Design of Wave Event/Return Periods Flashcards
Stat basics
P(x) is the probability that X<=x
-X is a random variable, continuous in the case of hydrodynamics
-CDF
p(x) = dP/dx = PDF
-used to get moments and expected values
mj = integral from -ve to +ve infinity of x^j.p(x) dx
*m0 = 1
*m1 = x bar
*m2 = mean square of X
mu_j = integral from -ve to +ve infinity of (x - x bar)^j .p(x) dx
*mu0 = 1
*mu1 = 0
*mu2 = variance
skewness
-defines the symmetry around x bar
-lambda
kurtosis
-defines the peaked-ness
- beta
statistically stationary
X(t) but p(x:t) = p(x)
-means all the statistics from the a problem can be taken from one sample
*short term oceans states are statistically steady (20mins to 3hrs) but long term not due to storm development and decay
temporal description
LRWT
-m1 = eta bar = 0
-m2 - sum to infinity of a_m^2/2 –> distribution of the wave components E summed over the f range
probabilistic description
-for a measured eta(t),the time that eta approxiates eta_j is T(eta_j)
- pdf p(eta_j) is lim as T -> 0 of T(eta_j)/T, T being the total time of the measured sample
*approxiamtely gaussian or normal but taken as gaussian and independent of eta bar since it is zero and X=eta in this case
spectral description
mean of eta(t) square = sum of a_m^2/2, the sum of the variance of its wave components = sigma_eta^2
-total E is the sum of the E for each wave
-since variance is proportional to the avergae E per unit ____
S_etaeta(omega) = mean of eta(t) square /delta omega = omni directional spectral density function
E(omega) = rho.g.S_etaeta(omega) = wave E spectra
sigma_eta^2 = S_etaeta(omega) . delta omega = a_m^2/2 for small delta omega
- rearrange for eta(x,t)
spectral bandwidth
-epsilon
-gives the relative width of the spectrum
-most sea states are approximated as narrow banded
-defines the number of f involved
linear wave statistics
for random waves the periods are taken at zero up crossings and the wave heights as the difference between the max and the min surface elevation
-alternative is taking period based on local crests
as spectral bandwidth tends to 1, crests occur above and below the SWL (gaussian)
as spectral bandwidth tends to 0, crests occur above and troughs below the SWL (rayleigh)
-assumes linear waves
- rayleigh vs gaussian can be plotted on a Pp(eta/sigma) to eta/sigma graph
significant wave height
Hs = H_1/3
-the expected value for the highest 1/3 of the waves
-related to what is observed as the highest wave
second order wave stats
-no effect on H since the crests get higher while troughs get shallower
-rayleigh cant capture it because specrtal bandwidth doesnt tend to 0
Empirical fits
*forristall 1978:
- for H
- 2 parameter weibull fit
- based on measure hurricane data of the gulf of mexico, only 116 hrs
*forristall 2000:
-captures second order crest height change
-simulated a wide range of seas using second order theory
-2 parameter weibull as well
-captures directionality which can be crucial
spectral representation of directionality
treated as independent from frequency
S_etaeta(omega,t) = D(theta).S_etaeta(omega)
-same directional distribution applied to all f but incorrect due to variation in swell and wind waves
-hard to separate swell and wind waves
notes on spectral directionality solutions
Normal
-A is a normalising coefficient so integral from zero to 2pi of D(omega,t) = 1
-sigma_theta approximately 10 to 20 degrees for a north sea storm
Mitsuyasm:
- s=150 for small spread and 4 for large spread but typically 7 to 15 in the northern sea storms
-cos raised to the s instead of the 2s is for the distribution of variance/energy
General
-theta_m is the mean direction
-all the ones in the notes give similar solutions
effect of directionality
-max elevation @ a point is unchanged linearly but nonlinearly the wave slope changes
-important for structures with large or multiple legs
-unidirectional waves are long crested with constant heights
-multidirectional waves are short crested with varying length, lengths decrease with increased directionality
linear directionality solutions
eta(x,y,t) = sum over f, sum over dir. of am.cos(omega_m.t - (km in x).x - (km in y).y + alpha_m)
-wave number split into x/y components
-wave particle kinematics are effected
for u
- sum the linear solution over f/dir. and split km again
-in the cosh/sinh use km^2 = sum of the two km components, both squared
-still suffers from high f contamination
* velocity reduction factor avoids this, added to the unidirectional velocities to account for directionality, usually 0.8-0.9
-design assuming unidirectional is conservative for fixed structures but nit for floating ones since they are under rolling motion
-directional second order random wave theories exist as do full nonlinear directional wave models