Estimation method of penetration depth using acoustic signals during laser welding

Abstract

To achieve an estimating method for a weld penetration depth utilizing an acoustic signal, preferably in real time, at a site during laser welding, whose technique and technical approach are not known at all conventionally, a determining method for welding defect using the estimating method, and a determining device or the like, and to provide an estimating technique for a weld penetration depth using a microphone which is lower in cost than an OCT and can easily and quickly sense and estimate a weld penetration depth.SOLUTION: An estimating method for a weld penetration depth of molten pool during laser welding includes a step of estimating a weld penetration depth in real time, on the basis of a correlation between strength of an acoustic signal with specific frequencies from molten pool and a weld penetration depth, which has previously been learned, from strength of an acoustic signal with the specific frequencies detected during laser welding.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for estimating penetration depth, preferably in real time on-site, using acoustic signals generated during laser welding by a laser welder or the like, and a method and device for determining defective welding using the estimation method. [Background technology]

[0002] In recent years, laser welding has become increasingly popular in production lines in the automotive and secondary battery industries to increase strength and automation. Laser welding achieves much deeper penetration than conventional arc welding, enabling high-strength welding even on thick metal plates. However, because the melting phenomenon during laser welding is highly volatile and unstable, it has recently become essential to use a testing device that constantly assesses the quality of the weld.

[0003] Previously, such devices were primarily used to determine burn-through and poor weld bead quality that occurred during welding based on surface heat and images before, during, and after welding, but users also needed technology to detect the penetration depth during welding. The main purpose of depth detection is that penetration depth is a parameter directly linked to strength, and can also be used as a guide for when you want to weld deeply without allowing the laser beam to penetrate the weld material and affect the underlying parts.

[0004] In order to grasp the quality of the welding state and the quality of the welding results during laser welding, attempts have been made in the past to monitor the welding state and grasp the quality of defects by detecting sound signals from the vicinity of the welding point, processing and analyzing these signals in addition to or instead of detecting reflected light / radiated light, vibration or temperature.

[0005] For example, Patent Document 1 listed below states, "In step 360 of FIG. 7(B), each sound profile is integrated over a time interval of 0.0 ms to 0.5 ms, and in step 362 the maximum value for each waveform is found. In step 364, the waveform with the largest integrated value is selected as the waveform due to welding at the focused position. In step 370 of FIG. 7(C), each sound waveform is integrated over a time interval of 5 ms to 11.5 ms, and in step 372 the minimum value for each waveform is found. In step 374, the waveform with the smallest minimum value is selected as the waveform due to welding at the focused position. In step 380 of FIG. 7(D), the slope of each sound waveform is read over a time interval of 0.25 ms to 0.4 ms, and in step 382 the maximum value for each waveform is found. In step 384, the waveform with the largest maximum value is selected as the waveform due to welding at the focused position" (paragraph

[0053] , etc.).

[0006] Furthermore, Patent Document 2 listed below states that "FIG. 2(a) shows the frequency characteristics when welding is performed well, and FIG. 2(b) shows the frequency characteristics when a welding defect occurs. As is clear from these drawings, when there is a welding defect (FIG. 2(b)), the intensity of the sound signal is weaker in a frequency band of approximately 4 to 7 kHz compared to when welding is performed well (FIG. 2(a)). Therefore, if only sound signals in this frequency band are acquired and used to determine whether there is a defect, a more accurate determination can be made than if sound signals in the entire frequency band were used to determine whether there is a defect. In the processing example shown in FIG. 2, features appear in the frequency band of 4 to 7 kHz, but the frequency band in which features appear varies depending on the processing conditions, so it is preferable to set the frequency band to be acquired appropriately" (paragraph

[0024] , etc.).

[0007] Furthermore, Patent Document 3 describes that "when laser welding is performed, the intensity of the sound during laser welding is measured by sensor 14. If the intensity of this sound is weaker than a predetermined level g, discrimination unit 16 determines that there is a defective weld, and memory unit 17 stores the coordinates of the defective weld. Then, if there is a defective weld after welding has been completed up to the end point of welding, the defective weld stored in memory unit 17 is re-welded."

[0008] Furthermore, Patent Document 4 below discloses that digital signal processing technology using FFT is employed in analyzing sound signals. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-066340 [Patent Document 2] Japanese Patent Application Publication No. 11-216583 [Patent Document 3] Japanese Patent Application Laid-Open No. 2001-321972 [Patent Document 4] Japanese Patent Application Laid-Open No. 2003-334679 Summary of the Invention [Problem to be solved by the invention]

[0010] It is difficult to measure the penetration depth described above using only information from the surface of the conventional welding material. In recent years, a technology called OCT (Optical Coherence Tomography) has been established that can estimate the depth with a certain degree of accuracy by irradiating a distance-measuring laser beam into a depression in the molten liquid called a keyhole that occurs in the center of the irradiated area during laser welding. However, this technology is very expensive (system price is around 10 to 20 million yen), and its structure makes it large in size. Furthermore, it has many limitations, such as restrictions on processing speed, so it cannot yet be said to be in widespread use.

