Open Access
Add to BibSonomy
Abstract
The shape of an instance hole (keyhole) created via a high-power laser was measured using a low-coherence interferometer with the following parameters: repetition rate, 10 MHz; center wavelength, 1550 nm; absolute spatial resolution, 10 µm; and measurement range, 5 mm. The keyhole was created on a 3-mm-thick stainless-steel plate using a high-power laser with 8-kW peak power and 1070-nm center wavelength. The cross-sectional area of the keyhole was measured to be 0.42 mm × 0.78 mm (width × depth) using the interferometer, and its side dimension was 0.46 mm × 0.78 mm (width × depth).
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
A sustainable society requires efficient and stable systems, including carbon-neutral energy [1], highly robust financial [2], semantic sensing [3], controllable fusion [4], narrative/antinarrative communication [5,6], and digital transformation (DX) [7] systems. DX systems include laser technology, which is a highly controllable tool. In particular, laser-based welding technology has been studied vigorously owing to its higher controllability than the typical plasma-based welding technology [8,9]. The laser-welding process induces three remarkable physical phenomena: a tiny hole (keyhole), which is created during the submillisecond high-power laser irradiation [10]; a metal–liquid (molten pool) phase transition [11]; and a metal–plasma phase transition [12]. Although these phenomena have been investigated by several researchers, the in-process distribution measurement of generated keyholes has not been demonstrated [13–16].
Optical measurement techniques can be applied to the in-process measurement of physical phenomena occurring in short periods. Generally, optical coherence tomography based on low-coherence interference (LCI) can be employed to measure the path length distribution of scattered or reflected light from an object. Three types of LCI exist: time-domain LCI (TD-LCI), swept-source LCI (SS-LCI), and spectral-domain LCI (SD-LCI) [17,18]. TD-LCI is suitable for low-repetition operations whereas SD-LCI and SS-LCI are applied to high-repetition measurements. The number of repetitions is limited by the sweep frequency of the light source in the case of SS-LCI and by the acquisition speed of the spectrometer in the case of SD-LCI. Recently, the time-stretch dispersion Fourier transform (TS-DFT) technique was proposed for performing high-repetition-rate measurements above 1 MHz [19,20]. To overcome the shortcomings of LCI, TS-DFT-based LCI has been studied [21,22]. In this method, the conversion from a time axis into a wavenumber axis and the measurement of group delay and chirp rate are required. In Ref. [21], the group velocity dispersion was measured with an expensive device, i.e., an optical spectrum analyzer, before measuring the sample; subsequently, the wavelength was calibrated. However, in LCI using TS-DFT, measuring both the spectrum and spectral interferogram is necessary. In Ref. [22], a single interferogram was used to calibrate and correct the nonlinearity of the chirp rate. Recently, we proposed an approximate calculation method that does not require expensive equipment and applied this method to TS-DFT [23]. We named this unique calculation method, “Applied TS-DFT Micrometer Length Measurable Light Detection and Ranging (MicroLiDAR)” [23]. The MicroLiDAR technique has not been applied to the in-process measurement of keyholes.
In this study, we applied the MicroLiDAR technique to the in-process measurement of keyholes after the evaluation of the MicroLiDAR system. The repetition rate of the system was 10 MHz; the center wavelength, 1550 nm; the absolute spatial resolution, 10 µm; and the measurement range, 5 mm. Using the MicroLiDAR, we measured the dimensions of a keyhole formed by irradiating a 3-mm thick stainless-steel plate with a high-power laser having a central wavelength of 1070 nm and a peak power of 8 kW. The cross-sectional area of the keyhole was measured to be 0.42 mm × 0.78 mm (width × depth), and its side-view dimension was 0.46 mm × 0.78 mm (width × depth).
