stationarity and degree of stationarity norden huang research center for adaptive data analysis...

55
Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Post on 19-Dec-2015

223 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Stationarity and

Degree of Stationarity

Norden Huang

Research Center for Adaptive Data Analysis

National Central University

Page 2: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Need to define the Degree Stationarity

• Traditionally, stationarity is taken for granted; it is given; it is an article of faith.

• All the definitions of stationarity are too restrictive.

• All definitions of stationarity are qualitative.• Good definition need to be quantitative to

give a Degree of Stationarity

Page 3: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Definition : Strictly Stationary

2

1 2 n 1 2 n

For a random var iable x( t ), if

x( t ) , x( t ) m, and that

x( t ), x( t ), ... x( t ) and x( t ), x( t ), ... x( t )

have the same joi nt distribution for all .

Page 4: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Definition : Wide Sense Stationary

2

1 2 1 2

1 2 1 2

For any random var iable x( t ), if

x( t ) , x( t ) m, and that

x( t ), x( t ) and x( t ), x( t )

have the same joi nt distribution for all .

Therefore , x( t ) x( t ) C( t t ) .

Page 5: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Definition : Statistically Stationary

• If the stationarity definitions are satisfied with certain degree of averaging.

• All averaging involves a time scale. The definition of this time scale is problematic.

Page 6: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Degree of Stationarity

t

2T

0

For a time frequency distribution, H( ,t ),

1n( ) H( ,t ) dt ;

T

1 H( ,t )DS( ) 1 dt .

T n( )

Page 7: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Degree of Statistical Stationarity

t

2Tt

0

For a time frequency distribution, H( ,t ),

1n( ) H( ,t ) dt ;

T

H( ,t )1DS( , t ) 1 dt .

T n( )

Page 8: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

An Example

Ocean Wind Wave Data

Page 9: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves

• Water waves are nonlinear.

• Crests of breaking waves need many harmonics to fit

• Waves are nonstationary

• Spectrum full of Harmonics; it is hard to separate free from bound wave energy

Page 10: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves : data

Page 11: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves : IMF

Page 12: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves : Hilbert Spectrum

Page 13: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves : Hilbert Spectrum x10

Page 14: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves : Hilbert Spectrum x100

Page 15: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Ocean Waves : Degree of Stationarity

Page 16: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

An Example

Earthquake DataChi-Chi, Taiwan

September 21, 1999

Huang, N. E. , et al. 2001 : A new spectral representation of earthquake data: Hilbert Spectral analysis of station TCU129, Chi-Chi, Taiwan, 21 September 1999, Bulletin of the Seismological Society of America, Volume 91, pp 1310-1338.

Page 17: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Earthquake

• Earthquake is definitely transient; therefore, nonstationary.

• For near field locations, the earth motion is also highly nonlinear.

• Traditional treatment of earthquake data by response spectral analysis is not adequate.

Page 18: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Response Spectrum

The response spectrum of a earthquake signal is defined through the maximum displacement of a linear single degree of freedom system with predetermined damping driven by the given earthquake signal. The displacement is given by the Duhamel Integral:

n

t( t )

n dd 0

n

2 1 / 2d n

1( t , , ) a( )e sin ( t )d ,

where

a( ) is the earthquake acceleration signal ,

is the undamped natural system frequency,

is the damping factor , and

( 1 ) is the damped system feequency.

Page 19: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Response Spectrum

n n

t

n nn 0

ti t i

n 0

n nn

For 0 , we have

1( t ; ) a( ) sin ( t ) d

1Im e a( )e d

1F( , t ) sin( t ) .

Page 20: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Response Spectrum

n n max n e nn

e

2e n n max n

Therefore ,

1F( ) ( ) A ( ) ,

where A , the equivalent acceleration is defined as

A ( , ) ( , ) .

Page 21: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Response Spectrum

• As the Duhamel Integral gives a quantity with the dimension of velocity, the response spectrum is also known as the pseudo-velocity spectrum.

