Study on the ultrasonic guided wave and online visual monitoring for ultrasonic precise bonding

Ultrasonic precise bonding is an emerging technology in the application of polymer micro-assembly. The propagation of ultrasound changes with the interfacial polymer physical state in the ultrasonic bonding process. So the ultrasonic guided wave is an effective parameter to in-situ monitor the fusion degree. The time-frequency characteristics in the ultrasonic guided wave are analyzed by vibration analysis methods. The polymer interfacial fusion is online visual monitored by the high-speed HD camera. The fusion behavior of the thermal melt interface and the time-frequency characteristics are analyzed and correlated. Results indicate that the change of the interfacial thermal melt state is related to the time-frequency characteristics of the ultrasonic guided wave. The generation of the melting zone, the fusion of the melting zone, the rotation of the micro-device, the generation or disappearance of local air bubbles all lead to the changing of the harmonic frequency and intensity in the ultrasonic bonding process.


Introduction
Since the 1990s, polymer micro-device has been widely used in the manufacture of MEMS devices such as microfluidic biochips, micro fuel cells, micro-valves, and micro-pumps due to high forming efficiency and lightweight [1,2]. As the improvement of the application technology of micro-devices, the demand for multi-functional integrated polymer microdevices increased. The precise bonding of polymer microdevices is the key in micro-nano manufacturing. Several bonding techniques have been applied in micro-nano manufacturing such as thermal bonding [3], laser bonding [4], adhesive bonding [5], microwave bonding [6], and ultrasonic bonding [7]. Compared with other bonding methods, ultrasonic bonding is more efficient. Ultrasonic bonding is suitable for most thermoplastic polymer materials without the additional implantation of energy-absorbing media. The complexity of the bonding process is reduced. The heataffected zone is controlled in a small area near the interface by the reasonable interface structure and process parameter configuration. The structural thermal deformation is little.
Ultrasonic bonding technology uses high-frequency periodic mechanical force to move two parts relative to each other and generates heat locally by interface friction and periodic energy loss of materials. Wu et al. [8] studied the interfacial friction heating mechanism through experiments and numerical schemes. The results indicate that the interfacial friction heating of polymer granules is a transient process with an inhomogeneous temperature field around the interface. The interfacial friction heating only acts in the initial stage of the ultrasonic bonding process. The flow rate of interfacial friction heating increases with the increase of ultrasonic amplitude. Zhang et al. [9] measured the interface temperature to study the heating mechanism. Results of simulation and experiment reveal that viscoelastic heat provides the most required heat in the bonding process when the temperature reaches Tg (glass transition temperature) of the material.
Ultrasonic bonding was first performed in the mid-1960s. Matheny et al. [10] introduced the process principle of ultrasonic metal bonding of softer alloys of copper, aluminum, and nickel, and the advantages and disadvantages of the bonding process. Benatar et al. [11] studied the difference between near and far-field ultrasonic bonding and discussed the weldability of thermoplastic materials. Truckenmuller et al. [12] first applied ultrasonic bonding technology to the bonding and packaging of microfluidic chips and the assembly of micro-pumps, micro-valves, and other small devices, which guided ultrasonic bonding technology in the field of MEMS in 2006. Jingmin Li et al. [13] designed an energy director structure and selfbalancing jig to improve the uniformity of the microchannel height and provide an effective method for large-scale ultrasonic bonding of microfluidic devices. For the bonding of metal and polymer materials, Kim Jongbaeg et al. [14] developed a miniature ultrasonic bonding device to realize the bonding of cellulose acetate micro-devices. Zhang et al. [15] verified the effectiveness of a new thermally-assisted ultrasonic bonding method for polymer Micro/NanoElectro-Mechanical Systems (M/NEMS) through bonding experiments on PMMA microfluidic chips. Sun et al. [16] proposed an ultrasonic precise bonding method based on ultrasonic propagation efficiency feedback to effectively improve the precision of ultrasonic bonding.
Online visual monitoring is introduced based on the ultrasonic guided wave feedback. The time-frequency characteristics are extracted by the vibration algorithm based on the timefrequency domain analysis method. The polymer interfacial thermal melt state is monitored and recorded by the highspeed HD camera online. The correlation between the timefrequency characteristics and the thermal melt state is analyzed on the same time scale. Compared with the traditional time, pressure, energy, and other parameter control methods, the monitoring of the thermal melt interface provides an effective method for the ultrasonic guided wave in this study. This paper is organized as follows. Section 2 introduces experimental materials and the control flow of the test rig. Section 3 introduces the vibration analysis methods. The correlation between the time-frequency characteristics and the thermal melt state change is studied in Section 4. Finally, conclusions are drawn in Section 5.

