Diseño de buques adaptado al comportamiento en la mar mediante inteligencia artificial

Resumen

Artificial intelligence (AI) is a field that has been developed over the past few decades, with an emphasis on intelligent systems for handwriting recognition, voice systems, autonomous systems, and so on. In the realm of naval architecture, the application of AI can have significant relevance, especially as predictive tools to assess the degree of compliance with design expectations. This Doctoral Thesis focuses on the application of machine learning (ML) tools to evaluate a specific aspect of naval architecture, such as the seakeeping of ships. For this, algorithms based on artificial neural networks (ANN) have been developed. Furthermore, these algorithms do not require the exact geometry of the hull, as they only depend on a limited number of dimensionless coefficients of the ship’s forms. This Doctoral Thesis will present the methodology used to obtain the best ANN architecture, generate the database of ships used for training, and for the subsequent data processing that allows the correct prediction of the objectives. A 3D potential flow has been used to solve the wave radiation-diffraction problem using the boundary element method (BEM), obtaining the added mass and damping matrices, and excitation forces for different wave directions and speeds. The main result of this Thesis is the development of an ANN capable of instantly predicting seakeeping with accuracy similar to that of BEM codes. In addition, two application cases of the algorithms developed within the framework of the Thesis are presented. In both application cases, the advantages of using AI over traditional methods to analyze seakeeping can be clearly appreciated, since it would not be possible to carry them out with the latter due to the enourmous computational resources required.