Ultrasonic bonding quality determination device and ultrasonic bonding system

Abstract

To provide an ultrasonic bonding quality determination device that is easy to handle and capable of performing quality determination with high accuracy.SOLUTION: An ultrasonic bonding quality determination device 2 is an ultrasonic bonding quality determination device that performs quality determination of a bonding state of a bonding target portion 16, which is bonded by applying ultrasonic vibration from an ultrasonic horn 11 to workpieces 14 and 15 placed on an anvil 13, and includes: an acoustic data acquisition unit 21 that acquires acoustic data indicating sound generated during an ultrasonic vibration application period from start to completion of ultrasonic vibration application; an analysis unit 23 that decomposes the acoustic data acquired by the acoustic data acquisition unit 21 into a product of a plurality of basis spectra and coefficients indicating a time change of intensity in the basis spectra by NMF; and a quality determination unit 26 that performs quality determination of the bonding state of the bonding portion 16 based on the coefficients of the plurality of basis spectra.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an ultrasonic bonding quality determination device and an ultrasonic bonding system. [Background technology]

[0002] Ultrasonic bonding machines have traditionally been used to bond similar or dissimilar inorganic materials, especially non-ferrous metals such as copper, aluminum, magnesium, etc. With this ultrasonic bonding machine, two materials to be bonded are placed on top of each other on the machine's anvil, and ultrasonic vibrations are applied while the machine's ultrasonic horn is pressed against the materials to be bonded, thereby bonding the two materials together.

[0003] In ultrasonic welding, welding can fail due to the influence of contamination such as oil and grease on the welding interface, and methods have been proposed to determine this and eliminate defective products. For example, Patent Document 1 discloses, as an example of a method for determining whether a welding is good or bad, a configuration in which at least one of the output waveform of an ultrasonic metal welding machine or the amount of sinking of the upper metal sheet of two metal sheets to be joined is detected from the start of ultrasonic vibration of an ultrasonic horn, and at least one of the output waveform of the ultrasonic metal welding machine or the amount of sinking of the metal sheets until a predetermined time has elapsed is compared with that during normal welding to determine whether the metal sheets have been joined.

[0004] Furthermore, Patent Document 2 discloses a configuration that determines whether the ultrasonic bonding strength is good or bad based on whether the amount of sinking detected by the detection means exceeds a sinking threshold, thereby preventing a good product from being judged as defective when contaminants such as oil and grease are attached to the metal plate.

[0005] On the other hand, Patent Document 3 discloses a configuration in which, in order to grasp the vibration state of the horn, which is an important factor in ultrasonic welding, the vibration state of the measurement point using laser light is measured at multiple locations on the tool at any point, the amplitude is measured simultaneously, the position of the vibration node of the horn is calculated, and the result is compared with a predetermined value to determine the quality of the welding. Also, Patent Document 4 discloses a configuration in which the amplitude and frequency are measured by capturing an image of the horn that emits ultrasonic waves with a CCD camera.

[0006] Furthermore, Patent Document 5 discloses a configuration for determining pass / fail based on vibrations detected by a vibration sensor, and discloses that the vibration sensor uses a piezoelectric element or a coil or a sensor that utilizes the principle of a microphone. Also, Patent Document 6 discloses a method for determining pass / fail based on either the amount of sinking of the joint or the state of ultrasonic output. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Japanese Patent Application Laid-Open No. 2010-184252 [Patent Document 2] Japanese Patent Application Laid-Open No. 2016-20003 [Patent Document 3] Japanese Patent Application Laid-Open No. 2007-142049 [Patent Document 4] Japanese Patent Application Laid-Open No. 2008-70158 [Patent Document 5] Japanese Patent Application Laid-Open No. 2004-58523 [Patent Document 6] Japanese Patent Application Laid-Open No. 2016-97425 Summary of the Invention [Problem to be solved by the invention]

[0008] However, in the configurations disclosed in Patent Documents 1 and 2, the quality of the ultrasonic bonded portion is judged based on the amount of sinking, but the amount of sinking may not reflect the quality of the ultrasonic bonded portion, so there is room for improvement in order to further improve the accuracy of the quality judgment. Furthermore, in the configuration disclosed in Patent Document 3, when vibration measurement is performed using a laser, piezoelectric, coil, etc., it may be difficult to secure space for installing the device or it may be difficult to handle. Furthermore, even a slight deviation in the installation position can significantly affect the measurement results, which may reduce the accuracy of the quality judgment.

[0009] Furthermore, in the configuration disclosed in Patent Document 5, if a microphone is used as a vibration sensor, all surrounding sounds are picked up. Therefore, if the sound waves related to the joining are small, they will be buried in other large sound waves, reducing the accuracy of the determination. Therefore, to improve the determination accuracy, it is necessary to make contact while suppressing the acquisition of other sound waves, which leaves problems such as the need to consider securing the installation space for the device, handling the device, and adjusting the installation position.

[0010] The present invention has been made in view of the above-mentioned problems, and aims to provide an ultrasonic bonding quality determining device that is easy to handle and can perform quality determination with high accuracy. [Means for solving the problem]

[0011] One aspect of the present invention is An ultrasonic bonding quality determination device that determines the quality of a bonding state of a workpiece to be bonded by applying ultrasonic vibrations from an ultrasonic horn to the workpiece to be bonded, the workpiece including the workpiece to be bonded being placed on an anvil, an acoustic data acquisition unit that acquires acoustic data indicating sounds generated during an ultrasonic vibration application period from the start to the end of application of the ultrasonic vibration; an analysis unit that decomposes the acoustic data acquired by the acoustic data acquisition unit into products of a plurality of basis spectra and coefficients that indicate time changes in intensity of the basis spectra by Non-negative Matrix Factorization (NMF); and a quality determining section that determines the quality of the bonded state of the parts to be bonded based on coefficients in the plurality of basis spectra. [Effects of the Invention]

[0012] In the ultrasonic bonding quality determination device of the above aspect, the device for acquiring acoustic data does not require highly accurate positioning, and the device is easy to handle. Furthermore, NMF decomposes the acoustic data into multiple basis spectra and their coefficients, and performs quality determination based on the coefficients of the multiple basis spectra. Therefore, by performing quality determination using the coefficients of the basis spectra corresponding to the sound related to bonding, even if the sound related to bonding is quiet, the sound waves related to bonding are prevented from being drowned out by other loud sounds, and the accuracy of quality determination can be improved.