[0011] For example, the invention disclosed in the above-mentioned Patent Document 1 discloses a technical idea of ​​performing calculations using the integrated value of a sound signal detected by a microphone during laser welding, but the integral calculation process requires a certain amount of processing time and lacks speed. Furthermore, the above-mentioned Patent Document 2 consistently explains that "the intensity of the sound signal weakens when there is a welding defect," and although there is a reference to frequency variation, there is no description or suggestion of sound characteristics other than "weakening of the sound signal intensity." Furthermore, Patent Document 3 describes that "the sound level decreases" when there is a poor weld, and discloses a technical idea of ​​determining that a weld is poor when the sound intensity is weaker than a predetermined level.

[0012] Furthermore, Patent Document 4 merely describes the processing of sound signals by digital processing. As can be understood from the above explanation, no method or technical idea for estimating or predicting penetration depth during laser welding from sound signals is known, and there have been no attempts or suggestions to do so, making it a technical point that has been completely overlooked in the past. On the other hand, in order to maintain stable and good welding quality, it is preferable to be able to quickly obtain knowledge about penetration depth during laser welding on the spot in real time.

[0013] Therefore, the present invention has been made in consideration of the above-mentioned problems, and aims to realize a method for estimating penetration depth using acoustic signals on-site during laser welding, preferably in real time, for which no techniques or technical approaches have been known until now, and a method and device for determining weld defects using this estimation method. Another aim of the present invention is to propose a penetration depth estimation technology using a microphone that is cheaper than OCT and that can detect and estimate penetration depth simply and quickly. [Means for solving the problem]

[0014] The method for estimating penetration depth during laser welding of the present invention is characterized by comprising a step of estimating penetration depth in real time from the intensity of an acoustic signal of a specific frequency detected during laser welding, based on a pre-learned correlation between the intensity of an acoustic signal of a specific frequency from a molten pool and the penetration depth.

[0015] In addition, in the method for estimating penetration depth during laser welding of the present invention, the pre-trained correlation is preferably such that the penetration depth is expressed by a linear expression of the intensity of an acoustic signal at a specific frequency. A method for estimating penetration depth during laser welding.

[0016] Furthermore, the method of estimating penetration depth during laser welding of the present invention is more preferably characterized in that the specific frequency is an acoustic signal in a frequency range of 5 to 9 kHz.

[0017] Furthermore, the method for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the intensity of the acoustic signal of a specific frequency during pre-learning is a value obtained by performing moving average processing on the absolute value of the detected acoustic signal.

[0018] Furthermore, the method for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the intensity of the acoustic signal of a specific frequency detected during laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool.

[0019] Furthermore, the method for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth.

[0020] The device for estimating penetration depth during laser welding of the present invention is characterized by comprising: a memory unit that stores a correlation between the intensity of an acoustic signal of a specific frequency from a molten pool and the penetration depth, which correlation is obtained in advance by pre-learning; and a calculation unit that calculates the penetration depth in real time from the intensity of the acoustic signal of the specific frequency detected during laser welding, based on the correlation stored in the memory unit.

[0021] The device for estimating penetration depth during laser welding of the present invention is preferably characterized in that the pre-trained correlation is expressed by a linear expression of the intensity of an acoustic signal at a specific frequency.

[0022] The device for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the specific frequency is an acoustic signal in a frequency range of 5 to 9 kHz.

[0023] The device for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the intensity of the acoustic signal of a specific frequency during pre-learning is a value obtained by performing moving average processing on the absolute value of the detected acoustic signal.

[0024] The device for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the intensity of the acoustic signal of a specific frequency detected during laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool.

[0025] The device for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth.

[0026] The device for determining poor welding performance during laser welding of the present invention is characterized by comprising: a calculation unit that calculates an estimated penetration depth in real time from the intensity of an acoustic signal of a specific frequency detected by a microphone during laser welding, based on the correlation between the intensity of an acoustic signal of a specific frequency from the molten pool and the penetration depth, which is pre-stored in a memory unit through advance learning; and a poor welding determination unit that determines that the welding is poor when the estimated penetration depth calculated by the calculation unit exceeds an upper threshold or falls below a lower threshold.

[0027] The apparatus for determining welding defects during laser welding according to the present invention is preferably characterized in that the pre-trained correlation is expressed by a linear expression of the penetration depth and the intensity of the acoustic signal at a specific frequency.

[0028] The device for determining welding defects during laser welding of the present invention is further preferably characterized in that the specific frequency is an acoustic signal in a frequency range of 5 to 9 kHz.

[0029] The device for determining welding defects during laser welding of the present invention is further preferably characterized in that, in calculating the correlation between the intensity of an acoustic signal of a specific frequency from the molten pool during pre-learning and the penetration depth, the intensity of the acoustic signal of the specific frequency is calculated by using a value obtained by performing moving average processing on the absolute value of the detected acoustic signal.

[0030] The device for determining welding defects during laser welding of the present invention is further preferably characterized in that the intensity of the acoustic signal of a specific frequency detected during laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool.

