Monitoring and repair of defects in ultrasonic additive manufacturing

Abstract

Ultrasonic additive manufacturing (UAM) involves ultrasonic welding of similar or dissimilar metal foils on top of a base substrate. UAM can produce solid consolidated structures under optimal processing conditions. However, inter-layer defects such as delamination/kissing bonds (type 1) and inter-track (type 2) defects are common. The authors previously developed an ultrasonic nondestructive evaluation (NDE) monitoring methodology to quantify layer-bonding stiffness modeled as an interfacial spring. In this study, ultrasonic NDE is used to monitor the evolution of type 1 defects in a UAM component divided into two zones. The first represents the base/build interface comprising of the first few layers on the base substrate, and the second region represents the bulk of the UAM stack. A mechanism for the formation and evolution of type 1 defects was proposed based on NDE and optical examination. Type 2 defects are often more catastrophic and are challenging to repair. In the present work, a novel solid-state repair technique using friction stir processing (FSP) was used to repair typical UAM defects. The use of FSP ensures that the microstructural advantages of UAM are retained while improving the part quality. Two modes of FSP were designed—FSP from above for repair of inter-track (type 2) defects and FSP from below the base for the repair of base/build (type 1a) defects. The results of this study pave the way towards the development of an integrated solid-state additive manufacturing system with UAM as the primary bonding mechanism and FSP as an enhancement and repair tool.

Appendices

Appendix 1. Bond quality inversion of UAM components using in situ ultrasonic NDE

A method for a model-based bond quality inversion has been presented in Nadimpalli et al. [19]. Due to the change in the experimental setup from Nadimpalli et al., we discuss the methodology here briefly. Fig. 19 shows the typical signals captured during in situ monitoring using a 10-MHz ultrasonic delay line transducer. One of the two signals selected is the good-quality component from Fig. 2, while the other one is a lower-quality component. The primary difference between the current setup and the one discussed in Nadimpalli et al. [19] is the presence of a coupling oil with an immersion mode instead of a contact mode measurement. The immersion mode measurement is more robust under the vibration of the base structure in comparison with the contact mode measurement.

Fig. 19

Ultrasonic signals of UAM components from in situ monitoring

To compare signals from various layers and components, the signal is phase-shifted and normalized so that the front surface signals are identical. The two distinct features of interest, base/build delamination, and the signal from the top of the stack are gated, as shown. The base/build interface gate is fixed in time, while the stack gate moves with changing layer number. Both the base/build and stack gates are chosen with the smallest possible overlap so that the interference effects between the two signals are minimized during bond quality evaluation. Using the signal from the base as a reference, the phase delay and attenuation are calculated utilizing a Fourier transformation. To estimate the phase velocity (${c}_{\mathsf{p}}$), the average thickness of each component was measured with digital calipers, and the layer thickness was calculated. The attenuation coefficient (α) was calculated after diffraction correction.

Under the influence of high-frequency ultrasonic vibration, the modulus of the UAM stack is better represented with a complex dynamic modulus ${\overset{\sim }{C}}_{\mathsf{stack}}$ having both real and imaginary components that represent the storage and loss moduli, respectively. After some tedious algebra, the complex stack impedance can be estimated as

$$ {\overset{\sim }{C}}_{\mathrm{stack}}=\rho {c}_{\mathrm{p}}^2\left(\ 1+2 i\psi \right) $$

(1)

where ψ is a real dimensionless quantity representing the ratio between the imaginary (c_i) and real (c_r) parts of the phase velocity. It can also be related to the experimentally measured attenuation (α) and phase velocity (c_p) as

$$ \frac{\alpha\;{c}_{\mathrm{p}}}{\omega }=\frac{c_{\mathrm{i}}}{c_{\mathrm{r}}}=\uppsi $$

(2)

Thus, the dynamic modulus of the UAM stack can be estimated from experimentally measured phase velocity and attenuation. The modulus of the UAM stack can be thought of as the combination of the set of imperfectly bonded layers one on top of the other. Hence, for a component of N layers with one base/build interface, the modulus is estimated as

$$ \frac{N}{{\overset{\sim }{C}}_{\mathrm{stack}}}=\frac{1}{\eta_1\;{C}_{\mathrm{layer}}}+\frac{N-1}{\eta\;{C}_{\mathrm{layer}}}+\frac{N}{C_{\mathrm{layer}}} $$