[0013] As described above, according to the present invention, it is possible to provide an ultrasonic bonding quality determining device that is easy to handle and can perform quality determination with high accuracy. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a conceptual diagram showing the configuration of an ultrasonic bonding system including an ultrasonic bonding device and a functional block diagram of an ultrasonic bonding quality determination device according to the first embodiment. [Figure 2] FIG. 3 is a diagram showing an example of acoustic data acquired by an acoustic data acquisition unit in the first embodiment. [Figure 3] FIG. 3 is a diagram showing an example of a spectrogram created by short-time Fourier transform in the first embodiment. [Figure 4] FIG. 2 is a diagram showing an example of a basis spectrum separated by NMF in the first embodiment. [Figure 5] FIG. 4 is a diagram showing an example of coefficients of a basis spectrum separated by NMF in the first embodiment. [Figure 6] FIG. 4 is a diagram showing examples of coefficients of a plurality of basis spectra used to set reference values ​​of coefficient intensities in the first embodiment. [Figure 7]FIG. 4 is a diagram showing an example of coefficients of a specific base spectrum used to set reference values ​​of coefficient intensities in the first embodiment. [Figure 8] FIG. 4 is a flowchart of a bonding state determination process in the first embodiment. [Figure 9] FIG. 10 is another flowchart of the bonding state determination process in the first embodiment. [Figure 10] FIG. 4 is a diagram showing an example of coefficients of a specific base spectrum used to set a reference value for the integral value of the coefficient in the first embodiment. [Figure 11] FIG. 10 is a diagram showing a first verification result in a verification test for determining pass / fail in the second embodiment. [Figure 12] FIG. 10 is a diagram showing a second verification result in a verification test for determining pass / fail in the second embodiment. [Figure 13] FIG. 10 is a flowchart of a bonding state determination process in the second embodiment. [Figure 14] FIG. 10 is another flowchart of the bonding state determination process in the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0015] In the above aspect, it is preferable that the quality determination unit determines the quality of the joining state of the parts to be joined based on a specific coefficient that is a coefficient of at least one specific base spectrum identified from the plurality of base spectra. In this case, by selecting a specific base spectrum that corresponds to the sound related to joining, even if the sound related to joining is small, the sound wave related to joining can be prevented from being drowned out by other large sounds, and the accuracy of the quality determination can be further improved.

[0016] In the above aspect, the welding apparatus may further include a reference value storage unit in which a reference value to be compared with the specific coefficient is stored in advance, and the analysis unit may determine whether the welding state of the parts to be welded is good or bad based on a comparison result between the specific coefficient and the reference value. In this case, the quality of the welding state of the parts to be welded can be easily determined.

[0017] In the above aspect, the reference value storage unit may store, as the reference value, a value for comparison with the maximum value of the intensity for the specific coefficient during the ultrasonic vibration application period, and the analysis unit may determine whether the joining state of the parts to be joined is good or bad based on a comparison result between the maximum value of the intensity for the specific coefficient and the reference value. In this case, the joining state of the parts to be joined can be determined easily and with higher accuracy.

[0018] In the above aspect, the reference value storage unit stores, as the reference value, a value to be compared with an integral value of the ultrasonic vibration application period when the specific coefficient is plotted on a graph with the intensity at the specific coefficient on the vertical axis and the elapsed time from the application on the horizontal axis, and the analysis unit can determine whether the joining state of the parts to be joined is good or bad based on a comparison result between the integral value of the ultrasonic vibration application period at the specific coefficient and the reference value. In this case, it is possible to easily determine whether the joining state of the parts to be joined is good or bad, and to perform the determination with even greater accuracy.

[0019] In the above aspect, a learning unit that performs machine learning in advance to set the reference value may be provided, and the reference value storage unit may store the reference value set by the learning unit. In this case, the quality of the welding state of the parts to be welded can be determined with even higher accuracy.

[0020] In the above aspect, it is preferable that the quality determining unit determines the quality of the bonded state of the parts to be joined based on a plurality of coefficients in the plurality of basis spectra, which can improve the accuracy of the quality determination compared to when the quality of the bonded state of the parts to be joined is determined based on a single coefficient.

[0021] In the above aspect, it is preferable that the pass / fail determination unit determines the pass / fail of the bonding state of the parts to be bonded based on all coefficients in the plurality of basis spectra. In this case, since there is no need to extract a specific basis spectrum, it is possible to save the effort of extraction and further improve the accuracy of the pass / fail determination by reducing arbitrariness.

[0022] In the above aspect, the pass / fail judgment unit can make a pass / fail judgment based on the data distances in the plurality of coefficients. The pass / fail judgment unit can also make a pass / fail judgment based on the data distances in all of the coefficients. It is preferable to calculate the Mahalanobis distance as the data distance. By using the Mahalanobis distance, it is possible to improve the judgment accuracy by using all of the coefficients, and it is also possible to prevent pass / fail information from being buried in similar information that is highly correlated, thereby improving the accuracy of the pass / fail judgment.

[0023] In the above aspect, it is preferable that the quality determining section calculates the average or maximum value of the squares of the Mahalanobis distances as the inter-data distance, which can further improve the accuracy of the determination.

[0024] In the above aspect, it is preferable that the upper limit of the frequency of the sound acquired by the acoustic data acquisition unit is at least 1.2 times the frequency of the ultrasonic vibration, thereby enabling the quality of the joining state of the parts to be joined to be determined with even higher accuracy.

