Examining random uncertainty using a newly developed power spectrum based method

In towing tank facilities such as found at MARIN, a wide range of ship design performance tests can be carried out. For example, the ships resistance at full scale can be predicted from model tests in a towing tank. In order to generate reliable results, steady-state conditions during measurements are required. A system in steady state condition shows behavior that can be extrapolated into the future, i.e. its properties are unchanging in time. Measurement times in research facilities, whether in towing tanks or in computer simulations, are finite due to time and/or geometry constraints. The finite measurement time introduces a random uncertainty in the mean value of a measured signal since it is only an estimate of the (hypothetically) infinite process' mean [2]. This uncertainty becomes evident when performing repeat tests. Each repeated estimate for the mean will be slightly different from the other. However, all of them should be in the same region around the true/infinite process' mean. During resistance towing tests this type of uncertainty is often not negligible. As in most experiments the mean value of a measured signal is used for further analysis, it is important to identify the magnitude of the uncertainty.