(3)

where η₁ is the base/build interfacial stiffness coefficient, and η is the stack stiffness coefficient. For a large enough number of layers (N > 5) the stack stiffness coefficient can be estimated as

$$ \eta \approx \frac{{\overset{\sim }{C}}_{\mathrm{stack}}}{C_{\mathrm{layer}}-{\overset{\sim }{C}}_{\mathrm{stack}}} $$

(4)

This forms the first estimate of η, from which we can calculate the dispersive stack impedance (${Z}_{\mathsf{stack}}$) and thus the base/build stiffness coefficient η₁. In the case of an aluminum 6061 base and Al6061 UAM stack, the displacement reflection coefficient would be

$$ R=\frac{\frac{Z_0-{Z}_{\mathrm{stack}}}{Z_0+{Z}_{\mathrm{stack}}}+\frac{i\;\omega\;{Z}_{\mathrm{p}}}{2\;{\kappa}_1}}{1+\frac{i\;\omega\;{Z}_{\mathrm{p}}}{2\;{\kappa}_1}} $$

(5)

Z₀ and Z_stackare the acoustic impedances of the base and the stack, respectively; ω is the angular frequency; κ₁ is the normal interfacial stiffness; and Z_p is a combination of the two acoustic impedances.

$$ {Z}_{\mathrm{p}}=\frac{2\;{Z}_0\;{Z}_{\mathrm{stack}}}{Z_0+{Z}_{\mathrm{stack}}} $$

(6)

The reflection coefficient can also be obtained as a ratio of the Fourier transform of the base/build gate signal to that of the reference. The corresponding complex interfacial stiffness estimate is

$$ {\kappa}_1=\frac{i\;\omega\;{Z}_{\mathrm{p}}}{2}\;\frac{1-R}{R-{R}_0} $$

(7)

where R₀ is the reflection coefficient from the perfectly bounded base/build interface,

$$ {R}_0=\frac{Z_0-{Z}_{\mathrm{stack}}}{Z_0+{Z}_{\mathrm{stack}}} $$

(8)

Equation 7 is the first estimate of κ₁, which can be used to recalculate the stack interfacial stiffness κ according to Eq. 3. The process is iteratively performed until it reaches convergence, which usually takes 3–4 cycles of the loop shown in Fig. 20.

Fig. 20

Schematic of the inversion model to calculate η₁, η

Appendix 2. Calibration of UAM vibration amplitude

The UAM process utilizes a 20-kHz transducer to vibrate the sonotrode during bonding. The vibration amplitudes range from 20––50 μm and can be measured using a laser vibrometer. It is essential to calibrate the UAM system before welding to ensure that the vibration amplitude setting [0–100 %] corresponds to the intended vibration amplitude. A laser vibrometer was used to measure the velocity of the sonotrode, as shown in Fig. 21.

Fig. 21

a Laser interferometer focused on the UAM sonotrode. b Laser vibrometer setup outside a Fabrisonic R200 UAM system for non-contact measurement

The harmonic signal obtained during vibration in free air at a 70% amplitude setting is shown in Fig. 22. We can use this signal to calculate the frequency and amplitude of vibration based on the sensitivity of the vibrometer. The frequency of vibration is close to 20 kHz and remained constant throughout testing. The vibration amplitude was changed from a 20–100% setting, and the physical amplitude was calculated from the velocity profiles. The R200 system and the R7200 system were calibrated in 2012 at a sensitivity of 5000 mm/s/V. In 2015, we calibrated the same R200 system. The results of the calibration are shown in Fig. 23. The standard deviation in the vibration amplitude was low (< 1 %), indicating an excellent control by the DUKANE iQ welder. There is a small change in the vibration characteristics at higher amplitudes of the R200 system between 2012 and 2015. This can be attributed to the retexturing operation that was performed periodically to improve sonotrode roughness. There are differences between the R200 and R7200 systems based on the amplitude setting parameter. Hence, all the data included in our paper has been calculated from the calibration curves so that the vibration amplitudes can be compared.

Fig. 22

Change in the velocity of sonotrode vibrating at 20-kHz frequency at a 70% amplitude setting

Fig. 23

UAM calibration of the physical vibration amplitude of UAM systems