[0031] The device for determining welding defects during laser welding of the present invention is further preferably characterized in that the linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. [Effects of the Invention]

[0032] The present invention makes it possible to realize a method for estimating penetration depth using acoustic signals on-site during laser welding, preferably in real time, and a method and device for determining weld defects using the estimation method.The present invention also makes it possible to realize a penetration depth estimation technology using a microphone that is cheaper than OCT and that can detect and estimate penetration depth simply and quickly. [Brief explanation of the drawings]

[0033] [Figure 1] 1 is a schematic diagram illustrating the configuration of an apparatus according to an embodiment of the present invention. [Figure 2] FIG. 1 is a diagram illustrating an acoustic signal waveform of a microphone irradiated under the conditions shown in Table 1. [Figure 3] FIG. 10 is a diagram illustrating a spectrogram obtained by frequency analysis of a microphone waveform obtained in pre-learning. [Figure 4] (a) is a photograph of the surface of the laser weld, and (b) is a diagram illustrating the cross-sectional photograph. [Figure 5] This figure explains the results of extracting only 5 to 9 kHz from the output waveform of the microphone shown in Figure 2 using a filter circuit, correcting the waveform that oscillates on the negative side to the positive side using an absolute value circuit, and further smoothing the waveform using moving average processing (moving average between 2000 points for 5 μs sampling). [Figure 6] The penetration depth (vertical direction) of the cross section shown in FIG. 4(b) was measured at intervals of 0.1 mm in the horizontal direction, and this is a graph illustrating the penetration depth (mm) against the horizontal axis converted into time (s). [Figure 7] 7 is a diagram illustrating a scatter diagram graph in which the horizontal axis represents microphone amplitude intensity and the vertical axis represents penetration depth (mm) by combining the results of FIGS. 5 and 6. FIG. [Figure 8] 7 is a graph illustrating the estimated penetration depth (mm) calculated by substituting the microphone amplitude in FIG. 5 into the derived linear approximation formula, and the actually measured penetration depth (mm) in FIG. 6 , superimposed on each other. [Figure 9](a) is a micrograph showing the surface state after irradiating a sample with a laser at a constant intensity for 0.5 seconds under each of the experimental conditions shown in Table 2, and (b) is a diagram illustrating the cut surface observed under a microscope after cutting at the “intermediate cut” point shown in Figure 9(a). [Figure 10] This figure explains the results of the verification experiment. The horizontal axis represents the consecutive number of the samples, with sample numbers 1 to 3 corresponding to 400 W, sample numbers 4 to 6 corresponding to 450 W, sample numbers 7 to 9 corresponding to 500 W, ..., and sample numbers 25 to 27 corresponding to 800 W. The vertical axis on the left represents the penetration depth, and the right axis represents the difference in percentage between the actual measured depth and the depth estimated by calculation. DETAILED DESCRIPTION OF THE INVENTION

[0034] (Pre-learning) FIG. 1 is a schematic diagram illustrating the configuration of the apparatus of this embodiment. In FIG. 1, 110 denotes a calculation device, 120 denotes an absolute value circuit, 130 denotes a frequency filter circuit, 140 denotes a microphone, 150 denotes a laser processing head, 160 denotes a laser beam, and 170 denotes a weld metal (sample). A microphone with an integrated preamplifier (model: TYPE 4158N) manufactured by ACO Corporation with a frequency sensitivity range of 20 Hz to 100 kHz was used. The microphone was positioned approximately 80 mm apart from the processing point to the tip of the microphone, and was fixed at an angle of 45° from the vertical. The distance and angle at which the microphone was installed were determined by prior verification by the inventors, and the microphone was positioned so as to be spaced from the weld (processing point) to a degree that would not interfere with welding, and to be highly sensitive.

[0035] The measured microphone signal was passed through a frequency filter circuit developed by the company to extract only specific acoustic frequency bands, and then the waveform was shaped to the positive side by an absolute value circuit. An arithmetic unit calculated and displayed an estimated penetration depth to determine pass / fail, and the judgment thresholds, etc. were also set and recorded.

[0036] [Pre-learning experimental conditions] [Table 1]

[0037] Table 1 explains the experimental conditions for the pre-learning experiment according to the embodiment. As shown in Table 1, a multimode fiber laser processing machine with a laser light wavelength of 1070 mm was used. The processing head of the laser processing machine had a collimating lens f100 and a focusing lens f120. The laser output during the pre-learning (simulation) was set to a power increase variation with a linear slope that increased from 400 W to 800 W over 0.5 s. The relative scanning speed between the sample and the laser light was 20 mm / s, the irradiation time was 0.5 s, and the target sample was SUS304 (thickness t = 5 mm).

[0038] Figure 2 shows the acoustic signal waveform of the microphone when irradiated under the conditions shown in Table 1. The graph shown in Figure 2 was recorded at 200 kHz sampling. In Figure 2, the horizontal axis represents the laser light irradiation time (seconds) and the vertical axis represents the amplitude of the microphone output. Figure 2 shows that the waveform amplitude gradually increases as the welding progresses (as the graph moves to the right). At this time, the inventors speculate that the welding sound generated from the molten pool during welding is the sound of metal vapor and molten liquid being ejected from a hole called a keyhole that appears in the center of the molten liquid during irradiation.

[0039] Figure 3 shows a spectrogram obtained by frequency analysis of the microphone waveform obtained in this pre-training. The horizontal axis is frequency [Hz], the vertical axis is time [s], and the color intensity represents signal strength. From the graph shown in Figure 3, it can be seen that at the beginning of irradiation, i.e., at the bottom of the graph, the band intensity in the relatively high frequency range around 15 kHz is relatively high (approximately region A in Figure 3), but as time passes (i.e., as you move upward on the graph), the intensity of the frequency band of around 5 to 9 kHz gradually increases (approximately region B in Figure 3).

