Publications — NavalAI

Featured Paper International Parametric Yacht Analysis & Naval Architecture Review

From Principal Dimensions to Full Hydrodynamic Characterization: A Machine Learning Methodology

Poorya Khorsandy, Seyed Saeed Hayati

A cascading data-driven framework capable of mapping a minimal baseline array of three principal parameters (Length, Beam, and Draft) directly into a complete generation loop of over 30 geometric profiles and advanced residuary resistance constraints.

Abstract Extract

The preliminary design phase in naval architecture is characterized by a trade-off between rapid, low-fidelity empirical methods and time-consuming, high-fidelity physics-based simulations. This creates a significant gap for a tool that can facilitate comprehensive and rapid design space exploration. This study introduces a novel, data-driven framework designed to bridge this gap. By leveraging a foundational, public-domain dataset derived from systematic model testing, we develop a comprehensive, cascading predictive tool. The proposed framework is capable of generating a full suite of over 30 geometric and hydrodynamic parameters from a minimal set of three principal dimensions (Length, Beam, and Draft).

Cascade Architecture & Performance

Geometric properties are traced using layered regressors including CatBoost, LightGBM, and Random Forest matrices, while residuary resistance targets ($C_r$) are handled via high-fidelity XGBoost estimators tracking along ITTC-1957 standard formulations.

Residuary Resistance Fit ($R^2$ Tracking)

XGBoost

0.998

CatBoost

0.982

Empirical

0.745

Read Full Text → Download DOCX

Conference Paper Khorramshahr University of Marine Science

Ship Residuary Resistance Prediction with Machine Learning

Poorya Khorsandy, Seyed Saeed Hayati

An investigation into state-of-the-art machine learning algorithms applied to the Delft yacht hydrodynamics dataset to bypass resource-intensive CFD iterative loops and achieve instant regression maps.

Abstract Extract

This paper explores machine learning techniques as a cost-effective, accurate, and relatively fast alternative to traditional methods for yacht residuary resistance calculation. Instead of solving CFD equations for every individual ship geometry, we propose applying machine learning algorithms to historical experiment arrays.

Algorithmic Benchmarks

Using 308 full-scale matrix test runs mapping 22 different hull parameters against specific Froude numbers, several models were tested. The ensemble architectures proved highly capable of capturing complex boundary-layer coefficients automatically.

Error Distribution Metric (RMSE)

XGBoost

0.54

CatBoost

0.82

SVR Engine

10.62

Read Full Paper → View PDF

Research Article NavalAI Engineering Series

RAG-Marine: An Explainable Hybrid Retrieval-Augmented Generation Framework for Maritime Classification Rules

Poorya Khorsandy, Seyed Saeed Hayati

An explainable Retrieval-Augmented Generation (RAG) system customized for maritime rulebooks, merging hybrid semantic/lexical embeddings with an audit-ready clause attribution layer.

Abstract Extract

Maritime classification rules are extensive technical documents that govern ship design and safety compliance, often exceeding several thousand pages. This study introduces RAG-Marine, an explainable Retrieval-Augmented Generation framework that integrates dense, sparse, and hybrid retrieval models with an explainability layer (RAG-Explain) to provide engineers with regulation-grounded, citation-aware answers.

Performance Highlights

By shifting to an accuracy-first architecture using a hybrid-rank retrieval system (re-ranking shortlist candidates via cross-encoders), the platform maintains perfect structural integrity over math formulas, text chunks, and classification tables.

Relative Context Precision (CP)

Hybrid-Rank

High

Dense Only

Med

Base LLM

Low

Read Full Text → Download DOCX