Add new post: Publication on LNG and LPG ship design using statistical and machine learning tools

Author: Daniel Precioso, PhD

_posts/2025-10-01-lng-lpg-design.md

@@ -0,0 +1,33 @@
+---
+title: "New publication on the design of LNG and LPG ships"
+date: 2025-10-15
+permalink: /news/2025/lng-lpg-design
+excerpt: "The Weather Navigation team publishes a new study on statistical and machine learning tools for the preliminary design of LNG and LPG carriers, based on a large-scale industrial dataset."
+featured_image: "https://danielprecioso.com/images/2025-10-15-marine-structures.jpg"
+tags:
+  - publications
+  - naval architecture
+  - data science
+  - machine learning
+  - marine structures
+---
+
+A new research paper co-authored by members of the Weather Navigation team has been published in *Marine Structures* (Elsevier, Vol. 104, October 2025). The study addresses a key challenge in modern ship design: how to improve the **preliminary design process for LNG and LPG ships** by leveraging statistical models and machine learning.
+
+The paper, authored by researchers from **Universidad de Cádiz** and **IE University**, analyzes a unique industrial dataset of 467 gas carriers built between 1979 and 2023 by Hyundai Heavy Industries. The database includes 145 LNG and 322 LPG ships, representing roughly **20% of the global fleet**, and offers a rare empirical foundation to explore the complex relationships between ship dimensions, deadweight, speed, and hull ratios.
+
+<img src="{{ page.featured_image }}" alt="LNG/LPG design publication" width="100%"/>
+
+The work explores how **simple and multiple regression analyses**, using functional forms such as linear, exponential, and logarithmic regressions, can be used to derive predictive formulas for key ship characteristics. These are benchmarked against **machine learning algorithms**, which show promise in capturing hidden, non-linear dependencies in the data, though with less interpretability compared to traditional methods.
+
+From a practical standpoint, the results provide **designers and engineers with updated predictive tools** for estimating main ship dimensions in the early design stage. This is especially valuable in an industry where prototypes are not constructed, and accurate early estimates are critical to avoid costly redesigns. The models help optimize design choices based on known constraints, such as port limits, energy consumption, and regulatory frameworks.
+
+This publication also contributes to the scientific effort to support **low-emission maritime transport**, given the rising role of LNG and LPG as transitional fuels. The improved design tools can guide more energy-efficient and sustainable vessel construction.
+
+<a href="https://doi.org/10.1016/j.marstruc.2025.103863" class="button">Read the new paper here</a>
+
+This research is supported by:
+
+- [BBVA Foundation](https://www.fbbva.es/) via the project "Mathematical optimization for a more efficient, safer and decarbonized maritime transport".
+
+- Spanish [Agencia Estatal de Investigación](https://www.aei.gob.es/) under grant  TED2021-129455B-I00, "Optimization of maritime routes with real time oceanographic and meteorological data".