[0040] Figure 4 also shows a surface photograph (a) of a laser weld and a cross-sectional photograph (b) of the same. Figure 4(a) is a surface photograph of the weld, showing the traces of the surface where laser light was irradiated onto SUS304 from left to right for 0.5 seconds, while Figure 4(b) is a photograph of the weld taken by cutting the center of the weld (shown as "center cut" in Figure 4(a)) horizontally, polishing the cross-section, and then etching to reveal the melted cross-section. The horizontal position in Figure 4(b) is consistent with the horizontal position in Figure 4(a), and the vertical direction in Figure 4(b) is the penetration depth direction.

[0041] That is, Figure 4 shows the surface (a) and cross section (b) traces of a laser beam scanned in 0.5 seconds with the output power increased linearly from 400 W to 800 W from left to right. From Figure 4(b), it can be seen that the initial melt depth on the left side is about 1 mm with a laser beam output of 400 W, and the final melt depth on the right side is about 2 mm with a laser beam output of 800 W.

[0042] 3 and 4, the inventors currently understand that in the early stages of laser irradiation, the penetration deepens rapidly and the keyhole is shallow, so the natural frequency is small and relatively high-pitched sounds predominate, and as the keyhole deepens, low-pitched sounds begin to appear.

[0043] Here, only the 5 to 9 kHz range was extracted from the microphone output waveform in Figure 2 using a filter circuit, and the waveform, which had negative amplitude, was corrected to the positive side using an absolute value circuit. Furthermore, the waveform was smoothed using moving average processing (moving average over 2000 points for 5 μs sampling), and the results are shown in Figure 5. In Figure 5, the horizontal axis represents time (a), and the vertical axis represents the microphone amplitude intensity processed as described above.

[0044] In addition, Figure 6 is a graph showing the penetration depth (vertical direction) of the cross section shown in Figure 4(b) measured at 0.1 mm intervals in the horizontal direction, converted into time (s) on the horizontal axis and illustrating the penetration depth (mm) as a function of time. Figures 5 and 6 show that the graph shapes of both graphs tend to rise to the right and are similar in terms of both, including the gradient.

[0045] Based on the results of Figures 5 and 6, these were combined to produce a scatter plot graph in Figure 7, where the horizontal axis represents microphone amplitude intensity and the vertical axis represents penetration depth (mm). A numerical calculation software was used to derive an approximation of a linear equation for the scatter plot shown in Figure 7. This linear approximation equation is:

[0046] γ=285.22x+0.0867...Equation (1)

[0047] where γ is the penetration depth (mm) and x is the microphone amplitude intensity.

[0048] Furthermore, Fig. 8 shows a graph that overlays the estimated penetration depth calculated by substituting the microphone amplitude in Fig. 5 into the above-mentioned linear approximation formula with the actual penetration depth in Fig. 6. In Fig. 8, the horizontal axis represents time (s). It can be seen from Fig. 8 that the estimated penetration depth and the actual penetration depth match to a certain extent using the linear approximation formula derived from Fig. 7. Using this linear approximation formula, new acoustic signal data can be obtained for a laser welding process under the same conditions, and the penetration depth during actual laser welding can be estimated.

[0049] The above-mentioned pre-learning can be performed each time (preferably from three or more data points) under the same conditions (laser beam intensity, type, irradiation conditions such as irradiation time, scanning speed, material properties, thickness, etc.) before the actual laser welding process is performed, and the linear equation can be derived. However, it is also possible to accumulate a large amount of data in a database over many years and use the linear equation that most closely matches the conditions. Furthermore, instead of or in addition to such a database, an AI (artificial intelligence) can store data in advance, and the AI ​​can select or derive the most appropriate linear equation from the collected data. This can then be automatically applied during laser processing, resulting in a laser processing machine that can present the penetration depth to the user based on sound. Even in this case, upper and lower thresholds can be set, and if the penetration depth deviates from a certain range, the laser processing machine can determine a poor weld, issue an alert, or stop the welding process. In a laser processing machine using AI, the AI ​​can automatically determine the processing material and its properties using image recognition.

[0050] (Verification results) Table 2 shows the conditions for the verification experiment to estimate penetration depth using sound during laser welding. Laser beam power was varied from 400 W to 800 W in 50 W increments, with n = 3 (three samples for each irradiation condition). The relationship between the actual penetration depth and the estimated penetration depth using microphone signals calculated based on the approximate formula (Equation (1)) shown in Figure 7 was then investigated. A cross section was cut perpendicular to the irradiated area at the midpoint of the irradiation position (Figure 9), and the actual penetration depth was measured. Figure 9(a) is a micrograph showing the surface condition of a sample after irradiating it with a laser at a constant intensity for 0.5 seconds under each of the experimental conditions shown in Table 2. Figure 9(b) illustrates the cross section observed under a microscope at the "mid-irradiation cut" point shown in Figure 9(a). 9(b), it can be seen that the weld mark is formed in two stages: a tapered portion A near the surface (approximately 0.5 mm deep from the surface in FIG. 9(b)) and a nearly vertical hole portion B (approximately 1 mm deep at a point deeper than A). Just to be clear, the penetration depth during laser welding in this invention is the sum of A and B.

