Uncertainty analysis of ship powering prediction methods using Monte Carlo simulation

Model towing tank testing is used to predict the powering required for full-scale vessels. The International Towing Tank Conference of 1978 developed a method that has been in use in its original or modified form in many institutions internationally. Originally the method was solved graphically, utilizing the experience of the analyst, however the use of computers and the automation of the method has impacted the reliability of this approach. Uncertainty analysis has up to this point focused on potential errors in the model data from the tests required to produce the data that is subsequently extrapolated using prediction methods. The overall sensitivity of predicted power from this and other prediction methods to variations in inputs from the model tests and from elements such as the frictional resistance coefficient, form factor or correlation allowance needs to be determined in order to properly interpret the results of an automated analysis. Rather than setting up a series of data reduction equations, a Monte Carlo simulation was used and the entire method was used as the data reduction equation in the uncertainty analysis. The levels of uncertainty in ship powering were obtained for assumed values of uncertainty in the experimental values from tests and for estimated uncertainty in the friction coefficients, wake fraction and wake scaling, thrust deduction fraction, form factor and correlation allowance. Several different extrapolation methods were studied to assess variation in powering prediction methods resulting from variation in input data and from the variation of the details of the extrapolation. These consisted of the ITTC 1978 method, variation that included extrapolation with and without a form factor, different friction lines and correlation allowances, different wake scale and thrust deduction fraction values, and a method that extrapolated from self-propulsion test data only analysed using the same procedure. Uncertainties in the predicted powering results of both methods were compared. Methods of reducing the uncertainty in the predicted power from the ITTC 1978 method were proposed