Abstract
This paper proposes an artificial neural network system for regressive estimation of wave making resistance, which is significantly important in preliminary design of high-speed ship. The neural network can explicitly realize nonlinear mapping between hull form and wave making resistance.
The system is composed of two kinds of neural networks ; Estimating Net and Descriptive Net. The Estimating Net learns the relation between hull form parameters and wave making resistance coefficients from a number of model-resistance test data. Consequently, when Froude Number, principal particular ratios, and area curve parameters of a hull form are given, the Estimating Net provides the wave making resistance coefficient. The Descriptive Net learns the density distribution of the learned data points in the hull form parameters' space. It provides the information about the density of the learned data at the input point in the above.parameters' space.
In this paper, the test data of 62 models : Series 60, are used for the construction of the system. The learning is successful and the results of playback calculation show good agreement with the original test data. Some applicable cases for non-learned hull forms are also explained. It is shown that the accuracy of the estimation is in accordance with the output of the Descriptive Net.
When other model test data are available, it is easy to modify the constructed system, taking advantage of learning ability of neural networks.
