Accepted Manuscript

Dynamics of Ultrasonic Additive Manufacturing

Adam Hehr, Marcelo J. Dapino

PII: S0041-624X(16)30142-1
DOI: http://dx.doi.org/10.1016/j.ultras.2016.08.009
Reference: ULTRAS 5349

To appear in: Ultrasonics

Please cite this article as: A. Hehr, M.J. Dapino, Dynamics of Ultrasonic Additive Manufacturing, Ultrasonics
(2016), doi: http://dx.doi.org/10.1016/j.ultras.2016.08.009

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Dynamics of Ultrasonic Additive Manufacturing

Adam Hehr, Marcelo J. Dapino∗

Center for Ultrasonic Additive Manufacturing
NSF I/UCRC on Smart Vehicle Concepts
Department of Mechanical and Aerospace Engineering
The Ohio State University
201 W 19th Ave. Columbus, Ohio 43210

Abstract

Ultrasonic additive manufacturing (UAM) is a solid-state additive manufacturing technology for

joining similar and dissimilar metal foils together near room temperature by scrubbing them
together with ultrasonic vibrations under pressure. Structural dynamics of the welding assembly
and work piece influence how energy is transferred during the process and ultimately, part quality.
To understand the impact of structural dynamics during UAM, a linear time-invariant model
is used to relate the inputs of shear force and electric current to resultant welder velocity and
voltage. Measured frequency response and operating performance of the welder under no load is
used to identify model parameters. Using this model and in-situ measurements, shear force and
welder efficiency are estimated to be near 2000 N and 80% when welding Al 6061-H18 weld
foil, respectively. Shear force and welder efficiency has never been estimated before in UAM.
The influence of processing conditions, i.e., welder amplitude, normal force, and weld speed, on
shear force and welder efficiency are also investigated. Welder velocity was found to strongly
influence the shear force magnitude and efficiency while normal force and weld speed showed
little to no influence. The proposed model is also used to describe high frequency harmonic
content in the velocity response of the welder during welding operations and coupling of the
UAM build with the welder.
Key words:
A. ultrasonic additive manufacturing (UAM), B. ultrasonic welding, C. LTI modeling, D.
frequency response function (FRF) estimation, E. piezoelectrics, F. in-situ measurements

∗ Corresponding author, E-mail: dapino.1@osu.edu, Phone: +(614) 688-3689

Preprint submitted to Ultrasonics June 22, 2016
1. Introduction

Ultrasonic additive manufacturing (UAM) or ultrasonic consolidation is a continuous solid-

state additive manufacturing process where thin foils of similar or dissimilar metals are ultrason-
ically welded together in a layer by layer process to form gapless, 3D metal parts [1, 2]. The
process begins by bringing a tool piece called the sonotrode or horn into contact with a metallic
foil under a controlled pressure or normal load. Then, the foil is joined to the pre-deposited
metal beneath via high power ultrasonic vibrations. The circular design of the sonotrode allows
it to rotate at a prescribed linear speed simultaneous to welding foil. Periodic CNC machining is
utilized intermittent of welding to create complex internal features, embed objects into the struc-
ture, and for net shaping welded components. The process has been improved upon by actuating
the sonotrode with two high power piezoelectric transducers [3]. This dual actuation increases
the delivered weld power from 1 kW in first generation systems to 9 kW. The increase in weld
power greatly improves bond quality and enables the joining of stronger and stiffer materials.
The sonotrode and high power piezoelectric transducer(s) are designed to resonate longitudi-
nally near 20 kHz. A schematic of the UAM process is shown in Figure 1 and a commercial 9
kW UAM system is shown in Figure 2.

Figure 1: Schematic representation of a 9 kW ultrasonic additive manufacturing welder (left) and the subtractive CNC
stage found in UAM systems (right).

The key mechanisms for ultrasonic metal welding include oxide fracture under pressure and
plastic deformation of surface asperities through shear [4, 5, 6]. Oxide dispersal allows nascent
metal surfaces to come into contact and form metallic bonds while surface asperity deformation
promotes dynamic recrystallization of the interface microstructure [7, 8, 9, 10]. The result is a
2
narrow weld region on the order of 10-20 µm in size and bulk temperatures far below metallic
melting temperatures [11]. Consequently, dissimilar metals can be welded together with minimal
intermetallic formation [12, 13, 14, 15, 16], metallurgical changes are minimized and highly lo-
calized to the interface region, and temperature sensitive metals or components can be combined
or built into metal structures [17, 18, 19, 20, 21].

Ultrasonic
Transducers

Horn/Sonotrode

Figure 2: Commercial 9 kW UAM system, Fabrisonic SonicLayer 4000. Reprinted with permission [22]

The aim of this paper is to describe UAM system dynamics and power conversion within the
welder using a linear time-invariant (LTI) model which explicitly specifies welder shear force
and electric current as system inputs. The outputs of the model are welder velocity and elec-
tric voltage. The model describes the conversion and transfer of electrical to mechanical power
within the welder. The model does not directly describe energy transferred to the weld, although
it can be used as part of a broader modeling framework to quantify the complete flow of energy
through the welder into the workpiece. This matter is discussed in Section 6. Conventional LTI
models for ultrasonic systems lump the influence of shear force or load into the motional feed-
back of the entire system for control purposes [23], i.e., Van Dyke system representation [24].
Because the focus of the paper is to describe the system dynamics of the welder, i.e., sonotrode
and transducers, it is required to explicitly express shear force as a system input. This alternative
LTI model can be used for improved control strategies and energy transfer analysis for the UAM
process.
The paper begins by describing how the process is controlled and how this control strat-
egy relates to UAM welder performance. Then, the welder is described in terms of impedance
relations in order to formulate the proposed system level LTI model. Experimental frequency
3
response functions (FRFs) and operating performance of the welding assembly are then mea-
sured to verify that the system follows LTI behavior and to identify model parameters. Using
the model combined with in-situ measurements, shear force and welder efficiency are estimated
for standard Al 6061 welding procedures. The influence of process variables, i.e., amplitude,
down force, and speed, on shear force and efficiency is then studied. Lastly, the high frequency
velocity response of the welder and how the welder couples with the UAM part is investigated
with the use of the LTI model.

2. UAM Control Background

UAM has been utilized for near a decade, yet the process uses ultrasonic metal welding con-
trol strategies developed long ago. Ultrasonic metal welding control strategies are designed for
single metal joints. In UAM, many joints are made and the build geometry changes throughout
component construction. Consequently, unwanted resonances can occur [25, 26] and the amount
of deformation at the weld interface decreases as the build becomes more compliant with added
layers [27, 28]. Due to less deformation occurring at the weld interface and because deforma-
tion is a leading mechanism for bonding, the bond quality degrades with additional layers. A
control strategy unique to UAM is needed to avoid or minimize undesired structural dynamics
and to maintain weld quality throughout component construction. Such a control strategy can be
developed with a reliable system level model of the UAM process.
Figure 3 demonstrates how the control dynamics of the UAM process change when welding
vs. actuating the sonotrode without load, i.e., no welding. In particular, the peak velocity of
the scrubbing motion decreases 10% (Figure 3(a)), the frequency of the welder increases 75 Hz
(Figure 3(b)), and the electric power draw of the piezoelectric transducers increases an order of
magnitude, see Figure 3(c). A customized ultrasonic generator for the UAM process is responsi-
ble for the control dynamics observed in Figure 3. The generator uses two closed loop controllers
which work simultaneously. The first controller uses a phase lock loop (PLL) algorithm to track
system resonance during welding by minimizing the phase angle between the applied voltage
and current [29, 30]. System resonance can change when welding due to added mass, stiffness,
and heat generation from the load [30, 31, 32]. This PLL algorithm is the reason for the upward
frequency shift in Figure 3(b), and this shift occurs due to the UAM build stiffening the system
during welding.
4
 The second controller works to maintain welder amplitude as the part is built, which under
open loop conditions would result in a decrease in amplitude. Welder amplitude in UAM is
maintained by controlling voltage to be constant. Voltage is controlled by varying the current to
maintain a set-point value for a given amplitude setting [33]. Further detail on voltage control
will be discussed in Section 4. Ultrasonic systems can also be controlled using electric current
in a similar manner [23, 30].
In order to accurately track the resonance of the system for both the PLL algorithm and ampli-
tude control, the mechanical motion of the welder needs to be measured. The most common way
of measuring mechanical motion is with the use of motional feedback methods. Motional feed-
back works by adding a reactive element in series or in parallel with the transducer to balance-out
its electrical impedance or admittance [23, 30, 31, 32]. For piezoelectric systems, this reactive
element is an inductor. By balancing out the electrical impedance of the transducer, the motional
impedance or motion of the transducer can be indirectly measured with applied current and volt-
age to the transducer. There are many different circuits utilized to implement motional feedback
control techniques [23, 32]. The ultrasonic generator used on UAM systems utilizes such a mo-
tional feedback method for resonance tracking and amplitude control. Because amplitude is not
measured directly in UAM, a decrease in vibration can occur if the reactive inductance element
does not sufficiently isolate the motional impedance of the transducer or if significant compli-
ance exists in the sonotrode, see Figure 3(a). This decrease in welder velocity and its relation
to motional feedback and sonotrode compliance will be discussed in more detail later. Lastly,
because voltage is controlled to be constant by increasing electric current during welding, the
average electric power draw increases substantially to maintain welder motion, see Figure 3(c).

