V - DIEES

Transcript

V - DIEES
Selective Measurement of volcanic
ash flow-rate
Methodologies of sensing for the volcanic ash monitoring
Bruno Andò, Salvatore Baglio, Vincenzo Marletta
DIEEI – University of Catania, Italy
[email protected]
May 6-9, 2013
The problem…
The atmospheric dispersion of ash produced by the explosive activity of the volcano Etna is a relevant factor of risk for
eastern Sicily and in particular for the Catania area. The ash fall-out following explosion activity of volcanoes represents
a factor of risk for people and a serious hazard for air traffic mainly when airports are close to active volcanoes. This is
the case of the Fontanarossa airport in Catania close to Mount Etna. During periods of activity ash clouds cause extensive
damages to roads, sanitation systems, agriculture, health and daily activities of people living in the countries on the
slopes of the volcano, but also a substantial hazard factor to air traffic.
The project…
Project
The
SECESTA: reti di SEnsori per il monitoraggio delle CEneri vulcaniche nella Sicurezza
del Trasporto Aereo
Is an italian acronym for “A sensor network for the monitoring of volcanic ash fall-out for
the safety of air transport”.
Regione Siciliana
ASSESSORATO REGIONALE DELLE ATTIVITA’
PRODUTTIVE
AVVISO PUBBLICO PER LA CONCESSIONE DELLE
AGEVOLAZIONI IN FAVORE
DELLA RICERCA, SVILUPPO ED INNOVAZIONE
PREVISTE DALL’ART 5 DELLA LEGGE REGIONALE
16.12.2008, N. 23
Linea di intervento 4.1.1.1 - POR FESR
Sicilia 2007-2013
Partners
Etna volcano
3340m sl
Rifugio Sapienza (mountain hut)
(1910m sl)
Nicolosi (700m sl)
Catania (7m sl)
Fontanarossa Airport
Etna volcano
3340m sl
Fontanarossa Airport
The project…
Goal of the project: Development of a early-warning distributed WSN of smart self powered
µcontroller based multisensor nodes for the monitoring of ash fall-out in
the area spreading from the crater of the volcano to the airport to the
purpose to provide information on the ash fall-out phenomena aimed to
foresight the time-space evolution of the phenomena.
DIEEI: development of methodologies of sensing for the monitoring of the following
characteristic parameters:
Ash detection (presence and discrimination from other sediments)
Ash flow rate
Ash average granulometry
Schematization of the prototype
Piezoelectric transducer
Wind sensor
On -board electronics
Rain sensor
100µF
Conditioning
electronic
Digital driver
Wi-Fi
µController
based electronic
Magnetometer
IR diode/phototransistor
Shutter
Permanent
magnet
Real view of the multisensor node prototype developed at the DIEEI
Multi Interface User Board (MuIn)
Microchip PIC18F2520 a 40 MHz
Volcanic ash flow rate measurement
Design of the flow rate sensor
Methodology: array of coupled IR diode – phototransistor to estimate the volume of ash in a
tank
∆V
φ=
[ ml / s ]
∆t
• Dimensions of the convoying structure
• Number of measurements/hour
