The location of trimaran side-hulls (amas) plays an important role in the wave-making resistance of the vessel. Constructive and destructive wave interference occurs between the hulls that either augments or reduces the wave making resistance of the vessel. In a FAST 2011 paper (Royce et al, 2011) the author presented an Artificial Neural Network (ANN) model for trimaran resistance based on numerous tests conducted at Webb Institute’s Robinson Model Basin over a two year period. The speed range covers Froude numbers from 0.12 – 0.5. This data is thought to be one of the most comprehensive sets of test data on side-hull placement for a single model.
In order to further evaluate the validity of the trimaran ANN, contour plots of the resistance were developed. The global minimum and maximums were determined for the domain covered by the model at a Froude number of 0.35. Experiments were then conducted at these two locations to compare with the predictions from the empirical model. The results of the experiments show that the ANN model tends to under-predict the resistance for the best case and over-predicts the worst case. This work demonstrates importance of the number of hidden neurons, much like the order of polynomials used in least square regressions.
