A Knowledge‐Based Computational Design Tool for Determining Preliminary Stability Particulars of Naval Ships

Abstract

The preliminary design stage of a ship constitutes an important base from which to develop the preliminary ship characteristics to be used as goals in the following stages of the ship design process. Thus, reliable and efficient design tools are required by ship designers to determine ship design particulars that can satisfy various performance measures of stability for safety requirements, as well as resistance, volume, and load capacity to set economic targets. In this study, a robust neural network (NN) structure is established and using principle design data from 22 naval ships (Bartholomew et al. 1992), a reliable design tool for ship designers for determining ship preliminary stability particulars and as well as load capacity is developed. In the NN structure the classical back‐propagation (CBA) (Rumelhart et al. 1986) and fast back‐propagation (FBA) (Karayiannis and Venetsanopoulas 1991, 1992, 1993) algorithms are employed. The user inputs a set of common ship hull form parameters, while the output from the NN consists of the transverse center of gravity above keel (KG), the transverse metacenter above keel (KM), and the transverse metacenter above center of buoyancy (BM). Further, the dead‐weight tonnage capacity (DWT) is also determined using the same NN structure. According to various tests of the NN structure it is demonstrated that the model is able to give highly sensitive outputs for naval vessels even using a more restricted input data set. Depending on the availability of the design data structure the proposed model can be reconstructed for other preliminary design objectives such as predicting the powering performance for different classes of ships.