2020 Volume 61 Issue 10 Pages 1994-2001
Miniaturization of bearing rollers used in autos and robots will require a manufacturing system that combines a deposition method that can fabricate thin jigs without defects and a non-destructive inspection method that can detect cracks on such jigs. Here, we are developing a system that uses directed energy deposition (DED), which is a 3D-printing (additive manufacturing, AM) process, to fabricate thin jigs, and then uses laser ultrasonics (LU) to inspect the jigs. Here, deposited layers having a 0.4 × 0.6 mm2 cross-section were fabricated using DED, and then non-destructively inspected using LU. However, using LU on such a small area has three problems: the effect of overlapping of the excitation and detection laser beams, difficulty in separating the multiple types of waves due to the simultaneous generation, and complexity of the acoustic field. Therefore, first, the acoustic field was examined using the finite element method (FEM), and then LU was used to inspect a small area of the deposited layer using complex discrete wavelet transform. Results show successfully detection of spontaneously occurring cracks, thus confirming the effectiveness of LU for non-destructive inspection of a thin jig.
Bearing rollers indispensable for automobiles and robots are smaller in diameter now, and in order to make them precision jigs required for polishing such rollers must be thinner. However, the conventional brazing method to deposit a sintered cemented carbide to a jig surface to prevent abrasion has a high failure rate due to cracking or peeling of this surface caused by thermal strain. Therefore, a 3D printing method (additive manufacturing, AM)1,2) called direct energy deposition (DED)3) is being developed to deposit cemented carbide directly on a thin jig substrate without defects. Advantages of the DED method include shorter delivery time, increased precision of the jig, and reduced jig maintenance cost. An advantage of DED over the commonly used powder bed fusion (PBF) method for metal AM is its suitability for molding large parts because the required amount of material (metal powder) is small and the molding speed is high.4)
The size of the jig which we are developing is about 300 × 20 × 10 mm3 and the size of cemented carbide is about 300 × 1 × 1 mm3. We are currently working on the development of DED technique and non-destructive inspection technique. Although X-ray CT is a non-destructive inspection method, its resolution diminishes as the inspection target becomes large (dm-scale).5) This method has difficulty in inspecting a jig of several tens of centimeter or more, and detecting a crack that has a small gap. The nondestructive inspection technique for AM using acoustic wave has been developed.6–15) Laser ultrasonics16) (LU) is a technique to irradiate a laser beam onto a sample and generate the acoustic wave by thermoelastic effect or ablation, which could be developed into the in-situ inspection of AM in the future.9–15) By using surface acoustic wave (SAW) with LU, Klein and Sears inspected a blind hole in a Ti 6–4 plate.9) By using 82 MHz SAW with LU, Clark et al. obtained 2D images of polished AM products that had voids.10) Cerniglia et al. proposed LU for in-line inspection of products fabricated using laser powder deposition, and successfully demonstrated this technique on deposition samples with induced flaws.11) Using bulk waves with LU, Lévesque et al. inspected AM samples that were manufactured using laser powder or electron beam wire deposition.12) Smith et al. generated SAW with LU using a projection mask, and then imaged the grains and pores from a sound velocity map.13,14) Davis et al. generated bulk waves with LU and detected millimeter-order standard defects in an AM sample.15) Based on all these studies, LU is the most promising method as an in-process evaluation method of metal AM to detect internal defects. The authors have developed a nondestructive inspection technique using laser ultrasound for the in-situ inspection of AM components, and showed that the delamination and pores of AM samples prepared by PBF method affect the propagation velocity and frequency characteristic of acoustic wave.17)
If a thick film of cemented carbide is deposited on a large jig, large defects will be occurred due to stress and other factors. The authors are developing on the optimization of cemented carbide powder and processes of DED by using small samples. In order to evaluate these techniques, it is also required to inspect a small crack in such a small sample. This study was conducted to confirm the detectability of visually undetectable micro-defects occurring in the small sample. This article is a more detailed study of what the authors reported at USE2019.18) Using the DED we are developing, the width of deposited layers is comparable with the wavelength of acoustic waves generated and detected by our equipment used in the experiments in this study (see Sec. 3). We therefore expected that the acoustic wave generated by LU will propagate as a guided wave.19–21) In this study, first, the finite element method (FEM)22) was used to determine what kind of acoustic waves are propagated by this LU method. Next, a sample was fabricated using DED to deposit layers on an SUS304 substrate. To deposit these layers without defects, Ni-based self-fluxing alloy powder was used here for DED. In our future plan, we plan to make jigs with WC in consideration of its durability, but this article we used Ni-based self-fluxing alloy because it is often used in thermal spraying and it is less prone to defects than WC. Finally, non-destructive inspection of the sample was performed by using LU.
