2025_programme: An Investigation of Machine Learning Capabilities for Cavitation Detection

• Day: June 17, Tuesday
    Location / Time: D. CHLOE at 17:00-17:20
• Last minutes changes: -
• Session: 06. Enhancing underwater acoustic sensing through machine learning
  Organiser(s): Cheng Chi
  Chairperson(s): Peng Xiao
• Lecture: An Investigation of Machine Learning Capabilities for Cavitation Detection
  Paper ID: 2358
  Author(s): Dale Smith, Oscar Carter
  Presenter: Dale Smith
  Abstract: Due to environmental concerns, there is a growing need to reduce the radiated noise of marine traffic. The noise from marine vessels comprise multiple sources, with propeller cavitation typically being the dominant source when present. Cavitation can be avoided by operating the vessel at lower speeds. However, the speed at which cavitation occurs may not be explicit and may change over the life of the vessel. Therefore, in order to ensure noise levels are minimised, there is a need to detect when cavitation occurs and take the necessary corrective action. In this contribution, the capability of machine learning in the detection of propeller cavitation is evaluated. Using time series signals as the input feature, the performance of the model is evaluated. A range of signal parameters including duration, sampling rate and signal to noise ratio are investigated to more fully understand the capabilities of the model and to determine if any signal conditioning is required to increase performance. A number of signal statistics are also evaluated to identify additional features that may accurately characterise the presence of cavitation. The present work enables a rapid detection of cavitation and may allow for vessels to avoid this noisy operating state and hence reduce the impact on the marine environment.
• Corresponding author: Dr Dale Smith
  Affiliation: QinetiQ
  Country: United Kingdom