Detecting mechanical vibrations in
superconducting magnets for quench detection
and diagnostics
M. Marchevsky,
Lawrence Berkeley National Laboratory
Credits: P. Roy, X. Wang, G. Sabbi, S. Prestemon,
LBNL
*supported by the Director, Office of Science, High Energy Physics, U.S. Department of Energy under contract
No. DE-AC02-05CH11231
M. Marchevsky --- WAMSDO, CERN 2013
1
Outline
 Motivation for mechanical vibration (acoustic) sensing and
earlier developments in the field
 Magnets as mechanical resonators
 Instrumentation for acoustic sensing
 Case study: correlation of acoustic and voltage imbalance
signals in the recent HQ magnet test (HQ01e3)
 Inductive sensing of mechanical vibrations and conductor
motion
 Future plans
M. Marchevsky --- WAMSDO, CERN 2013
2
Why acoustic sensing?
• Voltage taps: this approach is not optimal for longer magnets and may be not
viable in newer complex magnet geometries (multi-layers, etc.)
• Magnetic quench antennas: data requires significant post-processing; permanent
access to the bore or adaptation to the magnet geometry is needed
Advantages of sensing sounds for magnet diagnostics:
- Propagation velocity is large (several km/s), so that detection can be
accomplished on a time scale that is comparable (or faster) to other techniques
- Using sensor arrays, sound sources can be localized with a few cm accuracy
through triangulation
- Selectivity for different kinds of events, through frequency and phase analysis
- Outer surfaces sensor mounting for non-intrusive detection
- Immunity to magnetic fields
- Sensors and acquisition hardware are relatively inexpensive, portable and easily
adaptable to various magnet configurations
M. Marchevsky --- WAMSDO, CERN 2013
3
Earlier developments
Dislocation motion and micro-plasticity -> technical superconductors stability ->
superconducting magnets training -> active acoustic monitoring of SC magnets
•
•
•
•
•
•
•
•
P. P. Gillis, “Dislocation motion and acoustic emission”, ASTM STP 505, 20-29, 1972
“Dynamic stress effects in technical superconductors and the "training" problem of
superconducting magnets”, G. Pasztor and C. Schmidt, J. Appl. Phys. 49, 886 (1978)
H. Brechna and P. Turowski, “Training and degradation phenomena in superconducting magnets,”
Proc. 6th Intl. Conf. Magnet Tech. (MT6) (ALFA, Bratislava, Czechoslovakia) 597, (1978).
“Acoustic emission from NbTi superconductors during flux jump”, G. Pasztor and C. Schmidt,
Cryogenics 19, 608 (1979).
“Sources of acoustic emission in superconducting magnets”, O. Tsukamoto and Y. Iwasa, J. Appl.
Phys. 54, 997 (1983).
“Discussion on acoustic emission of a superconducting solenoid”, M. Pappe, IEEE Trans. on Magn.,
19, 1086 (1983)
“Acoustic emission monitoring results from a Fermi dipole”, O.O. Ige, A,D. Mclnturf and Y. Iwasa,
Cryogenics 26, 131, (1986)
“Mechanical Disturbances in Superconducting Magnets-A Review”, Y. Iwasa, IEEE Trans on Magn,
28 113 (1992)
M. Marchevsky --- WAMSDO, CERN 2013
4
Sound generation in superconducting magnets
Singular events
• Sudden mechanical motion of a cable portion or coil part
• Cracking / fracture of epoxy, de-laminations, etc...
Potentially, also:
• flux jump, as current re-distribution in the cable leads to the local
variation of the electromagnetic force
• quench development, as formation of a hot spot leads to the local
thermal expansion. It that leads to the change in local stress that
propagates away with a speed of sound
“Singular events” are mostly associated with well-localized sources. They generate
longitudinal (pressure) waves that propagate radially from the source with a speed of
sound. Wave fronts then gets partially reflected by the boundaries, converted into
resonant vibrational modes of the structure and into heat.
Continuous perturbations
• Mechanical vibrations (various flexural, hoop, “breathing” and other deformation
modes of coils, shell and support structures)
• Background noise (helium boiling, cryostat vibrations, etc.)
M. Marchevsky --- WAMSDO, CERN 2013
5
Coils as mechanical resonators
Long coils can be thought of as solid “bars” or “rods”
f1L
f2L
f3L
f4L
Longitudinal
(pressure wave)
modes:
Transverse
(flexural) modes
of “free” rods:
Dx
Vs
𝑉𝑠 =
where Y is Young
modulus and r is
the density
𝑓𝑛𝐿 = 𝑛𝑉𝑠 /2L
𝑓1𝑇 ≈ 1.028
f1T
𝑎 𝑌
𝐿2 𝜌
where L is the
length and a is the
cross-sectional area
f2T=2.76 f1T
f3T=5.40 f1T
Polarized
piezo-ceramics
𝑌
𝜌
𝑓𝑛𝐿 ≈ 0.441 𝑛 +
S1
1 2
𝑓1𝑇
2
S2
D
L
M. Marchevsky --- WAMSDO, CERN 2013
Vac
6
Localization of the sound source
1D
Sensor 1
Vs
Vs
t=0
t1
- L/2
Sensor 2
t2
Dx
0
Dt12=(t2-t1) = 2Dx/𝑽𝒔
L/2
2D
(x0-x1)2+(y0-y1)2=R12
Sensor 1
(x0-x2)2+(y0-y2)2=R22
(x0,y0)
R1
(x0-x3)2+(y0-y3)2=R32
R3
|R1-R2|=𝑉𝑠 Dt12
|R1-R3|=𝑉𝑠 Dt13
R2
Sensor 2
Think GPS !
