The installation of subsea cables for electrical interconnections and data transmission infrastructure is essential to ensuring both energy-system stability and efficiency and reliable telecommunications. Cable-laying operations are carried out using subsea ploughs, which excavate a trench into which the cable is laid directly on the seabed. The correct design of the towing process requires an estimate of the force needed to move the plough at the required rate. This force depends on several factors, including plough type, the hydro-mechanical properties of the soil, installation depth, and ploughing speed. Accurately predicting the required towing force for the effective design of installation operations is therefore a highly complex task. Recent studies have shown that existing methods for estimating towing force, generally based on analytical models supplemented by empirical correction factors, have limited predictive capability when applied to conditions different from those for which they were calibrated. The use of Machine Learning techniques, such as Support Vector Machines (SVM), offers a means of overcoming some of the limitations inherent in traditional approaches. In this paper, the authors assess the predictive potential of this approach for estimating towing force. The model is trained and validated using site data for granular soils, relying exclusively on installation depth, ploughing speed, and information obtained from standard site investigations, including grain-size characteristics and Cone Penetration Test (CPT) results.
L'installazione di cavi sottomarini per garantire il collegamento elettrico e per le infrastrutture di traffico dati è un processo fondamentale per assicurare non solo stabilità ed efficienza energetica, ma anche nel campo delle telecomunicazioni. Le operazioni di posa dei cavi avvengono per mezzo di aratri sottomarini che realizzano una trincea all’interno della quale viene adagiato il cavo, direttamente sul fondale. Per progettare correttamente il processo di traino è necessario stimare la forza necessaria per imporre il moto richiesto all’aratro, che dipende da vari fattori come la tipologia di aratro, le caratteristiche idro-meccaniche del terreno, la profondità di installazione e la velocità di avanzamento dell’aratro stesso: stimare accuratamente tale forza per la progettazione efficace delle operazioni è dunque un’operazione molto complessa. Recentemente è stato mostrato come i metodi disponibili per stimare la forza di traino basati su modelli analitici, con fattori di correzione empirici, hanno scarse capacità predittive se applicate a contesti differenti rispetto a quello di calibrazione. L'uso di tecniche di Machine Learning, quali il Support Vector Machine, permette di superare le limitazioni tipiche degli approcci tradizionali. In questa nota, gli autori valutano potenzialità predittive in termini di forza di traino di questo approccio, addestrato e validato attraverso dati di sito riguardanti terreni granulari, comprendenti esclusivamente la profondità di installazione, la velocità dell'aratro e dati provenienti dalla caratterizzazione standard del sito, quali le caratteristiche granulometriche e risultati di prove CPT.
Un approccio basato sull’Intelligenza Artificiale a supporto dell’installazione di cavi sottomarini
P. Marveggio;G. Della Vecchia
2025-01-01
Abstract
The installation of subsea cables for electrical interconnections and data transmission infrastructure is essential to ensuring both energy-system stability and efficiency and reliable telecommunications. Cable-laying operations are carried out using subsea ploughs, which excavate a trench into which the cable is laid directly on the seabed. The correct design of the towing process requires an estimate of the force needed to move the plough at the required rate. This force depends on several factors, including plough type, the hydro-mechanical properties of the soil, installation depth, and ploughing speed. Accurately predicting the required towing force for the effective design of installation operations is therefore a highly complex task. Recent studies have shown that existing methods for estimating towing force, generally based on analytical models supplemented by empirical correction factors, have limited predictive capability when applied to conditions different from those for which they were calibrated. The use of Machine Learning techniques, such as Support Vector Machines (SVM), offers a means of overcoming some of the limitations inherent in traditional approaches. In this paper, the authors assess the predictive potential of this approach for estimating towing force. The model is trained and validated using site data for granular soils, relying exclusively on installation depth, ploughing speed, and information obtained from standard site investigations, including grain-size characteristics and Cone Penetration Test (CPT) results.| File | Dimensione | Formato | |
|---|---|---|---|
|
1188-Malo_Romero_b.pdf
accesso aperto
:
Publisher’s version
Dimensione
1.23 MB
Formato
Adobe PDF
|
1.23 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.