5
 ;ĂͿ ;ďͿ ;ĐͿ

Average Electric Power (W)

Welder Frequency (kHz)
2 1000
19.85
Peak Velocity (m/s)

1.95
800
1.9 19.8
No Welding 600
1.85 Welding 19.75
1.8 400

1.75 19.7 200

1.7 0
0 0.5 1 0 1 0 0.5 1
0.5
Time (s) Time (s) Time (s)

Figure 3: UAM system dynamics and control when not welding, i.e., exciting welder in air without load, and welding
Al 6061 foil: (a) peak sonotrode or welder velocity; (b) excitation frequency; (c) average electric power draw from one
of the ultrasonic transducers. Data was collected using the welding process variables of 32.5 µm peak-peak prescribed
welder vibration, a down force of 5000 N, and a rolling speed of 5 m/min (200 in/min). Data was sampled at 50 kHz and
processed with a block size of 8192 points, welder vibration was measured with a non-contact laser vibrometer, welder
frequency was obtained using a short time windowed FFT, and power draw was sampled directly from the ultrasonic
generator.

3. Time-Invariant Linear Model of UAM Process

To describe the system dynamics of the welder for all frequencies (ω), the following LTI
system model with voltage-force as across variables and velocity-current as flow variables for
the system can be written,

    
 i( jω)   H ( jω) Hme ( jω)  V( jω) 
  e
 =   , (1)
  

δ̇( jω) H ( jω) em H ( jω)  F ( jω)
m s

where He is the FRF between applied voltage (V) and electric current (i), Hme is the FRF between
opposing shear force during welding (F s ) and current, Hem is the FRF between applied voltage
and velocity of the sonotrode (δ̇), and Hm is the FRF between shear force and velocity. Due to the
system being piezoelectric, the system follows the law of reciprocity [24, 34]. As a result, Hme
and Hem are equal in magnitude and phase. This LTI model with force as the across variable
and velocity as the flow variable is proposed initially due to piezoelectric systems conventionally
using this form [23, 24, 34], and because the system is feasible to characterize in this form. An
equivalent LTI model of the system using velocity as the across variable and force as the flow
variable will be presented at the end of this section because the motional feedback controller
utilizes this relationship. The equivalent form is derived using the principle of duality [34]. Both
6
LTI models can be used to describe the system behavior because the system behaves in a steady
manner during welding operations, as evidenced by the stable welding traces in Figure 3. The
system exhibits a small amount of variation when welding because the PLL algorithm moves
the voltage excitation frequency. This small frequency shift in the PLL controller may originate
from variations in the UAM build compliance and shear force character during welding. The
influence of UAM build stiffness on frequency and FRF magnitude will be discussed in more
detail in Section 8
To understand the relation between system physics and the FRF terms, equivalent circuit
analysis can be used to derive closed form FRF expressions near resonance. To model the
electro-mechanical coupling of the piezoelectric transducers, an ideal transformer can be uti-
lized [24, 35]. A schematic of the welding assembly along with its corresponding equivalent
circuits are shown in Figure 4. The equivalent circuit analysis is conducted assuming that the
welding assembly is geometrically symmetric in shape and properties and that the two trans-
ducers operate out of phase, i.e., push-pull configuration. The system can be assumed to be
geometrically symmetric because the sonotrode is precision machined, and the two transducers
used to drive the sonotrode exhibit near identical response [36]. Because the system is geomet-
rically symmetric and the transducers are operated out of phase, the transducers can be lumped
together as a single transducer driving the sonotrode. The out of phase actuation of the welder is
shown in Figure 4(a) with opposing voltage directions. The sonotrode in Figure 4(b) is modeled
as a black box 2-port electrical network because (i) the transducers actuate the sonotrode at a
different location than the shear force, and (ii) the design influence of the sonotrode on system
performance is not considered in this paper. Because sonotrode design is not an aspect of this
paper, Figure 4(b) can be simplified by lumping the transducer and sonotrode dynamics together,
see Figure 4(c). These two devices can be lumped together because both the transducers and
sonotrode are designed to resonate at the driving frequency of 20 kHz. The equivalent circuit in
Figure 4(c) is the Mason circuit representation for piezoelectric devices [24].

7
 ;ĂͿ ;ďͿ >ƵŵƉĞĚĐƚƵĂƚŽƌ ^ŽŶŽƚƌŽĚĞ
^ŽŶŽƚƌŽĚĞ ϭ͗Ɍ ϮD
 
ĐƚƵĂƚŽƌϭ ĐƚƵĂƚŽƌϮ ͬϮ Ϯ
‫ݔ‬ሶ ߜሶ ‫ݔ‬ሶ s ݅ Ɍ ‫ݑ‬ሶ ‫ݑ‬ሶ ‫ݔ‬ሶ   ߜሶ &

WŝĞnjŽ Ɍ ܸ ϭͬϮ<
000000000000000 0000000000000000
000000000000000 0000000000000000
000000000000000 0000000000000000
000000000000000 0000000000000000 ;ĐͿ
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
s >ƵŵƉĞĚ^LJƐƚĞŵ
000000000000000 0000000000000000 s
000000000000000 0000000000000000
ϭ͗Ɍ D  ϭͬ<
‫ݑ‬ሶ ‫ݑ‬ሶ
ŽŵƉƌĞƐƐŝŽŶ 
ZŽĚ
&
s݅ Ɍ ߜሶ Ɍ ܸ ߜሶ &

Figure 4: UAM weld assembly analysis: (a) schematic of assembly with components, voltages, forces, and degrees
of freedom detailed; (b) equivalent circuit of weld assembly with transducers and sonotrode modeled separately; (c)
simplified equivalent circuit of weld assembly by lumping the transducers and sonotrode dynamics together.

System FRFs can be derived from the equivalent circuit shown in Figure 4(c) by shorting out
one input at a time and relating system outputs [24],

i Φ2t
He = = jωCt + Kt
, (2)
V jωMt + Dt + jω

δ̇ Φt
Hem = = Kt
, (3)
V jωMt + Dt + jω

δ̇ 1
Hm = = Kt
. (4)
Fs jωMt + Dt + jω

where Ct is the capacitance of the transducers, Φt is the electro-mechanical transformer coef-

ficient or turns ratio of the system, Mt is the effective mass of the system, Kt is the effective
stiffness of the system, and Dt is the effective damping of the system.
Using the principle of duality, the equivalent circuit in Figure 4(c) can be redrawn using
velocity as the across variable and shear force as the flow variable, see Figure 5. This system form
is called the mechanical mobility equivalent circuit of the system [34], and the welder utilizes it
for control purposes. As mentioned earlier, motional feedback or purely electric signal feedback
is utilized by the ultrasonic generator to maintain welder amplitude or velocity during welding.
Motional feedback works by inserting an external inductor in series with the capacitive element
8
of the transducer, see Figure 5(a). If the impedance of the inductor is chosen to be equal to the
capacitor impedance at resonance, the electrical impedance of the transducer is will be balanced-
out or removed from the circuit, see Figure 5(b). By removing the capacitance of the transducer
near resonance, the motional impedance of the transducer can be isolated for control purposes,
hence the name motional feedback. Welder velocity is directly related to applied voltage too and
can be used for amplitude control purposes.

;ĂͿ Ɏ͗ϭ ;ďͿ Ɏ͗ϭ

>  ϭͬD < ϭͬ ϭͬD < ϭͬ
s ݅ Ɏߜሶ Ɏ݅ & ߜሶ s ݅ Ɏߜሶ Ɏ݅ & ߜሶ

Figure 5: Equivalent circuit for welder operation: (a) mobility form of system with added inductance to ’zero out’
transducer capacitance; (b) simplified mobility form of system without electrical influences from transducer, i.e. motional
feedback form.

An equivalent LTI model using current and shear force as inputs and velocity and voltage as
outputs can then be written,

    
V( jω)  H ∗ ( jω) Hme
∗
( jω)  i( jω) 
  e
 =   , (5)
  

δ̇( jω)  H ∗ ( jω)
em H ∗ ( jω)   F ( jω)
m s

where He∗ is the FRF between electric current and voltage (V), Hme
∗
is the FRF between shear
∗
force and voltage, Hem is the FRF between electric current and welder velocity, and Hm∗ is the
FRF between shear force and velocity. Hme
∗
and Hem
∗
are equal in magnitude and phase like the
other LTI model of the welder. These FRFs can written in terms of lumped system parameters
and assuming that the electrical impedance of the system is suppressed,

V Ψ2t
He∗ = = Kt
, (6)
i jωMt + Dt + jω

∗ δ̇ Ψt
Hem = = Kt
, (7)
i jωMt + Dt + jω

δ̇ 1
Hm∗ = = Kt
. (8)
Fs jωMt + Dt + jω

9
where Ψt is the electro-mechanical transformer coefficient of this equivalent form. The other
lumped paramters are assumed to not be dependent on the particular equivalent circuit form
because they influence the system resonance and system resonance is nearly identical between
the two forms. The shear force and applied electrical current in (5) are out of phase at the 20 kHz
driving frequency (ωo ) because shear force opposes welder vibration directly. This out of phase
shear force can be expressed with a minus sign,

δ̇( jωo ) = Hem ( jωo )i( jωo ) − Hm ( jωo )F s ( jωo ). (9)

Similarly for voltage,

V( jωo ) = He ( jωo )i( jωo ) − Hme ( jωo )F s ( jωo ). (10)

Harmonic excitation is assumed in relations (9) and (10) to describe system dynamics at reso-
nance. These relations will be used later in Section 5 of the paper to back calculate welder shear
force. Evidence supporting these relations will be presented in Section 7. Prior to using the
model to estimate shear force and welder efficiency, it is of interest to show that the presented
lumped system models are valid and can be used to describe UAM dynamics. Verification and
identification of model parameters is carried out in the next section utilizing experimental FRF
measurements and characterization techniques, respectively.

