Empirical mode decomposition for noise detection and filtration of ice-induced load measurements

Structure-ice interaction is a complex process. The related failure process may include intermediate failures of ice, e.g. flacking, that decreases the load level temporarily before larger failures like bending failure occurs. Simultaneously, loads on a frame move and can reach a level of MN. The loads are commonly determined from the structural response with strain gauge measurements where the noise may be few kNs. As the changes in load level related to intermediate failures may be at the same level as noise, it is difficult to separate noise from the measurements. Empirical Mode Decomposition is a relatively new method in engineering that considers the measurement signal to be composed of a set of functions. The method parses the measurement data into Intrinsic Mode Functions that represent the true data and noise. Removing the modes related to noise enables the use of the method as a filter. This study applies the described filtration technique to the full-scale measurements conducted onboard S.A. Agulhas II in the Baltic Sea. The results demonstrate a good applicability of the method for the filtration of the data that can be applied in the load identification from the measurement data in the future studies.