Quality-driven Volcanic Earthquake Detection using Wireless Sensor - - PowerPoint PPT Presentation

▶
quality driven volcanic earthquake detection using
SMART_READER_LITE
LIVE PREVIEW

Quality-driven Volcanic Earthquake Detection using Wireless Sensor - - PowerPoint PPT Presentation

Dec 17, 2023 27 likes •579 views

Quality-driven Volcanic Earthquake Detection using Wireless Sensor Networks Rui Tan 1 Guoliang Xing 1 Jinzhu Chen 1 Wen-Zhan Song 2 Renjie Huang 3 1 Michigan State University 2 Georgia State University 3 Washington State University Active Volcano

slide-1
SLIDE 1

Quality-driven Volcanic Earthquake Detection using Wireless Sensor Networks

Rui Tan1 Guoliang Xing1 Jinzhu Chen1 Wen-Zhan Song2 Renjie Huang3

1Michigan State University 2Georgia State University 3Washington State University

slide-2
SLIDE 2

Active Volcano Monitoring

Eruptions in Iceland, 2010 [Wikipedia] OASIS system on Mt St Helens [Song 2009]

  • Traditional broadband seismometers

– Expensive (~ $10K), difficult to install, small-scale

  • Wireless sensor networks on volcanoes

– Low cost (< $200), easy deployment, large-scale

slide-3
SLIDE 3

State-of-the-Art WSN Solutions

  • Centralized earthquake detection

– Energy-consuming data collection – Long latency

6 mins to transmit 1 min seismic data [Werner-Allen 2006]

  • Heuristic event-triggered data collection

– Low detection probability

Only 5% [Werner-Allen 2006]

slide-4
SLIDE 4

Goals

  • Quality-driven earthquake monitoring

– Assured false alarm rate & detection prob.

  • Real-time detection

– Temporal resolution: 1s

  • Long network lifetime

– Avoid raw data transmission

slide-5
SLIDE 5

Challenge 1: Spatial Diversity

  • Complicated physical process

Two earthquakes on Mt St Helens

slide-6
SLIDE 6

Challenge 1: Spatial Diversity

  • Complicated physical process

– Highly dynamic magnitude

Two earthquakes on Mt St Helens

slide-7
SLIDE 7

Challenge 1: Spatial Diversity

  • Complicated physical process

– Highly dynamic magnitude – Dynamic source location

Two earthquakes on Mt St Helens

slide-8
SLIDE 8

Challenge 2: Frequency Diversity

  • Responsive to P-wave within [1 Hz, 10 Hz]
slide-9
SLIDE 9

Challenge 2: Frequency Diversity

  • Responsive to P-wave within [1 Hz, 10 Hz]

[1 Hz, 5 Hz] Signal energy: X 10000

slide-10
SLIDE 10

Challenge 2: Frequency Diversity

  • Responsive to P-wave within [1 Hz, 10 Hz]

[1 Hz, 5 Hz] [5 Hz, 10 Hz] Signal energy: X 10000 X 100

slide-11
SLIDE 11

Challenge 2: Frequency Diversity

  • Responsive to P-wave within [1 Hz, 10 Hz]
  • Freq. spectrum changes with signal magnitude

[1 Hz, 5 Hz] [5 Hz, 10 Hz] Signal energy: X 10000 X 100

slide-12
SLIDE 12

Approach Overview

BS seismic sensor

slide-13
SLIDE 13

Approach Overview

BS seismic sensor

slide-14
SLIDE 14

Approach Overview

BS seismic sensor

slide-15
SLIDE 15

Approach Overview

  • Select sensors with best signal qualities

– FFT (computation-intensive)

BS

sensor selection

FFT FFT FFT

seismic sensor

slide-16
SLIDE 16

Approach Overview

  • Select sensors with best signal qualities

– FFT (computation-intensive)

  • Local detection

BS

sensor selection

seismic sensor

‘1’ ‘0’ ‘1’

slide-17
SLIDE 17

Approach Overview

  • Select sensors with best signal qualities

– FFT (computation-intensive)

  • Local detection
  • Decision fusion

BS

decision fusion system decision

seismic sensor

‘1’ ‘0’ ‘1’

slide-18
SLIDE 18

Approach Overview

  • Select sensors with best signal qualities

– FFT (computation-intensive)

