This paper presents novel techniques to improve the efficiency of numerical models for time-domain motion simulations and to reduce the number of experiments by using Artificial Neural Network (ANN) methods. A 'Static ANN' is successfully applied as an efficient way of updating the coefficients of the time domain motion equations for the instantaneous underwater hullform. The hydrodynamic coefficients used for training the ANN are obtained from a three-dimensional (3-D) frequency domain solver. A 'Dynamic ANN' is also successfully employed for the direct identification of highly non-linear model test responses in a short period of time. These techniques are applied to simulate the vertical motion response of a concept trimaran frigate that is under investigation by the Ministry of Defence (MoD). The results of the simulations are compared with the measured responses from a 6m model trimaran tested in the Ocean Basin at the Defence Evaluation and Research Agency (DERA), Haslar. The positive impact of the ANN techniques on the time-domain simulation methods and model testing procedures are discussed.
Efficient techniques in time-domain motion simulation based on artifical neural network
Effiziente Verfahren zur Zeitbereichs-Bewegungssimulation auf Basis eines künstlichen neuronalen Netzes
1998
19 Seiten, 13 Bilder, 3 Tabellen, 18 Quellen
Aufsatz (Konferenz)
Englisch
Efficient Techniques in Time-Domain Motion Simulation Based on Artificial Neural Network
British Library Conference Proceedings | 1998
|Plant Growth Model Using Artifical Neural Networks
SAE Technical Papers | 1997
|NTRS | 1961
|Helicopter track and balance with artifical neural nets
Tema Archiv | 1992
|Modeling of Shock Absorber Behavior using Artifical Neural Networks
SAE Technical Papers | 1994
|