Polymer ultrasonic precise bonding experiment
The bonding of connecting tube and substrate is carried out in experiments. PMMA is chosen to be the material for its high transparency, low price, easy processing. PMMA is a thermoplastic material, which is usually used as a substitute for glass. The glass transition temperature of PMMA is approximately 105°C. The PMMA connecting tube and PMMA substrate are shown in Fig. 1. The substrate is a rectangular parallelepiped of 8 mm×8 mm×2 mm, and the connecting pipe is a hollow cylinder with an outer diameter of 4 mm, an inner diameter of 1 mm, and a thickness of 2 mm.
The ultrasonic precise bonding test rig is established to study the correlation between the time-frequency characteristics in the ultrasonic guided wave and the thermal melt state change of the thermal melt interface during the polymer ultrasonic precise bonding process. The ultrasonic precise bonding test rig is shown in Fig. 2. The test rig is mainly composed of five parts: an ultrasonic system, a displacement-driven pressurizing mechanism, a leveling mechanism, a sensor detection module, and a high-speed HD camera.
The ultrasonic system includes a numerical control ultrasonic generator, a transducer, and an ultrasonic horn. The ultrasound is provided by the numerical control ultrasonic generator. The electrical energy is converted into mechanical energy by the transducer. Mechanical energy is transferred to the PMMA components by the ultrasonic horn. The movement of the ultrasonic horn is driven by the stepping motor and the linear motion guideway, which is a kind of displacement-driven pressure mode. Compared with the pneumatic pressure mode, it is more accurate and stable in the control of the pressure and deformation. The horn is driven to move longitudinally by the pressurizing mechanism to provide pre-tightening force for the microdevice. The sensor detection module includes CL-YD-301 piezoelectric dynamic force sensor produced by United Energy Electronics, EVT-20G static force sensor produced by Shanghai Youran Sensing Technology Co., Ltd., YE5852B type charge amplifier produced by Jiangsu Donghua Co., Ltd., and data acquisition card produced by Advantech Co., Ltd.. Leveling mechanism is designed to ensure that the horn and the bonding interface of the micro-device remain vertical in the experiment.
As shown in Fig. 3, the EVT-20G static force sensor produced by Shanghai Youran Sensing Technology Co., Ltd. is installed in the fixture of the ultrasonic horn. The pretightening force is measured by the static force sensor and adjusted by the linear motion guideway. In the ultrasonic bonding process, the components are compressed by the horn, and interfacial fusion is achieved under ultrasonic vibration. The vary of interfacial polymer physical state leads to the change of ultrasonic guided waves in the whole structure. The CL-YD-301 dynamic force sensor produced by United Energy Electronics is fixed under the static force sensor to measure the ultrasonic guided wave generated from the horn through the fixture. The high-speed HD camera produced by Shenzhen YVSION Technology Co., Ltd. is installed in the base to observe the interfacial fusion of polymer components. In this paper, the fusion of the thermal melt interface is online visual monitored by the ultrasonic guided waves. The ultrasonic frequency is 60 kHz, the ultrasonic amplitude is 2 μm, and the pre-tightening force is 1.6 N. The ultrasonic guided wave is detected by the dynamic force sensor, and the interfacial fusion process is online detected by the high-speed HD camera. The time-frequency characteristics in the ultrasonic guided wave that are analyzed by vibration analysis are correlated with the fusion behavior of the thermal melt interface.

Vibration analysis algorithm
The ultrasonic guided wave is analyzed by time domain, frequency domain, and time-frequency domain methods. The applied methods include Fast Fourier Transform (FFT) spectrum analysis, Wavelet Packet Decomposition (WPD), and envelope spectrum, respectively.