[0025] In the above aspect, it is preferable that the analysis unit performs a short-time Fourier transform on the acoustic data acquired by the acoustic data acquisition unit to create a spectrogram, and decompose the spectrogram into products of the plurality of basis spectra and the coefficients. In this case, since the acquired acoustic data changes significantly in a short time, by performing the short-time Fourier transform, it is possible to easily separate acoustic data attributable to bonding from acoustic data attributable to other causes, thereby enabling the acoustic data attributable to bonding to be extracted with high accuracy, and enabling the creation of highly accurate pass / fail judgment criteria.

[0026] Furthermore, an ultrasonic bonding system can be provided that includes the ultrasonic bonding quality determination device according to the above aspect, the anvil, the ultrasonic horn, and an ultrasonic bonding device that includes an ultrasonic vibration application unit that applies the ultrasonic vibrations to the parts to be bonded. In this case, an ultrasonic bonding system that is easy to handle and can perform quality determination with high accuracy can be provided.

[0027] (Embodiment 1) 1. Ultrasonic Bonding System 100 A first embodiment of an ultrasonic bonding system 100 and an ultrasonic bonding quality determination device 2 will be described with reference to Figs. 1 to 6. As shown in Fig. 1, the ultrasonic bonding system 100 includes an ultrasonic bonding device 1 and an ultrasonic bonding quality determination device 2. The ultrasonic bonding device 1 is a device for ultrasonically bonding parts to be bonded 16 of a pair of workpieces 14, 15. The ultrasonic bonding quality determination device 2 is a device for determining the quality of the bonded state of the parts to be bonded 16 ultrasonically bonded by the ultrasonic bonding device 1. Each device will be described in detail below.

[0028] 2.Ultrasonic bonding device 1 1, the ultrasonic bonding device 1 includes an ultrasonic horn 11, an ultrasonic oscillator 12, and an anvil 13. In the ultrasonic bonding device 1, workpieces 14 and 15 are placed on the anvil 13, and the ultrasonic horn 11 applies a load to parts to be bonded 16, while ultrasonic vibrations generated by the ultrasonic oscillator 12 are applied to the parts to be bonded 16 via the ultrasonic horn 11. The ultrasonic oscillator 12 constitutes an ultrasonic vibration application unit that applies ultrasonic vibrations to the parts to be bonded 16.

[0029] The ultrasonic horn 11 can be made of die steel, duralumin, Monel, titanium, phosphor bronze, etc., but die steel is suitable due to its low cost, ease of manufacture, and wide range of application to the materials to be joined. The frequency of the ultrasonic waves generated by the ultrasonic oscillator 12 is not limited, but can be in the range of 13 kHz to 70 kHz, and preferably in the range of 20 kHz to 48 kHz. The amplitude of the ultrasonic waves generated by the ultrasonic oscillator 12 is preferably in the range of 15 to 45 μm. The material of the anvil 13 is not particularly limited, and a known structure can be used.

[0030] 3. Ultrasonic bonding quality inspection device 2 As shown in FIG. 1, the ultrasonic bonding quality determining device 2 includes an acoustic data acquiring unit 21, an acoustic data storage unit 22, an analyzing unit 23, a reference value storage unit 24, a learning unit 25, and a quality determining unit 26.

[0031] 3-1. Acoustic data acquisition unit 21 The acoustic data acquisition unit 21 may be any unit that can acquire the sound associated with bonding in the ultrasonic bonding device 1 as acoustic data, and in this embodiment, a microphone that collects the sound associated with bonding is used as the acoustic data acquisition unit 21. There are no particular limitations on the type of microphone used as the acoustic data acquisition unit 21, and it may be a condenser microphone, a dynamic microphone, a tube microphone, a ribbon microphone, or the like. There are also no particular limitations on the directivity of the microphone, and it may be unidirectional, bidirectional, or omnidirectional, but it is preferable to use a unidirectional microphone with high directivity to avoid noise.

[0032] The installation position of the microphone serving as the acoustic data acquisition unit 21 is not particularly limited as long as it can clearly acquire sounds related to the joining, and does not necessarily have to be near the parts to be joined 16. The microphone is preferably installed at a position away from other devices to avoid noise. The number of microphones may be one or more. The upper limit of the frequency of the sound acquired by the acoustic data acquisition unit 21 is preferably at least 1.2 times the frequency of the ultrasonic vibrations generated by the ultrasonic oscillator 12.

[0033] 3-2. Acoustic data storage unit 22 The acoustic data storage unit 22 is configured with a rewritable memory and stores the acoustic data acquired by the acoustic data acquisition unit 21. The acoustic data storage unit 22 may store data converted into a digital signal via an A / D converter. The sampling rate and bit depth of the A / D converter can be set appropriately. For example, the acoustic data stored in the acoustic data storage unit 22 can be waveform data shown in FIG. 2.

[0034] 3-3.Analysis Department 23 The analysis unit 23 analyzes the acoustic data stored in the acoustic data storage unit 22. The analysis unit 23 is configured by a calculation device capable of executing a predetermined program. In this embodiment, as shown in FIG. 1, the analysis unit 23 includes a short-time Fourier transform processing unit 231 and an NMF processing unit 232.

[0035] The short-time Fourier transform processing unit 231 performs a short-time Fourier transform on the acoustic data stored in the acoustic data storage unit 22. In ultrasonic bonding, bonding often takes several seconds, or even less than one second. Therefore, vibrations occur for only a very short time, and the signal varies significantly within that short time. In such cases, the Fourier transform cannot accurately determine the bonding state, so a short-time Fourier transform is more suitable.

