WAFO - A MATLAB toolbox for random waves

and loads

Sofia Åberg

Mathematical Statistics
Centre for Mathematical Sciences
Lund University

Göteborg
August 15-19, 2005
Outline

Introduction

Overview
Random sea waves
Fatigue analysis
Extreme value analysis

Example from recent research

Environmental standards
Distribution of the maximum in bounded regions
rind
Results
WAFO briefly described

Wave Analysis for Fatigue and Oceanography

◮ Statistical analysis and simulation of random waves

◮ Calculation of theoretical distributions related to characteristic wave
parameters
◮ Applications to sea waves and fatigue analysis

Developed by the WAFO-group:

P.A. Brodtkorb, M. Frendahl, P. Johannesson, G. Lindgren, I. Rychlik, J.
Rydén, E. Sjö + others
WAFO - Philosophy

MAKE SCIENTIFIC COMPUTATIONS REPRODUCIBLE!!!

◮ Available free of charge at the Internet

◮ More than 250 routines organised in modules related to applications
◮ Easy to find routines, easy to add new ones
◮ Help pages in nice html-interface
◮ Tutorial with many examples

Important module: algorithms and code for generation of results in

selected articles

www.maths.lth.se/matstat/wafo/
Random sea waves

Modelling of sea waves as stationary transformed Gaussian processes.

◮ Extraction of wave characteristics from data

◮ Estimation of spectrum
◮ Spectral simulation
◮ Calculations of exact distributions for characteristic wave parameters
Example: Wave spectra

Estimation of spectrum from data.

>> Sest = dat2spec(data,200); >> wspecplot(Sest);

2
Spectral density
fp1 = 1.1 [rad/s]
0.18 fp2 = 0.58 [rad/s]
1.5

0.16

1
0.14
Surface elevation (m)

S(w) [m s / rad]
0.12
0.5

0.1

2
0
0.08

−0.5 0.06

0.04
−1
0.02

−1.5 0
0 50 100 150 200 250 0 1 2 3 4 5 6 7 8 9 10
Time (s)
Frequency [rad/s]
Joint distribution of wave characteristics

Definition of crest length and crest >> f = spec2thpdf(Sest,...);

amplitude.
Joint density of (Tc,Ac)
v=0

1.8 Level curves enclosing:

10
30
50
1.6 70
90
Ac
95
1.4 99

1.2

amplitude [m]
Tc
1

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6
period [s]
Routines related to random loads and fatigue

◮ Extraction of rainflow cycles from data

◮ Calculation of expected rainflow matrix
◮ Switching Markov loads
◮ Visualization of cycle counts etc.

Turning points in data Rainflow cycle counts Smoothed observed rainflow matrix
2
2 2

1.5 1.5 1.5

1 1 1
Surface elevation (m)

0.5 0.5
0.5
max

Max
0 0
0

−0.5 −0.5
−0.5
−1 −1
−1
−1.5 −1.5
−1.5
0 10 20 30 40 50 60 70 80
−2 −2
Time (s) −2 −1 0 1 2 −2 −1 0 1 2
min min
Extreme value analysis
WAFO contains a module for extreme value analysis. For example one
can
◮ estimate parameters in the Generalized Extreme Value distribution
and the Generalized Pareto distribution
◮ simulate from the GEV and GPD
◮ make probability and quantile plots

Gumbel Probability Plot Empirical and GEV estimated cdf (PWM method)
8 1

7 0.9

6 0.8

5 0.7
−log(−log(F))

4 0.6

F(x)
3 0.5

2 0.4

1 0.3

0 0.2

−1 0.1

−2 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
X x
Environmental standard for ozone

Ozone is an airpollutant which may cause severe lung damage. In the

U.S. the 1-hour air-quality standard for ozone is 0.12 ppm. If we get
concentrations above 0.18 ppm it is considered to be a serious violation
of the clean air act.

Can it be assured that people are not exposed to

dangerous concentrations even though the standard
is attained?
Experiment

Suppose that a measurement is obtained exactly at the standard level.

Define a region in the plane such that it consists of all points correlated
at least 0.7 with the monitoring site.

