Analysis of methods for determining ship speed during a sea trial

Abstract

The aim of this study is to evaluate the performance of two methods for determining the speed of a ship relative to the water during a so called speed trial.
The first method is called the Mean of means method and has been traditionally. The second is called the Iterative method and has been recently developed and is already used for speed trial analysis.
The methods have been evaluated by simulation.
A relationship between ship power and velocity was chosen based on a theoretical analysis, and this relationship is used to generate simulated speed trial data. Using the knowledge of the ’true’ relationship the performance of the methods could then be evaluated after implementation.
The findings are that the Iterative method needs a minimum of 8 measurements obtained by sailing back and forth 4 times in order to find the true relationship with acceptable
accuracy.
The Mean of means method needs a minimum of 12 measurements,
or 6 double runs.
The conclusion is that the Iterative method is more time efficient
for speed trial analysis.
In a test case assuming realistic measurement noise, 99:3% of found speeds had an error smaller than the simulated error margin of 0.1 knots, out of a sample size of 1000.
When comparing the Iterative to the Mean of means method under the same conditions, the Mean of means method made smaller errors for all observed cases.
From the results of this study has been concluded that the Iterative method is usable method for determining ship speed during a sea trial that has both advantages and disadvantages over the traditional Mean of means method.
Advantages are that the Iterative method requires less measurements than the Mean of means method, and it remains accurate even if measurements have a large or varying spacing in time.
A disadvantage is that it is more sensitive to measurement noise than the Mean of means method, making it prone to making larger errors.
A small adjustment to the current function the method assumes and uses for calculating current effects has been proposed.
The adjustment was to normalize the component of the current function that scales linearly with time.
This has been found to have a positive impact on the performance of the method.