[0036] Fig. 2 shows waveform data as acoustic data stored in acoustic data storage unit 22, and Fig. 3 shows a diagram of the acoustic data converted into a spectrogram (intensity distribution of each frequency component at each time) by short-time Fourier transform processing unit 231. The welding conditions were as follows: workpiece 14 was a magnesium alloy AZ31 plate material with a thickness of 1 mm, a width of 23 mm, and a length of 100 mm; workpiece 15 was an aluminum alloy A5052 plate material with a thickness of 1 mm and a width of 10 mm; these were overlapped by a length of 60 mm to form part to be welded 16, which was placed on anvil 13; ultrasonic horn 11 was pressed into part to be welded 16; when a load of 300 N was reached, ultrasonic waves were applied to ultrasonic horn 11 at a frequency of 20 kHz and an amplitude of 25 μm. The pressing speed was 0.47 mm / s, and the pressing depth was 0.7 mm, completing the welding.

[0037] Next, the NMF processing unit 232 performs NMF decomposition on the acoustic data processed by the short-time Fourier transform processing unit 231. NMF (Non-negative Matrix Factorization) is a method of approximating and expressing a non-negative matrix as the product of two non-negative matrices, and decomposes a decomposition source matrix X(n×m) into a basis matrix H(l×m) and a coefficient matrix W(n×l) in the following form:

[0038]

number

[0039] Any function can be set as the evaluation function for evaluating approximations in NMF, but NMF employs various functions such as Euclidean distance, KL divergence, Itakura-Saito pseudo-distance, etc. In this embodiment, the NMF processing unit 232 decomposes the acoustic data using NMF to separate sound waves originating from joints and sound wave information originating from other causes.

[0040] In this embodiment, by decomposing the acoustic data using NMF, multiple basis spectra, which are the basis matrix H, and their coefficients are obtained. There is no limit to the number of basis spectra to be decomposed, and any number can be used, but it is preferably 3 to 8, and more preferably 3 to 5. In this embodiment, by applying NMF to the acoustic data shown in Fig. 3, five basis spectra H1 to H5 shown in Figs. 4(a) to (e) and their coefficients W1 to W5 shown in Figs. 5(a) to (e) are obtained.

[0041] In this embodiment, the basis spectra H1 to H5 are represented by intensity distributions for each frequency component, as shown in Figures 4(a) to 4(e), and their coefficients W1 to W5 are represented by changes in intensity over time in the corresponding basis spectra, as shown in Figures 5(a) to 5(e). Figure 5 shows that the times at which the intensities of the coefficients W1 to W5 become high differ for the five basis spectra H1 to H5.

[0042] 3-4. Reference value storage unit 24 1 stores in advance a reference value to be compared with a specific coefficient in a quality determination unit 26, which will be described later. The specific coefficient is a coefficient of at least one specific basis spectrum identified from a plurality of basis spectra. The specific basis spectrum is identified from the plurality of basis spectra as one in which the change in intensity of the coefficient over time is closely related to the bonding state of the parts to be bonded 16.

[0043] In this embodiment, it is determined that the base spectra H1 and H2 shown in Figures 4(a) and 4(b) are the ultrasonic setting values ​​in the ultrasonic bonding apparatus 1, i.e., the frequencies of the ultrasonic vibrations generated by the ultrasonic oscillator 12, and do not correspond to the sound waves actually generated in the parts to be bonded 16. Therefore, it can be inferred that the base spectra H3, H4, and H5 shown in Figures 4(c) to 4(e) may be related to the bonding state of the parts to be bonded 16. Then, as will be described later, the coefficients W1 to W5 of each base spectrum H1 to H5 were obtained and compared under both conditions of good bonding and poor bonding, and it was found that the time change in the intensity of the coefficient W4 of the base spectrum H4 is closely related to the bonding state of the parts to be bonded 16. Therefore, in this embodiment, from the multiple base spectra H1 to H5, the base spectrum H4 is identified as a specific base spectrum, and its coefficient W4 is identified as a specific coefficient.

[0044] The reference value stored in the reference value storage unit 24 can be, for example, a value to be compared with the maximum value of the intensity at the specific coefficient during the ultrasonic vibration application period. Alternatively, the reference value can be a value to be compared with the integral value of the ultrasonic vibration application period when the specific coefficient is plotted on a graph with the intensity at the specific coefficient on the vertical axis and the elapsed time from the application on the horizontal axis.

[0045] 3-5. Learning Section 25 The learning unit 25 can set the reference value of the specific coefficient stored in the reference value storage unit 24 based on machine learning by the learning unit 25. The method of machine learning by the learning unit 25 is not limited, and may be machine learning with teacher data or machine learning without teacher data. In this embodiment, acoustic data for a total of 12 ultrasonic bonding operations under both conditions where the bonding state is good and conditions where the bonding state is poor is acquired, and the reference value of the specific coefficient is set by machine learning using this data as teacher data. The detailed good conditions and poor conditions are as follows.

[0046] The number of data points was set to 12, with the ultrasonic oscillation time being 0.5 to 3.0 seconds, the horn descending speed being 1.4 to 0.233 mm / s, the trigger pressure being 400 N, the ultrasonic amplitude being 85%, and the oscillation time x descending speed being 0.7 mm. The bonding condition was poor when the oscillation time was 0.7 seconds or less, and good when it was 1.0 seconds or more, so of the 12 acoustic data points, the three acoustic data points with an oscillation time of 0.7 seconds or less were used as training data for poor conditions, and the remaining nine acoustic data points with an oscillation time of 1.0 seconds or more were used as training data for good conditions.

[0047] The acoustic data for 12 times was subjected to a short-time Fourier transform as described above, and decomposed into basis spectra H1 to H5 and their coefficients W1 to W5 using NMF. The coefficients W1 to W5 were plotted with time on the horizontal axis and intensity on the vertical axis, as shown in Figures 6(a) to 6(e). From these, it was revealed that the time change in the intensity of the coefficient W4 with basis spectrum H4 shown in Figure 6(d) is closely related to the bonding state of the parts to be bonded 16.