4. Welder Characterization

4.1. Frequency Response Function Measurements

To describe welder boundary conditions for FRF estimation, the welder was characterized
inside the UAM machine, see Figure 6. To apply a controlled voltage and measure electric current
across the piezoelectric transducers, an AE Techron LVC 5050 was utilized. A Polytec PSV-
400 Doppler laser vibrometer was used to measure sonotrode motion in a non-contact manner.
The ability to measure welder vibration in a non-contact way is important because any added
mass and geometry will adversely influence the tuned resonance behavior of the system. A
high frequency modal hammer (PCB 086C30) was used to input a known force impulse into
the sonotrode. A modal hammer was utilized because mass loading on the structure would
be minimized compared to piezo reaction mass excitation. Lastly, a Data Physics QUATTRO
10
Dynamic Signal Analyzer was used to estimate H1 FRFs. The equipment used to characterize
the welder is shown in Figure 6 while a characterization schematic is shown in Figure 7.

ŶĂůLJƐŝƐ YhddZK
ŽŵƉƵƚĞƌ ŶĂůLJnjĞƌ

tĞůĚŝŶŐ
ƐƐĞŵďůLJ

>ĂƐĞƌsŝďƌŽŵĞƚĞƌ
/W
WŽǁĞƌŵƉůŝĨŝĞƌ ŵƉůŝĨŝĞƌ

Figure 6: Approach and equipment used to measure experimental FRF measurements of weld assembly. Characterization
was done inside the UAM machine to emulate operating boundary conditions.

;ĂͿ ŽůƚĞĚŝŶƚŽhDDĂĐŚŝŶĞ ;ďͿ ŽůƚĞĚŝŶƚŽhDDĂĐŚŝŶĞ

'ƌĂǀŝƚLJ 'ƌĂǀŝƚLJ
ĐƚƵĂƚŽƌϭ ĐƚƵĂƚŽƌϮ ĐƚƵĂƚŽƌϭ ĐƚƵĂƚŽƌϮ
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000 000
s ݅ 00000000 000000000 000000000 00000000
ƉƉůŝĞĚsŽůƚĂŐĞ KƉĞŶŝƌĐƵŝƚ ^ŚŽƌƚŝƌĐƵŝƚ ^ŚŽƌƚŝƌĐƵŝƚ

>ĂƐĞƌsŝďƌŽŵĞƚĞƌ >ĂƐĞƌsŝďƌŽŵĞƚĞƌ ߜሶ &

͗DŽĚĂů,ĂŵŵĞƌ/ŵƉĂĐƚ
ߜሶ

Figure 7: Conditions for assembly FRF measurements: (a) He and Hem measurements by driving one transducer and
leaving the other open; (b) Hm measurement by shorting both transducers with a low resistance stranded wire.

To measure He and Hem , one transducer was used to excite the system while the other was
put into an open circuit condition. This characterization approach is valid because the system is
symmetric, which makes the test conditions equivalent to driving the system with two transduc-
ers. Also, the open circuit condition is required to allow energy storage over the capacitor and to
prohibit current flow out of the passive transducer. Because the resonance of the welder is lightly
damped, swept sine excitation was utilized around the 20 kHz resonance. Frequency spacing
during the sweep was 0.5 Hz and measurements were taken until a minimum ordinary coherence
value of 0.99 was reached for the frequencies of concern. To measure Hm , both transducers were
11
short circuited with low resistance stranded wire while the modal hammer was used to excite the
sonotrode. The transducers were short circuited so that energy would not be stored within the
capacitive element of the transducers. FRF estimation utilized 10 hammer impacts for averaging
and an exponential processing window to minimize leakage error. Frequency resolution during
Hm estimation was near 2.4 Hz, so measuring the peak FRF value was difficult.
Empirical FRFs are compared in Figure 8(a) while the measurement of Hm for both transduc-
ers shorted, one transducer shorted, and no transducers shorted (open) is shown in Figure 8(b).
The FRF with the largest magnitude is He , followed by Hem , and then Hm in Figure 8(a). It is
shown in Figure 8(b) that the magnitude of Hm is similar for each electrical boundary condi-
tion, but the resonant frequency changes. This resonant frequency dependence on the electrical
boundary condition is typical for piezoelectric devices because the piezoelectric crystal stiffness
is different for an open and short circuit condition [37]. Consequently, the resonant frequency of
Hm does not coincide with Hem directly during testing. When the capacitive element is shorted,
it is analogous to making the system less stiff, which causes a downward frequency shift. On the
other hand, if the capacitive element is put into an open circuit state, the capacitance stiffens the
system and causes an upward frequency shift. During welder operation, the electrical boundary
conditions for the transducer are not shorted nor left open. Instead, it is a mixture of the two
boundary conditions, which coincides with the Hem resonance. Consequently, for comparison
purposes in Figure 8(a), Hm is manually moved to coincide with Hem .

12
 -25
;ĂͿ -35
;ďͿ Both Shorted
-30
-40 One Shorted
-35 Both Open

dB Magnitude (m/s/N)
He(A/V) -45
-40
dB Magnitude

Hem(m/s/V) -50
-45
Hm(m/s/N)
-50 -55
-55
-60
-60
-65
-65
-70 -70

-75 -75
19.6 19.65 19.7 19.75 19.8 19.85 19.9 19.6 19.65 19.7 19.75 19.8 19.85 19.9
Frequency (kHz) Frequency (kHz)

Figure 8: Experimental FRF measurements of weld assembly near 20 kHz: (a) comparison of He , Hem , and Hm FRFs;
(b) Hm resonance dependence on shorted and open circuit conditions. The system resonance is different for short and open
circuit conditions because the piezoelectric crystal stiffness is different between the two electric boundary conditions.

The Mason circuit model of the welder is compared to experimental FRFs in Figure 9. The
Mason circuit model was found using the measured FRF forms discussed earlier and with proce-
dures given for an ideal Mason circuit [24]. The He FRF was used to estimate the system capaci-
tance, mass, stiffness, and damping. The electro-mechanical transformer coefficient is found via
the ratio of the mobility loop diameters of Hem and He . As shown in the Figure, the model and
experiment demonstrate good correlation in magnitude with some error in phase. Phase error is
believed to originate in the linear amplifier used to make the voltage and current measurements
because 20 kHz is near the operating limit of the device. Consequently, phase error on the order
of 15 degrees is possible [38]. Lumped model parameters for the Mason circuit are listed in
Table 1.

13
 −26 −30 −30

Hem dB Magnitude (m/s/V)

H dB Magnitude (m/s/N)
H dB Magnitude (A/V)

−28 −40
−40
−30
−50
−32 −50
−60
−34 Exp.
−60
Fit −70
e

−36

m
−38 −70 −80
19.6 19.7 19.8 19.9 19.6 19.7 19.8 19.9 19.6 19.7 19.8 19.9
Frequency (kHz) Frequency (kHz) Frequency (kHz)

120 100 100

Hem Phase Angle (deg)

Hm Phase Angle (deg)

He Phase Angle (deg)

100 50
50
80 0
0
60 −50
−50
40 −100

20 −100 −150
19.6 19.7 19.8 19.9 19.6 19.7 19.8 19.9 19.6 19.7 19.8 19.9
Frequency (kHz) Frequency (kHz) Frequency (kHz)

Figure 9: Welder FRF comparison between measurement and system level model.

Model Variable Value

Ct (nF) 203.21
Mt (kg) 1.98
Kt (GN/m) 30.49
Dt (Ns/m) 82.05
Φt (N/V) 1.54

Table 1: Lumped parameter values for Mason circuit model of welder.

4.2. Welder Operation Measurements

Because the measured system FRFs agree well with a lumped parameter model, it can be in-
ferred that the mechanical mobility representation of the welder shown in Figure 5 is also valid.
To characterize this mechanical mobility representation of the welder, velocity, voltage, and av-
erage electric power measurements were made on the welder while it was operating without load.

14
These three variables were evaluated at different amplitude settings of the ultrasonic generator.
To measure welder velocity, the noncontact laser vibrometer was utilized. A Tektronix P6015A
voltage probe was used with an Agilent 54622A oscilloscope to measure RMS voltage and P-P
voltage of the welder. Power was measured by using the analog output channel on the ultrasonic
generator.
To obtain accurate estimates for system voltage and power, the welder was driven with one
transducer while the other was left in the open circuit condition. By driving the system with one
transducer, the ultrasonic generator doubles the applied current to achieve the same voltage and
welder velocity outputs [36]. Because current doubles and the voltage is constant, the power also
doubles. Peak welder velocity and average electric power draw for the amplitude levels of 40,
50, 60, 70, and 80% are shown in Figure 10(a) while peak voltage is shown in Figure 10(b). It
was found that the measured applied voltage was very similar to a sine wave, so the voltage in
Figure 14(b) was scaled by the RMS coefficient ( √12 ) for a sine wave. As shown in Figure 10(a),
peak welder velocity changes linearly with amplitude setting while average weld power changes
in a quadratic manner. Peak welder velocity is expected to change in a linear manner with applied
voltage when no shear force is present, see (5). Likewise, voltage also changes linearly with the
amplitude setting when there is no shear force present. Power, on the other hand is expected to
change quadratically because it is a function of current squared when shear force is not present.

3 400 1400
;ĂͿ 1300
;ďͿ
Average Electric Power (W)

2.5 300 1200

Peak Velocity (m/s)

Peak Voltage (V)

1100

2 200 1000

900

1.5 100 800

700

1 0 600
40 50 60 70 80 40 50 60 70 80
Amplitude Level (%) Amplitude Level (%)

Figure 10: Operation of welder operating under no load, i.e., not welding: (a) comparison of peak welder velocity and
average electric power draw as a function of amplitude setting; (b) applied voltage as a function of amplitude setting.

