Elsevier

Ocean Engineering

Volume 31, Issues 5–6, April 2004, Pages 761-777
Ocean Engineering

Technical note
Design of a robust neural network structure for determining initial stability particulars of fishing vessels

https://doi.org/10.1016/j.oceaneng.2003.08.002Get rights and content

Abstract

Stability problem is a vital issue as the total measure of the ship safety. Designers need to use reliable design tools for the definition of stability parameters during the preliminary design stage of ships. These tools are mostly built in the form of approximate expressions with some error level. In this study, a functional and reliable tool is proposed to ship designers for determining initial stability particulars of fishing vessels. It uses a robust neural network (NN) structure with different algorithms based on two fishing vessel databases containing the hull geometry and stability related parameters. The initial stability particulars of fishing vessels are almost exactly determined for an input set of ship data. With this method, using some sample ship data, the vertical center of gravity (KG), height of transverse metacenter above keel (KM) and vertical center of buoyancy (KB) are easily calculated. As a result, the designer can calculate transverse metacentric height (GM) and investigate a possible set of ship parameters affecting the ship’s intact stability.

Introduction

The importance of stability on the ship safety is evident. During the preliminary design stage of ships, the designers should have reliable and preferably practical design tools for defining various parameters of the ship. The stability parameters constitute a fundamental constraint for the preliminary design of ocean vehicles. The current literature proposing design tools for the definition of ship stability parameters mostly depend on approximate expressions based on regression analysis and therefore high error rates are encountered. As an alternative neural networks (NN) have been increasingly utilized in many disciplines as well as in ship design problems. One of the advantages of NN process is the capability of solving non-linear problems in which the convergence cannot be provided using linear approaches. In this study, two fishing vessel databases, previously used in a regression analysis based study (Yılmaz and Kükner, 1999), are taken into consideration within an NN structure for the purpose of developing a design tool identifying the initial stability parameters. It uses a robust NN structure with different algorithms based on the stability related parameters of the fishing vessel data. The initial stability particulars of fishing vessels are almost exactly determined. In the proposed method, the vertical center of gravity (KG), height of transverse metacenter above keel (KM) and vertical center of buoyancy (KB) values are calculated easily within high accuracy levels as compared to the actual ship data. As a result, the designer can calculate initial metacentric height (GM) which will help to define the best possible set of ship parameters affecting initial stability. Thus, a functional and accurate tool in determining initial stability particulars of fishing vessels is proposed to ship designers.

Section snippets

Neural networks (NN)

NNs are successfully used in areas such as control (Miller et al., 1995), early detection of machine faults as well as design problems including marine vehicles (Raya et al., 1996), the identification of ship coupled heave–pitch motions (Haddara and Jinsong, 1998) and the detection of the extent and location of ship damage (Zubaydi et al., 2002). The feed-forward NN is usually trained by a back-propagation training algorithm first proposed by Rumelhart, Hinton, and Williams in 1986. This was

Results and discussions

In the evaluation of the application results, some test vessels are taken into consideration. In Table 3, geometric properties of the test vessels belonging to Database-I are given to make comparisons between and NN computed values of the CBA and FBA algorithms. The comparison of computed KG values, as outputs of the CBA and FBA algorithms, are made with the actual values as shown in Table 4. The maximum absolute error rates in the KG computation of the test vessels are 1.25% for the CBA and

Conclusions

The proposed method of robust NN structures is given for the initial stability particulars (KG, GM) of fishing vessels having a range of small and large sizes. It enables the designer in the preliminary design stage to obtain vessel stability particulars and gives an almost exact idea for determining design parameters affecting stability. It can be easily seen from the comparison diagrams of actual and NN results in Fig. 2, Fig. 3, Fig. 8, Fig. 9.

Although CBA gives the output signal result in

References (9)

There are more references available in the full text version of this article.

Cited by (26)

  • Prediction of main particulars of container ships using artificial intelligence algorithms

    2022, Ocean Engineering
    Citation Excerpt :

    Identification of the damage has given a successful result while for the other two parameters that were sought which were extent and location more work is needed. Alkan et al. (2004), calculated the initial stability parameters of fishing vessels for which they developed two different ANN. Neural networks were developed based on two ship databases in order to estimate the first ANN’s vertical center of gravity and the second ANN the height of the transverse metacenter above the keel and the vertical center of buoyancy.

  • Determination of design formulas for container ships at the preliminary design stage using artificial neural network and multiple nonlinear regression

    2021, Ocean Engineering
    Citation Excerpt :

    In recent years, artificial neural networks have been used for ship design regression analysis. For example, Alkan et al. (2004) calculated the stability parameters of the fishing vessel through the use of neural networks. An Artificial neural network was developed to estimate the vertical centers of gravity and buoyancy and the transverse metacenter height using sample ship data.

  • An estimation of motor yacht light displacement based on design parameters using computational intelligence techniques

    2021, Ocean Engineering
    Citation Excerpt :

    Often in shipbuilding, several computational intelligent methods are used to estimate various ship characteristics. Alkan et al. (2004) used several algorithms to calculate fishing vessel stability parameters. Abramowski (2013) developed a model for the determination of effective ship power through the use of neural networks.

View all citing articles on Scopus
View full text