M. Marchevsky --- WAMSDO, CERN 2013
Sensor 3
7
Instrumentation
Cryogenic preamplifier
20 mm
Piezosensor
D
D
•
•
•
.
SM118 type piezoelectric ceramics,
polarized across thickness
OD 10 mm x ID 5 mm x Thickness 2 mm,
fr = (154 ± 4) kHz
DAQ
•
•
•
• GaAs MOSFET-based amplifier
• Linear bandwidth of 0-100 kHz
• 300 -1.9 K operation temperature range
Converts impedance down to ~1 kW, significantly
improves S/N ratio, allows use of regular “twisted pair”
connections in the cryostat instead of the coaxes
Software
Yokogawa WE7000
simultaneous multichannel DAQ system
100 mV-100 V range
up to 1 MHz speed
M. Marchevsky --- WAMSDO, CERN 2013
LabView-based software
for waveform analysis,
re-sampling and
location triangulation of
the sound source
8
Localization tests at RT using HQ Coil 14
R
L
Dt=0.24 ms
Dt=0.01 ms
Dt=0.18 ms
Sound speed: Vs~ 4.1-4.3 km/s
M. Marchevsky --- WAMSDO, CERN 2013
9
Installation on the HQ magnet
2
Sensor 2 is installed at the top plate
(bolted to the magnet shell)
~130 cm
1
“Available” (not optimized) locations were used
Sensor 1 is installed at the bottom
load plate (bolted to the axial rods)
M. Marchevsky --- WAMSDO, CERN 2013
10
Response to a shell excitation
Bottom sensor
Top sensor
0
-40
-80
0
20
40
60
80
100
rms
Power spectrum (mV
Bottom sensor
Top sensor
40
/Hz)
15
rms
(dB) /Hz)
spectrum
Powerspectrum
Power
ln(mV
HQ vibrations resonant spectrum (room temp.)
10
5
f (kHz)
0
2000
Response to a rod/load plate excitation
4000
6000
f (Hz)
/Hz)
rms
Power spectrum (mV
(dB)
Power
Power spectrum
spectrum (dB)
Bottom sensor
Top sensor
0
-40
-80
0
20
40
60
f (kHz)
80
10000
1800
Bottom sensor
Top sensor
1600
40
8000
100
1400
1200
1000
800
600
400
200
0
0
M. Marchevsky --- WAMSDO, CERN 2013
2000
4000
6000
f (Hz)
8000
10000
11
Extraction at 5.5 kA
Original sound
slowed down 10 times
When current is extracted from the magnet, sounds are recorded (step-like change of
elastic strain?), followed by a prolonged (0.5-1 s) “ringing” of the structure at its
resonance modes, with occasional “bursts” of mechanical activity (thermal relaxation?)
Magnet is a good mechanical resonator with Q~100!
M. Marchevsky --- WAMSDO, CERN 2013
12
HQ experimental setup
 Magnet imbalance signal is formed by subtracting negative half (Coils 5,7) from
the negative (Coils 8,9) of the magnet, then amplified x40.
 Sound signals from both sensors, magnet imbalance and magnet current are
recorded at 1 Ms/s; the time window is 0.2 s.
 Acquisition is triggered when either imbalance or sound is above the threshold
level.
Imag
Imag
Iq=10870 A
9 kA
75 A/s
t
75 A/s
t
Attempted current ramps
M. Marchevsky --- WAMSDO, CERN 2013
13
Events during an up-down current ramp
RR03: Ramp up to 9 kA and back down
Sound triggered
Imbalance triggered
Sound and imbalance
Sound triggered
Imbalance triggered 250
Sound and imbalance
10000
200
(mV)
max
6000
4000
150
100
U
I (A)
8000
50
2000
0
0
0
50
100
150
200
250
300
0
50
100
150
200
250
300
t (s)
t (s)
Current of the triggered events
Sound amplitude of the triggered events
Threshold settings:
• Sound: 5 mV
• Imbalance: 3 V (amplified; true imbalance threshold is ~75 mV)
M. Marchevsky --- WAMSDO, CERN 2013
14
Events during a current ramp into quench
RR04: Ramp to quench at 75 A/s
Iq=10870 A
Current at the triggered events
Sound amplitude at the triggered events
Sound triggered
Imbalance triggered
Sound and imbalance
12000
10000
500
400
, (mV)
300
200
max
6000
U
8000
I (A)
Sound triggered
Imbalance triggered
Sound and imbalance
4000
100
2000
0
0
0
100
200
300
400
0
100
t (s)
200
300
400
t (s)
Four possible scenarios are observed:
1. Imbalance variation without any associated sound (below 5 kA)
2. Imbalance variation associated with weak sound signals (below 5 kA)
3. Stronger sounds without association with imbalance variations (above 8.5 kA)
4. Stronger sounds associated with imbalance “spikes” (around 10-10.5 kA)
M. Marchevsky --- WAMSDO, CERN 2013
15
The 75A/s quench
C5
C9
C7
C8
Quench starts in the outer layer multi-turn of C5
M. Marchevsky --- WAMSDO, CERN 2013
16
Sounds of magnet (low current)
• Some imbalance variations at low currents are associated with (weak) sounds!