15
 As a result of the voltage and velocity changing linearly together in Figure 10, the coupling
constant for the mechanical mobility representation of the welder (Ψt ) can be found by fitting a
line between the two variables. As shown in (6)-(8), Ψt is scaled by the mechanical admittance
(Hm∗ ) to estimate Hem
∗
. As a result, applied peak current can be estimated by using Ψt , Hm∗ ,
and measured peak welder velocity. Estimated peak current can then be used with measured
RMS voltage to calculate average electric power, Pe,avg . Average electric power is defined as
the multiplication of RMS voltage, RMS current, and the cosine of the phase angle (θ) between
voltage and current,
Pe,avg = VRMS iRMS cos θ. (11)

The He∗ transfer function theoretically does not have a resonance and an anti-resonance in UAM
because motional feedback is used. Instead, the function will behave more like a single degree
of freedom model and have a phase angle of zero at resonance, see (6). Thus, the phase angle
between applied current and voltage can be assumed to be zero in (11). Using this phase angle
simplification, the average electric power can be estimated and compared to the measurement,
see Figure 11. The estimated power is higher than the measurement. To remedy this mismatch,
the peak magnitude value of Hm∗ was adjusted by increasing it 12%. As a result of the peak value
of Hm∗ being difficult to accurately measure during characterization, increasing its value is likely
more accurate as well. With this adjustment to Hm∗ , the power estimate shows closer agreement
with the measured value. The calculated constants utilized for this power calibration are listed in
Table 2.

16
 Average Power Comparison
350

300
Average Electric Power (W)

Measured
250
Estimate
Adj. Estimate
200

150

100

50
40 45 50 55 60 65 70 75 80
Amplitude Level (%)

Figure 11: Calibration of mechanical mobility model to average electric power.

Model Variable Value

Ψt (V*s/m) 515.3
Peak Hm∗ (m/s/N) 0.010
Adjusted Peak Hm∗ (m/s/N) 0.0112

Table 2: Coefficients used to validate mechanical mobility representation.

In addition to estimating power, the excitation frequency of the welder was compared to FRF
measurements. This excitation frequency was found using the measured welder velocity signal
and FFT algorithm within Matlab. The excitation frequency as a function of amplitude setting is
shown in Table 3. It was found that the excitation frequency of the system was identical or very
near the resonance of the system during FRF characterization. The small variation in excitation
frequency is believed to originate in the frequency resolution of the processing block size to
calculate the frequency.

17
 Amplitude Setting % Frequency (kHz)
40 19.760
50 19.757
60 19.757
70 19.754
80 19.754

Table 3: Frequency of welder velocity as a function of amplitude level. Frequency found via windowed FFT with
frequency resolution near 3 Hz. The median frequency of 19.757 kHz agrees well with FRF measurements. A Hamming
processing window was utilized to minimize leakage when estimating frequency.

5. Shear Force Estimates and Welder Efficiency

The welder calibration results in Figure 11 were calculated assuming that welder current,
voltage, and velocity signals were single frequency(ωo), i.e., sine waveforms. It was also as-
sumed that the phase angle between the applied current and voltage is zero due to the PLL
algorithm. Using these assumptions, (11) can be simplified to the following,

1
Pe,avg = Vi. (12)
2
Equation 12 is a reasonable simplification for average electric power draw considering that the
1 st harmonic of the weld velcocity is an order of magnitude greater than all other higher order
harmonics (see Figure 17 and Figure 21 in Section 7). Because applied voltage to the piezoelec-
tric transducers is controlled to be constant for a given amplitude set-point value, the ratio of
average electric power between two welding layers or states can be equated to the ratio of weld
currents,

P2 i2
= , (13)
P1 i1
where P2 and P1 are the average electric power for welding states 2 and 1, respectively. Likewise,
i2 and i1 are the applied peak electric current values for welding states 2 and 1. Using estimated
current from the model and measured average electric power for not welding (state 1), applied
current during welding can be estimated if average electric power draw is measured (state 2).
With this estimate for applied current at the excitation frequency, welder shear force can then be
18
estimated using the LTI model, assuming operation at resonance (9)-(10), and using measured
peak welder velocity,

Hem
∗
i − δ̇
Fs = . (14)
Hm∗
To illustrate the use of (14), the states of not welding and welding can be compared, see Fig-
ure 12. Along with estimated shear force between the two states, the measured average electric
power, measured peak welder velocity, and calculated excitation frequency are also shown in the
figure. In-situ measurements were sampled at 50 kHz (25 kHz Nyquist) with a National Instru-
ments DAQ module. Consequently, content in Figure 12 is from the excitation signal (near 20
kHz) and contains no higher order harmonic content. The average electric weld power increases
an order of magnitude, the welder excitation frequency increases 75 Hz, the peak welder velocity
decreases 10%, and shear force is near 1750 N during welding. Shear force during ultrasonic spot
welding has been measured previously [39], and found to be near 1800 N for the same aluminum
alloy with a similar sonotrode contact area.
The shear force of the welder exhibits similar behavior to the average electric power in Fig-
ure 12 since (13) is used to estimate applied current to the welder. The reported average electric
power in the figure was measured from a single piezoelectric transducer and doubled in value
because power draw is near symmetric between the two transducers on the sonotrode.

19
 Average Electric Power (W)

Weld Power Peak Welder Velocity

2000 2

Peak Velocity (m/s)

1500 1.9
No Welding
1000 1.8 Welding

500 1.7

0 1.6
0 0.5 1 0 0.5 1
Time (s) Time (s)
Excitation Frequency Peak Shear Force
Welder Frequency (kHz)

19.85 2000

Peak Force (N)

1500

19.8 1000

500

19.75 0
0 0.5 1 0 0.5 1
Time (s) Time (s)

Figure 12: Estimate of welder shear force from in-situ velocity and power measurements. Welder excitation frequency
is also compared to illustrate system stiffening when welding.

With the use of estimated peak shear force, measured peak welder velocity, and measured
average electric power draw, welder efficiency (e) can then be calculated,

1
Pm,avg δ̇F s
e= = 2 . (15)
Pe,avg Pe,avg
Welder efficiency is calculated using the ratio of average mechanical power (Pm,avg ) and average
electric power (Pe,avg ). This efficiency calculation was carried out with the assumption that the
majority of the applied electrical energy is concentrated near the excitation frequency as well.
Efficiency comparisons between the states of not welding and welding are shown in Table 4.
The efficiency is not exactly zero when not welding because there is a small residual shear force
from the calculation (14). Nontheless, the estimated welding efficiency is near estimates for ul-
trasonic metal welding systems [6] and below that of UAM piezoelectric transducers [40], which
is greater than 90%. This estimate is sensible considering efficiency cannot be greater than the
ultrasonic transducers themselves. Losses in the system which decrease efficiency are bushings,

20
bolted joints, material damping within the waveguide, and friction in the spring diaphragm ball
bearing. Additionally, some error in the efficiency and shear force measurement may be at-
tributed to higher order harmonics not being considered in the calculation. The effect of higher
order harmonics is discussed in Section 7.

No Welding Welding
Mean 4.31 83.66
Std. Dev. 0.50 0.44

Table 4: Welder efficiency comparison between the states of not welding and welding. A small efficiency is calcualted
for the state of not welding due to a small residual shear force being present from the calculation of shear force (14).

6. Influence of UAM Processing Conditions on Shear Force and Efficiency

In addition to comparing the shear force and efficiency estimates between the states of not
welding and welding, the influence of UAM processing conditions on these estimates can be
evaluated. The influence of welder amplitude or velocity on system dynamics is shown in Fig-
ure 13 while the efficiency for the amplitude levels of 40, 50, and 60% are listed in Table 5.
These measurements were made while all other processing variables were held constant and at
the same number of UAM weld layers. This same procedure was done when evaluating normal
force and speed.

21
 Average Electric Power (W)

Weld Power Peak Welder Velocity

2500 1.8

Peak Velocity (m/s)

2000 1.6 40%
1500 1.4 50%
60%
1000 1.2
500 1
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time (s) Time (s)
Excitation Frequency Peak Shear Force
Welder Frequency (kHz)

19.86 2000

Peak Force (N)

19.85
1500
19.84
1000
19.83

19.82 500
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time (s) Time (s)

Figure 13: Comparison of system dynamics as a function amplitude level. Other UAM processing conditions were held
constant in this study. In particular, normal load was 5000 N, linear weld speed was 5 m/min (200 in/min), and welds
were made at room temperature.

40% 50% 60%

Mean 92.4 86.2 83.7
Std. Dev. 0.53 0.43 0.47

Table 5: Welder efficiency comparison between different amplitude settings.

As shown in Figure 13, the the average electric power draw increases with amplitude level,
the peak welder velocity increases monotonically with amplitude setting, the excitation fre-
quency decreases with higher amplitude levels, and shear force increases with higher amplitude
levels. Power draw and shear force increases with amplitude level because enhanced motion in-
creases both elastic and plastic deformation of the system. This increase in deformation leads to
a corresponding increase in effort from the welder. The efficiency decreases with larger welder
amplitude levels because plastic deformation and some slip is present during UAM. The relation-
22
ship is made more clear by evaluating the closed form relation of efficiency in terms of welder
motion and shear force,

δ̇
e=1− . (16)
δ̇ + Hm∗ F s
The relation shows that if the shear force and welder motion are not linearly related, the efficiency
can show some variation. Since metal plasticity and slip are present in UAM, it is suspected that
their non-linearity is the reason for this efficiency variation in Table 5.
Average Electric Power (W)

Weld Power Peak Welder Velocity

2200 Peak Velocity (m/s) 1.75

2100 1.7

2000 1.65
4000 N
1900 1.6 5000 N
6000 N
1800 1.55
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time (s) Time (s)
Excitation Frequency Peak Shear Force
Welder Frequency (kHz)

19.84 2100
Peak Force (N)

19.83 2000

19.82 1900

19.81 1800

19.8 1700
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Time (s) Time (s)

Figure 14: Comparison of system dynamics as a function of normal force. Other UAM processing conditions were held
constant in this study. In particular, prescribed welder amplitude was 60% (32.5µm), linear weld speed was 5 m/min
(200 in/min), and welds were made at room temperature.

4000 N 5000 N 6000 N

Mean 82.48 81.34 80.56
Std. Dev. 0.60 0.36 0.36

Table 6: Welder efficiency comparison as a function of normal force.