Here, 2D-FEM (plane strain condition22)) and 3D-FEM were used to determine the type of acoustic wave propagating to the sample in the experiments. Figure 1 shows the 3D-FEmodel (finite element model), and Table 1 shows elastic parameters used in this simulation. x- and z-dimensions of 2D-FEmodel were same as 3D-FEmodel, and y-dimension of 2D-FEmodel was considered as infinity. Note that we used the elastic parameters of Inconel 600, which is a Ni alloy, because the elastic parameters of Ni-based self-fluxing alloy powder were unknown. The force
| \begin{equation} F(t) = -{\exp}\{-(t - 3\sigma )^{2}/\sigma^{2}\} \end{equation} | (1) |
| \begin{equation*} \sigma = 12.011224\,\text{ns} \end{equation*} |
3D Finite element model (3D-FEmodel). The 3D-FEmodel was divided into cubic elements whose dimensions were 0.01 × 0.01 × 0.01 mm3. x- and z-dimensions of the 2D-FEmodel were same as those of the 3D-FEmodel, and it was divided into square elements whose dimensions were 0.01 × 0.01 mm2.
Intensity modulation images of 3D-FEM. Displacements shown are magnified by a factor of 1000 times. Color indicates z-displacement (uz).
Position dependency of z-displacement of 3D-FEM.
Wavelet transforms of z-displacement. (a) uz at (0.8, 0) of 2D-FEmodel, (b) uz at (1.2, 0) of 2D-FEmodel, (c) uz at (0.8, 0, 0) of 3D-FEmodel, (b) uz at (1.2, 0, 0) of 3D-FEmodel. In 2D-FEM, (a) and (b) show propagation of a surface wave, but no velocity dispersion is observed. In 3D-FEM, (c) and (d) show velocity dispersion. Although it is like a guided wave, it is difficult to evaluate the propagation velocity for each frequency.
AM sample (specimen) to evaluate the non-destructive, non-contact inspection method using LU were fabricated as follows. A Ni-based self-fluxing alloy powder (Höganäs 1660-22) was deposited on a 2-mm-thick substrate (SUS304) via DED ALPION (MURATANI MACHINE MANUFACTURE Co., Ltd.). In DED, the wavelength of the direct diode laser was 975 nm, its power was 120 W in a continuous wave, its feed speed was 20 mm/s, feed rate of the powder was about 20 mg/s, and the spot diameter of the laser beam was 0.3 mm. Each deposited layer was 0.1-mm thick, the layer was deposited 6 times, and the cross-section of the deposited layers was 0.4 × 0.6 mm2.