[0051] The estimated depth value from equation (1) is the value at 0.25 s, which is the midpoint of the 0.5 s laser irradiation time. The microphone output intensity x at 0.25 s, which is substituted into equation (1), is a value obtained by extracting only the 5 to 9 kHz range from the raw microphone output data using a frequency filter, correcting all to positive outputs using an absolute value circuit, and smoothing it using moving average processing (2000 points moving average (0.01 s) with 5 μs sampling) in a calculation device. However, since this is all analog processing and no delay occurs, real-time estimation is possible. [Conditions for verification experiment of penetration depth estimation] [Table 2]

[0052] The results of the verification experiment are shown in Figure 10. The horizontal axis in Figure 10 represents the sequential number of the sample, with sample numbers 1 to 3 corresponding to 400 W, sample numbers 4 to 6 corresponding to 450 W, sample numbers 7 to 9 corresponding to 500 W, ..., and sample numbers 25 to 27 corresponding to 800 W. The vertical axis on the left side of Figure 10 represents the penetration depth, and the vertical axis on the right side represents the difference in percentage between the actual measured depth and the depth estimated by calculation. As can be seen from Figure 10, the difference between the microphone's estimated penetration value and the actual measured penetration value is within approximately 10%, indicating that the penetration depth can be estimated with high accuracy.

[0053] (summary) Using a microphone with sensitivity in the frequency range of 20Hz to 20kHz, and applying appropriate shaping processes to the obtained waveform, such as appropriate frequency filters, absolute value processing, and smoothing using moving averages, it was confirmed that the greater the microphone signal strength, the greater the penetration depth during laser welding, at a similar rate.

[0054] From such changes in microphone output strength and the actual measured penetration depth, it is possible to express the relationship between microphone signal strength and penetration depth as an approximation of a linear equation, etc., and thereby to estimate the penetration depth.

[0055] Derivation of a first-order approximation formula etc. can be done using the approximation formula function in Excel, but if, for example, the microphone amplitude intensity of three or more representative sample points and the penetration depth measurement values ​​at those times can be ascertained through advance learning etc., an approximation formula can be easily obtained by creating a graph with penetration depth on the vertical axis and microphone amplitude intensity on the horizontal axis based on those sample values, drawing a line passing through those three points etc., and creating a database of values ​​on the vertical and horizontal axes that match that line. Using this approximation formula, penetration depth can be calculated from acoustic data.

[0056] The method for estimating penetration depth during laser welding of the present invention is characterized by including a step of estimating penetration depth in real time from the intensity of an acoustic signal of a specific frequency detected during laser welding, based on a pre-learned correlation between the intensity of an acoustic signal of a specific frequency from a molten pool and the penetration depth.

[0057] The inventors have found that there is a certain correlation between the penetration depth of the molten pool during laser welding and the sound from the molten pool, particularly the sound intensity in a specific frequency range (typically 5 to 9 kHz). That is, they have found that the sound intensity (loudness) in a specific frequency range increases with the penetration depth.

[0058] This makes it possible to estimate the penetration depth of the molten pool, which cannot be visually or in-situ observed during actual laser welding, by detecting the sound emitted from the molten pool with a microphone.Since the penetration depth is closely related to the quality of laser welding, it is also possible to determine defects based on the sound.

[0059] The pre-learning can be performed each time prior to the laser welding process, using data obtained in advance depending on the characteristics of the material to be welded and the characteristics and strength of the laser. Alternatively, AI (artificial intelligence) can pre-learn a huge amount of data from laser welding accumulated in the past, and derive or present correlations applicable to the laser welding to be performed during the laser welding process, and use these. Alternatively, correlations stored in a database can be read from another storage device and used.

[0060] The method of estimating penetration depth during laser welding of the present invention is preferably characterized in that the pre-trained correlation is expressed by a linear expression of the intensity of an acoustic signal at a specific frequency.

[0061] According to the findings of the inventors based on experimental data, if the horizontal axis represents the intensity of the acoustic signal from the molten pool in a specific frequency range and the vertical axis represents the penetration depth of the molten pool, a linear relationship is established in which the higher the acoustic intensity, the greater the penetration depth. In other words, the estimated penetration depth can be expressed as a × acoustic signal intensity + b (where a and b are constants that depend on the welding characteristics).

[0062] Based on this relationship, if the welding material has the same characteristics (material, thickness, etc.) and the same laser irradiation conditions (intensity, pulse width, irradiation time, irradiation speed, etc.), it is possible to estimate the penetration depth of the molten pool in real time by measuring the intensity of the acoustic signal at a specific frequency.

[0063] The method for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the specific frequency is an acoustic signal in the frequency range of 5 to 9 kHz. As explained in the experimental results shown in Figure 3 and elsewhere, it has been found that in the early stages of laser irradiation when penetration is shallow (less than 1 mm, typically less than 0.2 mm), the acoustic signal intensity around 15 kHz is high, but as the laser irradiation progresses and penetration becomes deeper, the acoustic signal intensity around 5 to 9 kHz gradually increases. Therefore, by observing the acoustic signal intensity around 5 to 9 kHz, it is possible to efficiently and accurately detect and estimate a penetration depth of about 1 to 2 mm.

[0064] The method for estimating penetration depth during laser welding according to the present invention is further preferably characterized in that the intensity of the acoustic signal at a specific frequency during pre-learning is a value obtained by performing moving average processing on the absolute values ​​of the detected acoustic signal, thereby processing the microphone waveform, which has large positive and negative amplitudes, into a data format that makes it easy to grasp the overall trend of change over time.