23
 The next UAM processing variable to be evaluated is the normal or down force. Results for
the normal force levels of 4000, 5000, and 6000 N are shown in Figure 14 and Table 6. These
different normal force levels do not strongly influence UAM system dynamics from one level to
the next because it has been shown that frictional slip does not strongly influence UAM at these
force levels [11, 22]. Negligible magnitude difference is observed between the average electric
power draw and peak shear force for the different force levels. Peak welder velocity decreases
near 1.5% between each force level and excitation frequency decreases in an exponential manner
with normal force level. An explanation of this observed trend can be described by evaluating
the inductor impedance (Lt ) used in the motional impedance control strategy [23],

1
Lt = . (17)
ω2C t

If the inductance used to cancel out the electrical impedance of the transducer is fixed during
welding, velocity will not be held constant because excitation frequency changes. Down force
may influence the coupling of the welder to the build, which then ultimately influences the ex-
citation frequency of the welder. This welder amplitude influence would then make the system
less efficient since energy is being stored in the capacitor. This reasoning is supported by the
decrease in efficiency in Table 6.
The last UAM processing variable to evaluate is linear weld speed or how quickly the sonotrode
rolls along the surface of the UAM build during construction. Results for the speed levels of 3.75
(150), 5.00 (200), and 6.25 m/min (250 in/min) are shown in Figure 15 and Table 7. These weld
speed levels have little to no influence on welder performance or efficiency. This observation
corroborates other 9 kW UAM experimentation [22] and energy transfer models [28] because
slower weld speeds (Vt ) were found to linearly improve bond quality. In other words, the im-
parted energy to the weld interface (Ewelder ), and ultimately bond quality, is linear with speed
because imparted mechanical power (Pmech ) is not a function of welder speed within the investi-
gated speed range,
x
Ewelder = P mech , (18)
Vt
where x is the linear travel distance of the welder.

24
 Average Electric Power (W)

Weld Power Peak Welder Velocity

2400 2 3.75 m/min

Peak Velocity (m/s)

5.00 m/min
2200 1.8 6.25 m/min

2000 1.6

1800 1.4
0 0.5 1 1.5 0 0.5 1 1.5
Time (s) Time (s)
Excitation Frequency Peak Shear Force
Welder Frequency (kHz)

19.83 2200

19.82 Peak Force (N) 2100

2000
19.81
1900

19.8 1800
0 0.5 1 1.5 0 0.5 1 1.5
Time (s) Time (s)

Figure 15: Comparison of system dynamics as a function of weld speed. Other UAM processing conditions were held
constant in this study. In particular, prescribed welder amplitude was 60% (32.5µm), normal force was 5000 N, and
welds were made at room temperature.

3.5 m/min 5 m/min 6.5 m/min

Mean 82.40 82.14 81.81
Std. Dev. 0.61 0.62 1.10

Table 7: Welder efficiency comparison as a function of weld speed.

7. Welder Velocity Response during UAM

The following sections discuss the influence of shear force and build compliance on welder
velocity response during UAM. The first section discusses why welder shear force influences
harmonic content in the velocity spectrum of the welder. The second section examines the ef-
fect of UAM build compliance on the nominal excitation frequency and the harmonic content
magnitude.
25
7.1. Effect of Shear Force

As discussed in Section 2, the ultrasonic generator utilizes a phase lock loop algorithm com-
bined with motional feedback to maintain a prescribed welder vibration magnitude and to track
system resonance shifts during UAM. The relationship between build stiffness and excitation
frequency shifts were discussed in detail while the control framework for constant amplitude
control at the nominal excitation frequency was explained. It was not explained that additional
frequency content exists in the velocity response during welding, see Figure 16. Specifically, the
lateral scrubbing velocity response of the welder increases at all frequencies and exhibits content
at harmonic locations of the excitation frequency. This harmonic content has been observed prior
in UAM [29] and in ultrasonic spot welding [41], yet its origin has not been fully explained.

2 20

1.5
;ĂͿ ;ďͿ Sine Wave
0 Square Wave
1
No Welding
dB Velocity (m/s)

Sawtooth Wave
-20
Velocity (m/s)

0.5 Welding

0
-40

-0.5
-60
-1

-1.5 -80

-2
-100
14.82 14.84 14.86 14.88 14.90 14.92 14.94 14.96 14.98 15.00 0 20 40 60 80 100 120
Time (ms) Frequency (kHz)

Figure 16: Sonotrode or welder velocity character during UAM compared to actuating the welder without load, i.e. not
welding: (a) time history comparison of signals; (b) frequency domain comparison of signals. Signals were sampled at
512 kHz with an amplitude level of 60% (32.5 µm setpoint). The phase delay in (a) is of no significance. The small
decrease in welder amplitude during welding will be discussed in Section 8. Frequency data was obtained by averaging
the velocity signal magnitude with a block size of 81920 points and without windowing. Windowing was not used to
illustrate increased frequency content more clearly. Due to windowing not being used, some leakage may be present in
the figure.

The frequency response of the welder velocity is different during welding vs. not welding
because the shear force input function contains energy at all other frequencies in addition to the
nominal excitation signal. The exact excitation frequency of the welder depends on the effective
resonance of the sonotrode and transducers, as explained earlier. To explain why additional
26
frequency content exists in the welder velocity signal, the proposed LTI model expression (5)
can be used to describe the velocity response for all frequencies,

∗
δ̇( jω) = Hem i + Hm∗ F s . (19)

Equation (19) explains that welder velocity can be represented as a linear combination of the ap-
plied electric current and resultant shear force for all frequencies. It was established in Section 5
that the electric current used to actuate the welder could be approximated as a sine wave because
the measured welder output voltage was found to be nearly sinusoidal. This single tone excita-
tion is corroborated when analyzing the velocity response of the welder when operating under
no load because its response is also sinusoidal, Figure 4(b). Because the excitation signal is con-
centrated near resonance (ωo ), content at other frequencies must come from the shear force input
function. Further, the observed content in the velocity spectrum is originally passed through the
system transfer path or frequency response function (FRF) for a given system input. These FRFs
act as filters on the system, and they influence frequency content in the velocity spectrum. To
understand FRF influence on velocity and system response at high frequencies, these functions
can be measured.
Obtaining FRF measurements at frequencies above 20 kHz is difficult. For measuring Hm∗ ,
state-of-the-art high frequency modal hammers do not demonstrate the necessary frequency re-
sponse input character to reliably characterize the system upward of 20 kHz. For measuring
Hem
∗
, response above resonance is low and narrow band excitation techniques are required. De-
spite these measurement difficulties, the general character of these FRFs can be investigated, see
Figure 17. The FRFs in the figure were obtained using the procedures outlined in Section 4 and
using the built-in high frequency data acquisition system of the laser vibrometer. It should be
noted that Hem in Figure 17 is for the velcocity-voltage FRF (Hem ) and not the velocity-electric
current FRF (Hem
∗
). However, the two FRFs have similar behavior and can be used for compari-
son purposes.

27
 *
Hm
-20
Hem
Harmonics
-40
FRF dB Magnitude

-60

-80

-100

-120

-140
0 10 20 30 40 50 60 70 80 90 100
Frequency (kHz) 39.514 kHz

Figure 17: Hem and Hm∗ FRF estimates of the welder from 0 to 100 kHz. The nominal resonance of the welder is 19.757
kHz, and its harmonics are marked with astericks on the FRFs. The 2nd harmonic (39.514 kHz) is given as an example.
The locations of the harmonic frequencies were investigated to see if they coincided with any system resonances. No
harmonics showed correlation with system resonances, so amplification is not significant during welding. FRFs were
estimated using the built-in data acquisition of the Polytec laser vibrometer system.

The harmonic locations of the nominal resonance of 19.757 kHz are marked in Figure 17 with
asterisks. Aside from the excitation harmonic (1 st ), the harmonics of the excitation signal do not
coincide with any other system resonances. Consequently, significant amplification of harmonic
content is not suspected during welding, and the content in Figure 4(b) must be transmitted at
low response regions of the Hm∗ FRF. Because harmonic content is transmitted at low response
regions, the origin of the content must come from the shear force excitation function.

7.2. Effect of Build Compliance

Prior work by the authors [28] has shown that UAM build compliance influences the per-
formance of the process and ultimately part quality. To maintain performance and part quality
in UAM, the concept of power compensation was presented. Compensated power is when the
welder displacement is increased manually to maintain a constant electric power input into the
UAM build due to build compliance influencing the welder effort to reach a prescribed amplitude
limit. Uncompensated means that the welder amplitude is not adjusted during build construction,
28
i.e., default constant amplitude control of the welder is utilized. The concept of power compensa-
tion is shown schematically with corresponding measurements in Figure 18. Hehr et al. [28] did
not investigate welder velocity content during this study. Consequently, the impact of build com-
pliance and power compensation on the harmonic content in the velocity spectrum was studied
herein using in-situ welder velocity measurements as a function of build height.
Percent Average Electric Power (W/W)

&Ɛ 

^ŚĞĂƌ&ŽƌĐĞ;‫Ϳ ܨ‬
120
;ĂͿ EĞǁ>ĂLJĞƌ ;ďͿ ;ĐͿ ŽŵƉĞŶƐĂƚĞĚ
110
ZĞŐŝŽŶ
hD^ƚĂĐŬ 100 Uncompensated ݇
Compensated
ǁŝƚŚŬ  90
‫
ܨ‬
ƐƚŝĨĨŶĞƐƐ ݇ ݇
80

70

ߜ 60
0 5 10 15 20 25
ĂƐĞƉůĂƚĞ ߜ, ߜ, ߜ, ߜ, ŝƐƉ͘;ߜͿ
ߜ No. of Welded Layers

Figure 18: UAM build dynamics: (a) schematic of UAM build during welding showing areas of elastic and plastic
deformation; (b) comparison of measured average electric power draw percent for build with (compensated) and without
(uncompensated) power control, reprinted with permission [28]; (c) proposed bilinear shear force profile during welding.

