ISMMACS conference on Mathematical Modelling and Computer Simulation,
Bangalore, November 14-15, 2002
ESTIMATION OF EXTREME VALUE STATISTICS OF A
NON-GAUSIAN PROCESS.
Rakesh C.R*, K. Balaji Rao**, M.B.Anoop**
*Junior Research fellow,**Scientist
Structural Engineering Research Centre, Tharamani,
Chennai, India -600113
Abstract
The probability distribution of response of non-linear system is non-Gaussian, even though
input is a Gaussian random process. A significant feature of a non-Gaussian random process
is that the shape of the peak is sharp as contrasted to the round shape of the troughs.
Responses of some of the components of the certain offshore systems (for example Tension
Leg platforms) can be highly non-Gaussian. For example the response of marine riser is non-
Gaussian due to highly non-linear behaviour of the system. The estimation of statistical
information of peak (or trough) is an important necessary step in the design of such non-
linear systems. This would help in calculation of design forces due to extreme wave action.
In literature, probability density function applicable for peaks, troughs and peak to trough
excursion of a non-Gaussian random process have been developed. It has been found that the
probability density function applicable to peaks (and troughs) is equivalent to the density
function of the envelope of a random process consisting of the sum of a narrow-band
Gaussian process and sine wave having the same frequency. To obtain the probability density
function, non-Gaussian process is expressed as a sum of a normal process and its square
multiplied by a constant. The magnitude of the constant represents this intensity of non-linear
characterstics. As the magnitude of the constant decreases, the density function of peaks (and
troughs) can be approximately represented in the form of a Rayleigh distribution. It is noted
that the actual form of pdf and other statistical properties have not been determined. In this
paper, an attempt has been made to deter mine these.
Considering the envelope process as a ergodic process, the autocorrelation function and the
power spectral density (PSD) of the process have been estimated. And it has been observed
that the plot of auto correlation function of peaks, troughs and peak to trough varies
approximately linearly with time lag. Also, from the power spectral density of the process it is
noted that the processes corresponding to peak, trough and peak to trough are narrow banded.
From the comparison of autocorrelation and PSD of the processes, it is felt that the envelope
process may be gamma.
To check whether the pdf of the above distribution can be fitted with gamma distribution,
K-S Goodness-of- fit test have been performed. It has been found that the pdf of peaks,
roughs and peak to trough pass the K-S test at 5% significance level. From the test results and
observations of autocorrelation function and PSD it has been concluded that the pdf of peak,
troughs and peak to trough excursion of a non-Gaussian random process will follow the
gamma distribution for different values of non-linearity parameter. Based on the results
obtained recommendations have been made for determination of statistical properties (viz.,
mean, cov) of extreme values of the process.
Area(s) for consideration: Structural Design and Ocean Structures, Ocean Engineering.