[0065] The method for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the intensity of the acoustic signal of a specific frequency detected during laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool. In the above-mentioned experiment, a microphone (with an integrated preamplifier) ​​with a frequency sensitivity of 20 Hz to 100 kHz was placed 80 mm away from the molten pool and used. However, this is not limited to this, and any microphone with sensitivity in the 20 Hz to 20 kHz frequency range is considered to be sufficient for application of the present invention. The acoustic signal value detected by the microphone may be used as raw data. However, as in the case of pre-training, typically, only the frequency range of 5 to 9 kHz is extracted using a filter circuit, the negative amplitude is inverted to the positive side using an absolute value circuit, and the value smoothed using a moving average process is used as the microphone output value (all analog processing), making data processing easier and faster. Furthermore, since it is most important to extract and monitor only a specific frequency range using a filter circuit, correction using an absolute value circuit and smoothing processing (moving average processing) can be adopted in either or both ways as appropriate. This is purely a process for improving the convenience of data handling.

[0066] The method for estimating penetration depth during laser welding of the present invention is further characterized in that the linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. The linear equation, which expresses the correlation between the intensity of acoustic signals of a specific frequency from the molten pool and the penetration depth, can be derived from pre-training if there is a relationship between the intensity of acoustic signals of a specific frequency from at least two points on the molten pool and the penetration depth. That is, the constants a and b can be determined as follows: Estimated penetration depth = a × acoustic signal intensity + b (where a and b are constants depending on the welding characteristics). However, to obtain a more reliable estimate, it is preferable to determine the constants a and b with higher accuracy and precision. In this sense, the constants a and b may be determined from three, or more preferably three, points of pre-training data.

[0067] The device for estimating penetration depth during laser welding of the present invention is characterized by comprising: a memory unit that stores the correlation between the intensity of an acoustic signal of a specific frequency from a molten pool and the penetration depth, which correlation is obtained in advance by pre-learning; and a calculation unit that calculates the penetration depth in real time from the intensity of the acoustic signal of the specific frequency detected during laser welding, based on the correlation stored in the memory unit.

[0068] The inventors have found that there is a certain correlation between the penetration depth of the molten pool during laser welding and the sound from the molten pool, particularly the sound intensity in a specific frequency range (typically 5 to 9 kHz). That is, they have found that the sound intensity (loudness) in a specific frequency range increases with the penetration depth.

[0069] This makes it possible to realize a device that can estimate the penetration depth of the molten pool, which cannot be observed visually or in situ during actual laser welding, by detecting the sound emitted from the molten pool with a microphone.Since penetration depth is closely related to poor quality of laser welding, it can also be used to determine defects based on the sound.

[0070] The pre-learning can be performed each time prior to the laser welding process, using data obtained in advance depending on the characteristics of the material to be welded and the characteristics and strength of the laser. Alternatively, AI (artificial intelligence) can pre-learn a huge amount of data from laser welding accumulated in the past, and derive or present correlations applicable to the laser welding to be performed during the laser welding process, and use these. Alternatively, correlations stored in a database by AI or the like can be read from another storage device and used.

[0071] The device for estimating penetration depth during laser welding of the present invention is preferably characterized in that the pre-trained correlation is expressed by a linear expression of the intensity of an acoustic signal at a specific frequency.

[0072] According to the findings of the inventors based on experimental data, if the horizontal axis represents the intensity of the acoustic signal from the molten pool in a specific frequency range and the vertical axis represents the penetration depth of the molten pool, a linear relationship is established in which the higher the acoustic intensity, the greater the penetration depth. In other words, the estimated penetration depth can be expressed as a × acoustic signal intensity + b (where a and b are constants that depend on the welding characteristics).

[0073] Based on this relationship, if the welding material has the same characteristics (material, thickness, etc.) and the same laser irradiation conditions (intensity, pulse width, irradiation time, irradiation speed, etc.), it is possible to realize a device that can estimate the penetration depth of the molten pool in real time by measuring the intensity of an acoustic signal at a specific frequency.

[0074] The device for estimating penetration depth during laser welding of the present invention is further preferably characterized in that the specific frequency is an acoustic signal in the frequency range of 5 to 9 kHz. As shown in Figure 3 and other experimental results, it has been found that the acoustic signal intensity around 15 kHz is high in the initial stage of laser irradiation when penetration is shallow (less than 1 mm, typically less than 0.2 mm). However, as the laser irradiation time increases and the laser penetration becomes deeper, the acoustic signal intensity around 5 to 9 kHz gradually increases. Therefore, by observing the acoustic signal intensity typically around 5 to 9 kHz, a device can be realized that efficiently and reliably detects and estimates a penetration depth of 1 mm or more, preferably about 1 to 2 mm, with high accuracy.

[0075] The laser welding penetration depth estimation device of the present invention is further preferably characterized in that the intensity of the acoustic signal at a specific frequency during pre-learning is a value obtained by performing moving average processing on the absolute values ​​of the detected acoustic signal. This allows the raw microphone output data waveform (e.g., as shown in Figure 2) with its large positive and negative amplitudes to be processed into a data format (e.g., as shown in Figure 5) that makes it easy to grasp the overall trend of its change over time.

[0076] The laser welding penetration depth estimation device of the present invention is further characterized in that the intensity of the acoustic signal of a specific frequency detected during laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool. In the above-described experiment, a microphone (with an integrated preamplifier) ​​with a frequency sensitivity of 20 Hz to 100 kHz was placed 80 mm away from the molten pool and used. However, this is not a limitation; any microphone with sensitivity in the 20 Hz to 20 kHz frequency range is considered sufficient for application of the present invention. The acoustic signal value detected by the microphone may be used as raw data. However, as in the case of pre-training, typically, only the frequency range of 5 to 9 kHz is extracted using a filter circuit, the negative amplitude is inverted to the positive side using an absolute value circuit, and the value smoothed using a moving average process is used as the microphone output value (all analog processing), making data processing easier and faster.