2440 A
1883 A
M. Marchevsky --- WAMSDO, CERN 2013
17
Sounds at higher magnet current
Much stronger sounds are observed,
that are either not correlated with any
imbalance variations:
9628 A
or, occasionally, are correlated with a
short “spike” in the imbalance signal:
M. Marchevsky --- WAMSDO, CERN 2013
10036 A
18
Origin of the sounds?
2440 A
10036 A
10
12
Sensor 1
Sensor 2
Imbalance
5
10
10
0.4
U
-5
2
0.0
0.5
1.0
1.5
2
-0.2
0
-0.4
-2
0.631 ms
-10
0.0
4
U
snd
snd
(V)
0.2
(V)
4
6
imb
6
8
U
, (mV)
8
0
Sensor 1
Sensor 2
Imbalance
0.6
0.112 ms
0
2.0
-0.6
0.5
1.0
1.5
-4
t (ms)
t (ms)
The 0.63 ms delay corresponds to the
~2.6 m distance, which would be outside
of the magnet length. The sound is likely
produced during the (long) imbalance
variation, but not at its onset.
Current re-distribution in the cable triggers
sound?
The 0.11 ms delay corresponds to ~0.46 m
distance => the sound is produced within
the magnet length.
Mechanical motion event is triggering
the imbalance?
M. Marchevsky --- WAMSDO, CERN 2013
19
Frequency of the sounds
10248 A
0.8
60
0.7
50
Sensor 1 (bottom)
Sensor 2 (top)
2/Hz
rms
30
mV
f (kHz)
40
20
Sensor 1 (bottom)
Sensor 2 (top)
0.6
0.5
0.4
0.3
0.2
10
0.1
0
0
100
200
300
400
0.0
0
10
t (s)
20
30
40
50
f (kHz)
60
70
80
Very high frequency sound is detected at
Imag>10 kA
(Sound slowed down 10x)
M. Marchevsky --- WAMSDO, CERN 2013
20
Location of the sound sources
10036 A
0.30 ms
X= - 63 cm (at the bottom end)
It appears that the sources of strong sound generated in HQ01e3 ramps above 9 kA are
located near the bottom (return end) of the magnet
M. Marchevsky --- WAMSDO, CERN 2013
21
“Sound” from the Quench Antennas
HQ01d magnet, ramping at 100 A/s
Flux jumps
Mechanical vibrations
Inductive quench antenna is an electromagnetic
microphone! It picks up vibrations of the
current-carrying (or magnetized) structures
M. Marchevsky et al., ASC 2012 presentation
By correlating EM QA and piezo-sensor signals,
one can potentially differentiate between flux
jumps, conductor motion and other mechanical
motion in magnets
M. Marchevsky --- WAMSDO, CERN 2013
22
Future testing opportunities
Upcoming test of the high-field dipole magnet HD3 at LBL:
We plan to have both, inductive QA and the piezo-sensors installed.
QA
PS
Proposed positioning of the piezosensors on the magnet: four wedges that
are in direct mechanical contact with the windings
Upcoming test of the LARP HQ02 magnet:
At least two acoustic sensors can be installed on the endplates; some kind of
inductive pickup QA may also be installed, t.b.d.
M. Marchevsky --- WAMSDO, CERN 2013
23
Next steps / challenges
• Filtering out the resonant modes and improving selectivity for small
signals
• Developing microphone arrays and algorithms for precise localization
• Quantifying mechanical energy release and conductor motion
amplitudes observed with piezo and EM sensors
• Acoustic quench detection system?
M. Marchevsky --- WAMSDO, CERN 2013
24
Conclusions
• Amplified piezosensors, in combination with cryo-electronics, modern data
acquisition and processing techniques show good potential for real-time
characterization of various mechanical events in superconducting magnets
during ramping, quench and recovery
• HQ magnet produces increased acoustic emissions (seemingly unrelated to FJ)
and high-frequency (>50 kHz) vibration “bursts” when energized above 9kA.
The latter are occasionally correlated with the short imbalance spikes and most
likely caused by stick-slip motion of the conductor
• Inductive pickups sensors they provide a unique insight into conductor motion;
can be developed and used in conjunction with acoustic devices to improve
selectivity for the specific mechanical and electrical events
Listening to magnets sounds like fun!
M. Marchevsky --- WAMSDO, CERN 2013
25