Figure 5(a) shows an outline of the experimental setup, and Fig. 5(b) shows a schematic of the sample to be inspected. To generate an acoustic wave, a solid state laser was used (Laser-export Co. ltd., TECH-527 basic) whose wavelength was 523 nm and pulse width was 5 ns. The laser beam was expanded by a factor of 10 via a laser beam expander, and then focused to a spot diameter of about 0.1 mm by a convex lens whose focal length was 800 mm. The frequency of excitation was 4 kHz, and the energy was 8 µJ. This laser beam then scanned the sample surface at a scanning distance of about 1.5 mm by using a Galvanometer scanner (Cambridge Technology, a Novanta company, M3 scanner 20mmXY Head YAG). To detect the acoustic waves, a laser Doppler interferometer was used (Polytec GmbH OFV-3001 with OFV-303 and OVD-030). The laser beam irradiated from the interferometer head (OFV-303) was a continuous wave with a wavelength of 633 nm, and the laser output power was less than 1 mW, and the beam diameter was about 100 µm. This interferometer was designed to detect displacements on a flat surface, and the upper limit of frequency of displacement was 20 MHz. The laser beam of the interferometer was fixed at one point on the sample (Fig. 5(b)). A total of 51 positions of laser beam irradiation on the specimen were inspected, and the output of 65536 readings of the interferometer was averaged using a digital oscilloscope (Keysight MSOX3014A) and then output to a PC. The total time to obtain 51 experimental data was about 34 minutes. For clarity in this article, the laser beam that generates acoustic waves is simply called “laser beam”, and the laser beam that detects acoustic waves is simply called “probe”.
Experimental setup. (a) Overview. A pulse laser was used to generate acoustic waves, and a laser Doppler interferometer was used to detect the acoustic waves. Laser beam of the laser Doppler interferometer (probe) was fixed, and only the laser beam of the pulse laser (laser beam) was scanned using the Galvanometer scanner. (b) Schematic of the AM sample (specimen). Laser beam was applied to 51 points at intervals of about 0.03 mm.
Figure 6 shows displacements at 7 positions at which the laser beam irradiated the specimen surface. The origin (0 µs) of time (t) was determined the same as the peak time of the laser pulse. The origin (0 mm) of the x-position (laser beam irradiation position) was determined by the closest position to the probe. The displacements at x = 0.002 mm and t ∼ 0.23 µs seem to be affected by the overlap between the probe and the laser beam. Complex discrete wavelet transform using m = 4, 3 RI-Spline26–29) of the displacements (Fig. 7), revealed an effect of overlap under 10 MHz in the frequency domain (Fig. 7(a)). The peaks at x = −0.422 mm and t ∼ 1.5 µs and at x = −0.635 mm and t ∼ 2.2 µs were an acoustic wave. Its propagation velocity was about 300 m/s, indicating that this was an acoustic wave propagated in the air. The frequency range of this acoustic wave propagating in the air was 0.32–2.6 MHz, based on complex discrete wavelet transform using m = 4, 3 RI-Spline of the displacements at x = −0.422 mm (Fig. 7(b)). Based on the same analysis, the frequency range of the acoustic waves propagated in the sample was 5.1–20 MHz (Fig. 7).
Experimental results. Position of probe was set at x = 0. Displacement near the probe (x = 0.002 mm) shows the effect of the overlap of the probe and the laser beam. Further away from the probe (x = −0.422 mm and x = −0.635 mm), a low-frequency, slow wave was confirmed. Since its propagation time difference is 0.665 µs, the propagation velocity is estimated 320 m/s. A propagation velocity for this wave indicates that it was an acoustic wave propagating in air. When the laser beam was close to the probe, it will not be possible to distinguish between the acoustic wave propagating in the sample and the acoustic wave propagating in air. In addition, because the energy of the laser beam is small and the amount of the displacements is small, the acoustic wave propagating in the specimen is difficult to clarify.
Complex discrete wavelet transform using m = 4, 3 RI-Spline15–17) of the displacement. Horizontal axis indicates time, vertical axis does frequency, and color shows the norm of the synthetic wavelet $|d_{k}^{j}|$ which is defined as eq. (4) of Ref. 27). Blue shows small value, and red does large value. (a) Example of uz including the effect of overlap of the probe and laser beam. Overlapping effect existed below 10 MHz. (b) Example of uz including acoustic wave propagating in the air. Acoustic wave propagating in the air existed at t ≅ 2.2 µs, and within 0.3–5 MHz.