2. Principle
In optical analog computations, TS-DFT is considered an extremely simple method. Figure 1 shows a schematic of the TS-DFT process. When an ultrashort pulse passes through an optical fiber, the time waveform of the laser pulse is elongated owing to the effect of the wavelength dispersion, whereas the pulse propagates in the optical fiber. The dispersion coefficient of the optical fiber is expressed as
Figure 2 describes the principle of LCI with TS-DFT. The experimental system comprises an ultrashort pulse laser, a dispersion compensation optical fiber (DCF), a 50:50 optical fiber coupler, a reference mirror, a signal mirror, a photodiode, and an oscilloscope. At the photodetector, the electric fields of the reference light (${E_\textrm{r}}$) and the signal light (${E_\textrm{s}}$) are expressed as
Here, $\Delta l$ denotes the difference between the reference light and signal light path lengths, and $c\; $ represents the speed of light. ${E_{r0}}$ denotes the amplitude of the reference light, and ${E_{s0}}$ represents the amplitude of the signal light. These light components interfere in the detector, and the intensity, I, is expressed as
The relationship between the approximated and measured length $|{\mathrm{\Delta }l} |$ is calculated based on the third part of Eq. (6) as follows:
where f represents the peak frequency of the interference spectrum; D [ps/nm/km] denotes the dispersion coefficient of DCF; and ${l_0}$ [km] represents the length of the fiber.Further, we describe a procedure for calculating the distance using the approximation method [23]. Figure 3 shows the procedure, which can be divided into two main parts: calibration and measurement. In the calibration part, Fig. 3(a) shows the time series data ($Sa(T )$) measured using an oscilloscope under the condition that the difference between the reference and measurement path lengths ($\Delta l$) is $\Delta {l_1}$. In this case, the wavenumber monotonically decreases as the time increases. To synchronize the increasing direction of the time axis and that of the wavenumber axis, the time axis inverted timeseries, $Sa({T^{\prime}} )$, is calculated (Fig. 3(b)). On adding a time shift (${T_0}$) to $Sa({T^{\prime}} )$, the origin of the $S({T^{\prime} - {T_0}} )$ coincides with the central wavelength (Fig. 3(c)). The Hilbert transformation of $S({T^{\prime} - {T_0}} )$ affords its real and imaginary parts (Fig. 3(d)) from which the phase timeseries can be obtained (Fig. 3(e)). The phase data are approximated using a quadratic curve (Figs. 3(f) and (g)). The approximation equation is
where $\varphi (T )$ represents the phase data; $T$ denotes time; and ${B_1}$ and ${B_2}$ represent the first and second coefficients, respectively. The several ${B_2}$ and ${B_1}\; $ coefficients are measured by changing $\Delta l$ from $\Delta {l_2}$ to $\Delta {l_4}$. By plotting the dependence of ${B_2}$ on $\Delta l$, the approximated line is obtained. The approximated line about ${B_1}$ is also obtained as follows: where ${a_1}$ and ${a_2}$ denote coefficients. The time axis of the signal can be converted into the wavenumber with these coefficients (Fig. 3(h)), as follows: where ${\lambda _0}$ is the center wavelength of $S(t )$. Note that the relationship between $\Delta y$ and $\Delta l\; $ is $\Delta y = \Delta l/2$ when the path difference in the reference path is $\Delta y$ [m].
Fig. 3. Calibration and measurement procedure: (a) original measured waveform (Sa(T): time series of the waveform, $\Delta l$: path length difference, $\Delta {l_1}$: first path length difference, T: time), (b) time-reversal of Sa(T’) (T’: reversed time), (c) time shift for the origin setting (T0: shifted time, ${\lambda _0}$: center wavelength), (d) Hilbert converted waveform (Re(S): real part of the sample, Im(S): imaginary part of the sample), (e) phase time series (B2 [1/s2]: coefficient of T2, B1 [1/s]: coefficient of T), (f) B2 dependence on the path length difference (a2: coefficient of $\Delta l$, b2: constant), (g) B1 dependence on the path length difference (a1: coefficient of $\Delta l$, b1: constant), (h) k dependence on time (k0: center wavelength), (i) measured waveform dependent on k, (j) Fourier transformation of S(k) (F(S(k)): Fourier-transformed S(k)), and (k) measured distance dependence of $\Delta l$.