[0077] The laser welding penetration depth estimation device of the present invention is further characterized in that the linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. The linear equation, which shows the correlation between the intensity of acoustic signals of a specific frequency from the molten pool and the penetration depth, derived from pre-training, can be derived if there is data on the relationship between the intensity of acoustic signals of a specific frequency from at least two points on the molten pool and the penetration depth. That is, the constants a and b of the equation "Estimated penetration depth = a × acoustic signal intensity + b" (where a and b are constants depending on the welding characteristics) can be determined based on data from at least two points. However, to obtain a more reliable estimate, it is preferable to determine the constants a and b with higher accuracy and precision. In this sense, the constants a and b may be determined from pre-training data from three points, or preferably more.

[0078] The device for determining poor welding performance during laser welding of the present invention is characterized by comprising: a calculation unit that calculates an estimated penetration depth in real time from the intensity of an acoustic signal of a specific frequency detected by a microphone during laser welding, based on the correlation between the intensity of an acoustic signal of a specific frequency from the molten pool and the penetration depth, which is pre-stored in a memory unit through advance learning; and a poor welding determination unit that determines that the welding is poor when the estimated penetration depth calculated by the calculation unit exceeds an upper threshold or falls below a lower threshold.

[0079] That is, it is possible to realize a device for determining defective welding during laser welding that issues an alert (for example, an alert sound or a warning display on a monitor screen) when the estimated penetration depth value estimated using the penetration depth estimation method described above is not within a certain range between an upper threshold and a lower threshold and deviates from that range. As already shown in the experimental results as an example, the estimated penetration depth value is quite accurate, so that accurate determination and monitoring can be performed as if observing the spot, even though it is a real-time determination, making it possible to maintain high quality laser welding at low cost.

[0080] The apparatus for determining welding defects during laser welding according to the present invention is preferably characterized in that the pre-trained correlation is expressed by a linear expression of the penetration depth and the intensity of the acoustic signal at a specific frequency.

[0081] According to the findings of the inventors based on experimental data, if the intensity of the acoustic signal from the molten pool in a specific frequency range is plotted on the horizontal axis and the penetration depth of the molten pool is plotted on the vertical axis, a linear relationship is established in which the penetration depth increases as the acoustic intensity in the specific frequency range increases. In other words, the "estimated penetration depth" can be expressed as a × "acoustic signal intensity" + b (where a and b are constants that depend on the welding characteristics).

[0082] Based on this relationship, if the welding material has the same characteristics (material, thickness, etc.) and the same laser irradiation conditions (intensity, pulse width, irradiation time, irradiation speed, etc.), it is possible to realize a device that can estimate the penetration depth of the molten pool in real time and determine whether the welding is defective by measuring the intensity of the acoustic signal at a specific frequency.

[0083] The device for determining defective welds during laser welding according to the present invention is further preferably characterized in that the specific frequency is an acoustic signal in the frequency range of 5 to 9 kHz. As shown in Fig. 3 and other experimental results, it has been found that the acoustic signal intensity around 15 kHz is high in the initial stage of laser irradiation when the penetration is shallow (less than 1 mm, typically less than 0.2 mm). However, as the laser irradiation time increases and the laser penetration becomes deeper, the acoustic signal intensity around 5 to 9 kHz gradually increases. Therefore, by observing the acoustic signal intensity typically around 5 to 9 kHz, it is possible to realize a device for efficiently and reliably detecting and estimating a penetration depth of 1 mm or more, preferably about 1 to 2 mm, and determining defective welds.

[0084] The laser welding defect determination device of the present invention is further preferably characterized in that, in calculating the correlation between the intensity of a specific frequency acoustic signal from the molten pool and the penetration depth during pre-learning, the intensity of the specific frequency acoustic signal is calculated by performing a moving average process on the absolute values ​​of the detected acoustic signal. This allows the raw microphone output data waveform (e.g., as shown in Figure 2), which has large positive and negative amplitudes, to be processed into a data format (e.g., as shown in Figure 5) that makes it easy to grasp the overall trend of change over time, and an accurate penetration depth estimate based on this data can be used to determine weld defects.

[0085] The laser welding defect determination device of the present invention is further characterized in that the intensity of the acoustic signal of a specific frequency detected during laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool. In the above-described experiment, a microphone (with an integrated preamplifier) ​​with a frequency sensitivity of 20 Hz to 100 kHz was placed 80 mm away from the molten pool and used. Any microphone with sensitivity in the 20 Hz to 20 kHz frequency range is considered to be sufficient for the application of the present invention. However, this is not limited to this, and any microphone capable of detecting the acoustic signal in the specific wavelength range to be extracted can be used. As in the case of pre-training, the acoustic signal value detected by the microphone is typically extracted using a filter circuit, with the negative amplitude corrected to the positive side using an absolute value circuit, and the value smoothed using a moving average process can be used as the microphone output value.