Figure 18(a) illustrates why UAM build compliance impacts the process while measured av-
erage electric weld power is shown in Figure 18(b). Figure 18(c) illustrates the hypothesized
bilinear shear force input function and plasticity behavior of the weld foil during the UAM pro-
cess. The shear force or reaction force from the UAM build changes linearly with build stiffness
(k) until the yield strength (Fy ) of the foil is reached. Then, the weld foil plastically deforms
with a constant tangent stiffness (kT ) to the prescribed welder amplitude value or displacement
limit (δL ). Because the stiffness of the build decreases with build height, the amount of plastic
deformation also decreases due to it taking more elastic deformation (δE ) to initiate yielding.
More details on this idea can be found in previous work by the authors [28].
Power compensation is carried out manually by increasing the prescribed amplitude limit
each layer due to state-of-the-art UAM equipment not exhibiting a constant weld power feature.
The approach used to keep power constant is listed in Table 8 and is similar to the approach used
in previous work [28]. To measure average electric power draw as a function of weld height, the
analog output signal for weld power on the ultrasonic generator was synchronized with welder
29
velocity. Average electric power as a function of build height is shown in Figure 19(a) with first
standard deviation error bars about a mean value for a given layer. Because welder velocity is
measured, welder frequency as a function of build height can be estimated using Fourier analysis,
see Figure 19(b). Welder frequency was found by using the FFT function in Matlab, a Hamming
processing window to minimize leakage error, a 50% overlap, and a block size of 81920 (6.5 Hz
resolution). Because the ultrasonic generator changes the frequency slightly during welding, first
standard deviation error bars are used to show frequency variation about a mean value for each
weld layer.

Layer 1-2 3-4 5-6 7-8 9-10 11-15 16-20

Amp. (%) 60 61 62 63 64 65 66
P-P δL (µm) 32.5 33 33.6 34.1 34.7 35.2 35.8

Table 8: Power compensation approach by increasing prescribed peak-to-peak displacement (δL ) of the welder. Ampli-
tude percentage was input into the machine controller to achieve the given displacement.

1600 10
1550 ;ĂͿ Uncompensated
5
;ďͿ
Average Electric Power (W)

Compensated
1500
Frequency Shift (Hz)

0
1450

1400 -5

1350 -10

1300
-15
1250
-20
1200
-25
1150

1100 -30
0 5 10 15 20 0 5 10 15 20
No. of Layers No. of Layers

Figure 19: Power and excitation frequency shift during UAM stack construction: (a) average electric weld power vs.
number of layers; (b) excitation frequency shift vs. number of layers. The compensated power trace in (b) shows an
increased amount of frequency shift.

It is shown in Figure 19(b) that welder excitation frequency shifts downward with more layers
and with more displacement (compensated trace). The downward frequency shift for amplitude
30
control (uncompensated trace) corresponds to the UAM build becoming less stiff with more lay-
ers. Because the build becomes less stiff with more layers, the amount of plastic deformation
hardening decreases and leads to a decrease in the effective build stiffness. On the other hand,
the increased displacement for power compensatation increases this downward frequency shift.
This increased frequency shift occurs because the effective build stiffness decreases with more
plastic deformation. Additional evidence of this downward shift due to a larger amount of plas-
tic deformation is shown in Figure 13 when the influence of welder amplitude on shear force
magnitude was evaluated. These two frequency shift mechanisms are schematically detailed in
Figure 20.

 
&ŽƌĐĞ;‫Ϳ ܨ‬

&ŽƌĐĞ;‫Ϳ ܨ‬
;ĂͿ ;ďͿ

݇ ݇ 

݇ ݇

ŝƐƉ͘;ߜͿ ŝƐƉ͘;ߜͿ

Figure 20: Frequency shift mechanisms during UAM: (a) less plastic hardening decreases the effective build stiffness;
(b) more plastic deformation decreases the effective build stiffness.

In addition to analyzing the frequency shift of the excitation signal as a function of build
height, harmonic frequency content was examined. The dependence of this harmonic content
on prescribed displacement and build stiffness can give insight into shear force and plasticity
behavior of the weld foil during build construction. Averaged frequency content for the first three
harmonics are shown in Figure 21 for the compensated and uncompensated power conditions.
This data was obtained by utilizing a Hamming processing window to simultaneously minimize
leakage error and obtain accurate amplitude estimates. A 50% processing block overlap was
also used. The processing block size was 81920 points (6.5 Hz resolution), and this block size
provided an adequate number of averages to statistically separate the trends shown in the figure.
Because the processing frequency resolution was not extremely fine and the frequency is non-
31
stable during UAM, points at and around the harmonic locations were used to estimate content by
averaging them together for a given processing block. Three points were taken above and below
the harmonic peak for averaging. Thus, seven total points were averaged together to produce the
data shown in Figure 21 with the first standard deviation as error bars.

;ĂͿ 20 kHz Peak x 10

-3
;ďͿ 40 kHz Peak
-3
x 10 ;ĐͿ 60 kHz Peak
0.17 6 11
Uncompensated 10
0.16 Compensated 5
9
0.15 4 8
Velocity (m/s)

Velocity (m/s)

Velocity (m/s)
7
0.14 3
6

0.13 2 5

4
0.12 1
3

0.11 0 2
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
No. of Layers No. of Layers No. of Layers

Figure 21: Harmonic content during UAM stack construction for the approximate locations of 20, 40 ,and 60 kHz: (a) 1st
harmonic, 20 kHz; (b) 2nd harmonic, 40 kHz; (c) 3rd harmonic, 60 kHz. The influence of power compensation increases
the magnitude of the 1st harmonic, but does not influence the other harmonics. Due to the UAM build becoming more
compliant with construction, the magnitude of the 3rd harmonic decreases. Harmonic content was found using FFT
analysis using a Hamming processing window, a block size of 81920 (6.5 Hz resolution), and by averaging the harmonic
peak value with three points below and above it, i.e., seven total points total used to estimate content around the peak.
Seven points were used for averaging content because reliably and consistently estimating the peak harmonic response
is difficult with the processing block resolution.

The nominal excitation frequency content (20 kHz) is very steady for the uncompensated
trace in Figure 21(a). This steadiness is expected due to the ultrasonic generator attempting to
keep amplitude constant during build construction. The cyclic variation shown in the uncom-
pensated trace originates from periodically moving the measurement point of the laser vibrom-
eter every 5 layers to remain on the sonotrode combined with the deformation behavior of the
sonotrode. Sonotrode deformation behavior will be discussed in more detail in Section 8. The in-
creasing content in the compensated trace is expected due to the velocity or amplitude increasing
during build construction. It is shown in Figure 21(b) that the 2nd harmonic content (40 kHz) is
constant as a function of build height for both the compensated and uncompensated traces. This
constant content implies that the hardening behavior (sawtooth component) of the weld metal
does not change significantly with build height. On the other hand, the magnitude of the 3rd
32
harmonic (60 kHz) demonstrates a large amount of change with the number of weld layers, see
Figure 21(c). Like the excitation frequency shift in Figure 19(b), the compensated and uncom-
pensated traces show decay as a function of build height. The decay in the content is expected
because it is the first square wave harmonic and is the most sensitive to square wave changes,
i.e., non-constant build compliance. Consequently, the increasing build compliance of the UAM
stack with height influences the shear force profile.
The decreasing magnitude of the 3rd harmonic or square wave harmonic combined with
the prevalent high order odd harmonics in Figure 16(b) and transfer path character of Hm∗ (see
Figure 17) can be used to hypothesize the shear waveform behavior during welding, see Fig-
ure 22. The proposed forcing function exhibits predominant square waveform behavior because
the dominant high frequency harmonics in Figure 16(b) coincide with odd harmonic frequency
content (60 and 100 kHz peaks). It is suspected that some plastic deformation hardening does
exist in UAM as originally depicted in Figure 18(c) because the sawtooth wave or even harmonic
locations in Figure 16(b) have content, but this content is much less than the square wave con-
tent and shows no change with build height. The square wave content changes with build height
because the elastic stiffness of the build decreases and influences the shear force input function.
In particular, the forcing function becomes less square wave like due to the build becoming more
compliant with each added layer, which in turn reduces the magnitude of the 3rd harmonic with
each layer. Figure 22 illustrates the proposed force waveform behavior with k1 representing a
stiff UAM build while k2 represents a less stiff build. The k1 profile is more square like than k2 .

33
  

&ŽƌĐĞ;‫Ϳ ܨ‬

&ŽƌĐĞ;‫Ϳ ܨ‬
;ĂͿ ;ďͿ
‫ܨ‬
‫ܨ‬ ݇  ݇

ߜ ߜ ߜ ŝƐƉ͘;ߜͿ dŝŵĞ;ƐͿ

Figure 22: Proposed shear force profile during welding using evidence from welder velocity frequency response mea-
surements and Hm∗ : (a) force-displacement; (b) force-time. Due to the UAM build becoming more compliant with more
layers, the behavior of the forcing function becomes less square like with more layers.

8. Discussion

A linear time invariant model of the UAM process has been proposed to describe process
dynamics, see Figure 3. In particular, the model explicitly specifies welder shear force as a sys-
tem input to describe system dynamics and power conversion within the welder. This model
was explored through equivalent circuit analysis, empirical FRF estimation, and characterization
measurements to understand system behavior and to identify system parameters. With this LTI
model, shear force magnitude and welder efficiency were estimated for standard welding proce-
dures for Al 6061 foil. The impact of welding variables on shear force and efficiency was also
investigated. This model also provides insight into describing the high frequency harmonic con-
tent, the resonant frequency shift during welding, the decrease in welder amplitude under load,
and can be used to describe the influence of UAM build dynamics on welder behavior.