FFT spectrum analysis
Spectrum analysis is a common frequency domain analysis method. The signal spectrum is the total set of all frequency components that make up the signal. The signal is transformed from the time domain to the frequency domain by FFT to obtain the frequency components of the vibration signal.
FFT is an efficient and fast algorithm to deal with big data. For point N sequence x(n), the FFT pair is: The computation of N points X(k) requires N 2 complex multiplication and N(N-1) complex addition. Realizing a complex number multiplication requires 4 real number multiplications and 2 real number additions, realizing a complex number addition requires 2 real number additions. It is found that the amount of calculation is huge. The characteristics of FFT are as follows: According to the characteristics of FFT, the complex calculation is simplified, and the computational efficiency is greatly improved. The calculation amount of complex number multiplication of point N is reduced from N 2 to N 2 log N 2 times. The calculation amount is significantly reduced.

Envelope spectrum
Envelope spectrum is a frequency domain analysis method based on spectrum analysis, which is effective in analyzing the frequency of vibration signals. Firstly, the time domain signal is demodulated by Hilbert transform, and the extreme value is calculated. Then the envelope is obtained from the one-dimensional data. Finally, Fourier Transform (FT) is performed on the envelope. Unnecessary low-frequency interference is eliminated by envelope spectrum compared with FFT spectrum analysis. The harmonics are more effectively distinguished by the envelope spectrum.
The Hilbert transform of the signal x(t) is defined as: After the signal is demodulated by Hilbert transform, taking x(t) and b x t ð Þ as the real part and imaginary part, the analytical signal z t ð Þ is: The envelope curve equation of the signal is: The basic principle diagram of the polymer ultrasonic precise bonding test rig

Wavelet packet decomposition
As the ultrasonic guided waves in the ultrasonic bonding process are nonstationary time-domain signals, it is necessary to study both time-domain and frequency-domain in the process of vibration analysis. WPD is a multiscale time-frequency local analysis method, which is high resolution for the highfrequency and low-frequency components of signals. The consistency of the time scale is ensured by WPD. In wavelet multi-resolution analysis, L 2 (R)=⊕ j∈z Wj means that the space The decomposed subspace is defined as U k j , which represents the k-th wavelet packet subspace in j scale, the frequency factor is represented by k, k=0, 1, …, 2j −1. In this study, the ultrasonic guided wave is decomposed by the three levels wavelet packet, set j=3, and the expression of the decomposition is: The ultrasonic guided wave is decomposed into 2 3 =8 subspaces, and the frequency band corresponding to the third layer subspace is: f is the sampling frequency.
The sampling frequency of the ultrasonic guided wave is 500 kHz, the Nyquist frequency is 250 kHz, and the frequency band bandwidth of each subspace is 250 kHz/8=31.250 kHz.
Let di,m(r) be the amplitude of each discrete point after reconstruction, where i and m are the frequency factor and decomposition scale, r is the number of discrete sampling points. After the three levels wavelet packet, the energy of each subband signal is: The energy is normalized when it is large: The time-frequency characteristics are effectively extracted based on the above-mentioned vibration analysis methods. The harmonic of the ultrasonic guided wave is found by FFT spectrum analysis. The original time-domain signal is decomposed and reconstructed by WPD, and the main frequency band position is accurately located by the energy proportion. The harmonic generation (HG), second-harmonic generation (SHG), and third-harmonic generation (THG) generated by the main frequency band are extracted by spectrum envelope.