[0086] The laser welding defect detection device of the present invention is further characterized in that the linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. The linear equation, which shows the correlation between the intensity of acoustic signals of a specific frequency from the molten pool and the penetration depth, derived from pre-training, can be derived if there is data on the relationship between the intensity of acoustic signals of a specific frequency from at least two points on the molten pool and the penetration depth. That is, the constants a and b of the formula "estimated penetration depth = a × acoustic signal intensity + b" (where a and b are constants depending on the welding characteristics) can be determined based on data from at least two points. However, to obtain a more reliable estimate, it is preferable to determine the constants a and b with higher accuracy and precision. In this sense, the constants a and b may be determined from pre-training data from three points, or preferably more.

[0087] The device / system for determining welding defects during laser welding and its structure / method, etc., according to the present invention are not limited to the shapes, structures, methods, etc., described above and shown in the drawings, and may be adopted, transformed, arranged, combined, or arbitrarily modified using appropriate publicly known or well-known techniques, etc., known to those skilled in the art, within the scope of the present invention. [Explanation of symbols]

[0088] 110···Calculation device, 120···Absolute value circuit, 130···Frequency filter circuit, 140···Microphone, 150···Laser processing head, 160···Laser light, 170···Welding metal (sample).

Claims

1. In the method of estimating penetration depth during laser welding, and estimating a penetration depth estimate in real time from the intensity of the acoustic signal of a specific frequency detected during laser welding based on a pre-learned correlation between the intensity of the acoustic signal of the specific frequency from the molten pool and the penetration depth, The pre-trained correlation is such that the penetration depth estimate is expressed as a linear expression of the intensity of the acoustic signal at the specific frequency. A method for estimating penetration depth during laser welding.

2. The method for estimating penetration depth during laser welding according to claim 1, The linear expression is [Equation 1] where Y is the penetration depth estimate, X is the intensity of the acoustic signal at the specific frequency, and a and b are constants that depend on the welding characteristics. A method for estimating penetration depth during laser welding.

3. The method for estimating penetration depth during laser welding according to claim 1, The intensity of the acoustic signal of the specific frequency during the pre-learning is a value obtained by performing a moving average process on the absolute value of the detected acoustic signal. A method for estimating penetration depth during laser welding.

4. The method for estimating penetration depth during laser welding according to claim 1, The intensity of the acoustic signal of the specific frequency detected during the laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool. A method for estimating penetration depth during laser welding.

5. The method for estimating penetration depth during laser welding according to claim 1, The linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. A method for estimating penetration depth during laser welding.

6. In a laser welding penetration depth estimation device, a storage unit that stores a correlation between the intensity of an acoustic signal of a specific frequency from the molten pool and the penetration depth, the correlation being obtained in advance by prior learning; a calculation unit that calculates an estimated penetration depth in real time from the intensity of the acoustic signal of the specific frequency detected during laser welding based on the correlation stored in the storage unit, The pre-trained correlation is such that the penetration depth estimate is expressed as a linear expression of the intensity of the acoustic signal at the specific frequency. A device for estimating penetration depth during laser welding.

7. In the device for estimating penetration depth during laser welding according to claim 6, The linear expression is [Equation 2] where Y is the penetration depth estimate, X is the intensity of the acoustic signal at the specific frequency, and a and b are constants that depend on the welding characteristics. A device for estimating penetration depth during laser welding.

8. In the device for estimating penetration depth during laser welding according to claim 6, The intensity of the acoustic signal of the specific frequency during the pre-learning is a value obtained by performing a moving average process on the absolute value of the detected acoustic signal. A device for estimating penetration depth during laser welding.

9. In the device for estimating penetration depth during laser welding according to claim 6, The intensity of the acoustic signal of the specific frequency detected during the laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool. A device for estimating penetration depth during laser welding.

10. The device for estimating penetration depth during laser welding according to claim 6, The linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. A device for estimating penetration depth during laser welding.

11. In a device for determining welding defects during laser welding, a calculation unit that calculates an estimated penetration depth in real time from the intensity of the acoustic signal of a specific frequency detected by a microphone during laser welding, based on a correlation between the intensity of the acoustic signal of the specific frequency from the molten pool and the penetration depth, which correlation is stored in advance in a storage unit by prior learning; a defective welding determination unit that determines that the welding is defective when the penetration depth estimated value calculated by the calculation unit exceeds an upper threshold or falls below a lower threshold, The pre-trained correlation is such that the penetration depth estimate is expressed as a linear expression of the intensity of the acoustic signal at the specific frequency. A device for determining welding defects during laser welding.

12. The device for determining welding defects during laser welding according to claim 11, The linear expression is [Equation 3] where Y is the penetration depth estimate, X is the intensity of the acoustic signal at the specific frequency, and a and b are constants that depend on the welding characteristics. A device for determining welding defects during laser welding.

13. The device for determining welding defects during laser welding according to claim 11, In calculating the correlation between the intensity of the acoustic signal of a specific frequency from the molten pool during the pre-learning and the penetration depth, the intensity of the acoustic signal of the specific frequency is calculated by performing a moving average process on the absolute value of the detected acoustic signal. A device for determining welding defects during laser welding.

14. The device for determining welding defects during laser welding according to claim 11, The intensity of the acoustic signal of the specific frequency detected during the laser welding is the value of the acoustic signal detected by a microphone placed near the molten pool. A device for determining welding defects during laser welding.

15. The device for determining welding defects during laser welding according to claim 11, The linear equation is derived from the correlation between the intensity of acoustic signals of a specific frequency from at least three points on the molten pool and the penetration depth. A device for determining welding defects during laser welding.

Images