[0048] 7, the maximum strength of coefficient W4 was 0.2 or more under good conditions, and 0.15 or less under bad conditions. Based on this, a safety estimate was made, and a reference value of 0.2 for the specific coefficient was stored in the reference value storage unit 24. The criteria for determining whether a value equal to or greater than this reference value is a condition for a good bonding state was determined.

[0049] 3-6. Pass / fail judgment section 26 1 determines the quality of the bonded state of the parts to be bonded 16 based on the identifying coefficient. In this embodiment, the identifying coefficient W4 of the specific basis spectrum H4, out of a plurality of basis spectra H1 to H5 separated by NMF from acoustic data acquired during ultrasonic bonding of the workpieces 14 and 15 to be determined, is compared with a reference value stored in the reference value storage unit 24, and the quality of the bonded state of the parts to be bonded 16 is determined based on the comparison result. In this embodiment, if the identifying coefficient W4 is equal to or greater than the reference value, the bonded state is determined to be good, and if the identifying coefficient W4 is less than the reference value, the bonded state is determined to be poor.

[0050] Although not shown, the ultrasonic bonding device 1 may have a bonding condition adjustment unit that adjusts the bonding conditions of the ultrasonic bonding device 1 based on the judgment result of the quality judgment unit 26. For example, if the quality judgment unit 26 judges that the bonding state is poor, the ultrasonic bonding device 1 can change the bonding conditions so as to increase the oscillation time of the ultrasonic oscillator 12 by a predetermined time.

[0051] 4.Determining the bonding condition Next, the process of determining the bonding state by the ultrasonic bonding system 100 will be described in detail with reference to the flow chart shown in Fig. 8. First, in step S1 shown in Fig. 8, based on preset bonding conditions, in the ultrasonic bonding device 1, a load is applied to the bonding target parts 16 of the workpieces 14, 15 placed on the anvil 13 by the ultrasonic horn 11, and ultrasonic vibrations generated by the ultrasonic oscillator 12 are applied to the bonding target parts 16 via the ultrasonic horn 11. At the same time, the acoustic data acquisition unit 21 starts acquiring acoustic data. The acquired acoustic data is stored in the acoustic data storage unit 22.

[0052] Thereafter, when the timing to stop applying ultrasonic vibrations arrives, in step S2 shown in FIG. 8, the application of ultrasonic vibrations is terminated and the acoustic data acquisition unit 21 terminates the acquisition of acoustic data.

[0053] Then, in step S3, a spectrogram is created by short-time Fourier transforming the acoustic data by the short-time Fourier transform processing unit 231. After that, in step S4, the NMF processing unit 232 separates the spectrogram into a plurality of basis spectra H1 to H5 and their coefficients W1 to W4 by NMF.

[0054] Furthermore, in step S5, the quality determining unit 26 extracts a specific coefficient W4 in the specific base spectrum H4 from the multiple base spectra H1 to H5. Then, in step S6, the quality determining unit 26 determines whether the maximum intensity of the specific coefficient W4 is equal to or greater than a reference value stored in advance in the reference value storage unit 24.

[0055] If it is determined in step S6 that the maximum strength of the specific coefficient W4 is equal to or greater than the reference value, the process proceeds to Yes in step S6, and the quality determination unit 26 determines in step S7 that the bonding state of the parts to be welded 16 is good, thereby terminating the bonding state determination process. On the other hand, if it is determined in step S6 that the maximum strength of the specific coefficient W4 is not equal to or greater than the reference value, the process proceeds to No in step S6, and the quality determination unit 26 determines in step S8 that the bonding state of the parts to be welded 16 is poor, thereby terminating the bonding state determination process.

[0056] In the flow chart shown in Fig. 8, in step S6, the quality determination unit 26 determines whether or not the maximum intensity of the specific coefficient W4 is equal to or greater than the reference value previously stored in the reference value storage unit 24. Alternatively, after step S5 shown in Fig. 8, the process may proceed to step S60 shown in Fig. 9, where the quality determination unit 26 calculates the integral value of the specific coefficient, and in step S61, it may be determined whether or not the integral value of the specific coefficient is equal to or greater than the reference value previously stored in the reference value storage unit 24. After step S61, steps S7 and S8 onwards are performed, which are equivalent to those shown in Fig. 8.

[0057] The integral value in step S60 shown in FIG. 9 is the integral value during the ultrasonic vibration application period when the specific coefficient W4 is plotted on a graph with the vertical axis representing the coefficient strength and the horizontal axis representing the elapsed time since the application of the ultrasonic vibration.

[0058] The reference value to be compared with the integral in step S60 shown in FIG. 9 can be set to a safe value based on the integral of the specific coefficient W4 during the ultrasonic vibration application period when the bonded state is good and the integral of the specific coefficient W4 during the ultrasonic vibration application period when the bonded state is poor, as shown in FIG. 10. As shown in FIG. 10, the integral under good conditions, i.e., the area, is larger than the area that is the integral under poor conditions. Therefore, if the integral of the specific coefficient W4 is equal to or greater than the reference value in step S61, the process proceeds to Yes in step S61, and the bonded state is determined to be good in step S7. On the other hand, if the integral of the specific coefficient W4 is not equal to or greater than the reference value in step S61, the process proceeds to No in step S61, and the bonded state is determined to be poor in step S8.

[0059] In this embodiment, the acoustic data storage unit 22, the analysis unit 23, and the learning unit 25 shown in FIG. 1 are configured as one arithmetic device, but this is not limitative, and any of these may be configured as a separate arithmetic device.

[0060] 5. Effects According to the ultrasonic bonding quality determination device 2 of the first embodiment, the acoustic data acquisition unit 21 for acquiring acoustic data does not require highly accurate positioning, and the device is easy to handle. Furthermore, NMF is used to decompose the acoustic data into a plurality of basis spectra and their coefficients (for example, basis spectra H1 to H5 and their coefficients W1 to W5), and quality determination is performed based on the coefficients of the plurality of basis spectra. Therefore, by performing quality determination using the coefficients of the basis spectra corresponding to the sound associated with bonding, even if the sound associated with bonding is quiet, the sound waves associated with bonding are prevented from being drowned out by other loud sounds, and the accuracy of quality determination can be improved.