8.1. Welder Frequency and Amplitude

UAM system dynamics change during welding because the dynamics of the weld assembly
couple with the dynamics of the UAM build. The UAM build generates an opposing shear force
which acts on the sonotrode, and the build stiffens the system due to the resonant frequency
shifting upward. System stiffening can only occur if the sonotrode and the weld foil are not
34
slipping on preceeding layers when welding. Instead of slipping, it is believed that a psuedo-
steady stick condition occurs because a stable upward frequency shift is observed when welding,
Figure 3(c). If the sonotrode and weld tape did not stick to the previous layer, i.e. frictional
slip, the shear force would be fully decoupled from the build and no upward system resonance
shift would occur. It is believed that when the welder sticks to the UAM build, the surface
asperities of the foil at the interface undergo large amounts of plastic deformation through shear,
which then promotes bonding. Others have also noted low amounts of frictional slip in 9 kW
UAM when welding at higher loads (4-6 kN). Specifically, heat generation within the weld was
found to correlate well with plastic deformation heating and not frictional slip [11]. Also, normal
force was found to not be statistically significant on UAM weld strength [22]. This concept of
stick producing good bonds through shear deformation is contrary to the work of others [26].
However, this prior work did not include plastic deformation of the weld tape as a bonding
mechanism. Instead, the stick condition was assumed to purely limit interfacial sliding.
The 10% decrease in velocity during welding (see Figure 3(a)) occurs because (i) the mo-
tional feedback method employed by the ultrasonic generator may exhibit error when tracking
the amplitude and (ii) the sonotrode deforms slightly during welding operations. To explore
the motional feeback error, the presented LTI model can be used by evaluating the difference in
welder velocity between two different welding states, i.e., welding and not welding. Because
voltage is held constant between these two welding states, the voltage difference between the
two states is zero. Consequently, with the use of (5) it can then be found that the difference in
electric current is proportional to the difference in shear force,

He∗ ∆i = Hem
∗
∆F s . (20)

It is assumed in expression (20) that FRF magnitude is the same when welding or not welding
and that system energy is predominately concentrated at the excitation frequency (ωo ). FRF
magnitude dependence on welding will be discussed in the next section. Equation (20) can then
be used to explain the difference in welder velocity by relating it to the difference in shear force,

∗2
Hem
∆δ̇ = ( − Hm∗ )∆F s . (21)
He∗
Equation (21) demonstrates that shear force can influence welder velocity if the FRF term is not
zero. Ideally, this FRF term equates to zero if motional feedback completely removes the elec-
35
trical impedance influence of the transducer, see (6). However, if the electrical impedance of the
transducer is not completely removed, the FRF term will not equate to zero and welder velocity
will not be held constant when welding. The influence of remaining transducer impedance on
welder performance was discussed earlier in Section 6 when evaluating the influence of normal
force on welder effort.
To investigate the influence of sonotrode deformation on welder amplitude, a reduced fidelity
linear elastic solid model of the sonotrode was analyzed in COMSOL Multiphysics. The goal of
this simulation is not to fully describe the deformation behavior of the sonotrode, but to gain in-
sight into the magnitude and behavior of the deformation. The built-in High-Strength Alloy Steel
material option within COMSOL was used as the material in the model while the contact width
for the force was chosen to be 0.5 cm from experimental observations. The mesh of the model
utilized the physics controlled fine option within COMSOL. The reduced fidelity model along
with boundary conditions for the simulation are shown in Figure 23(a). Because the sonotrode is
constrained against the weld surface of the build with a prescribed normal force, a deformation
constraint was placed on the tool piece. The simulation output for a 9.65 cm diameter tool piece
with an applied 2000 N distributed load is shown in Figure 23(b). A 2000 N load was used be-
cause this is near the shear force estimate found in Section 5. Because the sonotrode is designed
to operate in the longitudinal mode of vibration near 20 kHz, the static assumption is valid due
to the self-exciting deformation modes of the sonotrode being sufficiently far away [36].

36
 ;ĂͿ (#0 - *1%)!- . ;ďͿ
(!') *!+,) #$ 2- .') *%!.) !. 3

- *&# !. /

!"#$ %&#'

Figure 23: Sonotrode deformation analysis from shear loading: (a) reduced model of sonotrode tool piece with assump-
tions; (b) deformation contour of 9.65 cm diameter tool piece with applied 2000 N distributed load.

It is shown in figure 23(b) that the maximum deformation of the tool is near 1.5 µm. This de-
formation magnitude is similar to the velocity decrease of 0.25 m/s in Figure 3(a). This velocity
decrease correlates to an amplitude magnitude estimate of near 2 µm. This estimate was found
by dividing the velocity decrease by frequency in rad/s, i.e., harmonic excitation assumption.
In practice, obtaining the maximum deformation of the tool piece is not feasible because the
laser vibrometer cannot reliably focus on the very tip of the tool piece. Nonetheless, this anal-
ysis shows that sonotrode deformation is likely a contributing factor to the decreased amplitude
during welding and may be a larger contributor at higher shear force loads.

8.2. Influence of UAM Build Dynamics

The influence of UAM build dynamics on system dynamics can be studied by replacing
welder shear force in Figure 5(b) with a prescribed forcing function or a load impedance function
(ZLD ) to describe energy transfer into the weld for UAM components. To describe UAM build
dynamics in a general way for an Al 6061 stack, a complex stiffness [42] load impedance can be
used,

ke f f
ZLD = (1 + jη), (22)
jω
37
where ke f f is the effective stiffness of the UAM build and η is the loss factor or energy transfer
efficiency of the process. This complex stiffness expression describes UAM build stiffness and
damping, independent of frequency. In other words, a hysteretic damping model was chosen
to model system losses because plastic deformation character would show little change near
the excitation frequency when the PLL controller moves the excitation frequency. Damping
from plastic deformation is typically described with a Coulomb-friction model [42], like bilinear
hysteresis [43, 44]. The purpose of the presented impedance function is not to describe the
elasto-plastic interaction during UAM but to describe the UAM build dynamics in a general way.
To describe the influence of the UAM build onto the system, the load impedance was im-
plemented by introducing effective build stiffness without damping initially (η = 0) and then
introducing a loss factor (η = 0.25). To estimate the effective build stiffness, system stiffness
was increased until a 75 Hz frequency shift was accomplished, see η = 0 traces in Figures 24-25.
This 75 Hz frequency shift was chosen because it was observed when welding an Al 6061 stack,
see Figure 3(b). The required build stiffness to cause this frequency shift is less than 1% of the
assembly stiffness. Also, because energy is being stored in the system, no change in system
response magnitude occurs. Because the welder is capable of moving the excitation frequency
during welding, it can be inferred that FRF response will not change due to this added stiffness.
For illustration purposes, a loss factor of 0.25 or a damping capacity of 25% was chosen to show
the behavior of system losses during UAM due to large amounts of plastic deformation occur-
ring. The η = 0.25 curves in Figures 24-25 illustrate that the system response decreases due to
energy no longer being stored within the welding assembly and being transferred to the weld
interface. A summary of key lumped parameter estimates for this analysis are shown in Table 9.

38
 20

15

10
dB Magnitude (m/s/A)

5
ZLD = 0
0 η=0
η = 0.25
−5

−10
em
H*

−15

−20

−25
19.5 19.6 19.7 19.8 19.9 20
Frequency (kHz)

Figure 24: Estimation of UAM build stiffness via a 75 Hz frequency shift and the investigation of UAM build stiffness
on FRF magnitude. Due to stiffness not influencing FRF magnitude, FRF magnitude would be consistent when system
resonance changes under load. Damping or energy leaving the system is shown to decrease system response.

39
 -3
2000 x 10
;ĂͿ 8 ;ďͿ
1500
ʘ Z =0 6 ʘ
LD
1000
η= 0

Hem Imag (m/s/V)

4
He Imag (V/A)

500 η = 0.25
2
0
0
*

-500
-2
-1000
-4
-1500
-6
-2000
0 1000 2000 3000 4000 0 5 10 15
H* Real (V/A) H Real (m/s/V) -3
e em x 10

Figure 25: Nyquist plots: (a) estimated motional impedance locus, He∗ ; (b) measured admittance locus, Hem . Resonant
frequencies for each locus are marked with an asterisk while adjacent locus points are marked with circles. Loading
the system without damping does not change the motional impedance locus due to the diameter remaining constant.
The decrease in locus diameter occurs only when damping is introduced [34]. The resonant frequency location in the
motional impedance locus occurs at 0 degrees because the electro-mechanical coupling coefficient is assumed to be real
valued [34]. To verify this assumption, the Hem admittance locus is plotted. The resonant peak has a phase location < 1
degrees, which supports the real valued coupling coefficient assumption of 0 degrees.

Model Variable Value

Kt (GN/m) 30.49
ke f f (GN/m) 0.02
η 0.25

Table 9: Key lumped parameter values used to describe UAM stack dynamics. Effective build stiffness (ke f f ) was chosen
by increasing the system resonance 75 Hz. A high loss factor (η) was chosen to illustrate the influence of the large
amount of plastic deformation on system response during UAM.

9. Concluding Remarks

A linear time invariant (LTI) model of the UAM process has been proposed to describe pro-
cess dynamics by explicitly specifying welder shear force as a system input. To verify and to
40
identify LTI model parameters, equivalent circuit analysis, empirical FRF estimation, and char-
acterization measurements, were utilized. These characterization techniques and models can
be applied to state-of-the-art UAM systems, can be used to improve controller design, and to
ensure welder consistency across various welder assemblies. Welder consistency is of utmost
importance as the use of UAM continues to expand.
Welder shear force and welder efficiency were estimated for the first time in UAM using this
LTI model. Shear force magnitude and efficiency estimates were found to be near 2000 N and
80%, respectively. The influence of welder amplitude, normal force, and weld speed on shear
force magnitude and welder efficiency was also evaluated. Normal force and weld speed show
little to no influence while welder amplitude showed significant influence. The strong influence
of welder amplitude occurs because of differences in elastic and plastic compliance between
amplitude levels. The LTI model, experimental FRF measurements, and high frequency in-
situ velocity measurements of the welder were also used to improve understanding of harmonic
content and frequency shift behaviors in UAM. From this harmonic content behavior, shear force
character can be estimated and understood as UAM build compliance changes.
An upward frequency shift on the order of 75 Hz occurs when welding Al 6061 because the
UAM build stiffens the overall system. System stiffening can only occur if the sonotrode sticks
to the UAM build. Otherwise, no system coupling would occur and no frequency shift would
be observed. Using a lumped parameter model of the welding assembly and the observed 75
Hz resonant frequency shift, the effective build stiffness was found, which is less than 1% of the
assembly stiffness.

