Ship model propulsion experiments analysis and random uncertainty. Discussion
暂无分享,去创建一个
In this paper two methods of evaluating ship model propulsion experiments using the Load Varying Method are described. They give model results that are quite close, but the nonlinear method is preferred and it has the advantage that it can provide the random uncertainty as well. The linear method, on the other hand, provides criteria or data consistency checks that the data should satisfy before it is accepted for analysis and prediction purposes; these checks are not comprehensive, however. The data used was obtained from the records of past work at the BMT Group Ltd, Vessel Hydrodynamics (formerly NMI and Ship Division, NPL) in the course of conducting good quality commercial tests. The sample size was reasonably large, involving 632 tank runs, covering 13 vessels of a variety of types, and 29 vessel conditions. Problems encountered in propulsion testing are discussed and propeller rotational speed is identified as the variable with the least random scatter present. The scatter is so low in comparison with the scatter in the other variables, excluding model speed, that the rotational speed can be assumed to be free from random error. The random error or uncertainty in the measurement of the towing force, on the jointly propelled and towed model, propagates throughout the analysis, and is central to the evaluation of the random uncertainties in the other variables and the derived results. Thus, models that have high scatter in the towing force have high random uncertainty in all the results. Large heavy models towed with conventional resistance dynamometers are susceptible to longitudinal model oscillation contaminating the towing force measurement. This problem is as old as model testing itself, but the development of ships has heightened it and is the reason for the widespread popularity in using the Load Varying Method of testing as a means of reducing uncertainty. In choosing the model size for a test, all the relevant factors should be considered if random uncertainty is to be minimised and not just the attainment of Reynolds numbers well in excess of threshold values. Large models can introduce greater uncertainty of measurement, thereby defeating the objective in using them. It appears that the random uncertainty in the model propulsion results produced at BMT is low, and probably of much less significance than the random and bias uncertainties introduced when performing full scale power-speed predictions from model results.