[0061] Furthermore, in this embodiment 1, the quality determining unit 26 determines the quality of the bonded state of the parts to be bonded 16 based on a specific coefficient, which is a coefficient of at least one specific base spectrum determined from a plurality of base spectra. By selecting a specific base spectrum that corresponds to the sound related to bonding, even if the sound related to bonding is small, the sound wave related to bonding is prevented from being drowned out by other large sounds, and the accuracy of the quality determination can be further improved.

[0062] Moreover, in the present embodiment 1, a reference value storage unit 24 is provided in which a reference value to be compared with the specific coefficient is stored in advance, and the analysis unit 23 judges whether the joining state of the parts to be joined 16 is good or bad based on the comparison result between the specific coefficient and the reference value. This makes it easy to judge whether the joining state of the parts to be joined 16 is good or bad.

[0063] In addition, in this embodiment 1, the reference value storage unit 24 stores, as a reference value, a value to be compared with the maximum value of the intensity in the specific coefficient during the ultrasonic vibration application period, and the analysis unit 23 judges the quality of the bonded state of the parts to be bonded 16 based on the comparison result between the maximum value of the intensity in the specific coefficient and the reference value. This makes it possible to easily judge the quality of the bonded state of the parts to be bonded 16 and to make the judgement with even higher accuracy.

[0064] Furthermore, in the present embodiment 1, the reference value storage unit 24 stores, as a reference value, a value to be compared with an integral value of the ultrasonic vibration application period when the specific coefficient is plotted on a graph with the intensity at the specific coefficient on the vertical axis and the time elapsed since the application of the ultrasonic vibration on the horizontal axis, and the analysis unit 23 can determine the quality of the bonded state of the parts to be bonded 16 based on the comparison result between the integral value of the ultrasonic vibration application period at the specific coefficient and the reference value. In this case as well, it is possible to easily determine the quality of the bonded state of the parts to be bonded 16 and to perform the determination with even greater accuracy.

[0065] Moreover, in this embodiment 1, a learning unit 25 that performs machine learning in advance to set a reference value is provided, and the reference value storage unit 24 stores the reference value set by the learning unit 25. This allows the quality of the joining state of the parts to be joined 16 to be determined with even higher accuracy.

[0066] In the first embodiment, the upper limit of the frequency of the acoustic data acquired by the acoustic data acquiring unit 21 is at least 1.2 times the frequency of the ultrasonic vibration, thereby enabling the quality of the bonding state of the parts to be bonded 16 to be determined with even higher accuracy.

[0067] In addition, in this embodiment 1, the analysis unit 23 performs a short-time Fourier transform on the acoustic data acquired by the acoustic data acquisition unit 21 to create a spectrogram, and decomposes the spectrogram into products of multiple basis spectra and coefficients. As a result, since the acquired acoustic data changes significantly in a short time, by performing a short-time Fourier transform, it is possible to easily separate acoustic data originating from bonding and acoustic data originating from other causes, and to extract acoustic data originating from bonding with high accuracy, thereby making it possible to create high-accuracy pass / fail judgment criteria.

[0068] Moreover, in this embodiment 1, the ultrasonic bonding system 100 includes an ultrasonic bonding apparatus 1 equipped with an ultrasonic bonding quality determination device 2, an anvil 13, an ultrasonic horn 11, and an ultrasonic oscillator 12 constituting an ultrasonic vibration application unit that applies ultrasonic vibrations to the parts to be bonded 16. This also makes the ultrasonic bonding system 100 easy to handle and enables high-precision quality determination.

[0069] As described above, according to the present invention, it is possible to provide an ultrasonic bonding quality determining device 2 that is easy to handle and can perform quality determination with high accuracy.

[0070] (Embodiment 2) 6. Ultrasonic Bonding Quality Judgment Device 2 and Ultrasonic Bonding System 100 In the ultrasonic bonding quality determination device 2 and ultrasonic bonding system 100 of the above-described first embodiment, the quality determination section 26 determines the quality of the bonded state of the parts to be bonded 16 based on a specific coefficient, which is a coefficient of at least one specific base spectrum identified from a plurality of base spectra, but instead, in the ultrasonic bonding quality determination device 2 and ultrasonic bonding system 100 of the second embodiment, the quality determination section 26 determines the quality of the bonded state of the parts to be bonded based on a plurality of coefficients in a plurality of base spectra. Note that the other configurations in the second embodiment are the same as those in the first embodiment shown in Fig. 1, and the same reference numerals as those in the first embodiment are used, and description thereof will be omitted.

[0071] 6-1. Pass / fail judgment section 26 In the second embodiment, the pass / fail judgement unit 26 judges pass / fail based on the inter-data distances of a plurality of coefficients in a plurality of basis spectra. In the pass / fail judgement, some or all of the coefficients in the plurality of basis spectra may be used. If all of the coefficients are used, the calculation load for the inter-data distance increases, but the calculation accuracy of the inter-data distance can be improved.

[0072] The types of inter-data distances calculated by the pass / fail judgement unit 26 include Mahalanobis distance, standard Euclidean distance, Manhattan distance, Chebyshev distance, and Minkowski distance, with Mahalanobis distance being preferred.

[0073] As shown in Figure 3 of the first embodiment, the ultrasonic bonding acoustic data is dominated by the ultrasonic oscillation frequency and its harmonic frequencies, and other frequencies originating from friction during bonding are weak in the spectrum. As shown in Figure 4, the basis spectra obtained by NMF decomposition all strongly reflect this, and the coefficients are not orthogonal to each other, and the relationships and correlations between the coefficients are generally strong. The pass / fail judgment in the first embodiment selects one specific coefficient that relatively strongly reflects pass / fail, and is nothing more than an attempt to make pass / fail judgment and its physical interpretation easier, even by ignoring the influence of correlation and accepting some information loss.