To evaluate only the acoustic waves propagated in the sample, Fig. 8 shows displacements whose frequency range was 10 to 20 MHz obtained by complex discrete wavelet transform using m = 4, 3 RI-Spline. This eliminated the acoustic wave that propagated in the air and eliminated the effect of overlap between the probe and laser beam. It appears that acoustic waves propagated in both the +x and −x directions. Figure 9 shows intensity modulation images of uz whose frequency was 10–20 MHz, and the horizontal axis indicates the x-position of the laser beam. The acoustic waves seem to propagate in the ±x-directions from x = 0 due to reciprocity30,31) (Fig. 9(a)–(b)), and reflected around x = −0.4 mm and x = 0.6 mm (Fig. 9(c)–(e)). However, the presence of the defect is not clearly seen because of the interference between the generated and reflected waves around x = −0.4 mm and x = 0.6 mm. If these figures are animated, the propagation and reflection of acoustic waves can be seen more easily, but an AI (artificial intelligence) which automatically detects the presence of defects from animation requires huge computational resources. In order to quantitatively evaluate the location of the defects, Fig. 10 was drawn that included the transient response. Figure 10 shows uz whose frequency was 10–20 MHz, and the horizontal axis indicates the x-position of the laser beam and the vertical axis indicates time. These results confirm that acoustic waves were reflected at x = −0.44 mm and x = 0.59 mm, which is a distance of 1.03 mm apart.
z-displacement uz of 10–20 MHz. To evaluate only the acoustic waves propagated in the sample, displacement in the 10 to 20 MHz frequency range was plotted based on complex discrete wavelet transform using m = 4, 3m RI-Spline.15–17) Results clarify that the acoustic wave was generated by LU and propagated through the sample. However, the acoustic field was complicated, thus making it difficult to understand what was happening inside the sample.
Intensity modulation images of experimental results. z-displacement is indicated by color, with the horizontal axis representing the position (51 locations) where the laser beam irradiated. (a)–(b) show the transient phenomenon. Due to reciprocity, acoustic waves propagated in ±x-directions from x = 0.
Image of z-displacement. Displacement is indicated by color, with the horizontal axis representing the position (51 locations) where the laser beam irradiated and the vertical axis representing time. Due to reciprocity, acoustic waves propagated from the probe and were reflected at x = −0.44 mm and at x = 0.59 mm.
Figure 11(a) shows an optical image of the specimen after the experiment. The surface of the deposited layers was very rough, thus making it difficult to detect cracks. However, two cracks were detected. Figure 11(b) shows the profile near a crack, showing a cross-sectional area of the deposited layers of 0.23 mm2 and Fig. 11(a) shows a distance of about 1.06 mm between cracks. This distance is similar to the estimated distance (1.03 mm), concluding that the cracks whose cross-sectional areas were about 0.23 mm2 in the AM sample were detected by LU. An optical image taken after the experiment (Fig. 11(a)) shows that there were no ablation marks, and thus the thermo-elastic mode was used to generate acoustic waves.
Optical images of sample after the experiment. (a) Sample external view. Presence of two cracks was confirmed and their spacing was 1.064 mm. (b) Profile near a left crack. Cross-sectional area of the deposited layers was about 0.23 mm2. These profile image and area were obtained by depth from defocus by the digital microscope (the accurate D.F.D method, Keyence corporation).