Experimental results and analysis
The precise bonding experiment was carried out based on the test rig. The time-domain waveform of the ultrasonic guided wave is shown in Fig. 4. The time range of 0~2s is the initial stage of ultrasonic loading, and the dynamic force amplitude changes significantly. First, the amplitude rises to 0.62 dB, Fig. 4 The time-domain waveform of the ultrasonic guided wave then attenuates to 0.16 dB, finally rises to the peak of 0.87 dB. The amplitude gradually attenuates to 0.2 dB after reaching the peak. In the time range of 2.5s~5.5s, multiple small peaks continuously appear, and the peaks increase with time. The amplitude is stable at 0.2 dB after 5.5s, and the ultrasonic loading is stopped at 9.5s.
The spectrum of the ultrasonic guided wave by FFT analysis is shown in Fig. 5. The frequency range of FFT spectrum analysis is from 0 to 125 kHz. The low-frequency noise generated in the experiment is mainly concentrated around 10 kHz. The main frequency is concentrated at 60 kHz, and the maximum amplitude reaches 1.3×10 5 dB.
The frequency characteristics of the 60 kHz main frequency band are studied. The frequency spectrum of the ultrasonic guided wave is processed by band-pass filtering to obtain the frequency spectrum of the 60 kHz frequency band. The frequency spectrum of the main frequency band around 60 kHz from 54 kHz to 66 kHz is shown in Fig. 6. The partial enlargement figure of the harmonic is shown in Fig. 7.
From Figs. 6 and 7, the main frequency of the ultrasonic guided wave is 59.8 kHz, and the frequency amplitude is as high as 127.6 kdB. The harmonic is generated by the 60 kHz main frequency band, the amplitude of which is smaller than the main frequency amplitude. The harmonic is symmetrical around the main frequency. The frequency of the left characteristic point is 55.49 kHz, the amplitude is 134.4 dB, and the right ones are 63.95 kHz and 578.8 dB.
The ultrasonic guided wave is decomposed by WPD step by step downwards. The number of decomposition levels is three levels, and a total of eight nodes in the third level. The tree decomposition after the three levels WPD is shown in Fig. 8. Each node corresponds to a different wavelet packet coefficient, which determines the The entire frequency-domain is divided into eight segments on average. The energy distribution proportion of each frequency band is shown in  Fig. 9 and Fig. 10. From the two waveforms, the amplitude change trends are the same. In the time range of 0~1.8s, the dynamic force amplitude increases sharply to about 1 dB and then attenuates to 0.35 dB. The reason is that the high-frequency vibration of the thermal melt interface occurs when ultrasound is loaded, the bonding interface is no longer in the glass state, and the amplitude changes significantly. In the time range of 1.8s~2.9s, the amplitude attenuates to 0.24 dB when the thermal melt interface is melted locally. In the time range of 2.9s~9.5s, the melting of the thermal melt interface requires continuous absorption of energy, and the amplitude increases slightly.
The envelope spectrum of the characteristic waveform of the 60 kHz main frequency band is shown in Fig. 11. The frequency range of the envelope spectrum is from 0 to 120 kHz.
From Fig. 11, it is found that a group of harmonic with obvious regularity, whose frequencies increase in multiples and amplitudes decrease in order. The harmonics are generated by the 60 kHz main frequency band, which are HG, SHG, and THG, respectively. The three frequencies from small to large are 5.393 kHz, 10.790 kHz, and 16.180 kHz, and the corresponding amplitudes from large to small are 45640 dB, 8218 dB, and 2574 dB. The frequency components between 20 kHz and 40 kHz have no regularity, and the frequency amplitudes are all below 2000 dB. The frequency components between 40 kHz and 60 kHz are relatively simple, and the corresponding amplitudes are extremely small. The frequencies larger than 60 kHz are not harmonics.
The HG, SHG, and THG time-frequency characteristics in the ultrasonic guided wave are processed by the vibration analysis algorithm. The three-dimensional diagram of the HG, SHG, and THG time-frequency characteristics is shown in Fig. 12. The x-axis is the time(s) of ultrasonic loading, the y-axis is the frequency (Hz), and the z-axis is the frequency intensity (dB). The HG, SHG, and THG positions are marked in Fig. 12. The ultrasonic loading time is 9.5s, and the frequency range is 0 to 20 kHz.
From Fig. 12, it is found that the trends of the HG, SHG, and THG are the same as the original waveform. They are all shown the trend of increasing-decreasing-increasing within 0~9.5s. In the time range of 0~2.8s, the HG, SHG, and THG frequency intensities vary significantly, which the peaks are To obtain the time-frequency characteristics in ultrasonic guided waves more clearly, the three-dimensional diagram is converted into two-dimensional diagrams. The time-harmonic frequency characteristics diagram and the time-harmonic frequency intensity characteristics diagram are shown in Figs. 13 and 14. The range of frequency intensity is truncated to 3000 dB to study the characteristics of Fig. 14 in more detail due to the HG maximum frequency intensity is 2886 dB.
From Fig. 13, it is found that the frequency trends of the HG and SHG are the same within 0~9.5s. In the time range of 0~2.8s, the frequency of HG and SHG shows a trend of decreasing first and then increasing, which attenuate to  Fig. 11 The envelope spectrum of the characteristic waveform in the 60 kHz main frequency band Fig. 12 The three-dimensional diagram of the HG, SHG, and THG time-frequency characteristics Fig. 13 The time-harmonic frequency characteristics diagram Fig. 14, it is found that the range of the HG frequency intensity is 757.7 dB to 2886 dB, the SHG is 106.2 dB to 552.8 dB, and the THG is 27.7 dB to 221.4 dB. The HG and THG frequency intensity is decreased at 0.9s and 3.6s, while the SHG is increased. The HG and THG frequency intensity is increased at 1.2s, while the SHG is decreased. The SHG and THG frequency intensity is increased at 4.2s, while the HG is decreased. The trends of HG, SHG, and THG frequency intensity are the same after 4.2s. Eight characteristic points are extracted according to the interfacial thermal melt state characteristics of the thermal melt interface during the ultrasonic bonding process. The thermal melt interface is unbonded at the start. The interface of the micro-device starts to change under the ultrasonic load. Eight characteristic points of the thermal melt interface are shown in Fig. 15.