Acknowledgments

This material is based upon work supported by the National Science Foundation, CMMI
Division Grant No. 1538275. Support for A.H. comes from a NSF Graduate Fellowship under
Grant No. 1102690 and from the member organizations of the Smart Vehicle Concepts Center
(www.SmartVehicleCenter.org), a National Science Foundation Industry/University Cooperative
Research Center. Any opinions, findings, and conclusions or recommendations expressed in this
material are those of the authors and do not necessarily reflect the views of the National Science
Foundation. The authors would like to thank I.V. Adamovich and B. Goldberg of OSU’s Non-
Equilibrium Thermodynamics Laboratory for helping with the high voltage measurements of the
41
welder.

References

References

[1] D. White, Ultrasonic consolidation of aluminum tooling, Advanced materials & processes 161 (1) (2003) 64–65.
[2] K. Graff, ASM Handbook: Volume 6A: Welding Fundamentals and Processes: Ultrasonic Additive Manufacturing,
American Society for Metals International, 2011.
[3] K. Graff, M. Short, M. Norfolk, Very High Power Ultrasonic Additive Manufacturing (VHP UAM), in: Interna-
tional Solid Freeform Fabrication Symposium, Austin, TX, 2011.
[4] E. Neppiras, Ultrasonic welding of metals, Ultrasonics 3 (3).
[5] H. Daniels, Ultrasonic welding of metals, Ultrasonics 3 (4).
[6] K. Graff, AWS Handbook 9th Edition: Volume3: Ultrasonic Welding of Metals, American Welding Society, 2001.
[7] K. Johnson, Interlaminar subgrain refinement in ultrasonic consolidation, Ph.D. thesis, Loughborough University,
Loughborough, UK (2008).
[8] R. Dehoff, S. Babu, Characterization of interfacial microstructures in 3003 aluminum alloy blocks fabricated by
ultrasonic additive manufacturing, Acta Materialia (2010) 1–12.
[9] M. Sriraman, S. Babu, M. Short, Bonding characteristics during very high power ultrasonic additive manufacturing
copper, Scripta Materialia 62 (2010) 560–563.
[10] H. Fujii, M. Sriraman, S. Babu, Quantitative evaluation of bulk and interface microstructures in Al-3003 alloy
builds made by very high power ultrasonic additive manufacturing, Metallurgical and Materials Transactions A:
Physical Metallurgy and Materials Science 42 (13) (2011) 4045–4055.
[11] M. Sriraman, M. Gonser, H. Fujii, S. Babu, M. Bloss, Thermal transients during processing of materials by very
high power ultrasonic additive manufacturing, Journal of Materials Processing Technology 211 (2011) 1650–1657.
[12] G. Janaki Ram, C. Robinson, Y. Yang, B. Stucker, Use of ultrasonic consolidation for fabrication of multi-material
structures, Rapid Prototyping Journal 13 (4) (2007) 226–235.
[13] J. Obielodan, A. Ceylan, L. Murr, B. Stucker, Multi-material bonding in ultrasonic consolidation, Rapid Prototyping
Journal 13 (3) (2010) 180–188.
[14] C. Hopkins, S. Fernandez, M. Dapino, Statistical Characterization of Ultrasonic Additive Manufacturing Ti/Al
Composites, Journal of Engineering Materials and Technology 132.
[15] A. Truog, Bond Improvement of Al/Cu Joints Created by Very High Power Ultrasonic Additive Manufacturing,
Master’s thesis, The Ohio State University, Columbus,OH (2012).
[16] J. M. Sietins, Exploring Diffusion of Ultrasonically Consolidated Aluminum and Copper Films Through Scanning
and Transmission Electron Microscopy, Ph.D. thesis, University of Delaware, Newark, DE, USA (2014).
[17] C. Kong, R. Soar, P. Dickens, Ultrasonic consolidation for embedding sma fibres within aluminium matrices,
Composite Structures 66 (2004) 421–427.
[18] C. Kong, R. Soar, Method for embedding optical fibers in an aluminum matrix by ultrasonic consolidation, Applied
Optics 44 (30) (2005) 6325–6333.

42
[19] E. Siggard, Investigative research into the structural embedding of electrical and mechanical systems using ultra-
sonic consolidation, Master’s thesis, Utah State University, Logan, UT (2007).
[20] X. Cheng, A. Datta, H. Choi, X. Zhang, X. Li, Study on Embedding and Integration of Microsensors Into Metal
Structures for Manufacturing Applications, Journal of Manufacturing Science and Engineering 129 (2007) 416–
424.
[21] R. Hahnlen, Characterization and Modeling of Active Metal-Matrix Composites with Embedded Shape Memory
Alloys, Ph.D. thesis, The Ohio State University, Columbus, OH (2012).
[22] P. Wolcott, A. Hehr, M. Dapino, Optimized welding parameters for Al 6061 ultrasonic additive manufactured
structures, Journal of Materials Research 29 (17).
[23] C. van der Burgt, H. Pijls, Motional Positive Feedback Systems for Ultrasonic Power Generators, IEEE Transac-
tions on Ultrasonics Engineering 10 (1).
[24] B. Richter, J. Twiefel, J. Wallaschek, Energy Harvesting Technologies: Ch4: Piezolectric Equivalent Circuit Mod-
els, Springer, 2009.
[25] J. M. Gibert, E. M. Austin, G. Fadel, Effect of height to width ratio on the dynamics of ultrasonic consolidation,
Rapid Prototyping Journal 16 (4).
[26] J. M. Gibert, G. Fadel, M. E. Daqaq, On the stick-slip dynamics in ultrasonic additive manufacturing, Journal of
Sound and Vibration 332 (2013) 4680–4695.
[27] C. Robinson, C. Zhang, G. Janaki-Ram, E. Siggard, B. Stucker, L. Li, Maximum height to width ratio of free-
standing structures built by ultrasonic consolidation, in: Proceedings of the 17th Solid Freeform Fabrication Sym-
posium, Austin, TX, 2006.
[28] A. Hehr, P. J. Wolcott, M. J. Dapino, Effect of Weld Power and Build Compliance on Ultrasonic Consolidation,
Rapid Prototyping Journal In Press.
[29] J. M. Gibert, Dynamics of Ultrasonic Consolidation, Ph.D. thesis, Clemson University, Clemson, SC, USA (2009).
[30] K. Graff, Power Ultrasonics: Ch6: Power ultrasonic transducers: principles and design, Elsevier, 2015.
[31] E. Neppiras, New Magnetostrictive Materials and Transducers: Part I, Journal of Sound and Vibration 8 (3).
[32] E. A. Neppiras, Motional feed-back systems for ultrasonic transducers, in: Ultrasonics, London, UK, 1971.
[33] Dukane Communication, private communication (2015).
[34] F. Hunt, Electroacoustics: The Analysis of Transduction and Its Historical Backaground, Cambridge: Harvard
University Press, 1954.
[35] W. P. Mason, Electromechanical Transducers and Wave Filters, D. Van Nostrand Company, Inc., 1942.
[36] A. J. Hehr, Process Control and Development for Ultrasonic Additive Manufacturing with Embedded Fibers, Ph.D.
thesis, The Ohio State University, Columbus, OH, USA (2016).
[37] IEEEl, IEEE Std 176-1987: Standard on Piezoelectricity (1988).
[38] AE Techron, Technical Manual: LVC5050 Power Supply Amplifier (2002).
[39] E. De Vries, Mechanics and mechanisms of ultrasonic metal welding, Ph.D. thesis, The Ohio State University,
Columbus, OH (2004).
[40] R. Friel, Power Ultrasonics: Ch13: Power ultrasonics for additive manufacturing and consolidation of materials,
Elsevier, 2015.
[41] F. Balle, G. Wagner, D. Eifler, Characterization of the Ultrasonic Welding Process through High-resolution Laser-

43
 Doppler Vibrometry, Polytec InFocus Magazine 1.
[42] C. de Silva, Vibration: Fundamentals and Practice, CRC Press LLC, 2000.
[43] T. K. Caughey, Sinusoidal Excitation of a System With Bilinear Hysteresis, Journal of Applied Mechanics 27 (4).
[44] W. Iwan, A Distributed-Element Model for Hysteresis and Its Steady-State Dynamic Response, Journal of Applied
Mechanics 33 (4).

44
Paper Highlights

• Presents a new linear time invariant (LTI) model for describing UAM system dynamics. The
model utilizes electroacoustics theory and incorporates welder shear force into the model.
• Characterization of welder with FRF estimation and in-situ measurements. These measurements
are used to calibrate the presented LTI model.
• Demonstration of model’s attributes by analyzing the observed frequency shift during welding,
sonotrode deformation behavior, and a generalized loading condition during welding.
• Use of LTI model to describe observed harmonic content and shear force behavior during UAM.
• Estimation of shear force magnitude and welder efficiency for welding Al 6061 foil. Shear force
magnitude and efficiency has never been estimated for the UAM process prior to this work.
• Analysis of the impact of changing UAM processing conditions, i.e., welder amplitude, down
force during welding, and welder rolling speed, on welder shear force and efficiency.
• Connection of welder shear force to average electric power draw.