[0074] In contrast, a method of increasing the number of basis spectra and using all of their coefficients is considered effective in avoiding information overlooks and improving judgment accuracy, but it may highlight similar information with strong correlations and bury information related to pass / fail. However, by calculating the Mahalanobis distance as the data distance calculated by the pass / fail judgment unit 26, the distance in the direction of variables with strong correlations is relatively short, so that even when all of the coefficients are used, the difference between pass / fail can be effectively expressed as a difference in distance. Therefore, the Mahalanobis distance can improve judgment accuracy by using all of the coefficients, and can prevent information related to pass / fail from being buried among similar information with strong correlations, so it is considered that the accuracy of pass / fail judgment can be improved compared to other data distances.

[0075] The Mahalanobis distance as the distance between data in the pass / fail judgement unit 26 can be expressed as follows: i1 ,x i2 ,···,x in ) is a reference sample X=(X1,X2,...,X p ) T Considering the average μ=(μ1,μ2,...,μ n ) and covariance matrix Σ=E[(X-μ)(X-μ) T ] for this sampling distribution with observation Y=(y1,y2,...,y n The Mahalanobis distance of

[0076]

number

[0077] In the second embodiment, the quality determining unit 26 determines the distance between data by using the Mahalanobis distance D M (t) is calculated and squared to give D M 2Based on the average value and maximum value of (t), the quality of the bonded state of the parts to be bonded 16 is judged. In the second embodiment, the reference value stored in the reference value storage unit 24 is the D calculated by the quality judging unit 26. M 2 (t) and the value to be compared with the average value of D calculated by the pass / fail judgment unit 26. M 2 The maximum value of (t) is stored in the memory 21. These values ​​can be set based on machine learning by the learning unit 25, as in the first embodiment.

[0078] 6-2. Verification test for pass / fail judgment In the second embodiment, the Mahalanobis distance D calculated by the quality determination unit 26 M 2 The following verification test was carried out for determining whether or not the joint is good based on the average value and maximum value of (t). For 50 pieces of data (26 good pieces and 24 bad pieces) of the joint object 16 that were joined under the same conditions, each piece of acoustic data was decomposed into 40 basis spectra based on the above formula (1) to obtain 40 coefficients (n=40) as samples, and the D M 2 The average and maximum values ​​of (t) were calculated, and the variations in each were normalized by the standard deviation of good products for performance comparison, and the values ​​are shown in Figures 11 and 12. Note that for the maximum values ​​shown in Figure 12, the normalized values ​​of 23 data points, excluding one of the defective product data points (x), were significantly greater than 200, and therefore are not shown.

[0079] As shown in Figures 11 and 12, M 2 In both cases, the average and maximum values ​​of (t) clearly distinguish between good and bad products, and the distance D M 2 It was shown that a high-accuracy pass / fail judgment can be made based on the average and maximum values ​​of (t). M 2 Compared with the average value of (t), the D M 2In this embodiment, the reference value stored in the reference value storage unit 25 can be a value between the good product group and the defective product group shown in FIGS.

[0080] 7.Determining the bonding condition 7-1. First junction state Next, the first bonding state determination process by the ultrasonic bonding system 100 in the second embodiment will be described in detail with reference to the flow chart shown in Fig. 13. First, the same steps as steps S1 to S4 shown in Fig. 8 in the first embodiment are performed. Then, in step S50 shown in Fig. 13, the pass / fail determination unit 26 calculates the Mahalanobis distance D M Then, the process proceeds to step S62, where the Mahalanobis distance D M (t) squared D M 2 The average value of (t) is calculated, and it is determined whether the average value is equal to or less than the reference value stored in the reference value storage unit 25.

[0081] In step S62, D M 2 If it is determined that the average value of (t) is equal to or less than the reference value, the process proceeds to Yes in step S62, and in step S7, as in the case shown in FIG. 8, it is determined that the bonding state is good, and the flow ends. M 2 If it is determined that the average value of (t) is not equal to or less than the reference value, the process proceeds to No in step S62, and the bonding state is determined to be poor in step S8, and the flow ends.

[0082] 7-2. Second bonding state Next, the second bonding state determination process by the ultrasonic bonding system 100 in the second embodiment will be described in detail with reference to the flow chart shown in Fig. 14. First, in the second bonding state determination process, step S50 is performed in the same manner as in the first bonding state determination process shown in Fig. 13, and the pass / fail determination unit 26 calculates the Mahalanobis distance D for all coefficients in a plurality of basis spectra. M Then, the process proceeds to step S63 shown in FIG. 14, where the Mahalanobis distance D M (t) squared D M 2 The maximum value of (t) is calculated, and it is determined whether the maximum value is equal to or less than the reference value stored in the reference value storage unit 25.

[0083] In step S62, D M 2 If it is determined that the maximum value of (t) is equal to or less than the reference value, the process proceeds to Yes in step S62, and in step S7, as in the case shown in FIG. 8, it is determined that the bonding state is good, and the flow ends. M 2 If it is determined that the maximum value of (t) is not equal to or less than the reference value, the process proceeds to No in step S62, and the bonding state is determined to be poor in step S8, and the flow ends.

[0084] 8. Action and Effects In the ultrasonic bonding quality determining device 2 of the second embodiment, the quality determining section 26 determines the quality of the bonding state of the parts to be bonded 16 based on a plurality of coefficients in a plurality of basis spectra. This makes it possible to improve the accuracy of the quality determination compared to when the quality of the bonding state of the parts to be bonded 16 is determined based on a single coefficient.

