Extrapolation of load spectra by optimized Kernel Density Estimators using Monte-Carlo-Simulation

Abstract In the present article a method based on Monte Carlo Simulation (MCS) is developed with which any spectrum extrapolation method can be statistically evaluated. Thereby, the duration of service life spectrum and the extrapolation factor can be specified arbitrarily. In addition to the evaluation of extrapolation methods, the simulation model allows to calculate statistically based safety factors for load assumption. The use of general, mostly conservative factors can thus be omitted. For the extrapolation of load spectra, a procedure with kernel density estimators using an adaptive bandwidth is developed. The new Kernel Density Estimation (KDE) method contains three parameters for controlling the adaptive bandwidth which are optimized using the simulation model. During optimization, mathematically describable spectra are used, which are completely described by the maximum amplitude of the spectrum, the spectrum range and the spectrum shape parameter.