From characteristics point 1, it is found that the melting points 1 and 2 appear in the random position on the thermal melt interface. In this case, it emerges from the bottom left of the thermal melt interface. The reason is that in the initial stage of ultrasonic loading, the heat generated by the friction of contacting surfaces under high-frequency vibration and is irregularly distributed due to the difference and random distribution of the surface roughness, melting points appear randomly. At characteristics point 2, the melting points 1 and 2 generate melting zone 1, and continue to spread. At characteristics point 3, a new melting zone 2 is generated, which is located at the upper right of the thermal melt interface. At Fig. 14 The time-harmonic frequency intensity characteristics diagram Fig. 15 The thermal melt interface of the micro-device characteristics points 4 and 5, the melting zones 1 and 2 are, respectively, fused to generate fusion zones 1 and 2. At characteristics point 6, dense air bubbles are generated on the thermal melt interface. The reason is that the high-frequency reciprocating motion of the thermal melt interface causes the temperature to rise, and the micro-device interface turns into a viscous flow state, resulting in the ultrasound cavitation effect. The thermal melt interface is fully fused at the end.
In the starting, the thermal melt interface is in a glassy state, and the temperature of the interface rises under the ultrasound load. The thermal melt interface is transformed from a glass state to a high elastic state, and melting zones appear when the temperature reaches the T g . The high elastic state is transformed to the viscous flow state, and the thermal melt interface melts in a large area when the temperature further rises to the viscous flow temperature (T f ).  The interfacial fusion process is studied together with the HG, SHG, and THG time-frequency characteristics in the ultrasonic guided wave. The correlation of thermal melt state and time-harmonic frequency is shown in Fig. 16. The correlation of thermal melt state and time-harmonic frequency intensity is shown in Fig. 17. The two diagrams are, respectively, divided into seven stages based on eight time-frequency characteristics points in Fig. 15.
From stage I in the range of 0~0.9s, it is found that the THG frequency changes from 15.900 kHz to 14.740 kHz at 0.4s, the SHG frequency intensity decreases from 286.8 dB to 232.3 dB. The micro-device starts to rotate at this time. The THG frequency changes from 14.620 kHz to 14.460 kHz at 0.9s, and the intensity increases from 120.8 dB to 189.3 dB. Melting points 1 and 2 are randomly generated on the thermal melt interface at 0.9s.
From stage II in the range of 0.9s~1.3s, the THG frequency rises to 15.710 kHz at 1.2s and decreases to 14.820 kHz at 1.3s. The SHG frequency intensity rises to a small peak of 396.2 dB at 1.2s. The THG frequency intensity rises to a small peak of 221.4 dB at 1.3s. The rotation process of the micro-device and the fusion process of melting zone 1 are shown in Figs. 18 and 19.
From Fig. 18, it is found that the micro-device remains motionless before 0.4s, and position 1 forms an angle of 26°w ith the centerline. The micro-device starts to rotate from 0.4s and stops rotating at 1.2s. Rotation with an angle of 3°with the centerline occurs in position 1. From Fig. 19, it is found that there are two small-zone melting points 1 and 2 on the left side of the thermal melt interface, which gradually spread to the surrounding under the ultrasonic load. Melting points 1 and 2 generate melting zone 1 at 1.3s.
From stage III in the range of 1.3s~1.8s, the melting zone 1 spreads uniformly around. The HG, SHG, and THG frequency is stable at 5.210 kHz, 10.420 kHz, and 15.670 kHz, and the frequency intensity is 900.1 dB, 136.4 dB, and 42.2 dB, respectively, at 1.8s.
From stage IV in the range of 1.8s~2.6s, the trends of the HG, SHG, and THG frequency are increased. The melting zone 2 is generated at the upper right of the thermal melt interface at 1.8s, and then the melting zones 1 and 2 spread around together. At 2.6s, the THG frequency attenuates sharply from Fig. 19 The fusion process of melting zone 1 Fig. 20 The local fusion process of melting zones 1 and 2 16.010 kHz to 13.510 kHz, and the HG and THG frequency intensity is the small peak. The melting zones 1 and 2 are locally fused. The local fusion process is shown in Fig. 20.
From Fig. 20, it is found that the melting zones 1 and 2 before fusion are located at the lower left and upper right of the thermal melt interface, respectively, and the area of melting zone 1 is larger than melting zone 2. They spread around at the same melting rate under ultrasonic load. The melting zones 1 and 2 are fused for the first time at the upper left of the thermal melt interface, and then the local fusion zone further spread from the center to the edge at 2.6s. There is another local unfused zone at the lower right of the thermal melt interface.
From stage V in the range of 2.6s~3.6s, the THG frequency attenuates from 16.190 kHz to 14.960 kHz at 3.6s. The HG and THG frequency intensity is decreased to 1109 dB and 100.5 dB, respectively, while the SHG increased to 176 dB. The melting zones 1 and 2 are completely fused at 3.6s.
From stage VI in the range of 3.6s~4.2s, a large area of the melting zone is generated on the thermal melt interface. The bonding process entered the mid-term. At 4.2s, the THG frequency rises from 14.970 kHz to 16.190 kHz. The HG frequency intensity is the peak, while the SHG and THG frequency intensity is the troughs. Dense air bubbles begin to appear on the thermal melt interface at this time.
From stage VII in the range of 4.2s~9.5s, which is the later stage of the bonding process. The HG, SHG, and THG frequency is stable at 5.400 kHz, 10.800 kHz, and 16.000 kHz, respectively. The THG frequency attenuates sharply to 13.870 kHz at 4.9s, and the intensity rises from 1477 dB to 1517 dB. The SHG frequency intensity attenuates from 225 dB to 215.5 dB, and the THG does not change obviously. The disappearance process of local air bubbles on the thermal melt interface is shown in Fig. 21.
From Fig. 21, it is found that some local air bubbles appear at the upper left of the thermal melt interface starting from 4.2s. Under the ultrasonic load, the number and diameter of local air bubbles are increased. At 4.9s, the air bubbles with increasing cavity area rupture, and disappear at the upper left of the thermal melt interface.
The thermal melt interface is completely fused at 8.5s, and the micro-device interface no longer changes. The ultrasonic loading is stopped at 9.5s, the micro-device bonding process is ended.