When the probe and laser beam overlap or are close to each other, an abnormality occurs in the displacement measured by the laser Doppler interferometer. There is no problem if the probe and laser beam are completely separated, but it is a problem when LU is used for evaluation of a small area such as less than several millimeters. Therefore, using complex discrete wavelet transform, we investigated the frequency band affected by the overlap between the probe and laser beam, and showed that, by limiting the frequency band, it is possible to evaluate displacement even in the region where the probe and the laser beam overlap (Figs. 7(a) and 8). The RI-Spline wavelet is effective for nonstationary signal analysis, and complex discrete wavelet transform by RI-Spline is better denoising method than conventional wavelet shrinkage using discrete wavelet transform.26,28) In addition, the effect of acoustic waves propagating in the air was eliminated by the same method. This result shows the effectiveness of complex discrete wavelet transform for LU, and that LU can be applied to the evaluation of a small area.
In this sample, cracks were present at 1.03-mm intervals, so the propagation distance of the acoustic wave was limited to that interval. Based on the FEM results, this experiment is considered to be a region in which the waves cannot propagate as a guided wave. Also, dispersion could not be obtained from the experimental results. In FEM, the acoustic wave was excited at the center of the symmetric model, and the result was only the symmetric mode. However, it is thought that the acoustic field was more complicated in the experiment than shown in FEM. In 3D-FEM, the beam diameter of the laser beam was simulated as a point source by using F(t), but in the experiment it was about 0.1 mm in diameter, and thus affected by the near field.32) This difference requires further analysis.
Since the distance between defects in the specimen and the distance at which the acoustic reflected waves were generated were almost equal, it can be said that defects with a maximum cross-sectional area of 0.23 mm2 were successfully detected. These defects were not defects created for evaluation, but were naturally occurring defects, and thus are considered to be valuable as examples of non-destructive inspection. Assuming that the propagation velocity of a surface wave in this sample was about 3000 m/s, the wavelength of this wave at 10 MHz was 0.3 mm. Since this wave appeared to be totally reflected (Fig. 10), the depth of each defect is considered to be 0.3 mm or more. The depth and shape of each defect can be verified by observing the cross-section of the defect. However, since the opening of such defects is narrow, it is difficult to observe the cross-section of each defect by shaving the sample, and verification is an issue for the future.
The goal is to inspect the finished product whose cemented carbide has been ground but this experiment was performed on the unpolished specimen (Fig. 11(a)). This is also in anticipation of future in-situ inspection, and Fig. 10 shows LU has possibility of in-situ inspection. The surface of the specimen was found to be roughness due to the self-fluxing alloy powder. The particle size is 20–53 µm (Table 2), which is smaller than the ultrasonic wavelength (about 300 µm). Therefore, the effect of the roughness on the propagation of the acoustic waves was also small, and the cracks could be clearly detected. In addition, as shown in Fig. 2, the acoustic waves spread over the entire width of the deposited layers and propagate in the x-direction. This makes it possible to inspect the entire deposited layers in 1D scanning. This is a feature of the guided-wave inspection technique. In this article, the authors have carried out experiments with a small specimen. However, since the acoustic wave propagates as a guided wave, it is thought that it is possible to inspect a larger specimen. Furthermore, in order to apply LU to the jig the motorized stage whose travel is several hundred mm is preparing.
McNutt classed cracking in laser metal deposited (same as DED) nickel superalloy CM247LC into solidification cracking, grain boundary liquation, ductility dip and strain age cracking.33) It is thought that the cracks in the specimen may be solidification cracking, but the verification is the future work.
Deposited layers (specimen) with a cross-sectional area of 0.23 mm2 were created by the directed energy deposition (DED), which is 3D printing (additive manufacturing, AM). The specimen was inspected with a non-destructive, non-contact method using laser ultrasonics (LU). Results showed that there was a portion where the acoustic wave was reflected, and the presence of a crack was confirmed at that location. Since these cracks are naturally occurring, the internal structure under the surface is unknown, but it can be said that cracks with a maximum cross-sectional area of 0.23 mm2 were successfully detected. In conclusion, the LU method is effective for non-destructive inspection of DED-deposited layers.
This work was supported by the Strategic Core Technology Advancement Program (Supporting Industry Program).
