To search for the extreme value of a measured quantity from the peak analysis of a time series it is common practice in the field of offshore engineering to perform the analysis by means of a Weibull distribution. Usually the peaks distribution is quite well described by a common two or at most three parameters Weibull. However, when severe sea states are investigated or the analyzed structure presents particular geometries, the representation on the Weibull plane can be nonlinear, even adopting a three parameters distribution. In such a case the standard Weibull distributions are no more suitable to describe the peaks population, which presents a multi-modal form. To overcome this problem, the Mixed Weibull distribution was here used to describe the populations and a procedure based on genetic algorithms has been established to fit the parameters of the multi-modal function. Thereafter the extreme value predictions by the developed procedure have been compared with the one coming from standard Weibull analysis.