The next section presents the measurement procedure. The time axis of arbitrary $S(t )$ is converted into a wavenumber axis with Eq. (11) (Fig. 3(i)). Employing the Fourier transformation of $S(k )$, the horizontal axis is changed to reflect the distance (Fig. 3(j)). The peak position of $S(k )$ indicates the measured distance. The peak position of $S(k )$ dependence on $\Delta l$ is theoretically determined to be 1.00. By calculating this dependence, the calibration can be evaluated (Fig. 3(k)).
For the practical application of the MicroLiDAR, the relationship between the measurement range, resolution, and repetition rate is explained. The measurement range is proportional to the bandwidth and chirp rate of the detector. The resolution is inversely proportional to the wavelength bandwidth of the incident light. To improve the performance of the MicroLiDAR in terms of resolution and measurement range, an incident light with a wide bandwidth and a time stretcher with a large chirp rate are necessary. The restriction condition of the repetition rate is the use of chirped pulses that do not overlap. Employing a laser source with a center wavelength of 1550 nm, wavelength bandwidth of 100 nm, repetition rate of 10 MHz, and chirp rate of 1000 ps/nm, the measurement resolution and range were calculated to be 10 µm and 40 mm, respectively.
3. Experimental setup and method
The experimental setup of the MicroLiDAR system is shown in Fig. 4. This system comprises two sections: the MicroLiDAR and welding head sections.
Fig. 4. (a) MicroLiDAR and welding head (BOA, wideband optical amplifier; BPD, balanced photodetector; GS, Galvano scanner; HNLF, highly nonlinear fiber; Laser Source 1, femtosecond laser; Laser Source 2, high-power laser; OSC, oscilloscope; PC, polarization controller; RA, robot arm; DM, dichroic Mirror, FL, focus lens), (b) sample 1 (mirror with an auto stage for displacement control), (c) sample 2 (step-shaped alumina bulk), and (d) sample 3 (stainless-steel plate).
Laser source 1 is a figure-eight laser with polarization-maintaining fibers. The configuration is similar to those in Ref. [23]. The repetition rate of the output was 28 MHz, the average peak power was 6 dBm, and the nonreputational phase shifter was used as the self-starting device of the femtosecond pulse. Its center wavelength was 1550 nm, the pulse width was 400 femtosecond, and a 100-m-long highly nonlinear fiber (HNLF) was used to increase the wavelength range of the pulse. A 20-km-long dispersion-composed fiber was used as a pulse-stretching device. The dispersion coefficient was 935 nm/s. The output of laser source 1 was amplified using a broadband semiconductor optical amplifier (BOA) after stretching in the DCF. The output was divided into two parts: one was used as the measurement light and the other as the reference light. The measurement light was launched toward the sample after adjusting its beam angle. The reflected light from the sample was coupled with the reference light, whose beams were amplified with the BOA. The beat signal was converted into an electrical signal using a balanced photodetector. The electrical signal was analyzed using an oscilloscope controlled through a personal computer. The sampling ratio of the oscilloscope was 20 GS/s, which corresponding to 7 GHz. The polarization state of the reference signal was optimized using a polarization controller to maximize the beat signal. The length of the reference light was optimized with a delay line comprising a silica single-mode fiber. The wavelength of laser source 2 was 1070 nm, which is 0.67 times higher than that of laser source 1. The peak-pulse power was 8 kW. The light was coupled with the output of laser source 1 in the dichroic mirror. Then the mixed light launched toward the sample after passing focus lens. The focus point was set to the surface of sample with 0.0345 mm spatial resolution. The laser spot diameter was 0.200 mm. The beam direction was controlled with a robot arm, which had three-dimensional freedom. A welding speed was 3 m/min. No shield gas was used. A mirror with an auto stage for controlling the position of the mirror was prepared as sample 1 (Fig. 4(b)). The displacement was changed from 0 µm to 40 µm with 10-µm steps using an ultrasonic vibration motor. Thereafter, a step-shaped aluminum bulk sample was prepared as sample 2 (Fig. 4(c)). The step interval was 1.0 mm, with five steps in total. A steel plate with a width of 3 mm (SUS304) was employed as sample 3 (Fig. 4(d)).