• The linear single degree of freedom system is a linear filter; therefore,

• There is a definitive relationship between the Fourier Spectrum and Response spectrum.

Page 22: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : Data

Page 23: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : F & RS ; E

Page 24: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : F & RS ; N

Page 25: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : F & RS ; Z

Page 26: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : Hilbert E

Page 27: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : Hilbert N

Page 28: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : Hilbert Z

Page 29: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : MH & F : E

Page 30: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : MH & F : N

Page 31: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : MH & F : Z

Page 32: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : Hilbert : E200

Page 33: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : Hilbert E1000

Page 34: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DS E

Page 35: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DS N

Page 36: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DS Z

Page 37: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DS All

Page 38: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DSS200 All

Page 39: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DSS1000 All

Page 40: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Chi-Chi Earthquake : DS

• Hilbert spectral analysis reveals a ‘damaging-causing’ low frequency band of energy not properly shown in the Fourier Analysis.

• The strongest component, EW, is also the most nonstationary one.

• The weakest component, Z, is also the most stationary one.

• The Hilbert and Fourier spectra agree well for the most stationary case.

Page 41: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Heart Rate Variability : HRV

Normal heart rate is chaotic

Page 42: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Quiz on physiologic dynamics

Heart Failure Heart Failure

Normal Atrial Fibrillation

• Loss of dynamical fluctuations is bad

• Not all dynamical fluctuations are good

Hea

rt R

ate

(bpm

)H

eart

Rat

e (b

pm)

Hea

rt R

ate

(bpm

)H

eart

Rat

e (b

pm)

Time (min) Time (min)

Page 43: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Heart Rate Variability : 8 hours

Page 44: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Degree of Stationarity

Page 45: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Data White Noise

Page 46: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Data White Noise

Page 47: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Degree of stationary for nonlinear data

Inter- and intra-wave modulations

Page 48: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip Data

0 200 400 600 800 1000-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

Duffing Chip Data

Time: Sec

Am

pli

tud

e

Page 49: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip : Hilbert ZCHilbert Spectrum : Duffing Chip ZC

Time: Sec

Fre

qu

en

cy

: C

ycle

/Se

c

0 200 400 600 800 10000

0.01

0.02

0.03

0.04

0.05

Page 50: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip : Hilbert QuadHilbert Spectrum : Duffing Chip Quad

Time: Sec

Fre

qu

en

cy

: C

ycle

/Se

c

0 200 400 600 800 10000

0.01

0.02

0.03

0.04

0.05

Page 51: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip : Hilbert HilbertHilbert Spectrum : Duffing Chip Hilbert

Time: Sec

Fre

qu

en

cy

: C

ycle

/Se

c

0 200 400 600 800 10000

0.01

0.02

0.03

0.04

0.05

Page 52: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip : Degree of Stationarity

10-4

10-3

10-2

10-1

100

101

102

103

104

Frequency : Cycle/Sec

De

gre

e o

f S

tati

on

ari

ty

Degree of Stationary : Duffing Chip

QuadHilbertZC

Page 53: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip : Degree of Stationarity

0 200 400 600 800 1000 12000

0.01

0.02

0.03

0.04

0.05

0.06

Time: Sec

Fre

qu

en

cy

: C

yc

le/S

ec

Duffing Chip : Instantaneous Frequency

QuadHilbertZC

Page 54: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Duffing Chip : Normalized Intra-wave Modulation

0 200 400 600 800 1000 1200-1

-0.5

0

0.5

1

1.5

Time: Sec

No

rma

lize

d I

ntr

a-w

av

e M

od

ula

tio

n

Normalized Intra-wave Modulation

QuadHilbert

Page 55: Stationarity and Degree of Stationarity Norden Huang Research Center for Adaptive Data Analysis National Central University

Conclusions

• The high frequency range of the spectrum is highly intermittent.

• Even the Statistical Degree of Stationarity cannot smooth the variations.

• Before invoke the stationarity assumption, we should check the Degree of Stationarity.