Machine learning based method for analyzing vibration and noise in large cruise ships.

Area of Science:

• Marine engineering
• Acoustics
• Data science

Background:

• Cruise ship comfort is crucial for passenger satisfaction and profitability.
• Vibration and noise are key indicators of cruise ship comfort.
• Existing analysis methods for vibration and noise data have limitations.

Purpose of the Study:

• To develop a machine learning-driven approach for analyzing cruise ship vibration and noise data.
• To identify essential noise frequencies associated with different ship decks.
• To create a classifier for distinguishing noise samples.

Main Methods:

• Utilized machine learning algorithms for feature ranking, selection, and classification.
• Analyzed 127 noise samples from China's first large-scale cruise ship.
• Set up a classification task with decks as labels and frequencies as features.
• Employed Random Forest and LightGBM for classification and feature selection.

Main Results:

• Identified one or two essential noise frequencies for most decks, partly validated by statistical methods.
• Developed a Random Forest classifier using eight optimal frequency features identified by LightGBM.
• Achieved a Matthews correlation coefficient of 0.3415 for the noise sample classifier.

Conclusions:

• The machine learning approach provides insightful analysis of ship vibration and noise.
• Findings aid in reducing and controlling vibration and noise levels on cruise ships.
• This study offers a novel direction for investigating ship acoustics and comfort.