4. Experimental results
After the evaluation of the MicroLiDAR system, we measured the keyhole in sample 3 using a laser-welding system including the MicroLiDAR. First, we evaluated the spatial resolution of the MicroLiDAR system. The position of the mirror on the auto stage was fixed to zero point. Subsequently, the interference signal was obtained.
The signal is shown in Fig. 5. The pulse width of the signal was 40 ns. The pulse shape was complex owing to the HNLF effect, which can be controlled using HNLFs with different ${\beta _3}$ values [24]. The position of the mirror was swept from zero point to 40 µm per 10 µm. The Fourier-transformed signal is shown in Fig. 6(a). Each peak is clearly observed. Figure 6(b) shows the dependence of the measured distance on the position of the mirror. The dependence was almost linear with a coefficient of 1.00, and the accuracy of this measurement was 0.21 µm; thus, it could be applied for practical keyhole measurements. Further, the measurement range was evaluated. The laser output was swept on sample 2. The Fourier-transformed signal is shown in Fig. 7(a). A difference of less than 5.0 mm in the distance was clearly detected. The dependence of the measured distance on the fixed distance in each step was linear, with a coefficient of 1.00, as shown in Fig. 7(b).
Fig. 6. Evaluation of the spatial resolution: (a) frequency domain analysis and (b) distance measurement.
Fig. 7. Evaluation of the measurement range: (a) frequency domain analysis and (b) distance measurement.
Finally, we demonstrated the in-process measurement of the keyhole using a laser-welding system equipped with the MicroLiDAR. Simultaneously, the output of laser source 2 was launched toward sample 3, and the output of the MicroLiDAR was swept around sample 3 by controlling the Galvano mirror. The output of laser source 2 was launched in the −X direction at 0.05 m/s using a robotic arm. The sweeping speed of the Galva’s mirror for laser source 1 was set to 18 m/s. The data at a distance of 3.6 mm were acquired during 200 µs. The sampling step was set to 640 nm. Figures 8(a) and 8(b) were obtained when the measurement light was manipulated in the direction of the weld progression (cross section of the keyhole) and its perpendicular direction (side section of the keyhole), respectively. The width of the cross section of the keyhole was 0.42 mm, and the depth was 0.78 mm. Scattering signals were detected at points with the relative positions of −0.07 and +0.12 mm, attributable to the scattered light from the evaporated metal. Figure 8(b) shows that the width of the side of the keyhole was 0.46 mm, and the depth was 0.78 mm. It had almost the same dimensions as the cross section of the keyhole. Moreover, a slope was present in the range of 0.39–1.6 mm; this could be considered a molten pool. Some measurement points in this range were missed due to the week reflected signal. As imaging analysis, we acquired the metallographic cross-section image of keyhole in Fig. 9. The welding depth of the keyhole was measured to be 0.81 mm, which was the almost same as that of measurement value with the MicroLiDAR.
5. Discussion
In this section, we discuss the accuracy and precision of the MicroLiDAR for the surface measurement by obtaining the shape of an artificial ellipse-shaped keyhole in Fig. 10. The keyhole has a 0.940 mm depth, a 0.317 mm major diameter along X direction (MDXD), and a 0.316 mm minor diameter along Y direction (MDYD) as true values, which was measured with an optical microscope.
Fig. 10. The artificial keyhole with a 0.940 mm depth, a 0.317 mm MDXD, and a 0.316 mm MDYD as true values.
We measured shape of the artificial ellipse-shaped keyhole with three times with the MicroLiDAR. The three-dimensional plotted shape of the keyhole in first acquisition was shown in Fig. 11. All measured value and true values were expressed in a Table 1.