  • Local detection
  • Decision fusion

BS

decision fusion system decision

seismic sensor

‘1’ ‘0’ ‘1’

avoid raw data transmission

slide-19
SLIDE 19

Outline

  • Motivation
  • Frequency-based local detection model

– Preserve essential features – Accurately detect earthquakes

  • Sensor selection for decision fusion
  • Evaluation
  • Conclusion
slide-20
SLIDE 20

Multi-scale Frequency Model

seismic data signal energy frequency spectrum X magnitude scale p log10( ) detection decision

slide-21
SLIDE 21

Multi-scale Frequency Model

  • Gaussian models

– Earthquake happens

... 2, 1, ), , ( ~  p C m N X

p p p

seismic data signal energy frequency spectrum X magnitude scale p log10( ) detection decision

slide-22
SLIDE 22

Multi-scale Frequency Model

  • Gaussian models

– Earthquake happens

... 2, 1, ), , ( ~  p C m N X

p p p

mean vector covariance matrix seismic data signal energy frequency spectrum X magnitude scale p log10( ) detection decision

slide-23
SLIDE 23

Multi-scale Frequency Model

  • Gaussian models

– Earthquake happens – No earthquake

) , ( ~ C m N X

... 2, 1, ), , ( ~  p C m N X

p p p

mean vector covariance matrix seismic data signal energy frequency spectrum X magnitude scale p log10( ) detection decision

slide-24
SLIDE 24

Local Detection at Sensor

  • Minimum error rate detection

– If gp(X) > g0(x), decide 1; otherwise, decide 0

decision function gi(X|mi, Ci)

slide-25
SLIDE 25

Local Detection at Sensor

  • Minimum error rate detection

– If gp(X) > g0(x), decide 1; otherwise, decide 0

decision function gi(X|mi, Ci)

  • Error rate decreases with p
  • Sensors receive different p’s

spatial/frequency diversities

slide-26
SLIDE 26

Outline

  • Motivation
  • Frequency-based local detection model
  • Sensor selection for decision fusion

– Avoid unnecessary FFT

  • Evaluation
  • Conclusion
slide-27
SLIDE 27

Decision Fusion at BS

  • Extended majority rule

> threshold, decide 1 # of positive local decisions total # of sensors

slide-28
SLIDE 28

Decision Fusion at BS

  • Extended majority rule
  • Closed-form detection performance

> threshold, decide 1 # of positive local decisions total # of sensors PF = f ( PF1, PF2, …, PFN ) PD = f ( PD1, PD2, …, PDN )

PFi / PDi : false alarm rate / detection prob. of sensor i

slide-29
SLIDE 29

Sensor Selection For Decision Fusion

Given {PFi, PDi | i=1, …, N}, find a sensor subset S

   

D F

P P S s.t. imize min

slide-30
SLIDE 30

Sensor Selection For Decision Fusion

  • Exclude sensors w/ low signal qualities

– Avoid unnecessary FFT

Given {PFi, PDi | i=1, …, N}, find a sensor subset S

   

D F

P P S s.t. imize min

slide-31
SLIDE 31

Sensor Selection For Decision Fusion

  • Exclude sensors w/ low signal qualities

– Avoid unnecessary FFT

  • Configurable system detection performance

Given {PFi, PDi | i=1, …, N}, find a sensor subset S

   

D F

P P S s.t. imize min

slide-32
SLIDE 32

Sensor Selection Algorithm

  • Select sensor every detection period
  • Brutal-force search: O(2N)

– Long latency

A fast near-optimal algorithm?

slide-33
SLIDE 33

Sensor Selection Algorithm

  • Select sensor every detection period
  • Brutal-force search: O(2N)

– Long latency

PD increases with Σi(PDi-PFi) w.h.p.

              

  

 i Di Di i Fi Di i Fi Fi D

P P P P P P Q Q P

2 2 1

) ( ) (

slide-34
SLIDE 34

Sensor Selection Algorithm

  • Select sensor every detection period
  • Brutal-force search: O(2N)

– Long latency

  • Sort sensors by (PDi-PFi), include one by one

PD increases with Σi(PDi-PFi) w.h.p.