• Number of IR couples
• Distance between contiguous IR couples
• Time between consecutive tank emptying
Etna
3340m sl
Rifugio Sapienza (1910m sl)
5 – 10 km
0.130mm ÷ 4mm
0.450 kg/m2 /h
Nicolosi (700m sl)
15-20 km
0.060mm ÷ 0.5mm
0.120 kg/m2 /h
Catania (7m sl)
25 – 30 km
0.030mm ÷ 0.125mm
0.090 kg/m2 /h
Design of the flow rate sensor
Rifugio Sapienza
(5 – 10 km)
Nicolosi
(15 – 20 km)
Catania
(25-30 km)
0.130mm ÷ 4mm
0.060mm ÷ 0.5mm
0.030mm ÷ 0.125mm
Given the granulometry g we set:
Prototype of Ergotronica
d2=34 mm
Prototype of DIEEI
d2=10 mm
d2
A2
A1
d1
d 2 ≥ 10 ∗ g
Specs on the ash mass flow (P)
(by INGV)
Rifugio Sapienza
(5 – 10 km)
Nicolosi
(15 – 20 km)
Catania
(25-30 km)
0.450 kg/m2 /h
0.120 kg/m2 /h
0.090 kg/m2 /h
ρ = 1000 kg/m3
Φ=P*A1 [kg/h]
tmax [h]
Vtmax=Φ*tmax [Kg]
A1 =
Vtmax= A2*h*ρ [Kg]
A2 ⋅ h ⋅ ρ
[m 2 ]
P ⋅ tmax
Design of the flow rate sensor
Tank
Height=10 cm
d2= 1 cm
h = 5mm
tmax = 20 s
IR diode
SEP8736 by Honeywell
AlGaAs Infrared Emitting diode
10° (nominal) beam angle
880 nm wavelength
Phototransistor
SDP8436 by Honeywell
NPN silicon phototransistor
18° (nominal) view angle
Design of the flow rate sensor
1.26
1
1.13
0.75
0.98
0.5
0.80
0.25
0.56
1.75
0.1
0.2
0.3
Flow rate [kg/m2*h]
0.4
0.5
9.6
8.4
7.2
3.03
6
2.76
4.8
2.47
3.6
2.14
2.4
1.75
1.2
0.589
0
0
0
0.6
0.1
0.2
0.12
h=10e-3 [m]
5.64
10s
20s 5.35
30s
1m 5.05
3m
5m 4.72
10m
25
22.5
20
17.5
15
4.37
12.5
3.99
10
3.57
7.5
3.09
5
2.52
2.5
1.78
0
0
0.1
0.2
0.3
Flow rate [kg/m2*h]
0.4
0.5
0
0.6
Collector diameter [m]
0
0
Top collector area [m2]
Top collector area [m2]
2
10.8
Top collector area [m2]
1.25
2.25
Top collector diameter [m]
1.5
10s
20s 1.69
30s
1m 1.60
3m
5m 1.49
10m
1.38
3.91
10s
20s 3.71
30s
1m 3.50
3m
5m 3.27
10m
12
1.78
0.3
Flow rate [kg/m2*h]
0.4
0.5
Collector diameter [m]
h=5e-3 [m]
h=1e-3 [m]
2.5
1.24
0.87
0
0.6
Design of the flow rate sensor: volume characterization
calibration diagram
5
4
medium ash size
φ = 0.3 ml/s
5
v = 0.7*x + 0.47
4
2,6
3
1,8
2
1,2
1
3.3
volume(ml)
volume (ml)
3,3
level x (n° of IR diode)
4
2.6
1.8
1.2
medium ash size
φ = 0.3 ml/s
0
0
2
4
6
8
time (s)
10
12
14
0
16
1
2
Uncertainty on the volume estimation worst case σ=0.11ml (~2%)
3
level x (n° of IR diode)
4
5
Design of the flow rate sensor: ash flow rate estimation
0.0312
Flow rate [ml/s]
0.0311
0.031
0.0309
0.0308
0.0307
0.0306
V1/t1
V2/t2
V3/t3
(V1+V2)/(t1+t2)
(V2+V3)/(t2+t3) (V1+V2+V3)/(t1+t2+t3)
The composite uncertainty on the fow-rate φ measurements can be evaluated by applying the propagation of
uncertainty law:
2
2
 ∂φ 
 ∂φ  2
uc2 (φ ) =   u2 (∆V ) + 
 u (∆t ) = 0.025e − 3 ml / s
 ∂V 
 ∂∆t 
Volcanic ash detection
(Selectivity)
Volcanic ash detection: the methodology
Methodology: exploitation of the paramagnetic properties of the volcanic ash
Low-power digital 3-axis magnetometer
MAG3110 by Freescale
• 1.95 V to 3.6 V supply voltage (VDD)
• 1.62 V to VDD IO Voltage (VDDIO)