Fig. 11. Measured shape of the artificial ellipse-shaped keyhole: (a) y direction and (b) x direction.
Table 1. Shape of an artificial ellipse-shaped keyhole and its true value.
In Table 1, the averaged value of MDXD was +0.019 mm larger than the true value of the same direction. The averaged value of MDYD was −0.001 mm larger than its true value. These differential values were related with accuracy of the MicroLiDAR for the surface measurement. The standard deviation values of all direction were less than 0.001 mm, which was the precision of the MicroLiDAR for the surface measurement.
6. Conclusions
The in-process measurement of a keyhole using a high-repetition-rate low-coherence interferometer was demonstrated. The interferometer had a repetition rate of 10 MHz and a center wavelength of 1550 nm. Its absolute spatial resolution and measurement range were evaluated to be 10 µm and 5 mm, respectively. Using the interferometer, we measured a keyhole created on a 3-mm-thick stainless-steel plate via a high-power laser with a center wavelength of 1070 nm and a peak power of 8 kW. The cross-sectional area of the keyhole was measured to be 0.42 mm × 0.78 mm (width × depth) using the interferometer, and its side-view dimension was 0.46 mm × 0.78 mm (width × depth). These findings will facilitate the development of high-performance laser-welding systems.
Funding
Adaptable and Seamless Technology Transfer Program through Target-Driven R and D (JPMJTR192F).
Acknowledgments
N. Hayashi thanks Keiko Watanabe (engineer) and Yozo Ashida (engineer) for help construction of a part of laser welding system for this work.
Disclosures
The authors declare that there's no conflicts of interest related to this article
Data availability
Underlying the results presented in this paper are not publicly available at this time but maybe obtained from the authors upon reasonable request.
References
1. A. Schoedel, Z. Ji, and O. M. Yaghi, “The role of metal–organic frameworks in a carbon-neutral energy cycle,” Nat. Energy 1(4), 16034 (2016). [CrossRef]
2. I. Jiménez, A. M. Mora-Valencia, and J. Perote, “Risk quantification and validation for Bitcoin,” Oper. Res. Lett. 48(4), 534–541 (2020). [CrossRef]
3. L. Burks, N. Ahmed, I. Loefgren, L. Barbier, J. Muesing, J. McGinley, and S. Vunnam, “Collaborative human-autonomy semantic sensing through structured POMDP planning,” Robot. Auton. Syst. 140, 103753 (2021). [CrossRef]
4. Y. Mori, T. Sekine, O. Komeda, S. Nakayama, K. Ishii, R. Hanayama, K. Fujita, S. Okihara, N. Satoh, T. Kurita, T. Kawashima, H. Kan, N. Nakamura, T. Kondo, M. Fujine, H. Azuma, T. Hioki, M. Kakeno, T. Motohiro, Y. Nishimura, A. Sunahara, Y. Sentoku, and Y. Kitagawa, “1-Hz fast-heating fusion driver HAMA pumped by a 10-J green diode-pumped solid-state laser,” Nucl. Fusion 53(7), 073011 (2013). [CrossRef]
5. B. Roberts, S. M. Wright, C. Christmas, M. Robertson, and D. S. Wu, “COVID-19 pandemic response: Development of outpatient palliative care toolkit based on narrative communication,” Am. J. Hosp. Palliat. Care 37(11), 985–987 (2020). [CrossRef]
6. Y. Masuda, Chap.6, ‘Sei no Genba no “Katari’ to Doki no Sigaku (‘Katari’ in the field of life and the poetics of motives) Hitsuji Shobo (in Japanese), 128–310 (2013), ISDN, 978-4-89476-619-8.
