be detected.

  Moreover, in the said embodiment, although the variation | change_quantity and the existing amount were calculated | required as normalization data, you may obtain | require any one of these as normalization data.

It is the schematic which shows the structure of the laser welding system which concerns on one Embodiment of this invention. It is a figure for demonstrating construction of the unit space in MT system. It is a figure which shows an example of the orthogonal table in MT system. It is a figure which shows the result of the effectiveness analysis in MT system. It is the schematic which shows the structure of a neural network system. It is a flowchart of the process in the laser welding system of FIG. It is a flowchart in the process before welding of FIG. It is a flowchart in the welding main process of FIG. It is a diagram which shows an example of reference | standard SN ratio data. It is a diagram which shows another example of reference | standard SN ratio data. It is a graph which shows the abnormal element detection result by the laser welding system of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Laser welding system, 10A, 10B ... Workpiece (workpiece), 50 ... Neural network, D1, D2 ... SN ratio data for reference | standard, W ... Welding part.

Claims (4)

An anomaly detection method for detecting an anomaly in a laser welding system that forms a weld on a workpiece by laser beam irradiation,
A pre-welding process for constructing a neural network model for detecting abnormal elements;
A main welding process for detecting the abnormal element based on the neural network model constructed in the pre-welding process,
The pre-welding process includes
A first step of irradiating the reference workpiece with the laser beam to form the weld and obtaining reference state data relating to the weld state at that time;
A second step for obtaining reference normalization data by normalizing the acquired reference state data;
A third step of calculating the SN ratio of the MT system based on the obtained normalized data for reference, and applying this SN ratio to the neural network as teacher data;
A fourth step of constructing the neural network model by performing the first to third steps on a plurality of reference workpieces,
The welding main process includes
A fifth step of irradiating the workpiece to be processed with the laser beam to form the weld, and obtaining processing state data relating to the welding state at that time,
A sixth step of obtaining the machining normalization data by normalizing the obtained machining state data;
A seventh step of calculating an SN ratio of the MT system based on the obtained normalization data for processing, and detecting the abnormal element based on the SN ratio and the neural network model constructed in the pre-welding process; An abnormality detection method for a laser welding system, comprising:   2. The laser welding system abnormality according to claim 1, wherein the state data relates to at least one of laser output, plasma, reflected light, temperature, sound, vibration, assist gas pressure, and supply power. Detection method.   3. The laser welding system abnormality detection method according to claim 1, wherein the SN ratio is an SN ratio of a desired characteristic. There are a plurality of elements detected as the abnormal elements,
The abnormality detection method for a laser welding system according to any one of claims 1 to 3, wherein the SN ratio is calculated as SN ratio data including an SN ratio group divided for each of the plurality of elements. .

Images