              

  

 i Di Di i Fi Di i Fi Fi D

P P P P P P Q Q P

2 2 1

) ( ) (

slide-35
SLIDE 35

Outline

  • Motivation
  • Frequency-based local detection model
  • Dynamic sensor selection for decision fusion
  • Evaluation

– Testbed experiments – Trace-driven simulations

  • Conclusion
slide-36
SLIDE 36

Implementation

  • Testbed experiments in lab

– 24 TelosB motes

slide-37
SLIDE 37

Implementation

  • Testbed experiments in lab

– 24 TelosB motes

  • Data acquisition

– Seismic data from Mt St Helens -> mote flash – Real-time data acquisition @ 100 Hz

slide-38
SLIDE 38

Implementation

  • Testbed experiments in lab

– 24 TelosB motes

  • Data acquisition

– Seismic data from Mt St Helens -> mote flash – Real-time data acquisition @ 100 Hz

  • On-mote seismic processing

– FFT: < 250 ms over 100 data points

slide-39
SLIDE 39

Baseline Approaches

slide-40
SLIDE 40

Baseline Approaches

  • Centralized processing

– Data collection w/ compression – Up to 4-fold data volume reduction

slide-41
SLIDE 41

Baseline Approaches

  • Centralized processing

– Data collection w/ compression – Up to 4-fold data volume reduction

  • STA/LTA

– Heuristic seismic detection algorithm [Endo 1991] – ≥ 30% sensors decide ‘1’, download 1 min data [Werner-Allen 2006]

slide-42
SLIDE 42

Baseline Approaches

  • Centralized processing

– Data collection w/ compression – Up to 4-fold data volume reduction

  • STA/LTA

– Heuristic seismic detection algorithm [Endo 1991] – ≥ 30% sensors decide ‘1’, download 1 min data [Werner-Allen 2006]

  • Weighted decision fusion [Chair&Varshney 1990]

– Account for signal qualities – Sensor selection is not necessary

slide-43
SLIDE 43

Energy Usage

Total energy consumption 12 motes 10 minutes

slide-44
SLIDE 44

Energy Usage

Total energy consumption 12 motes 10 minutes Centralized process.: transmit seismic data

slide-45
SLIDE 45

Energy Usage

Total energy consumption 12 motes 10 minutes Weighted fusion: always do FFT

slide-46
SLIDE 46

Energy Usage

  • 6-fold reduction in energy consumption

Total energy consumption 12 motes 10 minutes

19 days 3.9 months

slide-47
SLIDE 47

Trace-driven Simulations

  • Data traces from 12 sensors on Mt St Helens
  • More than 128 earthquakes in 6 months
slide-48
SLIDE 48

Trace-driven Simulations

  • Data traces from 12 sensors on Mt St Helens
  • More than 128 earthquakes in 6 months

7 3

slide-49
SLIDE 49

Trace-driven Simulations

  • Data traces from 12 sensors on Mt St Helens
  • More than 128 earthquakes in 6 months

Configurable trade-off btw detection performance and energy consumption 7 3

slide-50
SLIDE 50

Conclusions

  • Quality-driven earthquake detection

– In-network collaborative signal processing – No raw data transmission

  • Near-optimal sensor selection algorithm

– Handle earthquake dynamics – Minimize energy consumption

  • Extensive evaluation

– 6-fold energy reduction – Comparable sensing performance

slide-51
SLIDE 51

Thank you!

slide-52
SLIDE 52

Dynamic Detection Performance

  • Spatial-temporal variation
slide-53
SLIDE 53

Dynamic Detection Performance

  • Spatial-temporal variation

– Sensors have diverse magnitude scales

slide-54
SLIDE 54

Dynamic Detection Performance

  • Spatial-temporal variation

– Sensors have diverse magnitude scales – Each sensor has unpredictable pattern of p

slide-55
SLIDE 55

Ongoing Work

  • Earthquake source localization

– Accurate node-level onset time

Onset times of an explosive event An existing p-wave front picking algorithm

  • millisec accuracy
  • 7KB ROM + 13KB RAM

(unavailable on TelosB)