• Ultra small 2 mm x 2 mm x 0.85 mm, 0.4 mm pitch, 10-pin package
• Full-scale range ±1000 µT
• Sensitivity of 0.10 µT
• Noise down to 0.25 µT rms
• Output Data Rates (ODR) up to 80 Hz
• 400 kHz Fast Mode compatible I2C interface
• Low-power, single-shot measurement mode
Volcanic ash detection
tank
magnetometer
permanent magnet
d2
d1
0.8ml - mag@6cm - ash@0cm
-3360
-3380
-3380
-3400
-3400
Magnetic field [µT]
Magnetic field [µT]
0.6ml - mag@6cm - ash@0cm
-3360
-3420
-3440
-3420
-3440
-3460
-3460
-3480
-3480
-3500
0
0.5
1
1.5
2
2.5
3
time [s]
3.5
4
4.5
5
5.5
-3500
0
0.5
1
1.5
2
2.5
3
time [s]
3.5
4
4.5
5
5.5
Volcanic ash detection
tank
permanent magnet
magnetometer
d2
d1
Volcanic ash detection
tank
magnetometer
permanent magnet
d2
0.8ml
ash ash
0.6
mlof of
80
d1
d1=10cm
d1=9cm
d1=6cm
d1=4cm
70
60
d1=10cm
d1=9cm
d1=6cm
d1=4cm
80
∆H [µT]
50
∆H [µT]
ashash
0.80.6ml
mlofof
100
40
30
60
40
20
20
10
0
0
1
2
3
Prototype of DIEEI
4
d2 [cm]
5
d1 = 5cm
6
7
8
d2 = 0 cm
0
0
1
2
3
4
d2 [cm]
5
6
7
8
Volcanic ash granulometry classification
Design of the granulometry classificator
Methodology: piezoelectric to transduce ash impacts in a voltage signal
Modeling
volcanic ash
45°
Ft
piezoelectric
transducer
Fp
7BB-35-3L0 by Murata
g33 = 22e-3 Vm/N
d = 0.23e-3 m
S=4.155e-4 m2
F
2
 1  5 2 65
5
Hertz law F p = 1.2644 ρ A 
 k + k  R A v A
A 
 P
3
Vout = A ⋅
e −ξωnt
1−ξ
2
A = g 33 F p d S
sin (2πf nt )
Conditioning electronic
Charge amplifier
Cf
Inverting amplifier
Rf
Model of the
piezoelectric
R2
VDD
7
2
+
1
Vout
4
4
3
+
5
R1
-
6
Ceq
-
Req
11
11
VDD
VSS
VSS
Bode Diagram
Magnitude (dB)
-30
-40
-50
-60
-70
•Req = 67 kΩ
•Ceq = 14 nF
Phase (deg)
-80
90
ft =
1
= 6 Hz
2πR f C f
45
0
-1
10
ft < f n
10
0
1
10
Frequency (Hz)
10
2
10
3
4
3
experimental
theoretical
15
experimental
theoretical
3
experimental
theoretical
10
2
1
0
-1
small size
Output voltage Vout (V)
Output voltage Vout (V)
Output voltage Vout (V)
2
1
0
-1
medium size
5
0
-5
big size
-2
-2
-10
-3
-3
0
0,002 0,004 0,006 0,008 0,01
time (s)
0,012 0,014 0,016 0,018
-4
0
0,002 0,004 0,006 0,008 0,01
time (s)
0,012 0,014 0,016 0,018
-15
0
0.8
experimental
theoretical
0.6
2.5
0,012 0,014 0,016 0,018
6
experimental
theoretical
2
0,002 0,004 0,006 0,008 0,01
time (s)
experimental
theoretical
4
1.5
0.2
0
-0.2
-0.4
0.5
0
-0.5
-1
2
0
-2
-1.5
-0.6
-0.8
1
Output voltage Vout (V)
Output voltage Vout (V)
Output voltage Vout (V)
0.4
-4
-2
small size
0,0062
0,0064
time (s)
0,0066
0,0068
0,007
0,0014 0,0016 0,0018
big size
medium size
-2.5
0,006
0,002
0,0022 0,0024 0,0026 0,0028
time (s)
0,003
-6
0.038 0.04 0.042 0.044 0.046 0.048 0.05 0.052 0.054 0.056
time (s)
By fitting model (2) to observed responses for different particle sizes (small, medium, big) the following parameters
have been estimated:
ξ = 0.0098,
fn = 3.49 kHz and
As = 2.30 V, Am = 3.20 V, Ab = 11.5 V for the small, medium and big particles, respectively.