[0085] Furthermore, in the second embodiment, the quality determining section 26 determines the quality of the bonding state of the parts to be bonded 16 based on all the coefficients in the plurality of basis spectra. This eliminates the need to extract a specific basis spectrum, thereby saving the effort of extraction and further improving the accuracy of quality determination by reducing arbitrariness.

[0086] Furthermore, in the second embodiment, the pass / fail judgment unit 26 can make a pass / fail judgment based on the data distances in a plurality of coefficients, and in the present embodiment, the pass / fail judgment is made based on the data distances in all the coefficients. Then, the Mahalanobis distance is calculated as the data distance. By using the Mahalanobis distance, it is possible to improve the judgment accuracy by using all the coefficients, and it is also possible to prevent information related to pass / fail from being buried in similar information with a strong correlation, thereby improving the accuracy of the pass / fail judgment.

[0087] Furthermore, in the second embodiment, the quality determining section 26 calculates the average or maximum value of the squares of the Mahalanobis distances as the distance between the data, thereby further improving the accuracy of the determination.

[0088] In addition, the second embodiment also provides the same effects as the first embodiment, except for the effects provided by the configuration of the quality determining section 26 in the first embodiment.

[0089] The present invention is not limited to the above-described embodiments and modifications, and can be applied to various embodiments without departing from the spirit of the present invention. [Explanation of symbols]

[0090] 1 Ultrasonic bonding equipment 2. Ultrasonic bonding quality inspection device 100 Ultrasonic Bonding System 11 Ultrasonic Horn 12 Ultrasonic oscillator 13 Anvil 14, 15 Work 21 Acoustic data acquisition unit 22 Acoustic data storage unit 23 Analysis Department 231 Short-time Fourier transform processing section 232 NMF processing section 24 Reference value storage section 25 Learning Department 26. Good / bad judgement section H1~H5 base spectrum W1~W5 coefficients H4 specific basis spectrum W4 Specific Coefficient

Claims

1. An ultrasonic bonding quality determination device that applies ultrasonic vibrations from an ultrasonic horn to a workpiece including a part to be bonded placed on an anvil to determine the quality of the bonding state of the part to be bonded, an acoustic data acquisition unit that acquires acoustic data indicating sounds generated during an ultrasonic vibration application period from the start to the end of application of the ultrasonic vibration; an analysis unit that decomposes the acoustic data acquired by the acoustic data acquisition unit into products of a plurality of basis spectra and coefficients that indicate time changes in intensity of the basis spectra by NMF (Non-negative Matrix Factorization); and a quality determining unit that determines the quality of the bonded state of the parts to be bonded based on coefficients in the plurality of basis spectra.

2. 2. The ultrasonic bonding quality determination device according to claim 1, wherein the quality determination unit determines the quality of the bonding state of the parts to be bonded based on a specific coefficient that is a coefficient of at least one specific basis spectrum identified from the plurality of basis spectra.

3. a reference value storage unit in which a reference value to be compared with the specific coefficient is stored in advance; The ultrasonic bonding quality determining device according to claim 2 , wherein the quality determining unit determines whether the bonding state of the parts to be bonded is good or bad based on a comparison result between the specific coefficient and the reference value.

4. the reference value storage unit stores, as the reference value, a value to be compared with the maximum value of the intensity for the specific coefficient during the ultrasonic vibration application period; 4. The ultrasonic bonding quality determining device according to claim 3, wherein the quality determining unit determines the quality of the bonded state of the parts to be bonded based on a result of comparison between a maximum value of strength in the specific coefficient and the reference value.

5. the reference value storage unit stores, as the reference value, a value to be compared with an integral value of the ultrasonic vibration application period when the specific coefficient is plotted on a graph in which the vertical axis represents the intensity at the specific coefficient and the horizontal axis represents the elapsed time from the application of the ultrasonic vibration, 4. The ultrasonic bonding quality determination device according to claim 3, wherein the quality determination unit determines the quality of the bonding state of the parts to be bonded based on a comparison result between an integral value of the ultrasonic vibration application period at the specific coefficient and the reference value.

6. a learning unit that performs machine learning in advance to set the reference value, The ultrasonic bonding quality determining device according to claim 3 , wherein the reference value storage unit stores the reference value set by the learning unit.

7. The ultrasonic bonding quality determining device according to claim 1 , wherein the quality determining section determines whether the bonding state of the parts to be bonded is good or bad based on a plurality of coefficients in the plurality of basis spectra.

8. The ultrasonic bonding quality determining device according to claim 7 , wherein the quality determining unit determines quality based on distances between data in the plurality of coefficients.

9. The ultrasonic bonding quality determining device according to claim 3 , wherein the quality determining section determines the quality of the bonded state of the parts to be bonded based on all coefficients in the plurality of basis spectra.

10. The ultrasonic bonding quality determining device according to claim 9 , wherein the quality determining unit determines quality based on distances between data in all of the coefficients.

11. The ultrasonic bonding quality determining device according to claim 10 , wherein the quality determining unit calculates a Mahalanobis distance as the inter-data distance.

12. The ultrasonic bonding quality determining device according to claim 10 , wherein the quality determining unit calculates an average value or a maximum value of squares of Mahalanobis distances as the inter-data distance.

13. 13. The ultrasonic bonding quality determination device according to claim 1, wherein an upper limit of the frequency of the sound acquired by the acoustic data acquisition unit is at least 1.2 times the frequency of the ultrasonic vibration.

14. The ultrasonic bonding quality determination device according to any one of claims 1 to 12, wherein the analysis unit creates a spectrogram by performing a short-time Fourier transform on the acoustic data acquired by the acoustic data acquisition unit, and decomposes the spectrogram into products of the plurality of basis spectra and the coefficients using NMF based on the spectrogram.

15. The ultrasonic bonding quality determination device according to any one of claims 1 to 12, an ultrasonic bonding system including the anvil, the ultrasonic horn, and an ultrasonic bonding device including an ultrasonic vibration application unit that applies the ultrasonic vibration to the parts to be bonded.

Images