7. C. Ebert and C. H. C. Duarte, “Digital transformation,” IEEE Softw. 35(4), 16–21 (2018). [CrossRef]
8. S. Katayama, Fundamentals and Details of Laser Welding (Springer Singapore, 2020), Chap. 2.1.
9. S. Tani, Y. Aoyagi, and Y. Kobayashi, “Neural-network-assisted in situ processing monitoring by speckle pattern observation,” Opt. Express 28(18), 26180–26188 (2020). [CrossRef]
10. A. Matsunawa and V. Semak, “The simulation of front keyhole wall dynamics during laser welding,” J. Phys. D: Appl. Phys. 30(5), 798–809 (1997). [CrossRef]
11. V. V. Semak, J. A. Hopkins, M. H. McCay, and T. D. McCay, “Melt pool dynamics during laser welding,” J. Phys. D: Appl. Phys. 28(12), 2443–2450 (1995). [CrossRef]
12. G. Chen, M. Zhang, Z. Zhao, Y. Zhang, and S. Li, “Measurements of laser-induced plasma temperature field in deep penetration laser welding,” Opt. Laser Technol. 45, 551–557 (2013). [CrossRef]
13. Y. Feng, X. Gao, Y. Zhang, C. Peng, X. Gui, Y. Sun, and X. Xiao, “Simulation and experiment for dynamics of laser welding keyhole and molten pool at different penetration status,” Int. J. Adv. Manuf. Technol. 112(7-8), 2301–2312 (2021). [CrossRef]
14. R. Cunningham, C. Zhao, N. Parab, C. Kantzos, J. Pauza, K. Fezzaa, T. Sun, and A. D. Rollett, “Keyhole threshold and morphology in laser melting revealed by ultrahigh-speed x-ray imaging,” Science 363(6429), 849–852 (2019). [CrossRef]
15. T. R. Allen, W. Huang, J. R. Tanner, W. Tan, J. M. Fraser, and B. J. Simonds, “Energy-coupling mechanisms revealed through simultaneous keyhole depth and absorptance measurements during laser-metal processing,” Phys. Rev. Appl. 13(6), 064070 (2020). [CrossRef]
16. J. J. Blecher, C. M. Galbraith, C. Van Vlack, T. A. Palmer, J. M. Fraser, P. J. L. Webster, and T. DebRoy, “Real time monitoring of laser beam welding keyhole depth by laser interferometry,” Sci. Technol. Weld. Joining 19(7), 560–564 (2014). [CrossRef]
17. P. H. Tomlins and R. K. Wang, “Theory, developments and applications of optical coherence tomography,” J. Phys. D: Appl. Phys. 38(15), 2519–2535 (2005). [CrossRef]
18. W. Drexler, M. Liu, A. Kumar, T. Kamali, A. Unterhuber, and R. A. Leitgeb, “Optical coherence tomography today: Speed, contrast, and multimodality,” J. Biomed. Opt. 19(7), 071412 (2014). [CrossRef]
19. K. Goda and B. Jalali, “Dispersive Fourier transformation for fast continuous single-shot measurements,” Nat. Photonics 7(2), 102–112 (2013). [CrossRef]
20. K. Goda, K. K. Tsia, and B. Jalali, “Serial time-encoded amplifed imaging for real-time observation of fast dynamic phenomena,” Nature 458(7242), 1145–1149 (2009). [CrossRef]
21. S. Moon and D. Y. Kim, “Ultra-high-speed optical coherence tomography with a stretched pulse supercontinuum source,” Opt. Express 14(24), 11575–11584 (2006). [CrossRef]
22. J. Kang, P. Feng, X. Wei, E. Y. Lam, K. K. Tsia, and K. K. Y. Wong, “102-nm, 44.5-MHz inertial-free swept source by mode-locked fiber laser and time stretch technique for optical coherence tomography,” Opt. Express 26(4), 4370–4381 (2018). [CrossRef]
23. M. Hoshikawa, K. Ishii, T. Makino, T. Hashimoto, H. Furukawa, and N. Wada, “Low-coherence interferometer with 10 MHz repetition rate and compensation of nonlinear chromatic dispersion,” Opt. Rev. 27(2), 246–251 (2020). [CrossRef]
24. G. P. Agrawal, Nonlinear Fiber Optics (Academic Press, 1995), Chap. 3, pp. 84–90.