Conclusion
The correlation between the time-frequency characteristics in the ultrasonic guided wave and the interfacial thermal melt state during the polymer ultrasonic bonding process is studied. The interfacial thermal melt state change process is monitored online by the high-speed HD camera, and the time-frequency characteristics are extracted by the vibration analysis algorithm based on FFT spectrum analysis, WPD, and envelope spectrum. The results indicate that the generation and fusion of melting zones, the rotation of micro-devices, and the generation or disappearance of air bubbles are all accompanied by changes in the frequency and intensity of harmonics. The generation and fusion of the melting zone lead to the constant fluctuation of the harmonic frequency, the rotation of the micro-devices leads to the instantaneous decrease of the harmonic frequency, and the change of the bubbles causes the SHG and THG frequency intensities to be greatly reduced. The characteristic components of ultrasonic guided waves are effective parameters for in-situ monitoring of the fusion degree of the thermal melt interface.
General program funding for the China Postdoctoral Science Foundation (2019M651103).
The National Science Foundation of China under Grant (52005071).

Data Availability
The authors declare that the data and the materials of this study are available within the article.

Declarations
Ethics approval Not applicable.
Consent to participate Informed consent was obtained from all individual participants included in the study.

Consent for publication
The participants have consented to the submission of the case report to the journal.