In the paper the method of signal processing is offered. It was developed specially for processing of ice loads observations made in ice basin. The proposed method includes numerical search of the signal oscillatory components parameters by minimization of these components covariance and model discrepancy variance. In the method the clustering procedure is used for the final frequencies estimation. Uncorrelated random processes which parameters are estimated this way appear to be useful to describe the greater part of initial signal variance. The components of particular interest are stationary stochastic processes with determined frequency and random phase. The discrepancy for such model is in general a non-stationary process, but with variance much less then variance of initial signal. The characteristic of this method is that it uses not large time intervals, for example 1 second. The selection of time interval length depends on existing noise characteristics. In the paper the offered method is applied to explain the ice loads measurements that were got in experiments with tanker and ice-breaker hulls in ice basin. In the last section of the paper some statistics values for estimated parameters and ice load signal discrepancy are given. Comparing to spectral investigation, this method do not use necessarily orthogonal trigonometric functions to describe oscillatory components. But at the same time numerically found frequencies conform to the general spectral investigation results.
展开▼
