Comparison of Different Models for Residuary Resistance Prediction

Abstract

The paper presents several unconventional models of residuary resistance based on fuzzy logic and neural network techniques. First, two fuzzy models are built based on different hull parameters and different Froude numbers. These models are identified by a modification of Sugeno and Yasukawa identification algorithm. Next, a neuro-fuzzy model of residuary resistance is build, based on statistical learning theory. The model presents a fuzzy inference system of Takagi and Sugeno type that uses an extended relevance vector machine for learning its parameters and number of fuzzy rules. Finally, a neural network approach is applied to build four different models of residuary resistance. Two of the neural models apply classic extreme learning machine, and the other two implement incremental extreme learning machine philosophies. The obtained models are validated for their generalization and approximation performance, and although they all possess excellent approximation capabilities, our neural models based on extreme learning machine have shown the best simulation results.

Data-Driven Models for Yacht Hull Resistance Optimization: Exploring Geometric Parameters Beyond the Boundaries of the Delft Systematic Yacht Hull Series

“…In this section we will describe the data that we will exploit in this study. In particular, we leverage the DSYHS database [44] (available upon request to the Delft University of Technology Ship Hydromechanics Laboratory 3 ) which has been used in a number of works [26]- [29], [45].…”

“…In this setting, using CFD results is impractical due to its computational requirements [8], [9], [25]. For this reason, recently, Data-Driven Models (DDMs) are attracting the attention of the industry and academia for their ability to accurately surrogate complex experimental (e.g., EFD) [26]- [29] or numerical (e.g., CFD) [1], [14], [16], [17], [19], [21], [22] procedures based on a historical collection of their inputs and outputs, with a function that is computationally expensive to construct but computationally inexpensive to use. Consequently, DDMs can be included directly both in a human-driven optimization loop reducing the computational requirements (i.e., time) between design iterations or developing a fully automated optimization loop requiring minimal human intervention, enabling the exploration of a wider design space [2], [30].…”

“…Then, in order to further minimize the human intervention in the parametrization phase, we decoupled the parametrization exploited to define the geometry and to define the optimization parameters from the features necessary to predict the hull resistance based on DDMs. In particular, from a hull geometry defined by a particular configuration of parameters, we exploit the Nautilus code 1 which is able to automatically extract a series of features able to cover and extrapolate over multiple parent hulls while being informative enough to allow for effective and physically plausible predictions of the resistance associated with a particular hull [26]- [29]. This decoupling is a fundamental and key contribution to our approach.…”