Given the observation, what is then the distribution of the maximum over
this region?
One dimensional counterpart
Let {X (t), t ∈ [0, T ]} be a random process. Then

P( max X (t) > u) = P(X (0) > u) + P(X (0) ≤ u, max X (t) > u)
t∈[0,T ] t∈[0,T ]

= P(X (0) > u) + P(X (0) ≤ u, NT+ (u) ≥ 1),

where NT+ (u) is the number of up-crossings of level u of the process

X (t) in the interval [0, T ].

first up−crossing of u

τ
Rice bounds

Classical Rice upper bound:

P(X (0) ≤ u, NT+ (u) > 0) ≤ E (NT+ (u))

Z T
= E (X ′ (t)+ |X (t) = u)fX (t) (u)dt
0

Let τ be the first time the process crosses level u. Then by using the
first passage density

P(X (0) ≤ u, max X (t) > u) = P(τ ∈ [0, T ])

t∈[0,T ]
Z T
= E (X ′ (t)+ {X (s) < u, ∀s < t}|X (t) = u)fX (t) (u)dt
0
Two dimensions

Let W (x), x ∈ R2 be a random field and let S be a bounded region in R2

with boundary ∂S. Then

P(max W (x) > u)

x∈S
= P(max W (x) > u) + P(max W (x) ≤ u, max W (x) > u).
x∈∂S x∈∂S x∈S

Needs an analogue to the up-crossings in one dimension!!

Analogue to up-crossings

Use up-crossings in the x-direction.

These points satisfy

W (x) = u, W01 (x) = 0, W02 < 0, W10 > 0

Denote the number of such points by NS (u).

Upper bound in two dimensions

As in one dimension bound the probability that the number of such

points is greater than one with an expectation. This gives, after some
calculations :-)

P(max W (x) ≤ u, max W (x) > u) = P(max W (x) ≤ u, NS (u) ≥ 1)

x∈∂S x∈S x∈∂S
Z
+
≤ E (W02 (x)− W10 {W (s) ≤ u, ∀s ∈ ∂S}|W (x) = u, W01 (x) = 0)fW (x),W01 (x) (u, 0)dx
S
rind

To compute an upper bound for P(maxx∈S W (x) > u) we need to

compute multivariate normal expectations of the form

E (|Xd(1) · . . . · Xd(Nd)|{ai < Xd(i) < bi , cj < Xt(j) < dj }|Xc = xc) fXc (xc)

The WAFO-function rind is custom made for these type of calculations

under the Gaussian assumption!! The input is just the mean and
covariance matrix of the variables.
Ozone example revisited

Model and assumptions

◮ ξ(x) square root of true underlying ozone field. Assumed to be
Gaussian.
◮ observation z(x0 ) = ξ(x0 ) + ǫ(x0 )
◮ conditional process W (x) = {ξ(x)|ξ(x0 ) + ǫ(x0 ) = z(x0 )}
◮ E (ξ(x)) = 0.235, Var (ξ(x)) = 0.0642, Var (ǫ(x0 )) = 0.0322
◮ Covariance of ξ(x) is of the squared exponential type
Intensities computed by rind

Intensity on the boundary: Intensity within the region:

0.45 0.5
0.9

0.4 0.4
0.8

0.35 0.3
0.7

0.3 0.2
0.6
Intensity

0.1
0.25
0.5
0
0.2
0.4
−0.1
0.15
0.3
−0.2
0.1

−0.3 0.2
0.05

−0.4 0.1
0
0 1 2 3 4 5 6 7
θ −0.5 0
−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5
Distribution of the maximum

Ditribution of the maximum in the 0.7-correlation region on a linear and

log-scale respectively:

0.7 0
10

0.6 −2
10

0.5
−4
10
P(M (W)>u)

P(MS(W)>u)
0.4
−6
10
S

0.3

−8
10
0.2

−10
10
0.1

−12
0 10
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
u (ppm) u (ppm)

Note that there is a 5% risk of serious violation!!

Reminder!

The homepage of WAFO on the internet

http://www.maths.lth.se/matstat/wafo/

Versions to download are available for Windows and Unix.