The minimization paradigm adopted is the Nelder–Mead optimization algorithm with the following functional J:
2
 Σ N (V − V
out , pred )
 i out
J= ∑ 
N
size =1 

3




size :
1 = small
2 = medium
3 = big
Functional J is the sum of root mean squares of residuals between observed responses
for the three considered classes of particle size.
Vout
and predicted responses
Vout , pred
35
30
medium
small
big
Th1
Th2
Frequency
25
20
15
10
5
0
0
2
4
6
8
10
Output Voltage (V)
12
14
16
ROC curves theory
The binary classification problem
Let us consider a two-class prediction problem (binary classification)
There are four possible outcomes, given a value for the discrimination threshold :
True Negative, True Positive, False Positive, False Negative
True Positive Rate (TPR) =
True Positives
= Sensitivity
True Positives + False negatives
False Positive Rate (FPR) =
False Positives
= 1 − Specificity
True Negatives + False Positives
The “ideal” binary classification problem
The “real” binary classification problem
There is at least one value for the discrimination
threshold that gives:
TPR = 1, FPR = 0 Sensitivity = Specificity = 1 !
Trade-off between Sensitivity – Specificity
Sometimes we need maximum sensitivity (security)
Sometimes we need maximum specificity (money)
“How can I chose the value of the threshold that gives me the performance I need?”
Sensitivity & Specificity
ROC curves
ROC curves theory
ROC Curves: definition
In signal detection theory, a receiver operating characteristic (ROC), or
simply ROC curve, is a graphical plot of the TPR (sensitivity) vs. FPR (1 - specificity) for
a binary classifier system as its discrimination threshold is varied.
Negatives
Positives
Procedure to draw a ROC curve
1) Fixing a target and collecting the output
signal evolution;
N
P
2) Plotting the output signal distribution in the
presence and in the absence of target;
3) For different values of the discrimination
threshold, estimating:
the TPR, as the portion of the output signal
distribution, in the presence of a target, above
the threshold.
the FPR, as the portion of the output signal
distribution, in the absence of a target, above
the threshold.
ROC curves theory
ROC Curves: two particular cases
Negatives
Positives
N
AUC = 1
P
RTD
Negatives
Positives
RTD
AUC = 0.6351
ROC analysis as a methodology for Ash granulometry classification
1
0.9
True Positive Rate (sensitivity)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
ROC curve
no discrimination line
0.1
0
0
TP
TPR =
= Sensitivit y
TP + FN
0.2
FPR =
0.4
0.6
False Positive Rate (1-specificity)
0.8
FP
= 1 − Specificity
TN + FP
1
ROC curves for the volcanic ash granulometries classification
ROC curve for medium-large granulometries
1
0.9
0.9
0.8
0.8
True Positive Rate (Sensitivity)
True Positive Rate (Sensitivity)
ROC curve for small-medium granulometries
1
0.7
0.6
0.5
0.4
0.3
small-medium ash size
worst case
0.2
0.1
0
0
0.7
0.6
0.5
0.4
0.3
0.2
medium-large ash size
worst case
0.1
0.1
0.2
0.3
0.4
0.5
0.6
0.7
False Positive Rate (1-Specificity)
0.8
0.9
1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
False Positive Rate (1-Specificity)
0.8
0.9
1
Thank You for Your Kind Attention!